diff options
Diffstat (limited to 'drivers/soc/fsl')
44 files changed, 16293 insertions, 0 deletions
diff --git a/drivers/soc/fsl/Kconfig b/drivers/soc/fsl/Kconfig new file mode 100644 index 000000000..fcec6ed83 --- /dev/null +++ b/drivers/soc/fsl/Kconfig @@ -0,0 +1,55 @@ +# SPDX-License-Identifier: GPL-2.0-only +# +# NXP/Freescale QorIQ series SOC drivers +# + +menu "NXP/Freescale QorIQ SoC drivers" + +source "drivers/soc/fsl/qbman/Kconfig" +source "drivers/soc/fsl/qe/Kconfig" + +config FSL_GUTS + bool + select SOC_BUS + help + The global utilities block controls power management, I/O device + enabling, power-onreset(POR) configuration monitoring, alternate + function selection for multiplexed signals,and clock control. + This driver is to manage and access global utilities block. + Initially only reading SVR and registering soc device are supported. + Other guts accesses, such as reading RCW, should eventually be moved + into this driver as well. + +config FSL_MC_DPIO + tristate "QorIQ DPAA2 DPIO driver" + depends on FSL_MC_BUS + select SOC_BUS + select FSL_GUTS + select DIMLIB + help + Driver for the DPAA2 DPIO object. A DPIO provides queue and + buffer management facilities for software to interact with + other DPAA2 objects. This driver does not expose the DPIO + objects individually, but groups them under a service layer + API. + +config DPAA2_CONSOLE + tristate "QorIQ DPAA2 console driver" + depends on OF && (ARCH_LAYERSCAPE || COMPILE_TEST) + default y + help + Console driver for DPAA2 platforms. Exports 2 char devices, + /dev/dpaa2_mc_console and /dev/dpaa2_aiop_console, + which can be used to dump the Management Complex and AIOP + firmware logs. + +config FSL_RCPM + bool "Freescale RCPM support" + depends on PM_SLEEP && (ARM || ARM64) + help + The NXP QorIQ Processors based on ARM Core have RCPM module + (Run Control and Power Management), which performs all device-level + tasks associated with power management, such as wakeup source control. + Note that currently this driver will not support PowerPC based + QorIQ processor. +endmenu diff --git a/drivers/soc/fsl/Makefile b/drivers/soc/fsl/Makefile new file mode 100644 index 000000000..906f1cd8a --- /dev/null +++ b/drivers/soc/fsl/Makefile @@ -0,0 +1,12 @@ +# SPDX-License-Identifier: GPL-2.0-only +# +# Makefile for the Linux Kernel SOC fsl specific device drivers +# + +obj-$(CONFIG_FSL_DPAA) += qbman/ +obj-$(CONFIG_QUICC_ENGINE) += qe/ +obj-$(CONFIG_CPM) += qe/ +obj-$(CONFIG_FSL_RCPM) += rcpm.o +obj-$(CONFIG_FSL_GUTS) += guts.o +obj-$(CONFIG_FSL_MC_DPIO) += dpio/ +obj-$(CONFIG_DPAA2_CONSOLE) += dpaa2-console.o diff --git a/drivers/soc/fsl/dpaa2-console.c b/drivers/soc/fsl/dpaa2-console.c new file mode 100644 index 000000000..53917410f --- /dev/null +++ b/drivers/soc/fsl/dpaa2-console.c @@ -0,0 +1,330 @@ +// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) +/* + * Freescale DPAA2 Platforms Console Driver + * + * Copyright 2015-2016 Freescale Semiconductor Inc. + * Copyright 2018 NXP + */ + +#define pr_fmt(fmt) "dpaa2-console: " fmt + +#include <linux/module.h> +#include <linux/of_device.h> +#include <linux/of_address.h> +#include <linux/miscdevice.h> +#include <linux/uaccess.h> +#include <linux/slab.h> +#include <linux/fs.h> +#include <linux/io.h> + +/* MC firmware base low/high registers indexes */ +#define MCFBALR_OFFSET 0 +#define MCFBAHR_OFFSET 1 + +/* Bit masks used to get the most/least significant part of the MC base addr */ +#define MC_FW_ADDR_MASK_HIGH 0x1FFFF +#define MC_FW_ADDR_MASK_LOW 0xE0000000 + +#define MC_BUFFER_OFFSET 0x01000000 +#define MC_BUFFER_SIZE (1024 * 1024 * 16) +#define MC_OFFSET_DELTA MC_BUFFER_OFFSET + +#define AIOP_BUFFER_OFFSET 0x06000000 +#define AIOP_BUFFER_SIZE (1024 * 1024 * 16) +#define AIOP_OFFSET_DELTA 0 + +#define LOG_HEADER_FLAG_BUFFER_WRAPAROUND 0x80000000 +#define LAST_BYTE(a) ((a) & ~(LOG_HEADER_FLAG_BUFFER_WRAPAROUND)) + +/* MC and AIOP Magic words */ +#define MAGIC_MC 0x4d430100 +#define MAGIC_AIOP 0x41494F50 + +struct log_header { + __le32 magic_word; + char reserved[4]; + __le32 buf_start; + __le32 buf_length; + __le32 last_byte; +}; + +struct console_data { + void __iomem *map_addr; + struct log_header __iomem *hdr; + void __iomem *start_addr; + void __iomem *end_addr; + void __iomem *end_of_data; + void __iomem *cur_ptr; +}; + +static struct resource mc_base_addr; + +static inline void adjust_end(struct console_data *cd) +{ + u32 last_byte = readl(&cd->hdr->last_byte); + + cd->end_of_data = cd->start_addr + LAST_BYTE(last_byte); +} + +static u64 get_mc_fw_base_address(void) +{ + u64 mcfwbase = 0ULL; + u32 __iomem *mcfbaregs; + + mcfbaregs = ioremap(mc_base_addr.start, resource_size(&mc_base_addr)); + if (!mcfbaregs) { + pr_err("could not map MC Firmware Base registers\n"); + return 0; + } + + mcfwbase = readl(mcfbaregs + MCFBAHR_OFFSET) & + MC_FW_ADDR_MASK_HIGH; + mcfwbase <<= 32; + mcfwbase |= readl(mcfbaregs + MCFBALR_OFFSET) & MC_FW_ADDR_MASK_LOW; + iounmap(mcfbaregs); + + pr_debug("MC base address at 0x%016llx\n", mcfwbase); + return mcfwbase; +} + +static ssize_t dpaa2_console_size(struct console_data *cd) +{ + ssize_t size; + + if (cd->cur_ptr <= cd->end_of_data) + size = cd->end_of_data - cd->cur_ptr; + else + size = (cd->end_addr - cd->cur_ptr) + + (cd->end_of_data - cd->start_addr); + + return size; +} + +static int dpaa2_generic_console_open(struct inode *node, struct file *fp, + u64 offset, u64 size, + u32 expected_magic, + u32 offset_delta) +{ + u32 read_magic, wrapped, last_byte, buf_start, buf_length; + struct console_data *cd; + u64 base_addr; + int err; + + cd = kmalloc(sizeof(*cd), GFP_KERNEL); + if (!cd) + return -ENOMEM; + + base_addr = get_mc_fw_base_address(); + if (!base_addr) { + err = -EIO; + goto err_fwba; + } + + cd->map_addr = ioremap(base_addr + offset, size); + if (!cd->map_addr) { + pr_err("cannot map console log memory\n"); + err = -EIO; + goto err_ioremap; + } + + cd->hdr = (struct log_header __iomem *)cd->map_addr; + read_magic = readl(&cd->hdr->magic_word); + last_byte = readl(&cd->hdr->last_byte); + buf_start = readl(&cd->hdr->buf_start); + buf_length = readl(&cd->hdr->buf_length); + + if (read_magic != expected_magic) { + pr_warn("expected = %08x, read = %08x\n", + expected_magic, read_magic); + err = -EIO; + goto err_magic; + } + + cd->start_addr = cd->map_addr + buf_start - offset_delta; + cd->end_addr = cd->start_addr + buf_length; + + wrapped = last_byte & LOG_HEADER_FLAG_BUFFER_WRAPAROUND; + + adjust_end(cd); + if (wrapped && cd->end_of_data != cd->end_addr) + cd->cur_ptr = cd->end_of_data + 1; + else + cd->cur_ptr = cd->start_addr; + + fp->private_data = cd; + + return 0; + +err_magic: + iounmap(cd->map_addr); + +err_ioremap: +err_fwba: + kfree(cd); + + return err; +} + +static int dpaa2_mc_console_open(struct inode *node, struct file *fp) +{ + return dpaa2_generic_console_open(node, fp, + MC_BUFFER_OFFSET, MC_BUFFER_SIZE, + MAGIC_MC, MC_OFFSET_DELTA); +} + +static int dpaa2_aiop_console_open(struct inode *node, struct file *fp) +{ + return dpaa2_generic_console_open(node, fp, + AIOP_BUFFER_OFFSET, AIOP_BUFFER_SIZE, + MAGIC_AIOP, AIOP_OFFSET_DELTA); +} + +static int dpaa2_console_close(struct inode *node, struct file *fp) +{ + struct console_data *cd = fp->private_data; + + iounmap(cd->map_addr); + kfree(cd); + return 0; +} + +static ssize_t dpaa2_console_read(struct file *fp, char __user *buf, + size_t count, loff_t *f_pos) +{ + struct console_data *cd = fp->private_data; + size_t bytes = dpaa2_console_size(cd); + size_t bytes_end = cd->end_addr - cd->cur_ptr; + size_t written = 0; + void *kbuf; + int err; + + /* Check if we need to adjust the end of data addr */ + adjust_end(cd); + + if (cd->end_of_data == cd->cur_ptr) + return 0; + + if (count < bytes) + bytes = count; + + kbuf = kmalloc(bytes, GFP_KERNEL); + if (!kbuf) + return -ENOMEM; + + if (bytes > bytes_end) { + memcpy_fromio(kbuf, cd->cur_ptr, bytes_end); + if (copy_to_user(buf, kbuf, bytes_end)) { + err = -EFAULT; + goto err_free_buf; + } + buf += bytes_end; + cd->cur_ptr = cd->start_addr; + bytes -= bytes_end; + written += bytes_end; + } + + memcpy_fromio(kbuf, cd->cur_ptr, bytes); + if (copy_to_user(buf, kbuf, bytes)) { + err = -EFAULT; + goto err_free_buf; + } + cd->cur_ptr += bytes; + written += bytes; + + kfree(kbuf); + return written; + +err_free_buf: + kfree(kbuf); + + return err; +} + +static const struct file_operations dpaa2_mc_console_fops = { + .owner = THIS_MODULE, + .open = dpaa2_mc_console_open, + .release = dpaa2_console_close, + .read = dpaa2_console_read, +}; + +static struct miscdevice dpaa2_mc_console_dev = { + .minor = MISC_DYNAMIC_MINOR, + .name = "dpaa2_mc_console", + .fops = &dpaa2_mc_console_fops +}; + +static const struct file_operations dpaa2_aiop_console_fops = { + .owner = THIS_MODULE, + .open = dpaa2_aiop_console_open, + .release = dpaa2_console_close, + .read = dpaa2_console_read, +}; + +static struct miscdevice dpaa2_aiop_console_dev = { + .minor = MISC_DYNAMIC_MINOR, + .name = "dpaa2_aiop_console", + .fops = &dpaa2_aiop_console_fops +}; + +static int dpaa2_console_probe(struct platform_device *pdev) +{ + int error; + + error = of_address_to_resource(pdev->dev.of_node, 0, &mc_base_addr); + if (error < 0) { + pr_err("of_address_to_resource() failed for %pOF with %d\n", + pdev->dev.of_node, error); + return error; + } + + error = misc_register(&dpaa2_mc_console_dev); + if (error) { + pr_err("cannot register device %s\n", + dpaa2_mc_console_dev.name); + goto err_register_mc; + } + + error = misc_register(&dpaa2_aiop_console_dev); + if (error) { + pr_err("cannot register device %s\n", + dpaa2_aiop_console_dev.name); + goto err_register_aiop; + } + + return 0; + +err_register_aiop: + misc_deregister(&dpaa2_mc_console_dev); +err_register_mc: + return error; +} + +static int dpaa2_console_remove(struct platform_device *pdev) +{ + misc_deregister(&dpaa2_mc_console_dev); + misc_deregister(&dpaa2_aiop_console_dev); + + return 0; +} + +static const struct of_device_id dpaa2_console_match_table[] = { + { .compatible = "fsl,dpaa2-console",}, + {}, +}; + +MODULE_DEVICE_TABLE(of, dpaa2_console_match_table); + +static struct platform_driver dpaa2_console_driver = { + .driver = { + .name = "dpaa2-console", + .pm = NULL, + .of_match_table = dpaa2_console_match_table, + }, + .probe = dpaa2_console_probe, + .remove = dpaa2_console_remove, +}; +module_platform_driver(dpaa2_console_driver); + +MODULE_LICENSE("Dual BSD/GPL"); +MODULE_AUTHOR("Roy Pledge <roy.pledge@nxp.com>"); +MODULE_DESCRIPTION("DPAA2 console driver"); diff --git a/drivers/soc/fsl/dpio/Makefile b/drivers/soc/fsl/dpio/Makefile new file mode 100644 index 000000000..b9ff24c76 --- /dev/null +++ b/drivers/soc/fsl/dpio/Makefile @@ -0,0 +1,8 @@ +# SPDX-License-Identifier: GPL-2.0 +# +# QorIQ DPAA2 DPIO driver +# + +obj-$(CONFIG_FSL_MC_DPIO) += fsl-mc-dpio.o + +fsl-mc-dpio-objs := dpio.o qbman-portal.o dpio-service.o dpio-driver.o diff --git a/drivers/soc/fsl/dpio/dpio-cmd.h b/drivers/soc/fsl/dpio/dpio-cmd.h new file mode 100644 index 000000000..2fbcb78cd --- /dev/null +++ b/drivers/soc/fsl/dpio/dpio-cmd.h @@ -0,0 +1,58 @@ +/* SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) */ +/* + * Copyright 2013-2016 Freescale Semiconductor Inc. + * Copyright 2016 NXP + * + */ +#ifndef _FSL_DPIO_CMD_H +#define _FSL_DPIO_CMD_H + +/* DPIO Version */ +#define DPIO_VER_MAJOR 4 +#define DPIO_VER_MINOR 2 + +/* Command Versioning */ + +#define DPIO_CMD_ID_OFFSET 4 +#define DPIO_CMD_BASE_VERSION 1 + +#define DPIO_CMD(id) (((id) << DPIO_CMD_ID_OFFSET) | DPIO_CMD_BASE_VERSION) + +/* Command IDs */ +#define DPIO_CMDID_CLOSE DPIO_CMD(0x800) +#define DPIO_CMDID_OPEN DPIO_CMD(0x803) +#define DPIO_CMDID_GET_API_VERSION DPIO_CMD(0xa03) +#define DPIO_CMDID_ENABLE DPIO_CMD(0x002) +#define DPIO_CMDID_DISABLE DPIO_CMD(0x003) +#define DPIO_CMDID_GET_ATTR DPIO_CMD(0x004) +#define DPIO_CMDID_RESET DPIO_CMD(0x005) +#define DPIO_CMDID_SET_STASHING_DEST DPIO_CMD(0x120) + +struct dpio_cmd_open { + __le32 dpio_id; +}; + +#define DPIO_CHANNEL_MODE_MASK 0x3 + +struct dpio_rsp_get_attr { + /* cmd word 0 */ + __le32 id; + __le16 qbman_portal_id; + u8 num_priorities; + u8 channel_mode; + /* cmd word 1 */ + __le64 qbman_portal_ce_addr; + /* cmd word 2 */ + __le64 qbman_portal_ci_addr; + /* cmd word 3 */ + __le32 qbman_version; + __le32 pad1; + /* cmd word 4 */ + __le32 clk; +}; + +struct dpio_stashing_dest { + u8 sdest; +}; + +#endif /* _FSL_DPIO_CMD_H */ diff --git a/drivers/soc/fsl/dpio/dpio-driver.c b/drivers/soc/fsl/dpio/dpio-driver.c new file mode 100644 index 000000000..5a2edc48d --- /dev/null +++ b/drivers/soc/fsl/dpio/dpio-driver.c @@ -0,0 +1,344 @@ +// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) +/* + * Copyright 2014-2016 Freescale Semiconductor Inc. + * Copyright NXP 2016 + * + */ + +#include <linux/types.h> +#include <linux/init.h> +#include <linux/module.h> +#include <linux/platform_device.h> +#include <linux/interrupt.h> +#include <linux/msi.h> +#include <linux/dma-mapping.h> +#include <linux/delay.h> +#include <linux/io.h> +#include <linux/sys_soc.h> + +#include <linux/fsl/mc.h> +#include <soc/fsl/dpaa2-io.h> + +#include "qbman-portal.h" +#include "dpio.h" +#include "dpio-cmd.h" + +MODULE_LICENSE("Dual BSD/GPL"); +MODULE_AUTHOR("Freescale Semiconductor, Inc"); +MODULE_DESCRIPTION("DPIO Driver"); + +struct dpio_priv { + struct dpaa2_io *io; +}; + +static cpumask_var_t cpus_unused_mask; + +static const struct soc_device_attribute ls1088a_soc[] = { + {.family = "QorIQ LS1088A"}, + { /* sentinel */ } +}; + +static const struct soc_device_attribute ls2080a_soc[] = { + {.family = "QorIQ LS2080A"}, + { /* sentinel */ } +}; + +static const struct soc_device_attribute ls2088a_soc[] = { + {.family = "QorIQ LS2088A"}, + { /* sentinel */ } +}; + +static const struct soc_device_attribute lx2160a_soc[] = { + {.family = "QorIQ LX2160A"}, + { /* sentinel */ } +}; + +static int dpaa2_dpio_get_cluster_sdest(struct fsl_mc_device *dpio_dev, int cpu) +{ + int cluster_base, cluster_size; + + if (soc_device_match(ls1088a_soc)) { + cluster_base = 2; + cluster_size = 4; + } else if (soc_device_match(ls2080a_soc) || + soc_device_match(ls2088a_soc) || + soc_device_match(lx2160a_soc)) { + cluster_base = 0; + cluster_size = 2; + } else { + dev_err(&dpio_dev->dev, "unknown SoC version\n"); + return -1; + } + + return cluster_base + cpu / cluster_size; +} + +static irqreturn_t dpio_irq_handler(int irq_num, void *arg) +{ + struct device *dev = (struct device *)arg; + struct dpio_priv *priv = dev_get_drvdata(dev); + + return dpaa2_io_irq(priv->io); +} + +static void unregister_dpio_irq_handlers(struct fsl_mc_device *dpio_dev) +{ + struct fsl_mc_device_irq *irq; + + irq = dpio_dev->irqs[0]; + + /* clear the affinity hint */ + irq_set_affinity_hint(irq->virq, NULL); +} + +static int register_dpio_irq_handlers(struct fsl_mc_device *dpio_dev, int cpu) +{ + int error; + struct fsl_mc_device_irq *irq; + + irq = dpio_dev->irqs[0]; + error = devm_request_irq(&dpio_dev->dev, + irq->virq, + dpio_irq_handler, + 0, + dev_name(&dpio_dev->dev), + &dpio_dev->dev); + if (error < 0) { + dev_err(&dpio_dev->dev, + "devm_request_irq() failed: %d\n", + error); + return error; + } + + /* set the affinity hint */ + if (irq_set_affinity_hint(irq->virq, cpumask_of(cpu))) + dev_err(&dpio_dev->dev, + "irq_set_affinity failed irq %d cpu %d\n", + irq->virq, cpu); + + return 0; +} + +static int dpaa2_dpio_probe(struct fsl_mc_device *dpio_dev) +{ + struct dpio_attr dpio_attrs; + struct dpaa2_io_desc desc; + struct dpio_priv *priv; + int err = -ENOMEM; + struct device *dev = &dpio_dev->dev; + int possible_next_cpu; + int sdest; + + priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL); + if (!priv) + goto err_priv_alloc; + + dev_set_drvdata(dev, priv); + + err = fsl_mc_portal_allocate(dpio_dev, 0, &dpio_dev->mc_io); + if (err) { + dev_dbg(dev, "MC portal allocation failed\n"); + err = -EPROBE_DEFER; + goto err_priv_alloc; + } + + err = dpio_open(dpio_dev->mc_io, 0, dpio_dev->obj_desc.id, + &dpio_dev->mc_handle); + if (err) { + dev_err(dev, "dpio_open() failed\n"); + goto err_open; + } + + err = dpio_reset(dpio_dev->mc_io, 0, dpio_dev->mc_handle); + if (err) { + dev_err(dev, "dpio_reset() failed\n"); + goto err_reset; + } + + err = dpio_get_attributes(dpio_dev->mc_io, 0, dpio_dev->mc_handle, + &dpio_attrs); + if (err) { + dev_err(dev, "dpio_get_attributes() failed %d\n", err); + goto err_get_attr; + } + desc.qman_version = dpio_attrs.qbman_version; + desc.qman_clk = dpio_attrs.clk; + + err = dpio_enable(dpio_dev->mc_io, 0, dpio_dev->mc_handle); + if (err) { + dev_err(dev, "dpio_enable() failed %d\n", err); + goto err_get_attr; + } + + /* initialize DPIO descriptor */ + desc.receives_notifications = dpio_attrs.num_priorities ? 1 : 0; + desc.has_8prio = dpio_attrs.num_priorities == 8 ? 1 : 0; + desc.dpio_id = dpio_dev->obj_desc.id; + + /* get the cpu to use for the affinity hint */ + possible_next_cpu = cpumask_first(cpus_unused_mask); + if (possible_next_cpu >= nr_cpu_ids) { + dev_err(dev, "probe failed. Number of DPIOs exceeds NR_CPUS.\n"); + err = -ERANGE; + goto err_allocate_irqs; + } + desc.cpu = possible_next_cpu; + cpumask_clear_cpu(possible_next_cpu, cpus_unused_mask); + + sdest = dpaa2_dpio_get_cluster_sdest(dpio_dev, desc.cpu); + if (sdest >= 0) { + err = dpio_set_stashing_destination(dpio_dev->mc_io, 0, + dpio_dev->mc_handle, + sdest); + if (err) + dev_err(dev, "dpio_set_stashing_destination failed for cpu%d\n", + desc.cpu); + } + + if (dpio_dev->obj_desc.region_count < 3) { + /* No support for DDR backed portals, use classic mapping */ + /* + * Set the CENA regs to be the cache inhibited area of the + * portal to avoid coherency issues if a user migrates to + * another core. + */ + desc.regs_cena = devm_memremap(dev, dpio_dev->regions[1].start, + resource_size(&dpio_dev->regions[1]), + MEMREMAP_WC); + } else { + desc.regs_cena = devm_memremap(dev, dpio_dev->regions[2].start, + resource_size(&dpio_dev->regions[2]), + MEMREMAP_WB); + } + + if (IS_ERR(desc.regs_cena)) { + dev_err(dev, "devm_memremap failed\n"); + err = PTR_ERR(desc.regs_cena); + goto err_allocate_irqs; + } + + desc.regs_cinh = devm_ioremap(dev, dpio_dev->regions[1].start, + resource_size(&dpio_dev->regions[1])); + if (!desc.regs_cinh) { + err = -ENOMEM; + dev_err(dev, "devm_ioremap failed\n"); + goto err_allocate_irqs; + } + + err = fsl_mc_allocate_irqs(dpio_dev); + if (err) { + dev_err(dev, "fsl_mc_allocate_irqs failed. err=%d\n", err); + goto err_allocate_irqs; + } + + priv->io = dpaa2_io_create(&desc, dev); + if (!priv->io) { + dev_err(dev, "dpaa2_io_create failed\n"); + err = -ENOMEM; + goto err_dpaa2_io_create; + } + + err = register_dpio_irq_handlers(dpio_dev, desc.cpu); + if (err) + goto err_register_dpio_irq; + + dev_info(dev, "probed\n"); + dev_dbg(dev, " receives_notifications = %d\n", + desc.receives_notifications); + dpio_close(dpio_dev->mc_io, 0, dpio_dev->mc_handle); + + return 0; + +err_dpaa2_io_create: + unregister_dpio_irq_handlers(dpio_dev); +err_register_dpio_irq: + fsl_mc_free_irqs(dpio_dev); +err_allocate_irqs: + dpio_disable(dpio_dev->mc_io, 0, dpio_dev->mc_handle); +err_get_attr: +err_reset: + dpio_close(dpio_dev->mc_io, 0, dpio_dev->mc_handle); +err_open: + fsl_mc_portal_free(dpio_dev->mc_io); +err_priv_alloc: + return err; +} + +/* Tear down interrupts for a given DPIO object */ +static void dpio_teardown_irqs(struct fsl_mc_device *dpio_dev) +{ + unregister_dpio_irq_handlers(dpio_dev); + fsl_mc_free_irqs(dpio_dev); +} + +static int dpaa2_dpio_remove(struct fsl_mc_device *dpio_dev) +{ + struct device *dev; + struct dpio_priv *priv; + int err = 0, cpu; + + dev = &dpio_dev->dev; + priv = dev_get_drvdata(dev); + cpu = dpaa2_io_get_cpu(priv->io); + + dpaa2_io_down(priv->io); + + dpio_teardown_irqs(dpio_dev); + + cpumask_set_cpu(cpu, cpus_unused_mask); + + err = dpio_open(dpio_dev->mc_io, 0, dpio_dev->obj_desc.id, + &dpio_dev->mc_handle); + if (err) { + dev_err(dev, "dpio_open() failed\n"); + goto err_open; + } + + dpio_disable(dpio_dev->mc_io, 0, dpio_dev->mc_handle); + + dpio_close(dpio_dev->mc_io, 0, dpio_dev->mc_handle); + + fsl_mc_portal_free(dpio_dev->mc_io); + + return 0; + +err_open: + fsl_mc_portal_free(dpio_dev->mc_io); + + return err; +} + +static const struct fsl_mc_device_id dpaa2_dpio_match_id_table[] = { + { + .vendor = FSL_MC_VENDOR_FREESCALE, + .obj_type = "dpio", + }, + { .vendor = 0x0 } +}; + +static struct fsl_mc_driver dpaa2_dpio_driver = { + .driver = { + .name = KBUILD_MODNAME, + .owner = THIS_MODULE, + }, + .probe = dpaa2_dpio_probe, + .remove = dpaa2_dpio_remove, + .match_id_table = dpaa2_dpio_match_id_table +}; + +static int dpio_driver_init(void) +{ + if (!zalloc_cpumask_var(&cpus_unused_mask, GFP_KERNEL)) + return -ENOMEM; + cpumask_copy(cpus_unused_mask, cpu_online_mask); + + return fsl_mc_driver_register(&dpaa2_dpio_driver); +} + +static void dpio_driver_exit(void) +{ + free_cpumask_var(cpus_unused_mask); + fsl_mc_driver_unregister(&dpaa2_dpio_driver); +} +module_init(dpio_driver_init); +module_exit(dpio_driver_exit); diff --git a/drivers/soc/fsl/dpio/dpio-service.c b/drivers/soc/fsl/dpio/dpio-service.c new file mode 100644 index 000000000..1d2b27e3e --- /dev/null +++ b/drivers/soc/fsl/dpio/dpio-service.c @@ -0,0 +1,898 @@ +// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) +/* + * Copyright 2014-2016 Freescale Semiconductor Inc. + * Copyright 2016-2019 NXP + * + */ +#include <linux/types.h> +#include <linux/fsl/mc.h> +#include <soc/fsl/dpaa2-io.h> +#include <linux/init.h> +#include <linux/module.h> +#include <linux/platform_device.h> +#include <linux/interrupt.h> +#include <linux/dma-mapping.h> +#include <linux/dim.h> +#include <linux/slab.h> + +#include "dpio.h" +#include "qbman-portal.h" + +struct dpaa2_io { + struct dpaa2_io_desc dpio_desc; + struct qbman_swp_desc swp_desc; + struct qbman_swp *swp; + struct list_head node; + /* protect against multiple management commands */ + spinlock_t lock_mgmt_cmd; + /* protect notifications list */ + spinlock_t lock_notifications; + struct list_head notifications; + struct device *dev; + + /* Net DIM */ + struct dim rx_dim; + /* protect against concurrent Net DIM updates */ + spinlock_t dim_lock; + u16 event_ctr; + u64 bytes; + u64 frames; +}; + +struct dpaa2_io_store { + unsigned int max; + dma_addr_t paddr; + struct dpaa2_dq *vaddr; + void *alloced_addr; /* unaligned value from kmalloc() */ + unsigned int idx; /* position of the next-to-be-returned entry */ + struct qbman_swp *swp; /* portal used to issue VDQCR */ + struct device *dev; /* device used for DMA mapping */ +}; + +/* keep a per cpu array of DPIOs for fast access */ +static struct dpaa2_io *dpio_by_cpu[NR_CPUS]; +static struct list_head dpio_list = LIST_HEAD_INIT(dpio_list); +static DEFINE_SPINLOCK(dpio_list_lock); + +static inline struct dpaa2_io *service_select_by_cpu(struct dpaa2_io *d, + int cpu) +{ + if (d) + return d; + + if (cpu != DPAA2_IO_ANY_CPU && cpu >= num_possible_cpus()) + return NULL; + + /* + * If cpu == -1, choose the current cpu, with no guarantees about + * potentially being migrated away. + */ + if (cpu < 0) + cpu = raw_smp_processor_id(); + + /* If a specific cpu was requested, pick it up immediately */ + return dpio_by_cpu[cpu]; +} + +static inline struct dpaa2_io *service_select(struct dpaa2_io *d) +{ + if (d) + return d; + + d = service_select_by_cpu(d, -1); + if (d) + return d; + + spin_lock(&dpio_list_lock); + d = list_entry(dpio_list.next, struct dpaa2_io, node); + list_del(&d->node); + list_add_tail(&d->node, &dpio_list); + spin_unlock(&dpio_list_lock); + + return d; +} + +/** + * dpaa2_io_service_select() - return a dpaa2_io service affined to this cpu + * @cpu: the cpu id + * + * Return the affine dpaa2_io service, or NULL if there is no service affined + * to the specified cpu. If DPAA2_IO_ANY_CPU is used, return the next available + * service. + */ +struct dpaa2_io *dpaa2_io_service_select(int cpu) +{ + if (cpu == DPAA2_IO_ANY_CPU) + return service_select(NULL); + + return service_select_by_cpu(NULL, cpu); +} +EXPORT_SYMBOL_GPL(dpaa2_io_service_select); + +static void dpaa2_io_dim_work(struct work_struct *w) +{ + struct dim *dim = container_of(w, struct dim, work); + struct dim_cq_moder moder = + net_dim_get_rx_moderation(dim->mode, dim->profile_ix); + struct dpaa2_io *d = container_of(dim, struct dpaa2_io, rx_dim); + + dpaa2_io_set_irq_coalescing(d, moder.usec); + dim->state = DIM_START_MEASURE; +} + +/** + * dpaa2_io_create() - create a dpaa2_io object. + * @desc: the dpaa2_io descriptor + * @dev: the actual DPIO device + * + * Activates a "struct dpaa2_io" corresponding to the given config of an actual + * DPIO object. + * + * Return a valid dpaa2_io object for success, or NULL for failure. + */ +struct dpaa2_io *dpaa2_io_create(const struct dpaa2_io_desc *desc, + struct device *dev) +{ + struct dpaa2_io *obj = kmalloc(sizeof(*obj), GFP_KERNEL); + u32 qman_256_cycles_per_ns; + + if (!obj) + return NULL; + + /* check if CPU is out of range (-1 means any cpu) */ + if (desc->cpu != DPAA2_IO_ANY_CPU && desc->cpu >= num_possible_cpus()) { + kfree(obj); + return NULL; + } + + obj->dpio_desc = *desc; + obj->swp_desc.cena_bar = obj->dpio_desc.regs_cena; + obj->swp_desc.cinh_bar = obj->dpio_desc.regs_cinh; + obj->swp_desc.qman_clk = obj->dpio_desc.qman_clk; + obj->swp_desc.qman_version = obj->dpio_desc.qman_version; + + /* Compute how many 256 QBMAN cycles fit into one ns. This is because + * the interrupt timeout period register needs to be specified in QBMAN + * clock cycles in increments of 256. + */ + qman_256_cycles_per_ns = 256000 / (obj->swp_desc.qman_clk / 1000000); + obj->swp_desc.qman_256_cycles_per_ns = qman_256_cycles_per_ns; + obj->swp = qbman_swp_init(&obj->swp_desc); + + if (!obj->swp) { + kfree(obj); + return NULL; + } + + INIT_LIST_HEAD(&obj->node); + spin_lock_init(&obj->lock_mgmt_cmd); + spin_lock_init(&obj->lock_notifications); + spin_lock_init(&obj->dim_lock); + INIT_LIST_HEAD(&obj->notifications); + + /* For now only enable DQRR interrupts */ + qbman_swp_interrupt_set_trigger(obj->swp, + QBMAN_SWP_INTERRUPT_DQRI); + qbman_swp_interrupt_clear_status(obj->swp, 0xffffffff); + if (obj->dpio_desc.receives_notifications) + qbman_swp_push_set(obj->swp, 0, 1); + + spin_lock(&dpio_list_lock); + list_add_tail(&obj->node, &dpio_list); + if (desc->cpu >= 0 && !dpio_by_cpu[desc->cpu]) + dpio_by_cpu[desc->cpu] = obj; + spin_unlock(&dpio_list_lock); + + obj->dev = dev; + + memset(&obj->rx_dim, 0, sizeof(obj->rx_dim)); + INIT_WORK(&obj->rx_dim.work, dpaa2_io_dim_work); + obj->event_ctr = 0; + obj->bytes = 0; + obj->frames = 0; + + return obj; +} + +/** + * dpaa2_io_down() - release the dpaa2_io object. + * @d: the dpaa2_io object to be released. + * + * The "struct dpaa2_io" type can represent an individual DPIO object (as + * described by "struct dpaa2_io_desc") or an instance of a "DPIO service", + * which can be used to group/encapsulate multiple DPIO objects. In all cases, + * each handle obtained should be released using this function. + */ +void dpaa2_io_down(struct dpaa2_io *d) +{ + spin_lock(&dpio_list_lock); + dpio_by_cpu[d->dpio_desc.cpu] = NULL; + list_del(&d->node); + spin_unlock(&dpio_list_lock); + + kfree(d); +} + +#define DPAA_POLL_MAX 32 + +/** + * dpaa2_io_irq() - ISR for DPIO interrupts + * + * @obj: the given DPIO object. + * + * Return IRQ_HANDLED for success or IRQ_NONE if there + * were no pending interrupts. + */ +irqreturn_t dpaa2_io_irq(struct dpaa2_io *obj) +{ + const struct dpaa2_dq *dq; + int max = 0; + struct qbman_swp *swp; + u32 status; + + obj->event_ctr++; + + swp = obj->swp; + status = qbman_swp_interrupt_read_status(swp); + if (!status) + return IRQ_NONE; + + dq = qbman_swp_dqrr_next(swp); + while (dq) { + if (qbman_result_is_SCN(dq)) { + struct dpaa2_io_notification_ctx *ctx; + u64 q64; + + q64 = qbman_result_SCN_ctx(dq); + ctx = (void *)(uintptr_t)q64; + ctx->cb(ctx); + } else { + pr_crit("fsl-mc-dpio: Unrecognised/ignored DQRR entry\n"); + } + qbman_swp_dqrr_consume(swp, dq); + ++max; + if (max > DPAA_POLL_MAX) + goto done; + dq = qbman_swp_dqrr_next(swp); + } +done: + qbman_swp_interrupt_clear_status(swp, status); + qbman_swp_interrupt_set_inhibit(swp, 0); + return IRQ_HANDLED; +} + +/** + * dpaa2_io_get_cpu() - get the cpu associated with a given DPIO object + * + * @d: the given DPIO object. + * + * Return the cpu associated with the DPIO object + */ +int dpaa2_io_get_cpu(struct dpaa2_io *d) +{ + return d->dpio_desc.cpu; +} +EXPORT_SYMBOL(dpaa2_io_get_cpu); + +/** + * dpaa2_io_service_register() - Prepare for servicing of FQDAN or CDAN + * notifications on the given DPIO service. + * @d: the given DPIO service. + * @ctx: the notification context. + * @dev: the device that requests the register + * + * The caller should make the MC command to attach a DPAA2 object to + * a DPIO after this function completes successfully. In that way: + * (a) The DPIO service is "ready" to handle a notification arrival + * (which might happen before the "attach" command to MC has + * returned control of execution back to the caller) + * (b) The DPIO service can provide back to the caller the 'dpio_id' and + * 'qman64' parameters that it should pass along in the MC command + * in order for the object to be configured to produce the right + * notification fields to the DPIO service. + * + * Return 0 for success, or -ENODEV for failure. + */ +int dpaa2_io_service_register(struct dpaa2_io *d, + struct dpaa2_io_notification_ctx *ctx, + struct device *dev) +{ + struct device_link *link; + unsigned long irqflags; + + d = service_select_by_cpu(d, ctx->desired_cpu); + if (!d) + return -ENODEV; + + link = device_link_add(dev, d->dev, DL_FLAG_AUTOREMOVE_CONSUMER); + if (!link) + return -EINVAL; + + ctx->dpio_id = d->dpio_desc.dpio_id; + ctx->qman64 = (u64)(uintptr_t)ctx; + ctx->dpio_private = d; + spin_lock_irqsave(&d->lock_notifications, irqflags); + list_add(&ctx->node, &d->notifications); + spin_unlock_irqrestore(&d->lock_notifications, irqflags); + + /* Enable the generation of CDAN notifications */ + if (ctx->is_cdan) + return qbman_swp_CDAN_set_context_enable(d->swp, + (u16)ctx->id, + ctx->qman64); + return 0; +} +EXPORT_SYMBOL_GPL(dpaa2_io_service_register); + +/** + * dpaa2_io_service_deregister - The opposite of 'register'. + * @service: the given DPIO service. + * @ctx: the notification context. + * @dev: the device that requests to be deregistered + * + * This function should be called only after sending the MC command to + * to detach the notification-producing device from the DPIO. + */ +void dpaa2_io_service_deregister(struct dpaa2_io *service, + struct dpaa2_io_notification_ctx *ctx, + struct device *dev) +{ + struct dpaa2_io *d = ctx->dpio_private; + unsigned long irqflags; + + if (ctx->is_cdan) + qbman_swp_CDAN_disable(d->swp, (u16)ctx->id); + + spin_lock_irqsave(&d->lock_notifications, irqflags); + list_del(&ctx->node); + spin_unlock_irqrestore(&d->lock_notifications, irqflags); + +} +EXPORT_SYMBOL_GPL(dpaa2_io_service_deregister); + +/** + * dpaa2_io_service_rearm() - Rearm the notification for the given DPIO service. + * @d: the given DPIO service. + * @ctx: the notification context. + * + * Once a FQDAN/CDAN has been produced, the corresponding FQ/channel is + * considered "disarmed". Ie. the user can issue pull dequeue operations on that + * traffic source for as long as it likes. Eventually it may wish to "rearm" + * that source to allow it to produce another FQDAN/CDAN, that's what this + * function achieves. + * + * Return 0 for success. + */ +int dpaa2_io_service_rearm(struct dpaa2_io *d, + struct dpaa2_io_notification_ctx *ctx) +{ + unsigned long irqflags; + int err; + + d = service_select_by_cpu(d, ctx->desired_cpu); + if (!unlikely(d)) + return -ENODEV; + + spin_lock_irqsave(&d->lock_mgmt_cmd, irqflags); + if (ctx->is_cdan) + err = qbman_swp_CDAN_enable(d->swp, (u16)ctx->id); + else + err = qbman_swp_fq_schedule(d->swp, ctx->id); + spin_unlock_irqrestore(&d->lock_mgmt_cmd, irqflags); + + return err; +} +EXPORT_SYMBOL_GPL(dpaa2_io_service_rearm); + +/** + * dpaa2_io_service_pull_fq() - pull dequeue functions from a fq. + * @d: the given DPIO service. + * @fqid: the given frame queue id. + * @s: the dpaa2_io_store object for the result. + * + * Return 0 for success, or error code for failure. + */ +int dpaa2_io_service_pull_fq(struct dpaa2_io *d, u32 fqid, + struct dpaa2_io_store *s) +{ + struct qbman_pull_desc pd; + int err; + + qbman_pull_desc_clear(&pd); + qbman_pull_desc_set_storage(&pd, s->vaddr, s->paddr, 1); + qbman_pull_desc_set_numframes(&pd, (u8)s->max); + qbman_pull_desc_set_fq(&pd, fqid); + + d = service_select(d); + if (!d) + return -ENODEV; + s->swp = d->swp; + err = qbman_swp_pull(d->swp, &pd); + if (err) + s->swp = NULL; + + return err; +} +EXPORT_SYMBOL(dpaa2_io_service_pull_fq); + +/** + * dpaa2_io_service_pull_channel() - pull dequeue functions from a channel. + * @d: the given DPIO service. + * @channelid: the given channel id. + * @s: the dpaa2_io_store object for the result. + * + * Return 0 for success, or error code for failure. + */ +int dpaa2_io_service_pull_channel(struct dpaa2_io *d, u32 channelid, + struct dpaa2_io_store *s) +{ + struct qbman_pull_desc pd; + int err; + + qbman_pull_desc_clear(&pd); + qbman_pull_desc_set_storage(&pd, s->vaddr, s->paddr, 1); + qbman_pull_desc_set_numframes(&pd, (u8)s->max); + qbman_pull_desc_set_channel(&pd, channelid, qbman_pull_type_prio); + + d = service_select(d); + if (!d) + return -ENODEV; + + s->swp = d->swp; + err = qbman_swp_pull(d->swp, &pd); + if (err) + s->swp = NULL; + + return err; +} +EXPORT_SYMBOL_GPL(dpaa2_io_service_pull_channel); + +/** + * dpaa2_io_service_enqueue_fq() - Enqueue a frame to a frame queue. + * @d: the given DPIO service. + * @fqid: the given frame queue id. + * @fd: the frame descriptor which is enqueued. + * + * Return 0 for successful enqueue, -EBUSY if the enqueue ring is not ready, + * or -ENODEV if there is no dpio service. + */ +int dpaa2_io_service_enqueue_fq(struct dpaa2_io *d, + u32 fqid, + const struct dpaa2_fd *fd) +{ + struct qbman_eq_desc ed; + + d = service_select(d); + if (!d) + return -ENODEV; + + qbman_eq_desc_clear(&ed); + qbman_eq_desc_set_no_orp(&ed, 0); + qbman_eq_desc_set_fq(&ed, fqid); + + return qbman_swp_enqueue(d->swp, &ed, fd); +} +EXPORT_SYMBOL(dpaa2_io_service_enqueue_fq); + +/** + * dpaa2_io_service_enqueue_multiple_fq() - Enqueue multiple frames + * to a frame queue using one fqid. + * @d: the given DPIO service. + * @fqid: the given frame queue id. + * @fd: the frame descriptor which is enqueued. + * @nb: number of frames to be enqueud + * + * Return 0 for successful enqueue, -EBUSY if the enqueue ring is not ready, + * or -ENODEV if there is no dpio service. + */ +int dpaa2_io_service_enqueue_multiple_fq(struct dpaa2_io *d, + u32 fqid, + const struct dpaa2_fd *fd, + int nb) +{ + struct qbman_eq_desc ed; + + d = service_select(d); + if (!d) + return -ENODEV; + + qbman_eq_desc_clear(&ed); + qbman_eq_desc_set_no_orp(&ed, 0); + qbman_eq_desc_set_fq(&ed, fqid); + + return qbman_swp_enqueue_multiple(d->swp, &ed, fd, NULL, nb); +} +EXPORT_SYMBOL(dpaa2_io_service_enqueue_multiple_fq); + +/** + * dpaa2_io_service_enqueue_multiple_desc_fq() - Enqueue multiple frames + * to different frame queue using a list of fqids. + * @d: the given DPIO service. + * @fqid: the given list of frame queue ids. + * @fd: the frame descriptor which is enqueued. + * @nb: number of frames to be enqueud + * + * Return 0 for successful enqueue, -EBUSY if the enqueue ring is not ready, + * or -ENODEV if there is no dpio service. + */ +int dpaa2_io_service_enqueue_multiple_desc_fq(struct dpaa2_io *d, + u32 *fqid, + const struct dpaa2_fd *fd, + int nb) +{ + struct qbman_eq_desc *ed; + int i, ret; + + ed = kcalloc(sizeof(struct qbman_eq_desc), 32, GFP_KERNEL); + if (!ed) + return -ENOMEM; + + d = service_select(d); + if (!d) { + ret = -ENODEV; + goto out; + } + + for (i = 0; i < nb; i++) { + qbman_eq_desc_clear(&ed[i]); + qbman_eq_desc_set_no_orp(&ed[i], 0); + qbman_eq_desc_set_fq(&ed[i], fqid[i]); + } + + ret = qbman_swp_enqueue_multiple_desc(d->swp, &ed[0], fd, nb); +out: + kfree(ed); + return ret; +} +EXPORT_SYMBOL(dpaa2_io_service_enqueue_multiple_desc_fq); + +/** + * dpaa2_io_service_enqueue_qd() - Enqueue a frame to a QD. + * @d: the given DPIO service. + * @qdid: the given queuing destination id. + * @prio: the given queuing priority. + * @qdbin: the given queuing destination bin. + * @fd: the frame descriptor which is enqueued. + * + * Return 0 for successful enqueue, or -EBUSY if the enqueue ring is not ready, + * or -ENODEV if there is no dpio service. + */ +int dpaa2_io_service_enqueue_qd(struct dpaa2_io *d, + u32 qdid, u8 prio, u16 qdbin, + const struct dpaa2_fd *fd) +{ + struct qbman_eq_desc ed; + + d = service_select(d); + if (!d) + return -ENODEV; + + qbman_eq_desc_clear(&ed); + qbman_eq_desc_set_no_orp(&ed, 0); + qbman_eq_desc_set_qd(&ed, qdid, qdbin, prio); + + return qbman_swp_enqueue(d->swp, &ed, fd); +} +EXPORT_SYMBOL_GPL(dpaa2_io_service_enqueue_qd); + +/** + * dpaa2_io_service_release() - Release buffers to a buffer pool. + * @d: the given DPIO object. + * @bpid: the buffer pool id. + * @buffers: the buffers to be released. + * @num_buffers: the number of the buffers to be released. + * + * Return 0 for success, and negative error code for failure. + */ +int dpaa2_io_service_release(struct dpaa2_io *d, + u16 bpid, + const u64 *buffers, + unsigned int num_buffers) +{ + struct qbman_release_desc rd; + + d = service_select(d); + if (!d) + return -ENODEV; + + qbman_release_desc_clear(&rd); + qbman_release_desc_set_bpid(&rd, bpid); + + return qbman_swp_release(d->swp, &rd, buffers, num_buffers); +} +EXPORT_SYMBOL_GPL(dpaa2_io_service_release); + +/** + * dpaa2_io_service_acquire() - Acquire buffers from a buffer pool. + * @d: the given DPIO object. + * @bpid: the buffer pool id. + * @buffers: the buffer addresses for acquired buffers. + * @num_buffers: the expected number of the buffers to acquire. + * + * Return a negative error code if the command failed, otherwise it returns + * the number of buffers acquired, which may be less than the number requested. + * Eg. if the buffer pool is empty, this will return zero. + */ +int dpaa2_io_service_acquire(struct dpaa2_io *d, + u16 bpid, + u64 *buffers, + unsigned int num_buffers) +{ + unsigned long irqflags; + int err; + + d = service_select(d); + if (!d) + return -ENODEV; + + spin_lock_irqsave(&d->lock_mgmt_cmd, irqflags); + err = qbman_swp_acquire(d->swp, bpid, buffers, num_buffers); + spin_unlock_irqrestore(&d->lock_mgmt_cmd, irqflags); + + return err; +} +EXPORT_SYMBOL_GPL(dpaa2_io_service_acquire); + +/* + * 'Stores' are reusable memory blocks for holding dequeue results, and to + * assist with parsing those results. + */ + +/** + * dpaa2_io_store_create() - Create the dma memory storage for dequeue result. + * @max_frames: the maximum number of dequeued result for frames, must be <= 32. + * @dev: the device to allow mapping/unmapping the DMAable region. + * + * The size of the storage is "max_frames*sizeof(struct dpaa2_dq)". + * The 'dpaa2_io_store' returned is a DPIO service managed object. + * + * Return pointer to dpaa2_io_store struct for successfully created storage + * memory, or NULL on error. + */ +struct dpaa2_io_store *dpaa2_io_store_create(unsigned int max_frames, + struct device *dev) +{ + struct dpaa2_io_store *ret; + size_t size; + + if (!max_frames || (max_frames > 32)) + return NULL; + + ret = kmalloc(sizeof(*ret), GFP_KERNEL); + if (!ret) + return NULL; + + ret->max = max_frames; + size = max_frames * sizeof(struct dpaa2_dq) + 64; + ret->alloced_addr = kzalloc(size, GFP_KERNEL); + if (!ret->alloced_addr) { + kfree(ret); + return NULL; + } + + ret->vaddr = PTR_ALIGN(ret->alloced_addr, 64); + ret->paddr = dma_map_single(dev, ret->vaddr, + sizeof(struct dpaa2_dq) * max_frames, + DMA_FROM_DEVICE); + if (dma_mapping_error(dev, ret->paddr)) { + kfree(ret->alloced_addr); + kfree(ret); + return NULL; + } + + ret->idx = 0; + ret->dev = dev; + + return ret; +} +EXPORT_SYMBOL_GPL(dpaa2_io_store_create); + +/** + * dpaa2_io_store_destroy() - Frees the dma memory storage for dequeue + * result. + * @s: the storage memory to be destroyed. + */ +void dpaa2_io_store_destroy(struct dpaa2_io_store *s) +{ + dma_unmap_single(s->dev, s->paddr, sizeof(struct dpaa2_dq) * s->max, + DMA_FROM_DEVICE); + kfree(s->alloced_addr); + kfree(s); +} +EXPORT_SYMBOL_GPL(dpaa2_io_store_destroy); + +/** + * dpaa2_io_store_next() - Determine when the next dequeue result is available. + * @s: the dpaa2_io_store object. + * @is_last: indicate whether this is the last frame in the pull command. + * + * When an object driver performs dequeues to a dpaa2_io_store, this function + * can be used to determine when the next frame result is available. Once + * this function returns non-NULL, a subsequent call to it will try to find + * the next dequeue result. + * + * Note that if a pull-dequeue has a NULL result because the target FQ/channel + * was empty, then this function will also return NULL (rather than expecting + * the caller to always check for this. As such, "is_last" can be used to + * differentiate between "end-of-empty-dequeue" and "still-waiting". + * + * Return dequeue result for a valid dequeue result, or NULL for empty dequeue. + */ +struct dpaa2_dq *dpaa2_io_store_next(struct dpaa2_io_store *s, int *is_last) +{ + int match; + struct dpaa2_dq *ret = &s->vaddr[s->idx]; + + match = qbman_result_has_new_result(s->swp, ret); + if (!match) { + *is_last = 0; + return NULL; + } + + s->idx++; + + if (dpaa2_dq_is_pull_complete(ret)) { + *is_last = 1; + s->idx = 0; + /* + * If we get an empty dequeue result to terminate a zero-results + * vdqcr, return NULL to the caller rather than expecting him to + * check non-NULL results every time. + */ + if (!(dpaa2_dq_flags(ret) & DPAA2_DQ_STAT_VALIDFRAME)) + ret = NULL; + } else { + prefetch(&s->vaddr[s->idx]); + *is_last = 0; + } + + return ret; +} +EXPORT_SYMBOL_GPL(dpaa2_io_store_next); + +/** + * dpaa2_io_query_fq_count() - Get the frame and byte count for a given fq. + * @d: the given DPIO object. + * @fqid: the id of frame queue to be queried. + * @fcnt: the queried frame count. + * @bcnt: the queried byte count. + * + * Knowing the FQ count at run-time can be useful in debugging situations. + * The instantaneous frame- and byte-count are hereby returned. + * + * Return 0 for a successful query, and negative error code if query fails. + */ +int dpaa2_io_query_fq_count(struct dpaa2_io *d, u32 fqid, + u32 *fcnt, u32 *bcnt) +{ + struct qbman_fq_query_np_rslt state; + struct qbman_swp *swp; + unsigned long irqflags; + int ret; + + d = service_select(d); + if (!d) + return -ENODEV; + + swp = d->swp; + spin_lock_irqsave(&d->lock_mgmt_cmd, irqflags); + ret = qbman_fq_query_state(swp, fqid, &state); + spin_unlock_irqrestore(&d->lock_mgmt_cmd, irqflags); + if (ret) + return ret; + *fcnt = qbman_fq_state_frame_count(&state); + *bcnt = qbman_fq_state_byte_count(&state); + + return 0; +} +EXPORT_SYMBOL_GPL(dpaa2_io_query_fq_count); + +/** + * dpaa2_io_query_bp_count() - Query the number of buffers currently in a + * buffer pool. + * @d: the given DPIO object. + * @bpid: the index of buffer pool to be queried. + * @num: the queried number of buffers in the buffer pool. + * + * Return 0 for a successful query, and negative error code if query fails. + */ +int dpaa2_io_query_bp_count(struct dpaa2_io *d, u16 bpid, u32 *num) +{ + struct qbman_bp_query_rslt state; + struct qbman_swp *swp; + unsigned long irqflags; + int ret; + + d = service_select(d); + if (!d) + return -ENODEV; + + swp = d->swp; + spin_lock_irqsave(&d->lock_mgmt_cmd, irqflags); + ret = qbman_bp_query(swp, bpid, &state); + spin_unlock_irqrestore(&d->lock_mgmt_cmd, irqflags); + if (ret) + return ret; + *num = qbman_bp_info_num_free_bufs(&state); + return 0; +} +EXPORT_SYMBOL_GPL(dpaa2_io_query_bp_count); + +/** + * dpaa2_io_set_irq_coalescing() - Set new IRQ coalescing values + * @d: the given DPIO object + * @irq_holdoff: interrupt holdoff (timeout) period in us + * + * Return 0 for success, or negative error code on error. + */ +int dpaa2_io_set_irq_coalescing(struct dpaa2_io *d, u32 irq_holdoff) +{ + struct qbman_swp *swp = d->swp; + + return qbman_swp_set_irq_coalescing(swp, swp->dqrr.dqrr_size - 1, + irq_holdoff); +} +EXPORT_SYMBOL(dpaa2_io_set_irq_coalescing); + +/** + * dpaa2_io_get_irq_coalescing() - Get the current IRQ coalescing parameters + * @d: the given DPIO object + * @irq_holdoff: interrupt holdoff (timeout) period in us + */ +void dpaa2_io_get_irq_coalescing(struct dpaa2_io *d, u32 *irq_holdoff) +{ + struct qbman_swp *swp = d->swp; + + qbman_swp_get_irq_coalescing(swp, NULL, irq_holdoff); +} +EXPORT_SYMBOL(dpaa2_io_get_irq_coalescing); + +/** + * dpaa2_io_set_adaptive_coalescing() - Enable/disable adaptive coalescing + * @d: the given DPIO object + * @use_adaptive_rx_coalesce: adaptive coalescing state + */ +void dpaa2_io_set_adaptive_coalescing(struct dpaa2_io *d, + int use_adaptive_rx_coalesce) +{ + d->swp->use_adaptive_rx_coalesce = use_adaptive_rx_coalesce; +} +EXPORT_SYMBOL(dpaa2_io_set_adaptive_coalescing); + +/** + * dpaa2_io_get_adaptive_coalescing() - Query adaptive coalescing state + * @d: the given DPIO object + * + * Return 1 when adaptive coalescing is enabled on the DPIO object and 0 + * otherwise. + */ +int dpaa2_io_get_adaptive_coalescing(struct dpaa2_io *d) +{ + return d->swp->use_adaptive_rx_coalesce; +} +EXPORT_SYMBOL(dpaa2_io_get_adaptive_coalescing); + +/** + * dpaa2_io_update_net_dim() - Update Net DIM + * @d: the given DPIO object + * @frames: how many frames have been dequeued by the user since the last call + * @bytes: how many bytes have been dequeued by the user since the last call + */ +void dpaa2_io_update_net_dim(struct dpaa2_io *d, __u64 frames, __u64 bytes) +{ + struct dim_sample dim_sample = {}; + + if (!d->swp->use_adaptive_rx_coalesce) + return; + + spin_lock(&d->dim_lock); + + d->bytes += bytes; + d->frames += frames; + + dim_update_sample(d->event_ctr, d->frames, d->bytes, &dim_sample); + net_dim(&d->rx_dim, dim_sample); + + spin_unlock(&d->dim_lock); +} +EXPORT_SYMBOL(dpaa2_io_update_net_dim); diff --git a/drivers/soc/fsl/dpio/dpio.c b/drivers/soc/fsl/dpio/dpio.c new file mode 100644 index 000000000..8ed606ffa --- /dev/null +++ b/drivers/soc/fsl/dpio/dpio.c @@ -0,0 +1,238 @@ +// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) +/* + * Copyright 2013-2016 Freescale Semiconductor Inc. + * Copyright 2016 NXP + * + */ +#include <linux/kernel.h> +#include <linux/fsl/mc.h> + +#include "dpio.h" +#include "dpio-cmd.h" + +/* + * Data Path I/O Portal API + * Contains initialization APIs and runtime control APIs for DPIO + */ + +/** + * dpio_open() - Open a control session for the specified object + * @mc_io: Pointer to MC portal's I/O object + * @cmd_flags: Command flags; one or more of 'MC_CMD_FLAG_' + * @dpio_id: DPIO unique ID + * @token: Returned token; use in subsequent API calls + * + * This function can be used to open a control session for an + * already created object; an object may have been declared in + * the DPL or by calling the dpio_create() function. + * This function returns a unique authentication token, + * associated with the specific object ID and the specific MC + * portal; this token must be used in all subsequent commands for + * this specific object. + * + * Return: '0' on Success; Error code otherwise. + */ +int dpio_open(struct fsl_mc_io *mc_io, + u32 cmd_flags, + int dpio_id, + u16 *token) +{ + struct fsl_mc_command cmd = { 0 }; + struct dpio_cmd_open *dpio_cmd; + int err; + + /* prepare command */ + cmd.header = mc_encode_cmd_header(DPIO_CMDID_OPEN, + cmd_flags, + 0); + dpio_cmd = (struct dpio_cmd_open *)cmd.params; + dpio_cmd->dpio_id = cpu_to_le32(dpio_id); + + err = mc_send_command(mc_io, &cmd); + if (err) + return err; + + /* retrieve response parameters */ + *token = mc_cmd_hdr_read_token(&cmd); + + return 0; +} + +/** + * dpio_close() - Close the control session of the object + * @mc_io: Pointer to MC portal's I/O object + * @cmd_flags: Command flags; one or more of 'MC_CMD_FLAG_' + * @token: Token of DPIO object + * + * Return: '0' on Success; Error code otherwise. + */ +int dpio_close(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token) +{ + struct fsl_mc_command cmd = { 0 }; + + /* prepare command */ + cmd.header = mc_encode_cmd_header(DPIO_CMDID_CLOSE, + cmd_flags, + token); + + return mc_send_command(mc_io, &cmd); +} + +/** + * dpio_enable() - Enable the DPIO, allow I/O portal operations. + * @mc_io: Pointer to MC portal's I/O object + * @cmd_flags: Command flags; one or more of 'MC_CMD_FLAG_' + * @token: Token of DPIO object + * + * Return: '0' on Success; Error code otherwise + */ +int dpio_enable(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token) +{ + struct fsl_mc_command cmd = { 0 }; + + /* prepare command */ + cmd.header = mc_encode_cmd_header(DPIO_CMDID_ENABLE, + cmd_flags, + token); + + return mc_send_command(mc_io, &cmd); +} + +/** + * dpio_disable() - Disable the DPIO, stop any I/O portal operation. + * @mc_io: Pointer to MC portal's I/O object + * @cmd_flags: Command flags; one or more of 'MC_CMD_FLAG_' + * @token: Token of DPIO object + * + * Return: '0' on Success; Error code otherwise + */ +int dpio_disable(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token) +{ + struct fsl_mc_command cmd = { 0 }; + + /* prepare command */ + cmd.header = mc_encode_cmd_header(DPIO_CMDID_DISABLE, + cmd_flags, + token); + + return mc_send_command(mc_io, &cmd); +} + +/** + * dpio_get_attributes() - Retrieve DPIO attributes + * @mc_io: Pointer to MC portal's I/O object + * @cmd_flags: Command flags; one or more of 'MC_CMD_FLAG_' + * @token: Token of DPIO object + * @attr: Returned object's attributes + * + * Return: '0' on Success; Error code otherwise + */ +int dpio_get_attributes(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token, + struct dpio_attr *attr) +{ + struct fsl_mc_command cmd = { 0 }; + struct dpio_rsp_get_attr *dpio_rsp; + int err; + + /* prepare command */ + cmd.header = mc_encode_cmd_header(DPIO_CMDID_GET_ATTR, + cmd_flags, + token); + + err = mc_send_command(mc_io, &cmd); + if (err) + return err; + + /* retrieve response parameters */ + dpio_rsp = (struct dpio_rsp_get_attr *)cmd.params; + attr->id = le32_to_cpu(dpio_rsp->id); + attr->qbman_portal_id = le16_to_cpu(dpio_rsp->qbman_portal_id); + attr->num_priorities = dpio_rsp->num_priorities; + attr->channel_mode = dpio_rsp->channel_mode & DPIO_CHANNEL_MODE_MASK; + attr->qbman_portal_ce_offset = + le64_to_cpu(dpio_rsp->qbman_portal_ce_addr); + attr->qbman_portal_ci_offset = + le64_to_cpu(dpio_rsp->qbman_portal_ci_addr); + attr->qbman_version = le32_to_cpu(dpio_rsp->qbman_version); + attr->clk = le32_to_cpu(dpio_rsp->clk); + + return 0; +} + +int dpio_set_stashing_destination(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token, + u8 sdest) +{ + struct fsl_mc_command cmd = { 0 }; + struct dpio_stashing_dest *dpio_cmd; + + cmd.header = mc_encode_cmd_header(DPIO_CMDID_SET_STASHING_DEST, + cmd_flags, token); + dpio_cmd = (struct dpio_stashing_dest *)cmd.params; + dpio_cmd->sdest = sdest; + + return mc_send_command(mc_io, &cmd); +} + +/** + * dpio_get_api_version - Get Data Path I/O API version + * @mc_io: Pointer to MC portal's DPIO object + * @cmd_flags: Command flags; one or more of 'MC_CMD_FLAG_' + * @major_ver: Major version of DPIO API + * @minor_ver: Minor version of DPIO API + * + * Return: '0' on Success; Error code otherwise + */ +int dpio_get_api_version(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 *major_ver, + u16 *minor_ver) +{ + struct fsl_mc_command cmd = { 0 }; + int err; + + /* prepare command */ + cmd.header = mc_encode_cmd_header(DPIO_CMDID_GET_API_VERSION, + cmd_flags, 0); + + err = mc_send_command(mc_io, &cmd); + if (err) + return err; + + /* retrieve response parameters */ + mc_cmd_read_api_version(&cmd, major_ver, minor_ver); + + return 0; +} + +/** + * dpio_reset() - Reset the DPIO, returns the object to initial state. + * @mc_io: Pointer to MC portal's I/O object + * @cmd_flags: Command flags; one or more of 'MC_CMD_FLAG_' + * @token: Token of DPIO object + * + * Return: '0' on Success; Error code otherwise. + */ +int dpio_reset(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token) +{ + struct fsl_mc_command cmd = { 0 }; + + /* prepare command */ + cmd.header = mc_encode_cmd_header(DPIO_CMDID_RESET, + cmd_flags, + token); + + /* send command to mc*/ + return mc_send_command(mc_io, &cmd); +} diff --git a/drivers/soc/fsl/dpio/dpio.h b/drivers/soc/fsl/dpio/dpio.h new file mode 100644 index 000000000..7fda44f0d --- /dev/null +++ b/drivers/soc/fsl/dpio/dpio.h @@ -0,0 +1,94 @@ +/* SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) */ +/* + * Copyright 2013-2016 Freescale Semiconductor Inc. + * Copyright 2016 NXP + * + */ +#ifndef __FSL_DPIO_H +#define __FSL_DPIO_H + +struct fsl_mc_io; + +int dpio_open(struct fsl_mc_io *mc_io, + u32 cmd_flags, + int dpio_id, + u16 *token); + +int dpio_close(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token); + +/** + * enum dpio_channel_mode - DPIO notification channel mode + * @DPIO_NO_CHANNEL: No support for notification channel + * @DPIO_LOCAL_CHANNEL: Notifications on data availability can be received by a + * dedicated channel in the DPIO; user should point the queue's + * destination in the relevant interface to this DPIO + */ +enum dpio_channel_mode { + DPIO_NO_CHANNEL = 0, + DPIO_LOCAL_CHANNEL = 1, +}; + +/** + * struct dpio_cfg - Structure representing DPIO configuration + * @channel_mode: Notification channel mode + * @num_priorities: Number of priorities for the notification channel (1-8); + * relevant only if 'channel_mode = DPIO_LOCAL_CHANNEL' + */ +struct dpio_cfg { + enum dpio_channel_mode channel_mode; + u8 num_priorities; +}; + +int dpio_enable(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token); + +int dpio_disable(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token); + +/** + * struct dpio_attr - Structure representing DPIO attributes + * @id: DPIO object ID + * @qbman_portal_ce_offset: offset of the software portal cache-enabled area + * @qbman_portal_ci_offset: offset of the software portal cache-inhibited area + * @qbman_portal_id: Software portal ID + * @channel_mode: Notification channel mode + * @num_priorities: Number of priorities for the notification channel (1-8); + * relevant only if 'channel_mode = DPIO_LOCAL_CHANNEL' + * @qbman_version: QBMAN version + * @clk: QBMAN clock frequency value in Hz + */ +struct dpio_attr { + int id; + u64 qbman_portal_ce_offset; + u64 qbman_portal_ci_offset; + u16 qbman_portal_id; + enum dpio_channel_mode channel_mode; + u8 num_priorities; + u32 qbman_version; + u32 clk; +}; + +int dpio_get_attributes(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token, + struct dpio_attr *attr); + +int dpio_set_stashing_destination(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token, + u8 dest); + +int dpio_get_api_version(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 *major_ver, + u16 *minor_ver); + +int dpio_reset(struct fsl_mc_io *mc_io, + u32 cmd_flags, + u16 token); + +#endif /* __FSL_DPIO_H */ diff --git a/drivers/soc/fsl/dpio/qbman-portal.c b/drivers/soc/fsl/dpio/qbman-portal.c new file mode 100644 index 000000000..0a3fb6c11 --- /dev/null +++ b/drivers/soc/fsl/dpio/qbman-portal.c @@ -0,0 +1,1853 @@ +// SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) +/* + * Copyright (C) 2014-2016 Freescale Semiconductor, Inc. + * Copyright 2016-2019 NXP + * + */ + +#include <asm/cacheflush.h> +#include <linux/io.h> +#include <linux/slab.h> +#include <linux/spinlock.h> +#include <soc/fsl/dpaa2-global.h> + +#include "qbman-portal.h" + +/* All QBMan command and result structures use this "valid bit" encoding */ +#define QB_VALID_BIT ((u32)0x80) + +/* QBMan portal management command codes */ +#define QBMAN_MC_ACQUIRE 0x30 +#define QBMAN_WQCHAN_CONFIGURE 0x46 + +/* CINH register offsets */ +#define QBMAN_CINH_SWP_EQCR_PI 0x800 +#define QBMAN_CINH_SWP_EQCR_CI 0x840 +#define QBMAN_CINH_SWP_EQAR 0x8c0 +#define QBMAN_CINH_SWP_CR_RT 0x900 +#define QBMAN_CINH_SWP_VDQCR_RT 0x940 +#define QBMAN_CINH_SWP_EQCR_AM_RT 0x980 +#define QBMAN_CINH_SWP_RCR_AM_RT 0x9c0 +#define QBMAN_CINH_SWP_DQPI 0xa00 +#define QBMAN_CINH_SWP_DQRR_ITR 0xa80 +#define QBMAN_CINH_SWP_DCAP 0xac0 +#define QBMAN_CINH_SWP_SDQCR 0xb00 +#define QBMAN_CINH_SWP_EQCR_AM_RT2 0xb40 +#define QBMAN_CINH_SWP_RCR_PI 0xc00 +#define QBMAN_CINH_SWP_RAR 0xcc0 +#define QBMAN_CINH_SWP_ISR 0xe00 +#define QBMAN_CINH_SWP_IER 0xe40 +#define QBMAN_CINH_SWP_ISDR 0xe80 +#define QBMAN_CINH_SWP_IIR 0xec0 +#define QBMAN_CINH_SWP_ITPR 0xf40 + +/* CENA register offsets */ +#define QBMAN_CENA_SWP_EQCR(n) (0x000 + ((u32)(n) << 6)) +#define QBMAN_CENA_SWP_DQRR(n) (0x200 + ((u32)(n) << 6)) +#define QBMAN_CENA_SWP_RCR(n) (0x400 + ((u32)(n) << 6)) +#define QBMAN_CENA_SWP_CR 0x600 +#define QBMAN_CENA_SWP_RR(vb) (0x700 + ((u32)(vb) >> 1)) +#define QBMAN_CENA_SWP_VDQCR 0x780 +#define QBMAN_CENA_SWP_EQCR_CI 0x840 +#define QBMAN_CENA_SWP_EQCR_CI_MEMBACK 0x1840 + +/* CENA register offsets in memory-backed mode */ +#define QBMAN_CENA_SWP_DQRR_MEM(n) (0x800 + ((u32)(n) << 6)) +#define QBMAN_CENA_SWP_RCR_MEM(n) (0x1400 + ((u32)(n) << 6)) +#define QBMAN_CENA_SWP_CR_MEM 0x1600 +#define QBMAN_CENA_SWP_RR_MEM 0x1680 +#define QBMAN_CENA_SWP_VDQCR_MEM 0x1780 + +/* Reverse mapping of QBMAN_CENA_SWP_DQRR() */ +#define QBMAN_IDX_FROM_DQRR(p) (((unsigned long)(p) & 0x1ff) >> 6) + +/* Define token used to determine if response written to memory is valid */ +#define QMAN_DQ_TOKEN_VALID 1 + +/* SDQCR attribute codes */ +#define QB_SDQCR_FC_SHIFT 29 +#define QB_SDQCR_FC_MASK 0x1 +#define QB_SDQCR_DCT_SHIFT 24 +#define QB_SDQCR_DCT_MASK 0x3 +#define QB_SDQCR_TOK_SHIFT 16 +#define QB_SDQCR_TOK_MASK 0xff +#define QB_SDQCR_SRC_SHIFT 0 +#define QB_SDQCR_SRC_MASK 0xffff + +/* opaque token for static dequeues */ +#define QMAN_SDQCR_TOKEN 0xbb + +#define QBMAN_EQCR_DCA_IDXMASK 0x0f +#define QBMAN_ENQUEUE_FLAG_DCA (1ULL << 31) + +#define EQ_DESC_SIZE_WITHOUT_FD 29 +#define EQ_DESC_SIZE_FD_START 32 + +enum qbman_sdqcr_dct { + qbman_sdqcr_dct_null = 0, + qbman_sdqcr_dct_prio_ics, + qbman_sdqcr_dct_active_ics, + qbman_sdqcr_dct_active +}; + +enum qbman_sdqcr_fc { + qbman_sdqcr_fc_one = 0, + qbman_sdqcr_fc_up_to_3 = 1 +}; + +/* Internal Function declaration */ +static int qbman_swp_enqueue_direct(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd); +static int qbman_swp_enqueue_mem_back(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd); +static int qbman_swp_enqueue_multiple_direct(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + uint32_t *flags, + int num_frames); +static int qbman_swp_enqueue_multiple_mem_back(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + uint32_t *flags, + int num_frames); +static int +qbman_swp_enqueue_multiple_desc_direct(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + int num_frames); +static +int qbman_swp_enqueue_multiple_desc_mem_back(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + int num_frames); +static int qbman_swp_pull_direct(struct qbman_swp *s, + struct qbman_pull_desc *d); +static int qbman_swp_pull_mem_back(struct qbman_swp *s, + struct qbman_pull_desc *d); + +const struct dpaa2_dq *qbman_swp_dqrr_next_direct(struct qbman_swp *s); +const struct dpaa2_dq *qbman_swp_dqrr_next_mem_back(struct qbman_swp *s); + +static int qbman_swp_release_direct(struct qbman_swp *s, + const struct qbman_release_desc *d, + const u64 *buffers, + unsigned int num_buffers); +static int qbman_swp_release_mem_back(struct qbman_swp *s, + const struct qbman_release_desc *d, + const u64 *buffers, + unsigned int num_buffers); + +/* Function pointers */ +int (*qbman_swp_enqueue_ptr)(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd) + = qbman_swp_enqueue_direct; + +int (*qbman_swp_enqueue_multiple_ptr)(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + uint32_t *flags, + int num_frames) + = qbman_swp_enqueue_multiple_direct; + +int +(*qbman_swp_enqueue_multiple_desc_ptr)(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + int num_frames) + = qbman_swp_enqueue_multiple_desc_direct; + +int (*qbman_swp_pull_ptr)(struct qbman_swp *s, struct qbman_pull_desc *d) + = qbman_swp_pull_direct; + +const struct dpaa2_dq *(*qbman_swp_dqrr_next_ptr)(struct qbman_swp *s) + = qbman_swp_dqrr_next_direct; + +int (*qbman_swp_release_ptr)(struct qbman_swp *s, + const struct qbman_release_desc *d, + const u64 *buffers, + unsigned int num_buffers) + = qbman_swp_release_direct; + +/* Portal Access */ + +static inline u32 qbman_read_register(struct qbman_swp *p, u32 offset) +{ + return readl_relaxed(p->addr_cinh + offset); +} + +static inline void qbman_write_register(struct qbman_swp *p, u32 offset, + u32 value) +{ + writel_relaxed(value, p->addr_cinh + offset); +} + +static inline void *qbman_get_cmd(struct qbman_swp *p, u32 offset) +{ + return p->addr_cena + offset; +} + +#define QBMAN_CINH_SWP_CFG 0xd00 + +#define SWP_CFG_DQRR_MF_SHIFT 20 +#define SWP_CFG_EST_SHIFT 16 +#define SWP_CFG_CPBS_SHIFT 15 +#define SWP_CFG_WN_SHIFT 14 +#define SWP_CFG_RPM_SHIFT 12 +#define SWP_CFG_DCM_SHIFT 10 +#define SWP_CFG_EPM_SHIFT 8 +#define SWP_CFG_VPM_SHIFT 7 +#define SWP_CFG_CPM_SHIFT 6 +#define SWP_CFG_SD_SHIFT 5 +#define SWP_CFG_SP_SHIFT 4 +#define SWP_CFG_SE_SHIFT 3 +#define SWP_CFG_DP_SHIFT 2 +#define SWP_CFG_DE_SHIFT 1 +#define SWP_CFG_EP_SHIFT 0 + +static inline u32 qbman_set_swp_cfg(u8 max_fill, u8 wn, u8 est, u8 rpm, u8 dcm, + u8 epm, int sd, int sp, int se, + int dp, int de, int ep) +{ + return (max_fill << SWP_CFG_DQRR_MF_SHIFT | + est << SWP_CFG_EST_SHIFT | + wn << SWP_CFG_WN_SHIFT | + rpm << SWP_CFG_RPM_SHIFT | + dcm << SWP_CFG_DCM_SHIFT | + epm << SWP_CFG_EPM_SHIFT | + sd << SWP_CFG_SD_SHIFT | + sp << SWP_CFG_SP_SHIFT | + se << SWP_CFG_SE_SHIFT | + dp << SWP_CFG_DP_SHIFT | + de << SWP_CFG_DE_SHIFT | + ep << SWP_CFG_EP_SHIFT); +} + +#define QMAN_RT_MODE 0x00000100 + +static inline u8 qm_cyc_diff(u8 ringsize, u8 first, u8 last) +{ + /* 'first' is included, 'last' is excluded */ + if (first <= last) + return last - first; + else + return (2 * ringsize) - (first - last); +} + +/** + * qbman_swp_init() - Create a functional object representing the given + * QBMan portal descriptor. + * @d: the given qbman swp descriptor + * + * Return qbman_swp portal for success, NULL if the object cannot + * be created. + */ +struct qbman_swp *qbman_swp_init(const struct qbman_swp_desc *d) +{ + struct qbman_swp *p = kzalloc(sizeof(*p), GFP_KERNEL); + u32 reg; + u32 mask_size; + u32 eqcr_pi; + + if (!p) + return NULL; + + spin_lock_init(&p->access_spinlock); + + p->desc = d; + p->mc.valid_bit = QB_VALID_BIT; + p->sdq = 0; + p->sdq |= qbman_sdqcr_dct_prio_ics << QB_SDQCR_DCT_SHIFT; + p->sdq |= qbman_sdqcr_fc_up_to_3 << QB_SDQCR_FC_SHIFT; + p->sdq |= QMAN_SDQCR_TOKEN << QB_SDQCR_TOK_SHIFT; + if ((p->desc->qman_version & QMAN_REV_MASK) >= QMAN_REV_5000) + p->mr.valid_bit = QB_VALID_BIT; + + atomic_set(&p->vdq.available, 1); + p->vdq.valid_bit = QB_VALID_BIT; + p->dqrr.next_idx = 0; + p->dqrr.valid_bit = QB_VALID_BIT; + + if ((p->desc->qman_version & QMAN_REV_MASK) < QMAN_REV_4100) { + p->dqrr.dqrr_size = 4; + p->dqrr.reset_bug = 1; + } else { + p->dqrr.dqrr_size = 8; + p->dqrr.reset_bug = 0; + } + + p->addr_cena = d->cena_bar; + p->addr_cinh = d->cinh_bar; + + if ((p->desc->qman_version & QMAN_REV_MASK) < QMAN_REV_5000) { + + reg = qbman_set_swp_cfg(p->dqrr.dqrr_size, + 1, /* Writes Non-cacheable */ + 0, /* EQCR_CI stashing threshold */ + 3, /* RPM: RCR in array mode */ + 2, /* DCM: Discrete consumption ack */ + 2, /* EPM: EQCR in ring mode */ + 1, /* mem stashing drop enable enable */ + 1, /* mem stashing priority enable */ + 1, /* mem stashing enable */ + 1, /* dequeue stashing priority enable */ + 0, /* dequeue stashing enable enable */ + 0); /* EQCR_CI stashing priority enable */ + } else { + memset(p->addr_cena, 0, 64 * 1024); + reg = qbman_set_swp_cfg(p->dqrr.dqrr_size, + 1, /* Writes Non-cacheable */ + 1, /* EQCR_CI stashing threshold */ + 3, /* RPM: RCR in array mode */ + 2, /* DCM: Discrete consumption ack */ + 0, /* EPM: EQCR in ring mode */ + 1, /* mem stashing drop enable */ + 1, /* mem stashing priority enable */ + 1, /* mem stashing enable */ + 1, /* dequeue stashing priority enable */ + 0, /* dequeue stashing enable */ + 0); /* EQCR_CI stashing priority enable */ + reg |= 1 << SWP_CFG_CPBS_SHIFT | /* memory-backed mode */ + 1 << SWP_CFG_VPM_SHIFT | /* VDQCR read triggered mode */ + 1 << SWP_CFG_CPM_SHIFT; /* CR read triggered mode */ + } + + qbman_write_register(p, QBMAN_CINH_SWP_CFG, reg); + reg = qbman_read_register(p, QBMAN_CINH_SWP_CFG); + if (!reg) { + pr_err("qbman: the portal is not enabled!\n"); + kfree(p); + return NULL; + } + + if ((p->desc->qman_version & QMAN_REV_MASK) >= QMAN_REV_5000) { + qbman_write_register(p, QBMAN_CINH_SWP_EQCR_PI, QMAN_RT_MODE); + qbman_write_register(p, QBMAN_CINH_SWP_RCR_PI, QMAN_RT_MODE); + } + /* + * SDQCR needs to be initialized to 0 when no channels are + * being dequeued from or else the QMan HW will indicate an + * error. The values that were calculated above will be + * applied when dequeues from a specific channel are enabled. + */ + qbman_write_register(p, QBMAN_CINH_SWP_SDQCR, 0); + + p->eqcr.pi_ring_size = 8; + if ((p->desc->qman_version & QMAN_REV_MASK) >= QMAN_REV_5000) { + p->eqcr.pi_ring_size = 32; + qbman_swp_enqueue_ptr = + qbman_swp_enqueue_mem_back; + qbman_swp_enqueue_multiple_ptr = + qbman_swp_enqueue_multiple_mem_back; + qbman_swp_enqueue_multiple_desc_ptr = + qbman_swp_enqueue_multiple_desc_mem_back; + qbman_swp_pull_ptr = qbman_swp_pull_mem_back; + qbman_swp_dqrr_next_ptr = qbman_swp_dqrr_next_mem_back; + qbman_swp_release_ptr = qbman_swp_release_mem_back; + } + + for (mask_size = p->eqcr.pi_ring_size; mask_size > 0; mask_size >>= 1) + p->eqcr.pi_ci_mask = (p->eqcr.pi_ci_mask << 1) + 1; + eqcr_pi = qbman_read_register(p, QBMAN_CINH_SWP_EQCR_PI); + p->eqcr.pi = eqcr_pi & p->eqcr.pi_ci_mask; + p->eqcr.pi_vb = eqcr_pi & QB_VALID_BIT; + p->eqcr.ci = qbman_read_register(p, QBMAN_CINH_SWP_EQCR_CI) + & p->eqcr.pi_ci_mask; + p->eqcr.available = p->eqcr.pi_ring_size; + + /* Initialize the software portal with a irq timeout period of 0us */ + qbman_swp_set_irq_coalescing(p, p->dqrr.dqrr_size - 1, 0); + + return p; +} + +/** + * qbman_swp_finish() - Create and destroy a functional object representing + * the given QBMan portal descriptor. + * @p: the qbman_swp object to be destroyed + */ +void qbman_swp_finish(struct qbman_swp *p) +{ + kfree(p); +} + +/** + * qbman_swp_interrupt_read_status() + * @p: the given software portal + * + * Return the value in the SWP_ISR register. + */ +u32 qbman_swp_interrupt_read_status(struct qbman_swp *p) +{ + return qbman_read_register(p, QBMAN_CINH_SWP_ISR); +} + +/** + * qbman_swp_interrupt_clear_status() + * @p: the given software portal + * @mask: The mask to clear in SWP_ISR register + */ +void qbman_swp_interrupt_clear_status(struct qbman_swp *p, u32 mask) +{ + qbman_write_register(p, QBMAN_CINH_SWP_ISR, mask); +} + +/** + * qbman_swp_interrupt_get_trigger() - read interrupt enable register + * @p: the given software portal + * + * Return the value in the SWP_IER register. + */ +u32 qbman_swp_interrupt_get_trigger(struct qbman_swp *p) +{ + return qbman_read_register(p, QBMAN_CINH_SWP_IER); +} + +/** + * qbman_swp_interrupt_set_trigger() - enable interrupts for a swp + * @p: the given software portal + * @mask: The mask of bits to enable in SWP_IER + */ +void qbman_swp_interrupt_set_trigger(struct qbman_swp *p, u32 mask) +{ + qbman_write_register(p, QBMAN_CINH_SWP_IER, mask); +} + +/** + * qbman_swp_interrupt_get_inhibit() - read interrupt mask register + * @p: the given software portal object + * + * Return the value in the SWP_IIR register. + */ +int qbman_swp_interrupt_get_inhibit(struct qbman_swp *p) +{ + return qbman_read_register(p, QBMAN_CINH_SWP_IIR); +} + +/** + * qbman_swp_interrupt_set_inhibit() - write interrupt mask register + * @p: the given software portal object + * @inhibit: whether to inhibit the IRQs + */ +void qbman_swp_interrupt_set_inhibit(struct qbman_swp *p, int inhibit) +{ + qbman_write_register(p, QBMAN_CINH_SWP_IIR, inhibit ? 0xffffffff : 0); +} + +/* + * Different management commands all use this common base layer of code to issue + * commands and poll for results. + */ + +/* + * Returns a pointer to where the caller should fill in their management command + * (caller should ignore the verb byte) + */ +void *qbman_swp_mc_start(struct qbman_swp *p) +{ + if ((p->desc->qman_version & QMAN_REV_MASK) < QMAN_REV_5000) + return qbman_get_cmd(p, QBMAN_CENA_SWP_CR); + else + return qbman_get_cmd(p, QBMAN_CENA_SWP_CR_MEM); +} + +/* + * Commits merges in the caller-supplied command verb (which should not include + * the valid-bit) and submits the command to hardware + */ +void qbman_swp_mc_submit(struct qbman_swp *p, void *cmd, u8 cmd_verb) +{ + u8 *v = cmd; + + if ((p->desc->qman_version & QMAN_REV_MASK) < QMAN_REV_5000) { + dma_wmb(); + *v = cmd_verb | p->mc.valid_bit; + } else { + *v = cmd_verb | p->mc.valid_bit; + dma_wmb(); + qbman_write_register(p, QBMAN_CINH_SWP_CR_RT, QMAN_RT_MODE); + } +} + +/* + * Checks for a completed response (returns non-NULL if only if the response + * is complete). + */ +void *qbman_swp_mc_result(struct qbman_swp *p) +{ + u32 *ret, verb; + + if ((p->desc->qman_version & QMAN_REV_MASK) < QMAN_REV_5000) { + ret = qbman_get_cmd(p, QBMAN_CENA_SWP_RR(p->mc.valid_bit)); + /* Remove the valid-bit - command completed if the rest + * is non-zero. + */ + verb = ret[0] & ~QB_VALID_BIT; + if (!verb) + return NULL; + p->mc.valid_bit ^= QB_VALID_BIT; + } else { + ret = qbman_get_cmd(p, QBMAN_CENA_SWP_RR_MEM); + /* Command completed if the valid bit is toggled */ + if (p->mr.valid_bit != (ret[0] & QB_VALID_BIT)) + return NULL; + /* Command completed if the rest is non-zero */ + verb = ret[0] & ~QB_VALID_BIT; + if (!verb) + return NULL; + p->mr.valid_bit ^= QB_VALID_BIT; + } + + return ret; +} + +#define QB_ENQUEUE_CMD_OPTIONS_SHIFT 0 +enum qb_enqueue_commands { + enqueue_empty = 0, + enqueue_response_always = 1, + enqueue_rejects_to_fq = 2 +}; + +#define QB_ENQUEUE_CMD_ORP_ENABLE_SHIFT 2 +#define QB_ENQUEUE_CMD_IRQ_ON_DISPATCH_SHIFT 3 +#define QB_ENQUEUE_CMD_TARGET_TYPE_SHIFT 4 +#define QB_ENQUEUE_CMD_DCA_EN_SHIFT 7 + +/* + * qbman_eq_desc_clear() - Clear the contents of a descriptor to + * default/starting state. + */ +void qbman_eq_desc_clear(struct qbman_eq_desc *d) +{ + memset(d, 0, sizeof(*d)); +} + +/** + * qbman_eq_desc_set_no_orp() - Set enqueue descriptor without orp + * @d: the enqueue descriptor. + * @respond_success: 1 = enqueue with response always; 0 = enqueue with + * rejections returned on a FQ. + */ +void qbman_eq_desc_set_no_orp(struct qbman_eq_desc *d, int respond_success) +{ + d->verb &= ~(1 << QB_ENQUEUE_CMD_ORP_ENABLE_SHIFT); + if (respond_success) + d->verb |= enqueue_response_always; + else + d->verb |= enqueue_rejects_to_fq; +} + +/* + * Exactly one of the following descriptor "targets" should be set. (Calling any + * one of these will replace the effect of any prior call to one of these.) + * -enqueue to a frame queue + * -enqueue to a queuing destination + */ + +/** + * qbman_eq_desc_set_fq() - set the FQ for the enqueue command + * @d: the enqueue descriptor + * @fqid: the id of the frame queue to be enqueued + */ +void qbman_eq_desc_set_fq(struct qbman_eq_desc *d, u32 fqid) +{ + d->verb &= ~(1 << QB_ENQUEUE_CMD_TARGET_TYPE_SHIFT); + d->tgtid = cpu_to_le32(fqid); +} + +/** + * qbman_eq_desc_set_qd() - Set Queuing Destination for the enqueue command + * @d: the enqueue descriptor + * @qdid: the id of the queuing destination to be enqueued + * @qd_bin: the queuing destination bin + * @qd_prio: the queuing destination priority + */ +void qbman_eq_desc_set_qd(struct qbman_eq_desc *d, u32 qdid, + u32 qd_bin, u32 qd_prio) +{ + d->verb |= 1 << QB_ENQUEUE_CMD_TARGET_TYPE_SHIFT; + d->tgtid = cpu_to_le32(qdid); + d->qdbin = cpu_to_le16(qd_bin); + d->qpri = qd_prio; +} + +#define EQAR_IDX(eqar) ((eqar) & 0x7) +#define EQAR_VB(eqar) ((eqar) & 0x80) +#define EQAR_SUCCESS(eqar) ((eqar) & 0x100) + +#define QB_RT_BIT ((u32)0x100) +/** + * qbman_swp_enqueue_direct() - Issue an enqueue command + * @s: the software portal used for enqueue + * @d: the enqueue descriptor + * @fd: the frame descriptor to be enqueued + * + * Please note that 'fd' should only be NULL if the "action" of the + * descriptor is "orp_hole" or "orp_nesn". + * + * Return 0 for successful enqueue, -EBUSY if the EQCR is not ready. + */ +static +int qbman_swp_enqueue_direct(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd) +{ + int flags = 0; + int ret = qbman_swp_enqueue_multiple_direct(s, d, fd, &flags, 1); + + if (ret >= 0) + ret = 0; + else + ret = -EBUSY; + return ret; +} + +/** + * qbman_swp_enqueue_mem_back() - Issue an enqueue command + * @s: the software portal used for enqueue + * @d: the enqueue descriptor + * @fd: the frame descriptor to be enqueued + * + * Please note that 'fd' should only be NULL if the "action" of the + * descriptor is "orp_hole" or "orp_nesn". + * + * Return 0 for successful enqueue, -EBUSY if the EQCR is not ready. + */ +static +int qbman_swp_enqueue_mem_back(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd) +{ + int flags = 0; + int ret = qbman_swp_enqueue_multiple_mem_back(s, d, fd, &flags, 1); + + if (ret >= 0) + ret = 0; + else + ret = -EBUSY; + return ret; +} + +/** + * qbman_swp_enqueue_multiple_direct() - Issue a multi enqueue command + * using one enqueue descriptor + * @s: the software portal used for enqueue + * @d: the enqueue descriptor + * @fd: table pointer of frame descriptor table to be enqueued + * @flags: table pointer of QBMAN_ENQUEUE_FLAG_DCA flags, not used if NULL + * @num_frames: number of fd to be enqueued + * + * Return the number of fd enqueued, or a negative error number. + */ +static +int qbman_swp_enqueue_multiple_direct(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + uint32_t *flags, + int num_frames) +{ + uint32_t *p = NULL; + const uint32_t *cl = (uint32_t *)d; + uint32_t eqcr_ci, eqcr_pi, half_mask, full_mask; + int i, num_enqueued = 0; + + spin_lock(&s->access_spinlock); + half_mask = (s->eqcr.pi_ci_mask>>1); + full_mask = s->eqcr.pi_ci_mask; + + if (!s->eqcr.available) { + eqcr_ci = s->eqcr.ci; + p = s->addr_cena + QBMAN_CENA_SWP_EQCR_CI; + s->eqcr.ci = qbman_read_register(s, QBMAN_CINH_SWP_EQCR_CI); + s->eqcr.ci &= full_mask; + + s->eqcr.available = qm_cyc_diff(s->eqcr.pi_ring_size, + eqcr_ci, s->eqcr.ci); + if (!s->eqcr.available) { + spin_unlock(&s->access_spinlock); + return 0; + } + } + + eqcr_pi = s->eqcr.pi; + num_enqueued = (s->eqcr.available < num_frames) ? + s->eqcr.available : num_frames; + s->eqcr.available -= num_enqueued; + /* Fill in the EQCR ring */ + for (i = 0; i < num_enqueued; i++) { + p = (s->addr_cena + QBMAN_CENA_SWP_EQCR(eqcr_pi & half_mask)); + /* Skip copying the verb */ + memcpy(&p[1], &cl[1], EQ_DESC_SIZE_WITHOUT_FD - 1); + memcpy(&p[EQ_DESC_SIZE_FD_START/sizeof(uint32_t)], + &fd[i], sizeof(*fd)); + eqcr_pi++; + } + + dma_wmb(); + + /* Set the verb byte, have to substitute in the valid-bit */ + eqcr_pi = s->eqcr.pi; + for (i = 0; i < num_enqueued; i++) { + p = (s->addr_cena + QBMAN_CENA_SWP_EQCR(eqcr_pi & half_mask)); + p[0] = cl[0] | s->eqcr.pi_vb; + if (flags && (flags[i] & QBMAN_ENQUEUE_FLAG_DCA)) { + struct qbman_eq_desc *eq_desc = (struct qbman_eq_desc *)p; + + eq_desc->dca = (1 << QB_ENQUEUE_CMD_DCA_EN_SHIFT) | + ((flags[i]) & QBMAN_EQCR_DCA_IDXMASK); + } + eqcr_pi++; + if (!(eqcr_pi & half_mask)) + s->eqcr.pi_vb ^= QB_VALID_BIT; + } + + /* Flush all the cacheline without load/store in between */ + eqcr_pi = s->eqcr.pi; + for (i = 0; i < num_enqueued; i++) + eqcr_pi++; + s->eqcr.pi = eqcr_pi & full_mask; + spin_unlock(&s->access_spinlock); + + return num_enqueued; +} + +/** + * qbman_swp_enqueue_multiple_mem_back() - Issue a multi enqueue command + * using one enqueue descriptor + * @s: the software portal used for enqueue + * @d: the enqueue descriptor + * @fd: table pointer of frame descriptor table to be enqueued + * @flags: table pointer of QBMAN_ENQUEUE_FLAG_DCA flags, not used if NULL + * @num_frames: number of fd to be enqueued + * + * Return the number of fd enqueued, or a negative error number. + */ +static +int qbman_swp_enqueue_multiple_mem_back(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + uint32_t *flags, + int num_frames) +{ + uint32_t *p = NULL; + const uint32_t *cl = (uint32_t *)(d); + uint32_t eqcr_ci, eqcr_pi, half_mask, full_mask; + int i, num_enqueued = 0; + unsigned long irq_flags; + + spin_lock_irqsave(&s->access_spinlock, irq_flags); + + half_mask = (s->eqcr.pi_ci_mask>>1); + full_mask = s->eqcr.pi_ci_mask; + if (!s->eqcr.available) { + eqcr_ci = s->eqcr.ci; + s->eqcr.ci = qbman_read_register(s, QBMAN_CINH_SWP_EQCR_CI); + s->eqcr.ci &= full_mask; + s->eqcr.available = qm_cyc_diff(s->eqcr.pi_ring_size, + eqcr_ci, s->eqcr.ci); + if (!s->eqcr.available) { + spin_unlock_irqrestore(&s->access_spinlock, irq_flags); + return 0; + } + } + + eqcr_pi = s->eqcr.pi; + num_enqueued = (s->eqcr.available < num_frames) ? + s->eqcr.available : num_frames; + s->eqcr.available -= num_enqueued; + /* Fill in the EQCR ring */ + for (i = 0; i < num_enqueued; i++) { + p = (s->addr_cena + QBMAN_CENA_SWP_EQCR(eqcr_pi & half_mask)); + /* Skip copying the verb */ + memcpy(&p[1], &cl[1], EQ_DESC_SIZE_WITHOUT_FD - 1); + memcpy(&p[EQ_DESC_SIZE_FD_START/sizeof(uint32_t)], + &fd[i], sizeof(*fd)); + eqcr_pi++; + } + + /* Set the verb byte, have to substitute in the valid-bit */ + eqcr_pi = s->eqcr.pi; + for (i = 0; i < num_enqueued; i++) { + p = (s->addr_cena + QBMAN_CENA_SWP_EQCR(eqcr_pi & half_mask)); + p[0] = cl[0] | s->eqcr.pi_vb; + if (flags && (flags[i] & QBMAN_ENQUEUE_FLAG_DCA)) { + struct qbman_eq_desc *eq_desc = (struct qbman_eq_desc *)p; + + eq_desc->dca = (1 << QB_ENQUEUE_CMD_DCA_EN_SHIFT) | + ((flags[i]) & QBMAN_EQCR_DCA_IDXMASK); + } + eqcr_pi++; + if (!(eqcr_pi & half_mask)) + s->eqcr.pi_vb ^= QB_VALID_BIT; + } + s->eqcr.pi = eqcr_pi & full_mask; + + dma_wmb(); + qbman_write_register(s, QBMAN_CINH_SWP_EQCR_PI, + (QB_RT_BIT)|(s->eqcr.pi)|s->eqcr.pi_vb); + spin_unlock_irqrestore(&s->access_spinlock, irq_flags); + + return num_enqueued; +} + +/** + * qbman_swp_enqueue_multiple_desc_direct() - Issue a multi enqueue command + * using multiple enqueue descriptor + * @s: the software portal used for enqueue + * @d: table of minimal enqueue descriptor + * @fd: table pointer of frame descriptor table to be enqueued + * @num_frames: number of fd to be enqueued + * + * Return the number of fd enqueued, or a negative error number. + */ +static +int qbman_swp_enqueue_multiple_desc_direct(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + int num_frames) +{ + uint32_t *p; + const uint32_t *cl; + uint32_t eqcr_ci, eqcr_pi, half_mask, full_mask; + int i, num_enqueued = 0; + + half_mask = (s->eqcr.pi_ci_mask>>1); + full_mask = s->eqcr.pi_ci_mask; + if (!s->eqcr.available) { + eqcr_ci = s->eqcr.ci; + p = s->addr_cena + QBMAN_CENA_SWP_EQCR_CI; + s->eqcr.ci = qbman_read_register(s, QBMAN_CINH_SWP_EQCR_CI); + s->eqcr.available = qm_cyc_diff(s->eqcr.pi_ring_size, + eqcr_ci, s->eqcr.ci); + if (!s->eqcr.available) + return 0; + } + + eqcr_pi = s->eqcr.pi; + num_enqueued = (s->eqcr.available < num_frames) ? + s->eqcr.available : num_frames; + s->eqcr.available -= num_enqueued; + /* Fill in the EQCR ring */ + for (i = 0; i < num_enqueued; i++) { + p = (s->addr_cena + QBMAN_CENA_SWP_EQCR(eqcr_pi & half_mask)); + cl = (uint32_t *)(&d[i]); + /* Skip copying the verb */ + memcpy(&p[1], &cl[1], EQ_DESC_SIZE_WITHOUT_FD - 1); + memcpy(&p[EQ_DESC_SIZE_FD_START/sizeof(uint32_t)], + &fd[i], sizeof(*fd)); + eqcr_pi++; + } + + dma_wmb(); + + /* Set the verb byte, have to substitute in the valid-bit */ + eqcr_pi = s->eqcr.pi; + for (i = 0; i < num_enqueued; i++) { + p = (s->addr_cena + QBMAN_CENA_SWP_EQCR(eqcr_pi & half_mask)); + cl = (uint32_t *)(&d[i]); + p[0] = cl[0] | s->eqcr.pi_vb; + eqcr_pi++; + if (!(eqcr_pi & half_mask)) + s->eqcr.pi_vb ^= QB_VALID_BIT; + } + + /* Flush all the cacheline without load/store in between */ + eqcr_pi = s->eqcr.pi; + for (i = 0; i < num_enqueued; i++) + eqcr_pi++; + s->eqcr.pi = eqcr_pi & full_mask; + + return num_enqueued; +} + +/** + * qbman_swp_enqueue_multiple_desc_mem_back() - Issue a multi enqueue command + * using multiple enqueue descriptor + * @s: the software portal used for enqueue + * @d: table of minimal enqueue descriptor + * @fd: table pointer of frame descriptor table to be enqueued + * @num_frames: number of fd to be enqueued + * + * Return the number of fd enqueued, or a negative error number. + */ +static +int qbman_swp_enqueue_multiple_desc_mem_back(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + int num_frames) +{ + uint32_t *p; + const uint32_t *cl; + uint32_t eqcr_ci, eqcr_pi, half_mask, full_mask; + int i, num_enqueued = 0; + + half_mask = (s->eqcr.pi_ci_mask>>1); + full_mask = s->eqcr.pi_ci_mask; + if (!s->eqcr.available) { + eqcr_ci = s->eqcr.ci; + s->eqcr.ci = qbman_read_register(s, QBMAN_CINH_SWP_EQCR_CI); + s->eqcr.ci &= full_mask; + s->eqcr.available = qm_cyc_diff(s->eqcr.pi_ring_size, + eqcr_ci, s->eqcr.ci); + if (!s->eqcr.available) + return 0; + } + + eqcr_pi = s->eqcr.pi; + num_enqueued = (s->eqcr.available < num_frames) ? + s->eqcr.available : num_frames; + s->eqcr.available -= num_enqueued; + /* Fill in the EQCR ring */ + for (i = 0; i < num_enqueued; i++) { + p = (s->addr_cena + QBMAN_CENA_SWP_EQCR(eqcr_pi & half_mask)); + cl = (uint32_t *)(&d[i]); + /* Skip copying the verb */ + memcpy(&p[1], &cl[1], EQ_DESC_SIZE_WITHOUT_FD - 1); + memcpy(&p[EQ_DESC_SIZE_FD_START/sizeof(uint32_t)], + &fd[i], sizeof(*fd)); + eqcr_pi++; + } + + /* Set the verb byte, have to substitute in the valid-bit */ + eqcr_pi = s->eqcr.pi; + for (i = 0; i < num_enqueued; i++) { + p = (s->addr_cena + QBMAN_CENA_SWP_EQCR(eqcr_pi & half_mask)); + cl = (uint32_t *)(&d[i]); + p[0] = cl[0] | s->eqcr.pi_vb; + eqcr_pi++; + if (!(eqcr_pi & half_mask)) + s->eqcr.pi_vb ^= QB_VALID_BIT; + } + + s->eqcr.pi = eqcr_pi & full_mask; + + dma_wmb(); + qbman_write_register(s, QBMAN_CINH_SWP_EQCR_PI, + (QB_RT_BIT)|(s->eqcr.pi)|s->eqcr.pi_vb); + + return num_enqueued; +} + +/* Static (push) dequeue */ + +/** + * qbman_swp_push_get() - Get the push dequeue setup + * @s: the software portal object + * @channel_idx: the channel index to query + * @enabled: returned boolean to show whether the push dequeue is enabled + * for the given channel + */ +void qbman_swp_push_get(struct qbman_swp *s, u8 channel_idx, int *enabled) +{ + u16 src = (s->sdq >> QB_SDQCR_SRC_SHIFT) & QB_SDQCR_SRC_MASK; + + WARN_ON(channel_idx > 15); + *enabled = src | (1 << channel_idx); +} + +/** + * qbman_swp_push_set() - Enable or disable push dequeue + * @s: the software portal object + * @channel_idx: the channel index (0 to 15) + * @enable: enable or disable push dequeue + */ +void qbman_swp_push_set(struct qbman_swp *s, u8 channel_idx, int enable) +{ + u16 dqsrc; + + WARN_ON(channel_idx > 15); + if (enable) + s->sdq |= 1 << channel_idx; + else + s->sdq &= ~(1 << channel_idx); + + /* Read make the complete src map. If no channels are enabled + * the SDQCR must be 0 or else QMan will assert errors + */ + dqsrc = (s->sdq >> QB_SDQCR_SRC_SHIFT) & QB_SDQCR_SRC_MASK; + if (dqsrc != 0) + qbman_write_register(s, QBMAN_CINH_SWP_SDQCR, s->sdq); + else + qbman_write_register(s, QBMAN_CINH_SWP_SDQCR, 0); +} + +#define QB_VDQCR_VERB_DCT_SHIFT 0 +#define QB_VDQCR_VERB_DT_SHIFT 2 +#define QB_VDQCR_VERB_RLS_SHIFT 4 +#define QB_VDQCR_VERB_WAE_SHIFT 5 + +enum qb_pull_dt_e { + qb_pull_dt_channel, + qb_pull_dt_workqueue, + qb_pull_dt_framequeue +}; + +/** + * qbman_pull_desc_clear() - Clear the contents of a descriptor to + * default/starting state + * @d: the pull dequeue descriptor to be cleared + */ +void qbman_pull_desc_clear(struct qbman_pull_desc *d) +{ + memset(d, 0, sizeof(*d)); +} + +/** + * qbman_pull_desc_set_storage()- Set the pull dequeue storage + * @d: the pull dequeue descriptor to be set + * @storage: the pointer of the memory to store the dequeue result + * @storage_phys: the physical address of the storage memory + * @stash: to indicate whether write allocate is enabled + * + * If not called, or if called with 'storage' as NULL, the result pull dequeues + * will produce results to DQRR. If 'storage' is non-NULL, then results are + * produced to the given memory location (using the DMA address which + * the caller provides in 'storage_phys'), and 'stash' controls whether or not + * those writes to main-memory express a cache-warming attribute. + */ +void qbman_pull_desc_set_storage(struct qbman_pull_desc *d, + struct dpaa2_dq *storage, + dma_addr_t storage_phys, + int stash) +{ + /* save the virtual address */ + d->rsp_addr_virt = (u64)(uintptr_t)storage; + + if (!storage) { + d->verb &= ~(1 << QB_VDQCR_VERB_RLS_SHIFT); + return; + } + d->verb |= 1 << QB_VDQCR_VERB_RLS_SHIFT; + if (stash) + d->verb |= 1 << QB_VDQCR_VERB_WAE_SHIFT; + else + d->verb &= ~(1 << QB_VDQCR_VERB_WAE_SHIFT); + + d->rsp_addr = cpu_to_le64(storage_phys); +} + +/** + * qbman_pull_desc_set_numframes() - Set the number of frames to be dequeued + * @d: the pull dequeue descriptor to be set + * @numframes: number of frames to be set, must be between 1 and 16, inclusive + */ +void qbman_pull_desc_set_numframes(struct qbman_pull_desc *d, u8 numframes) +{ + d->numf = numframes - 1; +} + +/* + * Exactly one of the following descriptor "actions" should be set. (Calling any + * one of these will replace the effect of any prior call to one of these.) + * - pull dequeue from the given frame queue (FQ) + * - pull dequeue from any FQ in the given work queue (WQ) + * - pull dequeue from any FQ in any WQ in the given channel + */ + +/** + * qbman_pull_desc_set_fq() - Set fqid from which the dequeue command dequeues + * @d: the pull dequeue descriptor to be set + * @fqid: the frame queue index of the given FQ + */ +void qbman_pull_desc_set_fq(struct qbman_pull_desc *d, u32 fqid) +{ + d->verb |= 1 << QB_VDQCR_VERB_DCT_SHIFT; + d->verb |= qb_pull_dt_framequeue << QB_VDQCR_VERB_DT_SHIFT; + d->dq_src = cpu_to_le32(fqid); +} + +/** + * qbman_pull_desc_set_wq() - Set wqid from which the dequeue command dequeues + * @d: the pull dequeue descriptor to be set + * @wqid: composed of channel id and wqid within the channel + * @dct: the dequeue command type + */ +void qbman_pull_desc_set_wq(struct qbman_pull_desc *d, u32 wqid, + enum qbman_pull_type_e dct) +{ + d->verb |= dct << QB_VDQCR_VERB_DCT_SHIFT; + d->verb |= qb_pull_dt_workqueue << QB_VDQCR_VERB_DT_SHIFT; + d->dq_src = cpu_to_le32(wqid); +} + +/** + * qbman_pull_desc_set_channel() - Set channelid from which the dequeue command + * dequeues + * @d: the pull dequeue descriptor to be set + * @chid: the channel id to be dequeued + * @dct: the dequeue command type + */ +void qbman_pull_desc_set_channel(struct qbman_pull_desc *d, u32 chid, + enum qbman_pull_type_e dct) +{ + d->verb |= dct << QB_VDQCR_VERB_DCT_SHIFT; + d->verb |= qb_pull_dt_channel << QB_VDQCR_VERB_DT_SHIFT; + d->dq_src = cpu_to_le32(chid); +} + +/** + * qbman_swp_pull_direct() - Issue the pull dequeue command + * @s: the software portal object + * @d: the software portal descriptor which has been configured with + * the set of qbman_pull_desc_set_*() calls + * + * Return 0 for success, and -EBUSY if the software portal is not ready + * to do pull dequeue. + */ +static +int qbman_swp_pull_direct(struct qbman_swp *s, struct qbman_pull_desc *d) +{ + struct qbman_pull_desc *p; + + if (!atomic_dec_and_test(&s->vdq.available)) { + atomic_inc(&s->vdq.available); + return -EBUSY; + } + s->vdq.storage = (void *)(uintptr_t)d->rsp_addr_virt; + if ((s->desc->qman_version & QMAN_REV_MASK) < QMAN_REV_5000) + p = qbman_get_cmd(s, QBMAN_CENA_SWP_VDQCR); + else + p = qbman_get_cmd(s, QBMAN_CENA_SWP_VDQCR_MEM); + p->numf = d->numf; + p->tok = QMAN_DQ_TOKEN_VALID; + p->dq_src = d->dq_src; + p->rsp_addr = d->rsp_addr; + p->rsp_addr_virt = d->rsp_addr_virt; + dma_wmb(); + /* Set the verb byte, have to substitute in the valid-bit */ + p->verb = d->verb | s->vdq.valid_bit; + s->vdq.valid_bit ^= QB_VALID_BIT; + + return 0; +} + +/** + * qbman_swp_pull_mem_back() - Issue the pull dequeue command + * @s: the software portal object + * @d: the software portal descriptor which has been configured with + * the set of qbman_pull_desc_set_*() calls + * + * Return 0 for success, and -EBUSY if the software portal is not ready + * to do pull dequeue. + */ +static +int qbman_swp_pull_mem_back(struct qbman_swp *s, struct qbman_pull_desc *d) +{ + struct qbman_pull_desc *p; + + if (!atomic_dec_and_test(&s->vdq.available)) { + atomic_inc(&s->vdq.available); + return -EBUSY; + } + s->vdq.storage = (void *)(uintptr_t)d->rsp_addr_virt; + if ((s->desc->qman_version & QMAN_REV_MASK) < QMAN_REV_5000) + p = qbman_get_cmd(s, QBMAN_CENA_SWP_VDQCR); + else + p = qbman_get_cmd(s, QBMAN_CENA_SWP_VDQCR_MEM); + p->numf = d->numf; + p->tok = QMAN_DQ_TOKEN_VALID; + p->dq_src = d->dq_src; + p->rsp_addr = d->rsp_addr; + p->rsp_addr_virt = d->rsp_addr_virt; + + /* Set the verb byte, have to substitute in the valid-bit */ + p->verb = d->verb | s->vdq.valid_bit; + s->vdq.valid_bit ^= QB_VALID_BIT; + dma_wmb(); + qbman_write_register(s, QBMAN_CINH_SWP_VDQCR_RT, QMAN_RT_MODE); + + return 0; +} + +#define QMAN_DQRR_PI_MASK 0xf + +/** + * qbman_swp_dqrr_next_direct() - Get an valid DQRR entry + * @s: the software portal object + * + * Return NULL if there are no unconsumed DQRR entries. Return a DQRR entry + * only once, so repeated calls can return a sequence of DQRR entries, without + * requiring they be consumed immediately or in any particular order. + */ +const struct dpaa2_dq *qbman_swp_dqrr_next_direct(struct qbman_swp *s) +{ + u32 verb; + u32 response_verb; + u32 flags; + struct dpaa2_dq *p; + + /* Before using valid-bit to detect if something is there, we have to + * handle the case of the DQRR reset bug... + */ + if (unlikely(s->dqrr.reset_bug)) { + /* + * We pick up new entries by cache-inhibited producer index, + * which means that a non-coherent mapping would require us to + * invalidate and read *only* once that PI has indicated that + * there's an entry here. The first trip around the DQRR ring + * will be much less efficient than all subsequent trips around + * it... + */ + u8 pi = qbman_read_register(s, QBMAN_CINH_SWP_DQPI) & + QMAN_DQRR_PI_MASK; + + /* there are new entries if pi != next_idx */ + if (pi == s->dqrr.next_idx) + return NULL; + + /* + * if next_idx is/was the last ring index, and 'pi' is + * different, we can disable the workaround as all the ring + * entries have now been DMA'd to so valid-bit checking is + * repaired. Note: this logic needs to be based on next_idx + * (which increments one at a time), rather than on pi (which + * can burst and wrap-around between our snapshots of it). + */ + if (s->dqrr.next_idx == (s->dqrr.dqrr_size - 1)) { + pr_debug("next_idx=%d, pi=%d, clear reset bug\n", + s->dqrr.next_idx, pi); + s->dqrr.reset_bug = 0; + } + prefetch(qbman_get_cmd(s, + QBMAN_CENA_SWP_DQRR(s->dqrr.next_idx))); + } + + p = qbman_get_cmd(s, QBMAN_CENA_SWP_DQRR(s->dqrr.next_idx)); + verb = p->dq.verb; + + /* + * If the valid-bit isn't of the expected polarity, nothing there. Note, + * in the DQRR reset bug workaround, we shouldn't need to skip these + * check, because we've already determined that a new entry is available + * and we've invalidated the cacheline before reading it, so the + * valid-bit behaviour is repaired and should tell us what we already + * knew from reading PI. + */ + if ((verb & QB_VALID_BIT) != s->dqrr.valid_bit) { + prefetch(qbman_get_cmd(s, + QBMAN_CENA_SWP_DQRR(s->dqrr.next_idx))); + return NULL; + } + /* + * There's something there. Move "next_idx" attention to the next ring + * entry (and prefetch it) before returning what we found. + */ + s->dqrr.next_idx++; + s->dqrr.next_idx &= s->dqrr.dqrr_size - 1; /* Wrap around */ + if (!s->dqrr.next_idx) + s->dqrr.valid_bit ^= QB_VALID_BIT; + + /* + * If this is the final response to a volatile dequeue command + * indicate that the vdq is available + */ + flags = p->dq.stat; + response_verb = verb & QBMAN_RESULT_MASK; + if ((response_verb == QBMAN_RESULT_DQ) && + (flags & DPAA2_DQ_STAT_VOLATILE) && + (flags & DPAA2_DQ_STAT_EXPIRED)) + atomic_inc(&s->vdq.available); + + prefetch(qbman_get_cmd(s, QBMAN_CENA_SWP_DQRR(s->dqrr.next_idx))); + + return p; +} + +/** + * qbman_swp_dqrr_next_mem_back() - Get an valid DQRR entry + * @s: the software portal object + * + * Return NULL if there are no unconsumed DQRR entries. Return a DQRR entry + * only once, so repeated calls can return a sequence of DQRR entries, without + * requiring they be consumed immediately or in any particular order. + */ +const struct dpaa2_dq *qbman_swp_dqrr_next_mem_back(struct qbman_swp *s) +{ + u32 verb; + u32 response_verb; + u32 flags; + struct dpaa2_dq *p; + + /* Before using valid-bit to detect if something is there, we have to + * handle the case of the DQRR reset bug... + */ + if (unlikely(s->dqrr.reset_bug)) { + /* + * We pick up new entries by cache-inhibited producer index, + * which means that a non-coherent mapping would require us to + * invalidate and read *only* once that PI has indicated that + * there's an entry here. The first trip around the DQRR ring + * will be much less efficient than all subsequent trips around + * it... + */ + u8 pi = qbman_read_register(s, QBMAN_CINH_SWP_DQPI) & + QMAN_DQRR_PI_MASK; + + /* there are new entries if pi != next_idx */ + if (pi == s->dqrr.next_idx) + return NULL; + + /* + * if next_idx is/was the last ring index, and 'pi' is + * different, we can disable the workaround as all the ring + * entries have now been DMA'd to so valid-bit checking is + * repaired. Note: this logic needs to be based on next_idx + * (which increments one at a time), rather than on pi (which + * can burst and wrap-around between our snapshots of it). + */ + if (s->dqrr.next_idx == (s->dqrr.dqrr_size - 1)) { + pr_debug("next_idx=%d, pi=%d, clear reset bug\n", + s->dqrr.next_idx, pi); + s->dqrr.reset_bug = 0; + } + prefetch(qbman_get_cmd(s, + QBMAN_CENA_SWP_DQRR(s->dqrr.next_idx))); + } + + p = qbman_get_cmd(s, QBMAN_CENA_SWP_DQRR_MEM(s->dqrr.next_idx)); + verb = p->dq.verb; + + /* + * If the valid-bit isn't of the expected polarity, nothing there. Note, + * in the DQRR reset bug workaround, we shouldn't need to skip these + * check, because we've already determined that a new entry is available + * and we've invalidated the cacheline before reading it, so the + * valid-bit behaviour is repaired and should tell us what we already + * knew from reading PI. + */ + if ((verb & QB_VALID_BIT) != s->dqrr.valid_bit) { + prefetch(qbman_get_cmd(s, + QBMAN_CENA_SWP_DQRR(s->dqrr.next_idx))); + return NULL; + } + /* + * There's something there. Move "next_idx" attention to the next ring + * entry (and prefetch it) before returning what we found. + */ + s->dqrr.next_idx++; + s->dqrr.next_idx &= s->dqrr.dqrr_size - 1; /* Wrap around */ + if (!s->dqrr.next_idx) + s->dqrr.valid_bit ^= QB_VALID_BIT; + + /* + * If this is the final response to a volatile dequeue command + * indicate that the vdq is available + */ + flags = p->dq.stat; + response_verb = verb & QBMAN_RESULT_MASK; + if ((response_verb == QBMAN_RESULT_DQ) && + (flags & DPAA2_DQ_STAT_VOLATILE) && + (flags & DPAA2_DQ_STAT_EXPIRED)) + atomic_inc(&s->vdq.available); + + prefetch(qbman_get_cmd(s, QBMAN_CENA_SWP_DQRR(s->dqrr.next_idx))); + + return p; +} + +/** + * qbman_swp_dqrr_consume() - Consume DQRR entries previously returned from + * qbman_swp_dqrr_next(). + * @s: the software portal object + * @dq: the DQRR entry to be consumed + */ +void qbman_swp_dqrr_consume(struct qbman_swp *s, const struct dpaa2_dq *dq) +{ + qbman_write_register(s, QBMAN_CINH_SWP_DCAP, QBMAN_IDX_FROM_DQRR(dq)); +} + +/** + * qbman_result_has_new_result() - Check and get the dequeue response from the + * dq storage memory set in pull dequeue command + * @s: the software portal object + * @dq: the dequeue result read from the memory + * + * Return 1 for getting a valid dequeue result, or 0 for not getting a valid + * dequeue result. + * + * Only used for user-provided storage of dequeue results, not DQRR. For + * efficiency purposes, the driver will perform any required endianness + * conversion to ensure that the user's dequeue result storage is in host-endian + * format. As such, once the user has called qbman_result_has_new_result() and + * been returned a valid dequeue result, they should not call it again on + * the same memory location (except of course if another dequeue command has + * been executed to produce a new result to that location). + */ +int qbman_result_has_new_result(struct qbman_swp *s, const struct dpaa2_dq *dq) +{ + if (dq->dq.tok != QMAN_DQ_TOKEN_VALID) + return 0; + + /* + * Set token to be 0 so we will detect change back to 1 + * next time the looping is traversed. Const is cast away here + * as we want users to treat the dequeue responses as read only. + */ + ((struct dpaa2_dq *)dq)->dq.tok = 0; + + /* + * Determine whether VDQCR is available based on whether the + * current result is sitting in the first storage location of + * the busy command. + */ + if (s->vdq.storage == dq) { + s->vdq.storage = NULL; + atomic_inc(&s->vdq.available); + } + + return 1; +} + +/** + * qbman_release_desc_clear() - Clear the contents of a descriptor to + * default/starting state. + * @d: the pull dequeue descriptor to be cleared + */ +void qbman_release_desc_clear(struct qbman_release_desc *d) +{ + memset(d, 0, sizeof(*d)); + d->verb = 1 << 5; /* Release Command Valid */ +} + +/** + * qbman_release_desc_set_bpid() - Set the ID of the buffer pool to release to + * @d: the pull dequeue descriptor to be set + * @bpid: the bpid value to be set + */ +void qbman_release_desc_set_bpid(struct qbman_release_desc *d, u16 bpid) +{ + d->bpid = cpu_to_le16(bpid); +} + +/** + * qbman_release_desc_set_rcdi() - Determines whether or not the portal's RCDI + * interrupt source should be asserted after the release command is completed. + * @d: the pull dequeue descriptor to be set + * @enable: enable (1) or disable (0) value + */ +void qbman_release_desc_set_rcdi(struct qbman_release_desc *d, int enable) +{ + if (enable) + d->verb |= 1 << 6; + else + d->verb &= ~(1 << 6); +} + +#define RAR_IDX(rar) ((rar) & 0x7) +#define RAR_VB(rar) ((rar) & 0x80) +#define RAR_SUCCESS(rar) ((rar) & 0x100) + +/** + * qbman_swp_release_direct() - Issue a buffer release command + * @s: the software portal object + * @d: the release descriptor + * @buffers: a pointer pointing to the buffer address to be released + * @num_buffers: number of buffers to be released, must be less than 8 + * + * Return 0 for success, -EBUSY if the release command ring is not ready. + */ +int qbman_swp_release_direct(struct qbman_swp *s, + const struct qbman_release_desc *d, + const u64 *buffers, unsigned int num_buffers) +{ + int i; + struct qbman_release_desc *p; + u32 rar; + + if (!num_buffers || (num_buffers > 7)) + return -EINVAL; + + rar = qbman_read_register(s, QBMAN_CINH_SWP_RAR); + if (!RAR_SUCCESS(rar)) + return -EBUSY; + + /* Start the release command */ + p = qbman_get_cmd(s, QBMAN_CENA_SWP_RCR(RAR_IDX(rar))); + + /* Copy the caller's buffer pointers to the command */ + for (i = 0; i < num_buffers; i++) + p->buf[i] = cpu_to_le64(buffers[i]); + p->bpid = d->bpid; + + /* + * Set the verb byte, have to substitute in the valid-bit + * and the number of buffers. + */ + dma_wmb(); + p->verb = d->verb | RAR_VB(rar) | num_buffers; + + return 0; +} + +/** + * qbman_swp_release_mem_back() - Issue a buffer release command + * @s: the software portal object + * @d: the release descriptor + * @buffers: a pointer pointing to the buffer address to be released + * @num_buffers: number of buffers to be released, must be less than 8 + * + * Return 0 for success, -EBUSY if the release command ring is not ready. + */ +int qbman_swp_release_mem_back(struct qbman_swp *s, + const struct qbman_release_desc *d, + const u64 *buffers, unsigned int num_buffers) +{ + int i; + struct qbman_release_desc *p; + u32 rar; + + if (!num_buffers || (num_buffers > 7)) + return -EINVAL; + + rar = qbman_read_register(s, QBMAN_CINH_SWP_RAR); + if (!RAR_SUCCESS(rar)) + return -EBUSY; + + /* Start the release command */ + p = qbman_get_cmd(s, QBMAN_CENA_SWP_RCR_MEM(RAR_IDX(rar))); + + /* Copy the caller's buffer pointers to the command */ + for (i = 0; i < num_buffers; i++) + p->buf[i] = cpu_to_le64(buffers[i]); + p->bpid = d->bpid; + + p->verb = d->verb | RAR_VB(rar) | num_buffers; + dma_wmb(); + qbman_write_register(s, QBMAN_CINH_SWP_RCR_AM_RT + + RAR_IDX(rar) * 4, QMAN_RT_MODE); + + return 0; +} + +struct qbman_acquire_desc { + u8 verb; + u8 reserved; + __le16 bpid; + u8 num; + u8 reserved2[59]; +}; + +struct qbman_acquire_rslt { + u8 verb; + u8 rslt; + __le16 reserved; + u8 num; + u8 reserved2[3]; + __le64 buf[7]; +}; + +/** + * qbman_swp_acquire() - Issue a buffer acquire command + * @s: the software portal object + * @bpid: the buffer pool index + * @buffers: a pointer pointing to the acquired buffer addresses + * @num_buffers: number of buffers to be acquired, must be less than 8 + * + * Return 0 for success, or negative error code if the acquire command + * fails. + */ +int qbman_swp_acquire(struct qbman_swp *s, u16 bpid, u64 *buffers, + unsigned int num_buffers) +{ + struct qbman_acquire_desc *p; + struct qbman_acquire_rslt *r; + int i; + + if (!num_buffers || (num_buffers > 7)) + return -EINVAL; + + /* Start the management command */ + p = qbman_swp_mc_start(s); + + if (!p) + return -EBUSY; + + /* Encode the caller-provided attributes */ + p->bpid = cpu_to_le16(bpid); + p->num = num_buffers; + + /* Complete the management command */ + r = qbman_swp_mc_complete(s, p, QBMAN_MC_ACQUIRE); + if (unlikely(!r)) { + pr_err("qbman: acquire from BPID %d failed, no response\n", + bpid); + return -EIO; + } + + /* Decode the outcome */ + WARN_ON((r->verb & 0x7f) != QBMAN_MC_ACQUIRE); + + /* Determine success or failure */ + if (unlikely(r->rslt != QBMAN_MC_RSLT_OK)) { + pr_err("qbman: acquire from BPID 0x%x failed, code=0x%02x\n", + bpid, r->rslt); + return -EIO; + } + + WARN_ON(r->num > num_buffers); + + /* Copy the acquired buffers to the caller's array */ + for (i = 0; i < r->num; i++) + buffers[i] = le64_to_cpu(r->buf[i]); + + return (int)r->num; +} + +struct qbman_alt_fq_state_desc { + u8 verb; + u8 reserved[3]; + __le32 fqid; + u8 reserved2[56]; +}; + +struct qbman_alt_fq_state_rslt { + u8 verb; + u8 rslt; + u8 reserved[62]; +}; + +#define ALT_FQ_FQID_MASK 0x00FFFFFF + +int qbman_swp_alt_fq_state(struct qbman_swp *s, u32 fqid, + u8 alt_fq_verb) +{ + struct qbman_alt_fq_state_desc *p; + struct qbman_alt_fq_state_rslt *r; + + /* Start the management command */ + p = qbman_swp_mc_start(s); + if (!p) + return -EBUSY; + + p->fqid = cpu_to_le32(fqid & ALT_FQ_FQID_MASK); + + /* Complete the management command */ + r = qbman_swp_mc_complete(s, p, alt_fq_verb); + if (unlikely(!r)) { + pr_err("qbman: mgmt cmd failed, no response (verb=0x%x)\n", + alt_fq_verb); + return -EIO; + } + + /* Decode the outcome */ + WARN_ON((r->verb & QBMAN_RESULT_MASK) != alt_fq_verb); + + /* Determine success or failure */ + if (unlikely(r->rslt != QBMAN_MC_RSLT_OK)) { + pr_err("qbman: ALT FQID %d failed: verb = 0x%08x code = 0x%02x\n", + fqid, r->verb, r->rslt); + return -EIO; + } + + return 0; +} + +struct qbman_cdan_ctrl_desc { + u8 verb; + u8 reserved; + __le16 ch; + u8 we; + u8 ctrl; + __le16 reserved2; + __le64 cdan_ctx; + u8 reserved3[48]; + +}; + +struct qbman_cdan_ctrl_rslt { + u8 verb; + u8 rslt; + __le16 ch; + u8 reserved[60]; +}; + +int qbman_swp_CDAN_set(struct qbman_swp *s, u16 channelid, + u8 we_mask, u8 cdan_en, + u64 ctx) +{ + struct qbman_cdan_ctrl_desc *p = NULL; + struct qbman_cdan_ctrl_rslt *r = NULL; + + /* Start the management command */ + p = qbman_swp_mc_start(s); + if (!p) + return -EBUSY; + + /* Encode the caller-provided attributes */ + p->ch = cpu_to_le16(channelid); + p->we = we_mask; + if (cdan_en) + p->ctrl = 1; + else + p->ctrl = 0; + p->cdan_ctx = cpu_to_le64(ctx); + + /* Complete the management command */ + r = qbman_swp_mc_complete(s, p, QBMAN_WQCHAN_CONFIGURE); + if (unlikely(!r)) { + pr_err("qbman: wqchan config failed, no response\n"); + return -EIO; + } + + WARN_ON((r->verb & 0x7f) != QBMAN_WQCHAN_CONFIGURE); + + /* Determine success or failure */ + if (unlikely(r->rslt != QBMAN_MC_RSLT_OK)) { + pr_err("qbman: CDAN cQID %d failed: code = 0x%02x\n", + channelid, r->rslt); + return -EIO; + } + + return 0; +} + +#define QBMAN_RESPONSE_VERB_MASK 0x7f +#define QBMAN_FQ_QUERY_NP 0x45 +#define QBMAN_BP_QUERY 0x32 + +struct qbman_fq_query_desc { + u8 verb; + u8 reserved[3]; + __le32 fqid; + u8 reserved2[56]; +}; + +int qbman_fq_query_state(struct qbman_swp *s, u32 fqid, + struct qbman_fq_query_np_rslt *r) +{ + struct qbman_fq_query_desc *p; + void *resp; + + p = (struct qbman_fq_query_desc *)qbman_swp_mc_start(s); + if (!p) + return -EBUSY; + + /* FQID is a 24 bit value */ + p->fqid = cpu_to_le32(fqid & 0x00FFFFFF); + resp = qbman_swp_mc_complete(s, p, QBMAN_FQ_QUERY_NP); + if (!resp) { + pr_err("qbman: Query FQID %d NP fields failed, no response\n", + fqid); + return -EIO; + } + *r = *(struct qbman_fq_query_np_rslt *)resp; + /* Decode the outcome */ + WARN_ON((r->verb & QBMAN_RESPONSE_VERB_MASK) != QBMAN_FQ_QUERY_NP); + + /* Determine success or failure */ + if (r->rslt != QBMAN_MC_RSLT_OK) { + pr_err("Query NP fields of FQID 0x%x failed, code=0x%02x\n", + p->fqid, r->rslt); + return -EIO; + } + + return 0; +} + +u32 qbman_fq_state_frame_count(const struct qbman_fq_query_np_rslt *r) +{ + return (le32_to_cpu(r->frm_cnt) & 0x00FFFFFF); +} + +u32 qbman_fq_state_byte_count(const struct qbman_fq_query_np_rslt *r) +{ + return le32_to_cpu(r->byte_cnt); +} + +struct qbman_bp_query_desc { + u8 verb; + u8 reserved; + __le16 bpid; + u8 reserved2[60]; +}; + +int qbman_bp_query(struct qbman_swp *s, u16 bpid, + struct qbman_bp_query_rslt *r) +{ + struct qbman_bp_query_desc *p; + void *resp; + + p = (struct qbman_bp_query_desc *)qbman_swp_mc_start(s); + if (!p) + return -EBUSY; + + p->bpid = cpu_to_le16(bpid); + resp = qbman_swp_mc_complete(s, p, QBMAN_BP_QUERY); + if (!resp) { + pr_err("qbman: Query BPID %d fields failed, no response\n", + bpid); + return -EIO; + } + *r = *(struct qbman_bp_query_rslt *)resp; + /* Decode the outcome */ + WARN_ON((r->verb & QBMAN_RESPONSE_VERB_MASK) != QBMAN_BP_QUERY); + + /* Determine success or failure */ + if (r->rslt != QBMAN_MC_RSLT_OK) { + pr_err("Query fields of BPID 0x%x failed, code=0x%02x\n", + bpid, r->rslt); + return -EIO; + } + + return 0; +} + +u32 qbman_bp_info_num_free_bufs(struct qbman_bp_query_rslt *a) +{ + return le32_to_cpu(a->fill); +} + +/** + * qbman_swp_set_irq_coalescing() - Set new IRQ coalescing values + * @p: the software portal object + * @irq_threshold: interrupt threshold + * @irq_holdoff: interrupt holdoff (timeout) period in us + * + * Return 0 for success, or negative error code on error. + */ +int qbman_swp_set_irq_coalescing(struct qbman_swp *p, u32 irq_threshold, + u32 irq_holdoff) +{ + u32 itp, max_holdoff; + + /* Convert irq_holdoff value from usecs to 256 QBMAN clock cycles + * increments. This depends on the QBMAN internal frequency. + */ + itp = (irq_holdoff * 1000) / p->desc->qman_256_cycles_per_ns; + if (itp > 4096) { + max_holdoff = (p->desc->qman_256_cycles_per_ns * 4096) / 1000; + pr_err("irq_holdoff must be <= %uus\n", max_holdoff); + return -EINVAL; + } + + if (irq_threshold >= p->dqrr.dqrr_size) { + pr_err("irq_threshold must be < %u\n", p->dqrr.dqrr_size - 1); + return -EINVAL; + } + + p->irq_threshold = irq_threshold; + p->irq_holdoff = irq_holdoff; + + qbman_write_register(p, QBMAN_CINH_SWP_DQRR_ITR, irq_threshold); + qbman_write_register(p, QBMAN_CINH_SWP_ITPR, itp); + + return 0; +} + +/** + * qbman_swp_get_irq_coalescing() - Get the current IRQ coalescing parameters + * @p: the software portal object + * @irq_threshold: interrupt threshold (an IRQ is generated when there are more + * DQRR entries in the portal than the threshold) + * @irq_holdoff: interrupt holdoff (timeout) period in us + */ +void qbman_swp_get_irq_coalescing(struct qbman_swp *p, u32 *irq_threshold, + u32 *irq_holdoff) +{ + if (irq_threshold) + *irq_threshold = p->irq_threshold; + if (irq_holdoff) + *irq_holdoff = p->irq_holdoff; +} diff --git a/drivers/soc/fsl/dpio/qbman-portal.h b/drivers/soc/fsl/dpio/qbman-portal.h new file mode 100644 index 000000000..b23883dd2 --- /dev/null +++ b/drivers/soc/fsl/dpio/qbman-portal.h @@ -0,0 +1,664 @@ +/* SPDX-License-Identifier: (GPL-2.0+ OR BSD-3-Clause) */ +/* + * Copyright (C) 2014-2016 Freescale Semiconductor, Inc. + * Copyright 2016-2019 NXP + * + */ +#ifndef __FSL_QBMAN_PORTAL_H +#define __FSL_QBMAN_PORTAL_H + +#include <soc/fsl/dpaa2-fd.h> + +#define QMAN_REV_4000 0x04000000 +#define QMAN_REV_4100 0x04010000 +#define QMAN_REV_4101 0x04010001 +#define QMAN_REV_5000 0x05000000 + +#define QMAN_REV_MASK 0xffff0000 + +struct dpaa2_dq; +struct qbman_swp; + +/* qbman software portal descriptor structure */ +struct qbman_swp_desc { + void *cena_bar; /* Cache-enabled portal base address */ + void __iomem *cinh_bar; /* Cache-inhibited portal base address */ + u32 qman_version; + u32 qman_clk; + u32 qman_256_cycles_per_ns; +}; + +#define QBMAN_SWP_INTERRUPT_EQRI 0x01 +#define QBMAN_SWP_INTERRUPT_EQDI 0x02 +#define QBMAN_SWP_INTERRUPT_DQRI 0x04 +#define QBMAN_SWP_INTERRUPT_RCRI 0x08 +#define QBMAN_SWP_INTERRUPT_RCDI 0x10 +#define QBMAN_SWP_INTERRUPT_VDCI 0x20 + +/* the structure for pull dequeue descriptor */ +struct qbman_pull_desc { + u8 verb; + u8 numf; + u8 tok; + u8 reserved; + __le32 dq_src; + __le64 rsp_addr; + u64 rsp_addr_virt; + u8 padding[40]; +}; + +enum qbman_pull_type_e { + /* dequeue with priority precedence, respect intra-class scheduling */ + qbman_pull_type_prio = 1, + /* dequeue with active FQ precedence, respect ICS */ + qbman_pull_type_active, + /* dequeue with active FQ precedence, no ICS */ + qbman_pull_type_active_noics +}; + +/* Definitions for parsing dequeue entries */ +#define QBMAN_RESULT_MASK 0x7f +#define QBMAN_RESULT_DQ 0x60 +#define QBMAN_RESULT_FQRN 0x21 +#define QBMAN_RESULT_FQRNI 0x22 +#define QBMAN_RESULT_FQPN 0x24 +#define QBMAN_RESULT_FQDAN 0x25 +#define QBMAN_RESULT_CDAN 0x26 +#define QBMAN_RESULT_CSCN_MEM 0x27 +#define QBMAN_RESULT_CGCU 0x28 +#define QBMAN_RESULT_BPSCN 0x29 +#define QBMAN_RESULT_CSCN_WQ 0x2a + +/* QBMan FQ management command codes */ +#define QBMAN_FQ_SCHEDULE 0x48 +#define QBMAN_FQ_FORCE 0x49 +#define QBMAN_FQ_XON 0x4d +#define QBMAN_FQ_XOFF 0x4e + +/* structure of enqueue descriptor */ +struct qbman_eq_desc { + u8 verb; + u8 dca; + __le16 seqnum; + __le16 orpid; + __le16 reserved1; + __le32 tgtid; + __le32 tag; + __le16 qdbin; + u8 qpri; + u8 reserved[3]; + u8 wae; + u8 rspid; + __le64 rsp_addr; +}; + +struct qbman_eq_desc_with_fd { + struct qbman_eq_desc desc; + u8 fd[32]; +}; + +/* buffer release descriptor */ +struct qbman_release_desc { + u8 verb; + u8 reserved; + __le16 bpid; + __le32 reserved2; + __le64 buf[7]; +}; + +/* Management command result codes */ +#define QBMAN_MC_RSLT_OK 0xf0 + +#define CODE_CDAN_WE_EN 0x1 +#define CODE_CDAN_WE_CTX 0x4 + +/* portal data structure */ +struct qbman_swp { + const struct qbman_swp_desc *desc; + void *addr_cena; + void __iomem *addr_cinh; + + /* Management commands */ + struct { + u32 valid_bit; /* 0x00 or 0x80 */ + } mc; + + /* Management response */ + struct { + u32 valid_bit; /* 0x00 or 0x80 */ + } mr; + + /* Push dequeues */ + u32 sdq; + + /* Volatile dequeues */ + struct { + atomic_t available; /* indicates if a command can be sent */ + u32 valid_bit; /* 0x00 or 0x80 */ + struct dpaa2_dq *storage; /* NULL if DQRR */ + } vdq; + + /* DQRR */ + struct { + u32 next_idx; + u32 valid_bit; + u8 dqrr_size; + int reset_bug; /* indicates dqrr reset workaround is needed */ + } dqrr; + + struct { + u32 pi; + u32 pi_vb; + u32 pi_ring_size; + u32 pi_ci_mask; + u32 ci; + int available; + u32 pend; + u32 no_pfdr; + } eqcr; + + spinlock_t access_spinlock; + + /* Interrupt coalescing */ + u32 irq_threshold; + u32 irq_holdoff; + int use_adaptive_rx_coalesce; +}; + +/* Function pointers */ +extern +int (*qbman_swp_enqueue_ptr)(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd); +extern +int (*qbman_swp_enqueue_multiple_ptr)(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + uint32_t *flags, + int num_frames); +extern +int (*qbman_swp_enqueue_multiple_desc_ptr)(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + int num_frames); +extern +int (*qbman_swp_pull_ptr)(struct qbman_swp *s, struct qbman_pull_desc *d); +extern +const struct dpaa2_dq *(*qbman_swp_dqrr_next_ptr)(struct qbman_swp *s); +extern +int (*qbman_swp_release_ptr)(struct qbman_swp *s, + const struct qbman_release_desc *d, + const u64 *buffers, + unsigned int num_buffers); + +/* Functions */ +struct qbman_swp *qbman_swp_init(const struct qbman_swp_desc *d); +void qbman_swp_finish(struct qbman_swp *p); +u32 qbman_swp_interrupt_read_status(struct qbman_swp *p); +void qbman_swp_interrupt_clear_status(struct qbman_swp *p, u32 mask); +u32 qbman_swp_interrupt_get_trigger(struct qbman_swp *p); +void qbman_swp_interrupt_set_trigger(struct qbman_swp *p, u32 mask); +int qbman_swp_interrupt_get_inhibit(struct qbman_swp *p); +void qbman_swp_interrupt_set_inhibit(struct qbman_swp *p, int inhibit); + +void qbman_swp_push_get(struct qbman_swp *p, u8 channel_idx, int *enabled); +void qbman_swp_push_set(struct qbman_swp *p, u8 channel_idx, int enable); + +void qbman_pull_desc_clear(struct qbman_pull_desc *d); +void qbman_pull_desc_set_storage(struct qbman_pull_desc *d, + struct dpaa2_dq *storage, + dma_addr_t storage_phys, + int stash); +void qbman_pull_desc_set_numframes(struct qbman_pull_desc *d, u8 numframes); +void qbman_pull_desc_set_fq(struct qbman_pull_desc *d, u32 fqid); +void qbman_pull_desc_set_wq(struct qbman_pull_desc *d, u32 wqid, + enum qbman_pull_type_e dct); +void qbman_pull_desc_set_channel(struct qbman_pull_desc *d, u32 chid, + enum qbman_pull_type_e dct); + +void qbman_swp_dqrr_consume(struct qbman_swp *s, const struct dpaa2_dq *dq); + +int qbman_result_has_new_result(struct qbman_swp *p, const struct dpaa2_dq *dq); + +void qbman_eq_desc_clear(struct qbman_eq_desc *d); +void qbman_eq_desc_set_no_orp(struct qbman_eq_desc *d, int respond_success); +void qbman_eq_desc_set_token(struct qbman_eq_desc *d, u8 token); +void qbman_eq_desc_set_fq(struct qbman_eq_desc *d, u32 fqid); +void qbman_eq_desc_set_qd(struct qbman_eq_desc *d, u32 qdid, + u32 qd_bin, u32 qd_prio); + + +void qbman_release_desc_clear(struct qbman_release_desc *d); +void qbman_release_desc_set_bpid(struct qbman_release_desc *d, u16 bpid); +void qbman_release_desc_set_rcdi(struct qbman_release_desc *d, int enable); + +int qbman_swp_acquire(struct qbman_swp *s, u16 bpid, u64 *buffers, + unsigned int num_buffers); +int qbman_swp_alt_fq_state(struct qbman_swp *s, u32 fqid, + u8 alt_fq_verb); +int qbman_swp_CDAN_set(struct qbman_swp *s, u16 channelid, + u8 we_mask, u8 cdan_en, + u64 ctx); + +void *qbman_swp_mc_start(struct qbman_swp *p); +void qbman_swp_mc_submit(struct qbman_swp *p, void *cmd, u8 cmd_verb); +void *qbman_swp_mc_result(struct qbman_swp *p); + +/** + * qbman_swp_enqueue() - Issue an enqueue command + * @s: the software portal used for enqueue + * @d: the enqueue descriptor + * @fd: the frame descriptor to be enqueued + * + * Return 0 for successful enqueue, -EBUSY if the EQCR is not ready. + */ +static inline int +qbman_swp_enqueue(struct qbman_swp *s, const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd) +{ + return qbman_swp_enqueue_ptr(s, d, fd); +} + +/** + * qbman_swp_enqueue_multiple() - Issue a multi enqueue command + * using one enqueue descriptor + * @s: the software portal used for enqueue + * @d: the enqueue descriptor + * @fd: table pointer of frame descriptor table to be enqueued + * @flags: table pointer of QBMAN_ENQUEUE_FLAG_DCA flags, not used if NULL + * @num_frames: number of fd to be enqueued + * + * Return the number of fd enqueued, or a negative error number. + */ +static inline int +qbman_swp_enqueue_multiple(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + uint32_t *flags, + int num_frames) +{ + return qbman_swp_enqueue_multiple_ptr(s, d, fd, flags, num_frames); +} + +/** + * qbman_swp_enqueue_multiple_desc() - Issue a multi enqueue command + * using multiple enqueue descriptor + * @s: the software portal used for enqueue + * @d: table of minimal enqueue descriptor + * @fd: table pointer of frame descriptor table to be enqueued + * @num_frames: number of fd to be enqueued + * + * Return the number of fd enqueued, or a negative error number. + */ +static inline int +qbman_swp_enqueue_multiple_desc(struct qbman_swp *s, + const struct qbman_eq_desc *d, + const struct dpaa2_fd *fd, + int num_frames) +{ + return qbman_swp_enqueue_multiple_desc_ptr(s, d, fd, num_frames); +} + +/** + * qbman_result_is_DQ() - check if the dequeue result is a dequeue response + * @dq: the dequeue result to be checked + * + * DQRR entries may contain non-dequeue results, ie. notifications + */ +static inline int qbman_result_is_DQ(const struct dpaa2_dq *dq) +{ + return ((dq->dq.verb & QBMAN_RESULT_MASK) == QBMAN_RESULT_DQ); +} + +/** + * qbman_result_is_SCN() - Check the dequeue result is notification or not + * @dq: the dequeue result to be checked + * + */ +static inline int qbman_result_is_SCN(const struct dpaa2_dq *dq) +{ + return !qbman_result_is_DQ(dq); +} + +/* FQ Data Availability */ +static inline int qbman_result_is_FQDAN(const struct dpaa2_dq *dq) +{ + return ((dq->dq.verb & QBMAN_RESULT_MASK) == QBMAN_RESULT_FQDAN); +} + +/* Channel Data Availability */ +static inline int qbman_result_is_CDAN(const struct dpaa2_dq *dq) +{ + return ((dq->dq.verb & QBMAN_RESULT_MASK) == QBMAN_RESULT_CDAN); +} + +/* Congestion State Change */ +static inline int qbman_result_is_CSCN(const struct dpaa2_dq *dq) +{ + return ((dq->dq.verb & QBMAN_RESULT_MASK) == QBMAN_RESULT_CSCN_WQ); +} + +/* Buffer Pool State Change */ +static inline int qbman_result_is_BPSCN(const struct dpaa2_dq *dq) +{ + return ((dq->dq.verb & QBMAN_RESULT_MASK) == QBMAN_RESULT_BPSCN); +} + +/* Congestion Group Count Update */ +static inline int qbman_result_is_CGCU(const struct dpaa2_dq *dq) +{ + return ((dq->dq.verb & QBMAN_RESULT_MASK) == QBMAN_RESULT_CGCU); +} + +/* Retirement */ +static inline int qbman_result_is_FQRN(const struct dpaa2_dq *dq) +{ + return ((dq->dq.verb & QBMAN_RESULT_MASK) == QBMAN_RESULT_FQRN); +} + +/* Retirement Immediate */ +static inline int qbman_result_is_FQRNI(const struct dpaa2_dq *dq) +{ + return ((dq->dq.verb & QBMAN_RESULT_MASK) == QBMAN_RESULT_FQRNI); +} + + /* Park */ +static inline int qbman_result_is_FQPN(const struct dpaa2_dq *dq) +{ + return ((dq->dq.verb & QBMAN_RESULT_MASK) == QBMAN_RESULT_FQPN); +} + +/** + * qbman_result_SCN_state() - Get the state field in State-change notification + */ +static inline u8 qbman_result_SCN_state(const struct dpaa2_dq *scn) +{ + return scn->scn.state; +} + +#define SCN_RID_MASK 0x00FFFFFF + +/** + * qbman_result_SCN_rid() - Get the resource id in State-change notification + */ +static inline u32 qbman_result_SCN_rid(const struct dpaa2_dq *scn) +{ + return le32_to_cpu(scn->scn.rid_tok) & SCN_RID_MASK; +} + +/** + * qbman_result_SCN_ctx() - Get the context data in State-change notification + */ +static inline u64 qbman_result_SCN_ctx(const struct dpaa2_dq *scn) +{ + return le64_to_cpu(scn->scn.ctx); +} + +/** + * qbman_swp_fq_schedule() - Move the fq to the scheduled state + * @s: the software portal object + * @fqid: the index of frame queue to be scheduled + * + * There are a couple of different ways that a FQ can end up parked state, + * This schedules it. + * + * Return 0 for success, or negative error code for failure. + */ +static inline int qbman_swp_fq_schedule(struct qbman_swp *s, u32 fqid) +{ + return qbman_swp_alt_fq_state(s, fqid, QBMAN_FQ_SCHEDULE); +} + +/** + * qbman_swp_fq_force() - Force the FQ to fully scheduled state + * @s: the software portal object + * @fqid: the index of frame queue to be forced + * + * Force eligible will force a tentatively-scheduled FQ to be fully-scheduled + * and thus be available for selection by any channel-dequeuing behaviour (push + * or pull). If the FQ is subsequently "dequeued" from the channel and is still + * empty at the time this happens, the resulting dq_entry will have no FD. + * (qbman_result_DQ_fd() will return NULL.) + * + * Return 0 for success, or negative error code for failure. + */ +static inline int qbman_swp_fq_force(struct qbman_swp *s, u32 fqid) +{ + return qbman_swp_alt_fq_state(s, fqid, QBMAN_FQ_FORCE); +} + +/** + * qbman_swp_fq_xon() - sets FQ flow-control to XON + * @s: the software portal object + * @fqid: the index of frame queue + * + * This setting doesn't affect enqueues to the FQ, just dequeues. + * + * Return 0 for success, or negative error code for failure. + */ +static inline int qbman_swp_fq_xon(struct qbman_swp *s, u32 fqid) +{ + return qbman_swp_alt_fq_state(s, fqid, QBMAN_FQ_XON); +} + +/** + * qbman_swp_fq_xoff() - sets FQ flow-control to XOFF + * @s: the software portal object + * @fqid: the index of frame queue + * + * This setting doesn't affect enqueues to the FQ, just dequeues. + * XOFF FQs will remain in the tenatively-scheduled state, even when + * non-empty, meaning they won't be selected for scheduled dequeuing. + * If a FQ is changed to XOFF after it had already become truly-scheduled + * to a channel, and a pull dequeue of that channel occurs that selects + * that FQ for dequeuing, then the resulting dq_entry will have no FD. + * (qbman_result_DQ_fd() will return NULL.) + * + * Return 0 for success, or negative error code for failure. + */ +static inline int qbman_swp_fq_xoff(struct qbman_swp *s, u32 fqid) +{ + return qbman_swp_alt_fq_state(s, fqid, QBMAN_FQ_XOFF); +} + +/* If the user has been allocated a channel object that is going to generate + * CDANs to another channel, then the qbman_swp_CDAN* functions will be + * necessary. + * + * CDAN-enabled channels only generate a single CDAN notification, after which + * they need to be reenabled before they'll generate another. The idea is + * that pull dequeuing will occur in reaction to the CDAN, followed by a + * reenable step. Each function generates a distinct command to hardware, so a + * combination function is provided if the user wishes to modify the "context" + * (which shows up in each CDAN message) each time they reenable, as a single + * command to hardware. + */ + +/** + * qbman_swp_CDAN_set_context() - Set CDAN context + * @s: the software portal object + * @channelid: the channel index + * @ctx: the context to be set in CDAN + * + * Return 0 for success, or negative error code for failure. + */ +static inline int qbman_swp_CDAN_set_context(struct qbman_swp *s, u16 channelid, + u64 ctx) +{ + return qbman_swp_CDAN_set(s, channelid, + CODE_CDAN_WE_CTX, + 0, ctx); +} + +/** + * qbman_swp_CDAN_enable() - Enable CDAN for the channel + * @s: the software portal object + * @channelid: the index of the channel to generate CDAN + * + * Return 0 for success, or negative error code for failure. + */ +static inline int qbman_swp_CDAN_enable(struct qbman_swp *s, u16 channelid) +{ + return qbman_swp_CDAN_set(s, channelid, + CODE_CDAN_WE_EN, + 1, 0); +} + +/** + * qbman_swp_CDAN_disable() - disable CDAN for the channel + * @s: the software portal object + * @channelid: the index of the channel to generate CDAN + * + * Return 0 for success, or negative error code for failure. + */ +static inline int qbman_swp_CDAN_disable(struct qbman_swp *s, u16 channelid) +{ + return qbman_swp_CDAN_set(s, channelid, + CODE_CDAN_WE_EN, + 0, 0); +} + +/** + * qbman_swp_CDAN_set_context_enable() - Set CDAN contest and enable CDAN + * @s: the software portal object + * @channelid: the index of the channel to generate CDAN + * @ctx:i the context set in CDAN + * + * Return 0 for success, or negative error code for failure. + */ +static inline int qbman_swp_CDAN_set_context_enable(struct qbman_swp *s, + u16 channelid, + u64 ctx) +{ + return qbman_swp_CDAN_set(s, channelid, + CODE_CDAN_WE_EN | CODE_CDAN_WE_CTX, + 1, ctx); +} + +/* Wraps up submit + poll-for-result */ +static inline void *qbman_swp_mc_complete(struct qbman_swp *swp, void *cmd, + u8 cmd_verb) +{ + int loopvar = 2000; + + qbman_swp_mc_submit(swp, cmd, cmd_verb); + + do { + cmd = qbman_swp_mc_result(swp); + } while (!cmd && loopvar--); + + WARN_ON(!loopvar); + + return cmd; +} + +/* Query APIs */ +struct qbman_fq_query_np_rslt { + u8 verb; + u8 rslt; + u8 st1; + u8 st2; + u8 reserved[2]; + __le16 od1_sfdr; + __le16 od2_sfdr; + __le16 od3_sfdr; + __le16 ra1_sfdr; + __le16 ra2_sfdr; + __le32 pfdr_hptr; + __le32 pfdr_tptr; + __le32 frm_cnt; + __le32 byte_cnt; + __le16 ics_surp; + u8 is; + u8 reserved2[29]; +}; + +int qbman_fq_query_state(struct qbman_swp *s, u32 fqid, + struct qbman_fq_query_np_rslt *r); +u32 qbman_fq_state_frame_count(const struct qbman_fq_query_np_rslt *r); +u32 qbman_fq_state_byte_count(const struct qbman_fq_query_np_rslt *r); + +struct qbman_bp_query_rslt { + u8 verb; + u8 rslt; + u8 reserved[4]; + u8 bdi; + u8 state; + __le32 fill; + __le32 hdotr; + __le16 swdet; + __le16 swdxt; + __le16 hwdet; + __le16 hwdxt; + __le16 swset; + __le16 swsxt; + __le16 vbpid; + __le16 icid; + __le64 bpscn_addr; + __le64 bpscn_ctx; + __le16 hw_targ; + u8 dbe; + u8 reserved2; + u8 sdcnt; + u8 hdcnt; + u8 sscnt; + u8 reserved3[9]; +}; + +int qbman_bp_query(struct qbman_swp *s, u16 bpid, + struct qbman_bp_query_rslt *r); + +u32 qbman_bp_info_num_free_bufs(struct qbman_bp_query_rslt *a); + +/** + * qbman_swp_release() - Issue a buffer release command + * @s: the software portal object + * @d: the release descriptor + * @buffers: a pointer pointing to the buffer address to be released + * @num_buffers: number of buffers to be released, must be less than 8 + * + * Return 0 for success, -EBUSY if the release command ring is not ready. + */ +static inline int qbman_swp_release(struct qbman_swp *s, + const struct qbman_release_desc *d, + const u64 *buffers, + unsigned int num_buffers) +{ + return qbman_swp_release_ptr(s, d, buffers, num_buffers); +} + +/** + * qbman_swp_pull() - Issue the pull dequeue command + * @s: the software portal object + * @d: the software portal descriptor which has been configured with + * the set of qbman_pull_desc_set_*() calls + * + * Return 0 for success, and -EBUSY if the software portal is not ready + * to do pull dequeue. + */ +static inline int qbman_swp_pull(struct qbman_swp *s, + struct qbman_pull_desc *d) +{ + return qbman_swp_pull_ptr(s, d); +} + +/** + * qbman_swp_dqrr_next() - Get an valid DQRR entry + * @s: the software portal object + * + * Return NULL if there are no unconsumed DQRR entries. Return a DQRR entry + * only once, so repeated calls can return a sequence of DQRR entries, without + * requiring they be consumed immediately or in any particular order. + */ +static inline const struct dpaa2_dq *qbman_swp_dqrr_next(struct qbman_swp *s) +{ + return qbman_swp_dqrr_next_ptr(s); +} + +int qbman_swp_set_irq_coalescing(struct qbman_swp *p, u32 irq_threshold, + u32 irq_holdoff); + +void qbman_swp_get_irq_coalescing(struct qbman_swp *p, u32 *irq_threshold, + u32 *irq_holdoff); + +#endif /* __FSL_QBMAN_PORTAL_H */ diff --git a/drivers/soc/fsl/guts.c b/drivers/soc/fsl/guts.c new file mode 100644 index 000000000..6bf3e6a98 --- /dev/null +++ b/drivers/soc/fsl/guts.c @@ -0,0 +1,279 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Freescale QorIQ Platforms GUTS Driver + * + * Copyright (C) 2016 Freescale Semiconductor, Inc. + */ + +#include <linux/io.h> +#include <linux/slab.h> +#include <linux/module.h> +#include <linux/of_fdt.h> +#include <linux/sys_soc.h> +#include <linux/of_address.h> +#include <linux/platform_device.h> +#include <linux/fsl/guts.h> + +struct fsl_soc_die_attr { + char *die; + u32 svr; + u32 mask; +}; + +struct fsl_soc_data { + const char *sfp_compat; + u32 uid_offset; +}; + +/* SoC die attribute definition for QorIQ platform */ +static const struct fsl_soc_die_attr fsl_soc_die[] = { + /* + * Power Architecture-based SoCs T Series + */ + + /* Die: T4240, SoC: T4240/T4160/T4080 */ + { .die = "T4240", + .svr = 0x82400000, + .mask = 0xfff00000, + }, + /* Die: T1040, SoC: T1040/T1020/T1042/T1022 */ + { .die = "T1040", + .svr = 0x85200000, + .mask = 0xfff00000, + }, + /* Die: T2080, SoC: T2080/T2081 */ + { .die = "T2080", + .svr = 0x85300000, + .mask = 0xfff00000, + }, + /* Die: T1024, SoC: T1024/T1014/T1023/T1013 */ + { .die = "T1024", + .svr = 0x85400000, + .mask = 0xfff00000, + }, + + /* + * ARM-based SoCs LS Series + */ + + /* Die: LS1043A, SoC: LS1043A/LS1023A */ + { .die = "LS1043A", + .svr = 0x87920000, + .mask = 0xffff0000, + }, + /* Die: LS2080A, SoC: LS2080A/LS2040A/LS2085A */ + { .die = "LS2080A", + .svr = 0x87010000, + .mask = 0xff3f0000, + }, + /* Die: LS1088A, SoC: LS1088A/LS1048A/LS1084A/LS1044A */ + { .die = "LS1088A", + .svr = 0x87030000, + .mask = 0xff3f0000, + }, + /* Die: LS1012A, SoC: LS1012A */ + { .die = "LS1012A", + .svr = 0x87040000, + .mask = 0xffff0000, + }, + /* Die: LS1046A, SoC: LS1046A/LS1026A */ + { .die = "LS1046A", + .svr = 0x87070000, + .mask = 0xffff0000, + }, + /* Die: LS2088A, SoC: LS2088A/LS2048A/LS2084A/LS2044A */ + { .die = "LS2088A", + .svr = 0x87090000, + .mask = 0xff3f0000, + }, + /* Die: LS1021A, SoC: LS1021A/LS1020A/LS1022A */ + { .die = "LS1021A", + .svr = 0x87000000, + .mask = 0xfff70000, + }, + /* Die: LX2160A, SoC: LX2160A/LX2120A/LX2080A */ + { .die = "LX2160A", + .svr = 0x87360000, + .mask = 0xff3f0000, + }, + /* Die: LS1028A, SoC: LS1028A */ + { .die = "LS1028A", + .svr = 0x870b0000, + .mask = 0xff3f0000, + }, + { }, +}; + +static const struct fsl_soc_die_attr *fsl_soc_die_match( + u32 svr, const struct fsl_soc_die_attr *matches) +{ + while (matches->svr) { + if (matches->svr == (svr & matches->mask)) + return matches; + matches++; + } + return NULL; +} + +static u64 fsl_guts_get_soc_uid(const char *compat, unsigned int offset) +{ + struct device_node *np; + void __iomem *sfp_base; + u64 uid; + + np = of_find_compatible_node(NULL, NULL, compat); + if (!np) + return 0; + + sfp_base = of_iomap(np, 0); + if (!sfp_base) { + of_node_put(np); + return 0; + } + + uid = ioread32(sfp_base + offset); + uid <<= 32; + uid |= ioread32(sfp_base + offset + 4); + + iounmap(sfp_base); + of_node_put(np); + + return uid; +} + +static const struct fsl_soc_data ls1028a_data = { + .sfp_compat = "fsl,ls1028a-sfp", + .uid_offset = 0x21c, +}; + +/* + * Table for matching compatible strings, for device tree + * guts node, for Freescale QorIQ SOCs. + */ +static const struct of_device_id fsl_guts_of_match[] = { + { .compatible = "fsl,qoriq-device-config-1.0", }, + { .compatible = "fsl,qoriq-device-config-2.0", }, + { .compatible = "fsl,p1010-guts", }, + { .compatible = "fsl,p1020-guts", }, + { .compatible = "fsl,p1021-guts", }, + { .compatible = "fsl,p1022-guts", }, + { .compatible = "fsl,p1023-guts", }, + { .compatible = "fsl,p2020-guts", }, + { .compatible = "fsl,bsc9131-guts", }, + { .compatible = "fsl,bsc9132-guts", }, + { .compatible = "fsl,mpc8536-guts", }, + { .compatible = "fsl,mpc8544-guts", }, + { .compatible = "fsl,mpc8548-guts", }, + { .compatible = "fsl,mpc8568-guts", }, + { .compatible = "fsl,mpc8569-guts", }, + { .compatible = "fsl,mpc8572-guts", }, + { .compatible = "fsl,ls1021a-dcfg", }, + { .compatible = "fsl,ls1043a-dcfg", }, + { .compatible = "fsl,ls2080a-dcfg", }, + { .compatible = "fsl,ls1088a-dcfg", }, + { .compatible = "fsl,ls1012a-dcfg", }, + { .compatible = "fsl,ls1046a-dcfg", }, + { .compatible = "fsl,lx2160a-dcfg", }, + { .compatible = "fsl,ls1028a-dcfg", .data = &ls1028a_data}, + {} +}; + +static int __init fsl_guts_init(void) +{ + struct soc_device_attribute *soc_dev_attr; + static struct soc_device *soc_dev; + const struct fsl_soc_die_attr *soc_die; + const struct fsl_soc_data *soc_data; + const struct of_device_id *match; + struct ccsr_guts __iomem *regs; + const char *machine = NULL; + struct device_node *np; + bool little_endian; + u64 soc_uid = 0; + u32 svr; + int ret; + + np = of_find_matching_node_and_match(NULL, fsl_guts_of_match, &match); + if (!np) + return 0; + soc_data = match->data; + + regs = of_iomap(np, 0); + if (!regs) { + of_node_put(np); + return -ENOMEM; + } + + little_endian = of_property_read_bool(np, "little-endian"); + if (little_endian) + svr = ioread32(®s->svr); + else + svr = ioread32be(®s->svr); + iounmap(regs); + of_node_put(np); + + /* Register soc device */ + soc_dev_attr = kzalloc(sizeof(*soc_dev_attr), GFP_KERNEL); + if (!soc_dev_attr) + return -ENOMEM; + + if (of_property_read_string(of_root, "model", &machine)) + of_property_read_string_index(of_root, "compatible", 0, &machine); + if (machine) { + soc_dev_attr->machine = kstrdup(machine, GFP_KERNEL); + if (!soc_dev_attr->machine) + goto err_nomem; + } + + soc_die = fsl_soc_die_match(svr, fsl_soc_die); + if (soc_die) { + soc_dev_attr->family = kasprintf(GFP_KERNEL, "QorIQ %s", + soc_die->die); + } else { + soc_dev_attr->family = kasprintf(GFP_KERNEL, "QorIQ"); + } + if (!soc_dev_attr->family) + goto err_nomem; + + soc_dev_attr->soc_id = kasprintf(GFP_KERNEL, "svr:0x%08x", svr); + if (!soc_dev_attr->soc_id) + goto err_nomem; + + soc_dev_attr->revision = kasprintf(GFP_KERNEL, "%d.%d", + (svr >> 4) & 0xf, svr & 0xf); + if (!soc_dev_attr->revision) + goto err_nomem; + + if (soc_data) + soc_uid = fsl_guts_get_soc_uid(soc_data->sfp_compat, + soc_data->uid_offset); + if (soc_uid) + soc_dev_attr->serial_number = kasprintf(GFP_KERNEL, "%016llX", + soc_uid); + + soc_dev = soc_device_register(soc_dev_attr); + if (IS_ERR(soc_dev)) { + ret = PTR_ERR(soc_dev); + goto err; + } + + pr_info("Machine: %s\n", soc_dev_attr->machine); + pr_info("SoC family: %s\n", soc_dev_attr->family); + pr_info("SoC ID: %s, Revision: %s\n", + soc_dev_attr->soc_id, soc_dev_attr->revision); + + return 0; + +err_nomem: + ret = -ENOMEM; +err: + kfree(soc_dev_attr->machine); + kfree(soc_dev_attr->family); + kfree(soc_dev_attr->soc_id); + kfree(soc_dev_attr->revision); + kfree(soc_dev_attr->serial_number); + kfree(soc_dev_attr); + + return ret; +} +core_initcall(fsl_guts_init); diff --git a/drivers/soc/fsl/qbman/Kconfig b/drivers/soc/fsl/qbman/Kconfig new file mode 100644 index 000000000..bdecb86bb --- /dev/null +++ b/drivers/soc/fsl/qbman/Kconfig @@ -0,0 +1,68 @@ +# SPDX-License-Identifier: GPL-2.0-only +menuconfig FSL_DPAA + bool "QorIQ DPAA1 framework support" + depends on ((FSL_SOC_BOOKE || ARCH_LAYERSCAPE) && ARCH_DMA_ADDR_T_64BIT) + select GENERIC_ALLOCATOR + help + The Freescale Data Path Acceleration Architecture (DPAA) is a set of + hardware components on specific QorIQ multicore processors. + This architecture provides the infrastructure to support simplified + sharing of networking interfaces and accelerators by multiple CPUs. + The major h/w blocks composing DPAA are BMan and QMan. + + The Buffer Manager (BMan) is a hardware buffer pool management block + that allows software and accelerators on the datapath to acquire and + release buffers in order to build frames. + + The Queue Manager (QMan) is a hardware queue management block + that allows software and accelerators on the datapath to enqueue and + dequeue frames in order to communicate. + +if FSL_DPAA + +config FSL_DPAA_CHECKING + bool "Additional driver checking" + help + Compiles in additional checks, to sanity-check the drivers and + any use of the exported API. Not recommended for performance. + +config FSL_BMAN_TEST + tristate "BMan self-tests" + help + Compile the BMan self-test code. These tests will + exercise the BMan APIs to confirm functionality + of both the software drivers and hardware device. + +config FSL_BMAN_TEST_API + bool "High-level API self-test" + depends on FSL_BMAN_TEST + default y + help + This requires the presence of cpu-affine portals, and performs + high-level API testing with them (whichever portal(s) are affine + to the cpu(s) the test executes on). + +config FSL_QMAN_TEST + tristate "QMan self-tests" + help + Compile self-test code for QMan. + +config FSL_QMAN_TEST_API + bool "QMan high-level self-test" + depends on FSL_QMAN_TEST + default y + help + This requires the presence of cpu-affine portals, and performs + high-level API testing with them (whichever portal(s) are affine to + the cpu(s) the test executes on). + +config FSL_QMAN_TEST_STASH + bool "QMan 'hot potato' data-stashing self-test" + depends on FSL_QMAN_TEST + default y + help + This performs a "hot potato" style test enqueuing/dequeuing a frame + across a series of FQs scheduled to different portals (and cpus), with + DQRR, data and context stashing always on. + +endif # FSL_DPAA diff --git a/drivers/soc/fsl/qbman/Makefile b/drivers/soc/fsl/qbman/Makefile new file mode 100644 index 000000000..811312ad5 --- /dev/null +++ b/drivers/soc/fsl/qbman/Makefile @@ -0,0 +1,13 @@ +# SPDX-License-Identifier: GPL-2.0 +obj-$(CONFIG_FSL_DPAA) += bman_ccsr.o qman_ccsr.o \ + bman_portal.o qman_portal.o \ + bman.o qman.o dpaa_sys.o + +obj-$(CONFIG_FSL_BMAN_TEST) += bman-test.o +bman-test-y = bman_test.o +bman-test-$(CONFIG_FSL_BMAN_TEST_API) += bman_test_api.o + +obj-$(CONFIG_FSL_QMAN_TEST) += qman-test.o +qman-test-y = qman_test.o +qman-test-$(CONFIG_FSL_QMAN_TEST_API) += qman_test_api.o +qman-test-$(CONFIG_FSL_QMAN_TEST_STASH) += qman_test_stash.o diff --git a/drivers/soc/fsl/qbman/bman.c b/drivers/soc/fsl/qbman/bman.c new file mode 100644 index 000000000..6cc1847e5 --- /dev/null +++ b/drivers/soc/fsl/qbman/bman.c @@ -0,0 +1,819 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "bman_priv.h" + +#define IRQNAME "BMan portal %d" +#define MAX_IRQNAME 16 /* big enough for "BMan portal %d" */ + +/* Portal register assists */ + +#if defined(CONFIG_ARM) || defined(CONFIG_ARM64) +/* Cache-inhibited register offsets */ +#define BM_REG_RCR_PI_CINH 0x3000 +#define BM_REG_RCR_CI_CINH 0x3100 +#define BM_REG_RCR_ITR 0x3200 +#define BM_REG_CFG 0x3300 +#define BM_REG_SCN(n) (0x3400 + ((n) << 6)) +#define BM_REG_ISR 0x3e00 +#define BM_REG_IER 0x3e40 +#define BM_REG_ISDR 0x3e80 +#define BM_REG_IIR 0x3ec0 + +/* Cache-enabled register offsets */ +#define BM_CL_CR 0x0000 +#define BM_CL_RR0 0x0100 +#define BM_CL_RR1 0x0140 +#define BM_CL_RCR 0x1000 +#define BM_CL_RCR_PI_CENA 0x3000 +#define BM_CL_RCR_CI_CENA 0x3100 + +#else +/* Cache-inhibited register offsets */ +#define BM_REG_RCR_PI_CINH 0x0000 +#define BM_REG_RCR_CI_CINH 0x0004 +#define BM_REG_RCR_ITR 0x0008 +#define BM_REG_CFG 0x0100 +#define BM_REG_SCN(n) (0x0200 + ((n) << 2)) +#define BM_REG_ISR 0x0e00 +#define BM_REG_IER 0x0e04 +#define BM_REG_ISDR 0x0e08 +#define BM_REG_IIR 0x0e0c + +/* Cache-enabled register offsets */ +#define BM_CL_CR 0x0000 +#define BM_CL_RR0 0x0100 +#define BM_CL_RR1 0x0140 +#define BM_CL_RCR 0x1000 +#define BM_CL_RCR_PI_CENA 0x3000 +#define BM_CL_RCR_CI_CENA 0x3100 +#endif + +/* + * Portal modes. + * Enum types; + * pmode == production mode + * cmode == consumption mode, + * Enum values use 3 letter codes. First letter matches the portal mode, + * remaining two letters indicate; + * ci == cache-inhibited portal register + * ce == cache-enabled portal register + * vb == in-band valid-bit (cache-enabled) + */ +enum bm_rcr_pmode { /* matches BCSP_CFG::RPM */ + bm_rcr_pci = 0, /* PI index, cache-inhibited */ + bm_rcr_pce = 1, /* PI index, cache-enabled */ + bm_rcr_pvb = 2 /* valid-bit */ +}; +enum bm_rcr_cmode { /* s/w-only */ + bm_rcr_cci, /* CI index, cache-inhibited */ + bm_rcr_cce /* CI index, cache-enabled */ +}; + + +/* --- Portal structures --- */ + +#define BM_RCR_SIZE 8 + +/* Release Command */ +struct bm_rcr_entry { + union { + struct { + u8 _ncw_verb; /* writes to this are non-coherent */ + u8 bpid; /* used with BM_RCR_VERB_CMD_BPID_SINGLE */ + u8 __reserved1[62]; + }; + struct bm_buffer bufs[8]; + }; +}; +#define BM_RCR_VERB_VBIT 0x80 +#define BM_RCR_VERB_CMD_MASK 0x70 /* one of two values; */ +#define BM_RCR_VERB_CMD_BPID_SINGLE 0x20 +#define BM_RCR_VERB_CMD_BPID_MULTI 0x30 +#define BM_RCR_VERB_BUFCOUNT_MASK 0x0f /* values 1..8 */ + +struct bm_rcr { + struct bm_rcr_entry *ring, *cursor; + u8 ci, available, ithresh, vbit; +#ifdef CONFIG_FSL_DPAA_CHECKING + u32 busy; + enum bm_rcr_pmode pmode; + enum bm_rcr_cmode cmode; +#endif +}; + +/* MC (Management Command) command */ +struct bm_mc_command { + u8 _ncw_verb; /* writes to this are non-coherent */ + u8 bpid; /* used by acquire command */ + u8 __reserved[62]; +}; +#define BM_MCC_VERB_VBIT 0x80 +#define BM_MCC_VERB_CMD_MASK 0x70 /* where the verb contains; */ +#define BM_MCC_VERB_CMD_ACQUIRE 0x10 +#define BM_MCC_VERB_CMD_QUERY 0x40 +#define BM_MCC_VERB_ACQUIRE_BUFCOUNT 0x0f /* values 1..8 go here */ + +/* MC result, Acquire and Query Response */ +union bm_mc_result { + struct { + u8 verb; + u8 bpid; + u8 __reserved[62]; + }; + struct bm_buffer bufs[8]; +}; +#define BM_MCR_VERB_VBIT 0x80 +#define BM_MCR_VERB_CMD_MASK BM_MCC_VERB_CMD_MASK +#define BM_MCR_VERB_CMD_ACQUIRE BM_MCC_VERB_CMD_ACQUIRE +#define BM_MCR_VERB_CMD_QUERY BM_MCC_VERB_CMD_QUERY +#define BM_MCR_VERB_CMD_ERR_INVALID 0x60 +#define BM_MCR_VERB_CMD_ERR_ECC 0x70 +#define BM_MCR_VERB_ACQUIRE_BUFCOUNT BM_MCC_VERB_ACQUIRE_BUFCOUNT /* 0..8 */ +#define BM_MCR_TIMEOUT 10000 /* us */ + +struct bm_mc { + struct bm_mc_command *cr; + union bm_mc_result *rr; + u8 rridx, vbit; +#ifdef CONFIG_FSL_DPAA_CHECKING + enum { + /* Can only be _mc_start()ed */ + mc_idle, + /* Can only be _mc_commit()ed or _mc_abort()ed */ + mc_user, + /* Can only be _mc_retry()ed */ + mc_hw + } state; +#endif +}; + +struct bm_addr { + void *ce; /* cache-enabled */ + __be32 *ce_be; /* Same as above but for direct access */ + void __iomem *ci; /* cache-inhibited */ +}; + +struct bm_portal { + struct bm_addr addr; + struct bm_rcr rcr; + struct bm_mc mc; +} ____cacheline_aligned; + +/* Cache-inhibited register access. */ +static inline u32 bm_in(struct bm_portal *p, u32 offset) +{ + return ioread32be(p->addr.ci + offset); +} + +static inline void bm_out(struct bm_portal *p, u32 offset, u32 val) +{ + iowrite32be(val, p->addr.ci + offset); +} + +/* Cache Enabled Portal Access */ +static inline void bm_cl_invalidate(struct bm_portal *p, u32 offset) +{ + dpaa_invalidate(p->addr.ce + offset); +} + +static inline void bm_cl_touch_ro(struct bm_portal *p, u32 offset) +{ + dpaa_touch_ro(p->addr.ce + offset); +} + +static inline u32 bm_ce_in(struct bm_portal *p, u32 offset) +{ + return be32_to_cpu(*(p->addr.ce_be + (offset/4))); +} + +struct bman_portal { + struct bm_portal p; + /* interrupt sources processed by portal_isr(), configurable */ + unsigned long irq_sources; + /* probing time config params for cpu-affine portals */ + const struct bm_portal_config *config; + char irqname[MAX_IRQNAME]; +}; + +static cpumask_t affine_mask; +static DEFINE_SPINLOCK(affine_mask_lock); +static DEFINE_PER_CPU(struct bman_portal, bman_affine_portal); + +static inline struct bman_portal *get_affine_portal(void) +{ + return &get_cpu_var(bman_affine_portal); +} + +static inline void put_affine_portal(void) +{ + put_cpu_var(bman_affine_portal); +} + +/* + * This object type refers to a pool, it isn't *the* pool. There may be + * more than one such object per BMan buffer pool, eg. if different users of the + * pool are operating via different portals. + */ +struct bman_pool { + /* index of the buffer pool to encapsulate (0-63) */ + u32 bpid; + /* Used for hash-table admin when using depletion notifications. */ + struct bman_portal *portal; + struct bman_pool *next; +}; + +static u32 poll_portal_slow(struct bman_portal *p, u32 is); + +static irqreturn_t portal_isr(int irq, void *ptr) +{ + struct bman_portal *p = ptr; + struct bm_portal *portal = &p->p; + u32 clear = p->irq_sources; + u32 is = bm_in(portal, BM_REG_ISR) & p->irq_sources; + + if (unlikely(!is)) + return IRQ_NONE; + + clear |= poll_portal_slow(p, is); + bm_out(portal, BM_REG_ISR, clear); + return IRQ_HANDLED; +} + +/* --- RCR API --- */ + +#define RCR_SHIFT ilog2(sizeof(struct bm_rcr_entry)) +#define RCR_CARRY (uintptr_t)(BM_RCR_SIZE << RCR_SHIFT) + +/* Bit-wise logic to wrap a ring pointer by clearing the "carry bit" */ +static struct bm_rcr_entry *rcr_carryclear(struct bm_rcr_entry *p) +{ + uintptr_t addr = (uintptr_t)p; + + addr &= ~RCR_CARRY; + + return (struct bm_rcr_entry *)addr; +} + +#ifdef CONFIG_FSL_DPAA_CHECKING +/* Bit-wise logic to convert a ring pointer to a ring index */ +static int rcr_ptr2idx(struct bm_rcr_entry *e) +{ + return ((uintptr_t)e >> RCR_SHIFT) & (BM_RCR_SIZE - 1); +} +#endif + +/* Increment the 'cursor' ring pointer, taking 'vbit' into account */ +static inline void rcr_inc(struct bm_rcr *rcr) +{ + /* increment to the next RCR pointer and handle overflow and 'vbit' */ + struct bm_rcr_entry *partial = rcr->cursor + 1; + + rcr->cursor = rcr_carryclear(partial); + if (partial != rcr->cursor) + rcr->vbit ^= BM_RCR_VERB_VBIT; +} + +static int bm_rcr_get_avail(struct bm_portal *portal) +{ + struct bm_rcr *rcr = &portal->rcr; + + return rcr->available; +} + +static int bm_rcr_get_fill(struct bm_portal *portal) +{ + struct bm_rcr *rcr = &portal->rcr; + + return BM_RCR_SIZE - 1 - rcr->available; +} + +static void bm_rcr_set_ithresh(struct bm_portal *portal, u8 ithresh) +{ + struct bm_rcr *rcr = &portal->rcr; + + rcr->ithresh = ithresh; + bm_out(portal, BM_REG_RCR_ITR, ithresh); +} + +static void bm_rcr_cce_prefetch(struct bm_portal *portal) +{ + __maybe_unused struct bm_rcr *rcr = &portal->rcr; + + DPAA_ASSERT(rcr->cmode == bm_rcr_cce); + bm_cl_touch_ro(portal, BM_CL_RCR_CI_CENA); +} + +static u8 bm_rcr_cce_update(struct bm_portal *portal) +{ + struct bm_rcr *rcr = &portal->rcr; + u8 diff, old_ci = rcr->ci; + + DPAA_ASSERT(rcr->cmode == bm_rcr_cce); + rcr->ci = bm_ce_in(portal, BM_CL_RCR_CI_CENA) & (BM_RCR_SIZE - 1); + bm_cl_invalidate(portal, BM_CL_RCR_CI_CENA); + diff = dpaa_cyc_diff(BM_RCR_SIZE, old_ci, rcr->ci); + rcr->available += diff; + return diff; +} + +static inline struct bm_rcr_entry *bm_rcr_start(struct bm_portal *portal) +{ + struct bm_rcr *rcr = &portal->rcr; + + DPAA_ASSERT(!rcr->busy); + if (!rcr->available) + return NULL; +#ifdef CONFIG_FSL_DPAA_CHECKING + rcr->busy = 1; +#endif + dpaa_zero(rcr->cursor); + return rcr->cursor; +} + +static inline void bm_rcr_pvb_commit(struct bm_portal *portal, u8 myverb) +{ + struct bm_rcr *rcr = &portal->rcr; + struct bm_rcr_entry *rcursor; + + DPAA_ASSERT(rcr->busy); + DPAA_ASSERT(rcr->pmode == bm_rcr_pvb); + DPAA_ASSERT(rcr->available >= 1); + dma_wmb(); + rcursor = rcr->cursor; + rcursor->_ncw_verb = myverb | rcr->vbit; + dpaa_flush(rcursor); + rcr_inc(rcr); + rcr->available--; +#ifdef CONFIG_FSL_DPAA_CHECKING + rcr->busy = 0; +#endif +} + +static int bm_rcr_init(struct bm_portal *portal, enum bm_rcr_pmode pmode, + enum bm_rcr_cmode cmode) +{ + struct bm_rcr *rcr = &portal->rcr; + u32 cfg; + u8 pi; + + rcr->ring = portal->addr.ce + BM_CL_RCR; + rcr->ci = bm_in(portal, BM_REG_RCR_CI_CINH) & (BM_RCR_SIZE - 1); + pi = bm_in(portal, BM_REG_RCR_PI_CINH) & (BM_RCR_SIZE - 1); + rcr->cursor = rcr->ring + pi; + rcr->vbit = (bm_in(portal, BM_REG_RCR_PI_CINH) & BM_RCR_SIZE) ? + BM_RCR_VERB_VBIT : 0; + rcr->available = BM_RCR_SIZE - 1 + - dpaa_cyc_diff(BM_RCR_SIZE, rcr->ci, pi); + rcr->ithresh = bm_in(portal, BM_REG_RCR_ITR); +#ifdef CONFIG_FSL_DPAA_CHECKING + rcr->busy = 0; + rcr->pmode = pmode; + rcr->cmode = cmode; +#endif + cfg = (bm_in(portal, BM_REG_CFG) & 0xffffffe0) + | (pmode & 0x3); /* BCSP_CFG::RPM */ + bm_out(portal, BM_REG_CFG, cfg); + return 0; +} + +static void bm_rcr_finish(struct bm_portal *portal) +{ +#ifdef CONFIG_FSL_DPAA_CHECKING + struct bm_rcr *rcr = &portal->rcr; + int i; + + DPAA_ASSERT(!rcr->busy); + + i = bm_in(portal, BM_REG_RCR_PI_CINH) & (BM_RCR_SIZE - 1); + if (i != rcr_ptr2idx(rcr->cursor)) + pr_crit("losing uncommitted RCR entries\n"); + + i = bm_in(portal, BM_REG_RCR_CI_CINH) & (BM_RCR_SIZE - 1); + if (i != rcr->ci) + pr_crit("missing existing RCR completions\n"); + if (rcr->ci != rcr_ptr2idx(rcr->cursor)) + pr_crit("RCR destroyed unquiesced\n"); +#endif +} + +/* --- Management command API --- */ +static int bm_mc_init(struct bm_portal *portal) +{ + struct bm_mc *mc = &portal->mc; + + mc->cr = portal->addr.ce + BM_CL_CR; + mc->rr = portal->addr.ce + BM_CL_RR0; + mc->rridx = (mc->cr->_ncw_verb & BM_MCC_VERB_VBIT) ? + 0 : 1; + mc->vbit = mc->rridx ? BM_MCC_VERB_VBIT : 0; +#ifdef CONFIG_FSL_DPAA_CHECKING + mc->state = mc_idle; +#endif + return 0; +} + +static void bm_mc_finish(struct bm_portal *portal) +{ +#ifdef CONFIG_FSL_DPAA_CHECKING + struct bm_mc *mc = &portal->mc; + + DPAA_ASSERT(mc->state == mc_idle); + if (mc->state != mc_idle) + pr_crit("Losing incomplete MC command\n"); +#endif +} + +static inline struct bm_mc_command *bm_mc_start(struct bm_portal *portal) +{ + struct bm_mc *mc = &portal->mc; + + DPAA_ASSERT(mc->state == mc_idle); +#ifdef CONFIG_FSL_DPAA_CHECKING + mc->state = mc_user; +#endif + dpaa_zero(mc->cr); + return mc->cr; +} + +static inline void bm_mc_commit(struct bm_portal *portal, u8 myverb) +{ + struct bm_mc *mc = &portal->mc; + union bm_mc_result *rr = mc->rr + mc->rridx; + + DPAA_ASSERT(mc->state == mc_user); + dma_wmb(); + mc->cr->_ncw_verb = myverb | mc->vbit; + dpaa_flush(mc->cr); + dpaa_invalidate_touch_ro(rr); +#ifdef CONFIG_FSL_DPAA_CHECKING + mc->state = mc_hw; +#endif +} + +static inline union bm_mc_result *bm_mc_result(struct bm_portal *portal) +{ + struct bm_mc *mc = &portal->mc; + union bm_mc_result *rr = mc->rr + mc->rridx; + + DPAA_ASSERT(mc->state == mc_hw); + /* + * The inactive response register's verb byte always returns zero until + * its command is submitted and completed. This includes the valid-bit, + * in case you were wondering... + */ + if (!rr->verb) { + dpaa_invalidate_touch_ro(rr); + return NULL; + } + mc->rridx ^= 1; + mc->vbit ^= BM_MCC_VERB_VBIT; +#ifdef CONFIG_FSL_DPAA_CHECKING + mc->state = mc_idle; +#endif + return rr; +} + +static inline int bm_mc_result_timeout(struct bm_portal *portal, + union bm_mc_result **mcr) +{ + int timeout = BM_MCR_TIMEOUT; + + do { + *mcr = bm_mc_result(portal); + if (*mcr) + break; + udelay(1); + } while (--timeout); + + return timeout; +} + +/* Disable all BSCN interrupts for the portal */ +static void bm_isr_bscn_disable(struct bm_portal *portal) +{ + bm_out(portal, BM_REG_SCN(0), 0); + bm_out(portal, BM_REG_SCN(1), 0); +} + +static int bman_create_portal(struct bman_portal *portal, + const struct bm_portal_config *c) +{ + struct bm_portal *p; + int ret; + + p = &portal->p; + /* + * prep the low-level portal struct with the mapped addresses from the + * config, everything that follows depends on it and "config" is more + * for (de)reference... + */ + p->addr.ce = c->addr_virt_ce; + p->addr.ce_be = c->addr_virt_ce; + p->addr.ci = c->addr_virt_ci; + if (bm_rcr_init(p, bm_rcr_pvb, bm_rcr_cce)) { + dev_err(c->dev, "RCR initialisation failed\n"); + goto fail_rcr; + } + if (bm_mc_init(p)) { + dev_err(c->dev, "MC initialisation failed\n"); + goto fail_mc; + } + /* + * Default to all BPIDs disabled, we enable as required at + * run-time. + */ + bm_isr_bscn_disable(p); + + /* Write-to-clear any stale interrupt status bits */ + bm_out(p, BM_REG_ISDR, 0xffffffff); + portal->irq_sources = 0; + bm_out(p, BM_REG_IER, 0); + bm_out(p, BM_REG_ISR, 0xffffffff); + snprintf(portal->irqname, MAX_IRQNAME, IRQNAME, c->cpu); + if (request_irq(c->irq, portal_isr, 0, portal->irqname, portal)) { + dev_err(c->dev, "request_irq() failed\n"); + goto fail_irq; + } + + if (dpaa_set_portal_irq_affinity(c->dev, c->irq, c->cpu)) + goto fail_affinity; + + /* Need RCR to be empty before continuing */ + ret = bm_rcr_get_fill(p); + if (ret) { + dev_err(c->dev, "RCR unclean\n"); + goto fail_rcr_empty; + } + /* Success */ + portal->config = c; + + bm_out(p, BM_REG_ISDR, 0); + bm_out(p, BM_REG_IIR, 0); + + return 0; + +fail_rcr_empty: +fail_affinity: + free_irq(c->irq, portal); +fail_irq: + bm_mc_finish(p); +fail_mc: + bm_rcr_finish(p); +fail_rcr: + return -EIO; +} + +struct bman_portal *bman_create_affine_portal(const struct bm_portal_config *c) +{ + struct bman_portal *portal; + int err; + + portal = &per_cpu(bman_affine_portal, c->cpu); + err = bman_create_portal(portal, c); + if (err) + return NULL; + + spin_lock(&affine_mask_lock); + cpumask_set_cpu(c->cpu, &affine_mask); + spin_unlock(&affine_mask_lock); + + return portal; +} + +static u32 poll_portal_slow(struct bman_portal *p, u32 is) +{ + u32 ret = is; + + if (is & BM_PIRQ_RCRI) { + bm_rcr_cce_update(&p->p); + bm_rcr_set_ithresh(&p->p, 0); + bm_out(&p->p, BM_REG_ISR, BM_PIRQ_RCRI); + is &= ~BM_PIRQ_RCRI; + } + + /* There should be no status register bits left undefined */ + DPAA_ASSERT(!is); + return ret; +} + +int bman_p_irqsource_add(struct bman_portal *p, u32 bits) +{ + unsigned long irqflags; + + local_irq_save(irqflags); + p->irq_sources |= bits & BM_PIRQ_VISIBLE; + bm_out(&p->p, BM_REG_IER, p->irq_sources); + local_irq_restore(irqflags); + return 0; +} + +int bm_shutdown_pool(u32 bpid) +{ + int err = 0; + struct bm_mc_command *bm_cmd; + union bm_mc_result *bm_res; + + + struct bman_portal *p = get_affine_portal(); + while (1) { + /* Acquire buffers until empty */ + bm_cmd = bm_mc_start(&p->p); + bm_cmd->bpid = bpid; + bm_mc_commit(&p->p, BM_MCC_VERB_CMD_ACQUIRE | 1); + if (!bm_mc_result_timeout(&p->p, &bm_res)) { + pr_crit("BMan Acquire Command timedout\n"); + err = -ETIMEDOUT; + goto done; + } + if (!(bm_res->verb & BM_MCR_VERB_ACQUIRE_BUFCOUNT)) { + /* Pool is empty */ + goto done; + } + } +done: + put_affine_portal(); + return err; +} + +struct gen_pool *bm_bpalloc; + +static int bm_alloc_bpid_range(u32 *result, u32 count) +{ + unsigned long addr; + + addr = gen_pool_alloc(bm_bpalloc, count); + if (!addr) + return -ENOMEM; + + *result = addr & ~DPAA_GENALLOC_OFF; + + return 0; +} + +static int bm_release_bpid(u32 bpid) +{ + int ret; + + ret = bm_shutdown_pool(bpid); + if (ret) { + pr_debug("BPID %d leaked\n", bpid); + return ret; + } + + gen_pool_free(bm_bpalloc, bpid | DPAA_GENALLOC_OFF, 1); + return 0; +} + +struct bman_pool *bman_new_pool(void) +{ + struct bman_pool *pool = NULL; + u32 bpid; + + if (bm_alloc_bpid_range(&bpid, 1)) + return NULL; + + pool = kmalloc(sizeof(*pool), GFP_KERNEL); + if (!pool) + goto err; + + pool->bpid = bpid; + + return pool; +err: + bm_release_bpid(bpid); + return NULL; +} +EXPORT_SYMBOL(bman_new_pool); + +void bman_free_pool(struct bman_pool *pool) +{ + bm_release_bpid(pool->bpid); + + kfree(pool); +} +EXPORT_SYMBOL(bman_free_pool); + +int bman_get_bpid(const struct bman_pool *pool) +{ + return pool->bpid; +} +EXPORT_SYMBOL(bman_get_bpid); + +static void update_rcr_ci(struct bman_portal *p, int avail) +{ + if (avail) + bm_rcr_cce_prefetch(&p->p); + else + bm_rcr_cce_update(&p->p); +} + +int bman_release(struct bman_pool *pool, const struct bm_buffer *bufs, u8 num) +{ + struct bman_portal *p; + struct bm_rcr_entry *r; + unsigned long irqflags; + int avail, timeout = 1000; /* 1ms */ + int i = num - 1; + + DPAA_ASSERT(num > 0 && num <= 8); + + do { + p = get_affine_portal(); + local_irq_save(irqflags); + avail = bm_rcr_get_avail(&p->p); + if (avail < 2) + update_rcr_ci(p, avail); + r = bm_rcr_start(&p->p); + local_irq_restore(irqflags); + put_affine_portal(); + if (likely(r)) + break; + + udelay(1); + } while (--timeout); + + if (unlikely(!timeout)) + return -ETIMEDOUT; + + p = get_affine_portal(); + local_irq_save(irqflags); + /* + * we can copy all but the first entry, as this can trigger badness + * with the valid-bit + */ + bm_buffer_set64(r->bufs, bm_buffer_get64(bufs)); + bm_buffer_set_bpid(r->bufs, pool->bpid); + if (i) + memcpy(&r->bufs[1], &bufs[1], i * sizeof(bufs[0])); + + bm_rcr_pvb_commit(&p->p, BM_RCR_VERB_CMD_BPID_SINGLE | + (num & BM_RCR_VERB_BUFCOUNT_MASK)); + + local_irq_restore(irqflags); + put_affine_portal(); + return 0; +} +EXPORT_SYMBOL(bman_release); + +int bman_acquire(struct bman_pool *pool, struct bm_buffer *bufs, u8 num) +{ + struct bman_portal *p = get_affine_portal(); + struct bm_mc_command *mcc; + union bm_mc_result *mcr; + int ret; + + DPAA_ASSERT(num > 0 && num <= 8); + + mcc = bm_mc_start(&p->p); + mcc->bpid = pool->bpid; + bm_mc_commit(&p->p, BM_MCC_VERB_CMD_ACQUIRE | + (num & BM_MCC_VERB_ACQUIRE_BUFCOUNT)); + if (!bm_mc_result_timeout(&p->p, &mcr)) { + put_affine_portal(); + pr_crit("BMan Acquire Timeout\n"); + return -ETIMEDOUT; + } + ret = mcr->verb & BM_MCR_VERB_ACQUIRE_BUFCOUNT; + if (bufs) + memcpy(&bufs[0], &mcr->bufs[0], num * sizeof(bufs[0])); + + put_affine_portal(); + if (ret != num) + ret = -ENOMEM; + return ret; +} +EXPORT_SYMBOL(bman_acquire); + +const struct bm_portal_config * +bman_get_bm_portal_config(const struct bman_portal *portal) +{ + return portal->config; +} diff --git a/drivers/soc/fsl/qbman/bman_ccsr.c b/drivers/soc/fsl/qbman/bman_ccsr.c new file mode 100644 index 000000000..cb24a08be --- /dev/null +++ b/drivers/soc/fsl/qbman/bman_ccsr.c @@ -0,0 +1,320 @@ +/* Copyright (c) 2009 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "bman_priv.h" + +u16 bman_ip_rev; +EXPORT_SYMBOL(bman_ip_rev); + +/* Register offsets */ +#define REG_FBPR_FPC 0x0800 +#define REG_ECSR 0x0a00 +#define REG_ECIR 0x0a04 +#define REG_EADR 0x0a08 +#define REG_EDATA(n) (0x0a10 + ((n) * 0x04)) +#define REG_SBEC(n) (0x0a80 + ((n) * 0x04)) +#define REG_IP_REV_1 0x0bf8 +#define REG_IP_REV_2 0x0bfc +#define REG_FBPR_BARE 0x0c00 +#define REG_FBPR_BAR 0x0c04 +#define REG_FBPR_AR 0x0c10 +#define REG_SRCIDR 0x0d04 +#define REG_LIODNR 0x0d08 +#define REG_ERR_ISR 0x0e00 +#define REG_ERR_IER 0x0e04 +#define REG_ERR_ISDR 0x0e08 + +/* Used by all error interrupt registers except 'inhibit' */ +#define BM_EIRQ_IVCI 0x00000010 /* Invalid Command Verb */ +#define BM_EIRQ_FLWI 0x00000008 /* FBPR Low Watermark */ +#define BM_EIRQ_MBEI 0x00000004 /* Multi-bit ECC Error */ +#define BM_EIRQ_SBEI 0x00000002 /* Single-bit ECC Error */ +#define BM_EIRQ_BSCN 0x00000001 /* pool State Change Notification */ + +struct bman_hwerr_txt { + u32 mask; + const char *txt; +}; + +static const struct bman_hwerr_txt bman_hwerr_txts[] = { + { BM_EIRQ_IVCI, "Invalid Command Verb" }, + { BM_EIRQ_FLWI, "FBPR Low Watermark" }, + { BM_EIRQ_MBEI, "Multi-bit ECC Error" }, + { BM_EIRQ_SBEI, "Single-bit ECC Error" }, + { BM_EIRQ_BSCN, "Pool State Change Notification" }, +}; + +/* Only trigger low water mark interrupt once only */ +#define BMAN_ERRS_TO_DISABLE BM_EIRQ_FLWI + +/* Pointer to the start of the BMan's CCSR space */ +static u32 __iomem *bm_ccsr_start; + +static inline u32 bm_ccsr_in(u32 offset) +{ + return ioread32be(bm_ccsr_start + offset/4); +} +static inline void bm_ccsr_out(u32 offset, u32 val) +{ + iowrite32be(val, bm_ccsr_start + offset/4); +} + +static void bm_get_version(u16 *id, u8 *major, u8 *minor) +{ + u32 v = bm_ccsr_in(REG_IP_REV_1); + *id = (v >> 16); + *major = (v >> 8) & 0xff; + *minor = v & 0xff; +} + +/* signal transactions for FBPRs with higher priority */ +#define FBPR_AR_RPRIO_HI BIT(30) + +/* Track if probe has occurred and if cleanup is required */ +static int __bman_probed; +static int __bman_requires_cleanup; + + +static int bm_set_memory(u64 ba, u32 size) +{ + u32 bar, bare; + u32 exp = ilog2(size); + /* choke if size isn't within range */ + DPAA_ASSERT(size >= 4096 && size <= 1024*1024*1024 && + is_power_of_2(size)); + /* choke if '[e]ba' has lower-alignment than 'size' */ + DPAA_ASSERT(!(ba & (size - 1))); + + /* Check to see if BMan has already been initialized */ + bar = bm_ccsr_in(REG_FBPR_BAR); + if (bar) { + /* Maker sure ba == what was programmed) */ + bare = bm_ccsr_in(REG_FBPR_BARE); + if (bare != upper_32_bits(ba) || bar != lower_32_bits(ba)) { + pr_err("Attempted to reinitialize BMan with different BAR, got 0x%llx read BARE=0x%x BAR=0x%x\n", + ba, bare, bar); + return -ENOMEM; + } + pr_info("BMan BAR already configured\n"); + __bman_requires_cleanup = 1; + return 1; + } + + bm_ccsr_out(REG_FBPR_BARE, upper_32_bits(ba)); + bm_ccsr_out(REG_FBPR_BAR, lower_32_bits(ba)); + bm_ccsr_out(REG_FBPR_AR, exp - 1); + return 0; +} + +/* + * Location and size of BMan private memory + * + * Ideally we would use the DMA API to turn rmem->base into a DMA address + * (especially if iommu translations ever get involved). Unfortunately, the + * DMA API currently does not allow mapping anything that is not backed with + * a struct page. + */ +static dma_addr_t fbpr_a; +static size_t fbpr_sz; + +static int bman_fbpr(struct reserved_mem *rmem) +{ + fbpr_a = rmem->base; + fbpr_sz = rmem->size; + + WARN_ON(!(fbpr_a && fbpr_sz)); + + return 0; +} +RESERVEDMEM_OF_DECLARE(bman_fbpr, "fsl,bman-fbpr", bman_fbpr); + +static irqreturn_t bman_isr(int irq, void *ptr) +{ + u32 isr_val, ier_val, ecsr_val, isr_mask, i; + struct device *dev = ptr; + + ier_val = bm_ccsr_in(REG_ERR_IER); + isr_val = bm_ccsr_in(REG_ERR_ISR); + ecsr_val = bm_ccsr_in(REG_ECSR); + isr_mask = isr_val & ier_val; + + if (!isr_mask) + return IRQ_NONE; + + for (i = 0; i < ARRAY_SIZE(bman_hwerr_txts); i++) { + if (bman_hwerr_txts[i].mask & isr_mask) { + dev_err_ratelimited(dev, "ErrInt: %s\n", + bman_hwerr_txts[i].txt); + if (bman_hwerr_txts[i].mask & ecsr_val) { + /* Re-arm error capture registers */ + bm_ccsr_out(REG_ECSR, ecsr_val); + } + if (bman_hwerr_txts[i].mask & BMAN_ERRS_TO_DISABLE) { + dev_dbg(dev, "Disabling error 0x%x\n", + bman_hwerr_txts[i].mask); + ier_val &= ~bman_hwerr_txts[i].mask; + bm_ccsr_out(REG_ERR_IER, ier_val); + } + } + } + bm_ccsr_out(REG_ERR_ISR, isr_val); + + return IRQ_HANDLED; +} + +int bman_is_probed(void) +{ + return __bman_probed; +} +EXPORT_SYMBOL_GPL(bman_is_probed); + +int bman_requires_cleanup(void) +{ + return __bman_requires_cleanup; +} + +void bman_done_cleanup(void) +{ + __bman_requires_cleanup = 0; +} + +static int fsl_bman_probe(struct platform_device *pdev) +{ + int ret, err_irq; + struct device *dev = &pdev->dev; + struct device_node *node = dev->of_node; + struct resource *res; + u16 id, bm_pool_cnt; + u8 major, minor; + + __bman_probed = -1; + + res = platform_get_resource(pdev, IORESOURCE_MEM, 0); + if (!res) { + dev_err(dev, "Can't get %pOF property 'IORESOURCE_MEM'\n", + node); + return -ENXIO; + } + bm_ccsr_start = devm_ioremap(dev, res->start, resource_size(res)); + if (!bm_ccsr_start) + return -ENXIO; + + bm_get_version(&id, &major, &minor); + if (major == 1 && minor == 0) { + bman_ip_rev = BMAN_REV10; + bm_pool_cnt = BM_POOL_MAX; + } else if (major == 2 && minor == 0) { + bman_ip_rev = BMAN_REV20; + bm_pool_cnt = 8; + } else if (major == 2 && minor == 1) { + bman_ip_rev = BMAN_REV21; + bm_pool_cnt = BM_POOL_MAX; + } else { + dev_err(dev, "Unknown Bman version:%04x,%02x,%02x\n", + id, major, minor); + return -ENODEV; + } + + /* + * If FBPR memory wasn't defined using the qbman compatible string + * try using the of_reserved_mem_device method + */ + if (!fbpr_a) { + ret = qbman_init_private_mem(dev, 0, &fbpr_a, &fbpr_sz); + if (ret) { + dev_err(dev, "qbman_init_private_mem() failed 0x%x\n", + ret); + return -ENODEV; + } + } + + dev_dbg(dev, "Allocated FBPR 0x%llx 0x%zx\n", fbpr_a, fbpr_sz); + + bm_set_memory(fbpr_a, fbpr_sz); + + err_irq = platform_get_irq(pdev, 0); + if (err_irq <= 0) { + dev_info(dev, "Can't get %pOF IRQ\n", node); + return -ENODEV; + } + ret = devm_request_irq(dev, err_irq, bman_isr, IRQF_SHARED, "bman-err", + dev); + if (ret) { + dev_err(dev, "devm_request_irq() failed %d for '%pOF'\n", + ret, node); + return ret; + } + /* Disable Buffer Pool State Change */ + bm_ccsr_out(REG_ERR_ISDR, BM_EIRQ_BSCN); + /* + * Write-to-clear any stale bits, (eg. starvation being asserted prior + * to resource allocation during driver init). + */ + bm_ccsr_out(REG_ERR_ISR, 0xffffffff); + /* Enable Error Interrupts */ + bm_ccsr_out(REG_ERR_IER, 0xffffffff); + + bm_bpalloc = devm_gen_pool_create(dev, 0, -1, "bman-bpalloc"); + if (IS_ERR(bm_bpalloc)) { + ret = PTR_ERR(bm_bpalloc); + dev_err(dev, "bman-bpalloc pool init failed (%d)\n", ret); + return ret; + } + + /* seed BMan resource pool */ + ret = gen_pool_add(bm_bpalloc, DPAA_GENALLOC_OFF, bm_pool_cnt, -1); + if (ret) { + dev_err(dev, "Failed to seed BPID range [%d..%d] (%d)\n", + 0, bm_pool_cnt - 1, ret); + return ret; + } + + __bman_probed = 1; + + return 0; +}; + +static const struct of_device_id fsl_bman_ids[] = { + { + .compatible = "fsl,bman", + }, + {} +}; + +static struct platform_driver fsl_bman_driver = { + .driver = { + .name = KBUILD_MODNAME, + .of_match_table = fsl_bman_ids, + .suppress_bind_attrs = true, + }, + .probe = fsl_bman_probe, +}; + +builtin_platform_driver(fsl_bman_driver); diff --git a/drivers/soc/fsl/qbman/bman_portal.c b/drivers/soc/fsl/qbman/bman_portal.c new file mode 100644 index 000000000..4d7b9caee --- /dev/null +++ b/drivers/soc/fsl/qbman/bman_portal.c @@ -0,0 +1,244 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "bman_priv.h" + +static struct bman_portal *affine_bportals[NR_CPUS]; +static struct cpumask portal_cpus; +static int __bman_portals_probed; +/* protect bman global registers and global data shared among portals */ +static DEFINE_SPINLOCK(bman_lock); + +static struct bman_portal *init_pcfg(struct bm_portal_config *pcfg) +{ + struct bman_portal *p = bman_create_affine_portal(pcfg); + + if (!p) { + dev_crit(pcfg->dev, "%s: Portal failure on cpu %d\n", + __func__, pcfg->cpu); + return NULL; + } + + bman_p_irqsource_add(p, BM_PIRQ_RCRI); + affine_bportals[pcfg->cpu] = p; + + dev_info(pcfg->dev, "Portal initialised, cpu %d\n", pcfg->cpu); + + return p; +} + +static int bman_offline_cpu(unsigned int cpu) +{ + struct bman_portal *p = affine_bportals[cpu]; + const struct bm_portal_config *pcfg; + + if (!p) + return 0; + + pcfg = bman_get_bm_portal_config(p); + if (!pcfg) + return 0; + + /* use any other online CPU */ + cpu = cpumask_any_but(cpu_online_mask, cpu); + irq_set_affinity(pcfg->irq, cpumask_of(cpu)); + return 0; +} + +static int bman_online_cpu(unsigned int cpu) +{ + struct bman_portal *p = affine_bportals[cpu]; + const struct bm_portal_config *pcfg; + + if (!p) + return 0; + + pcfg = bman_get_bm_portal_config(p); + if (!pcfg) + return 0; + + irq_set_affinity(pcfg->irq, cpumask_of(cpu)); + return 0; +} + +int bman_portals_probed(void) +{ + return __bman_portals_probed; +} +EXPORT_SYMBOL_GPL(bman_portals_probed); + +static int bman_portal_probe(struct platform_device *pdev) +{ + struct device *dev = &pdev->dev; + struct device_node *node = dev->of_node; + struct bm_portal_config *pcfg; + struct resource *addr_phys[2]; + int irq, cpu, err, i; + + err = bman_is_probed(); + if (!err) + return -EPROBE_DEFER; + if (err < 0) { + dev_err(&pdev->dev, "failing probe due to bman probe error\n"); + return -ENODEV; + } + + pcfg = devm_kmalloc(dev, sizeof(*pcfg), GFP_KERNEL); + if (!pcfg) { + __bman_portals_probed = -1; + return -ENOMEM; + } + + pcfg->dev = dev; + + addr_phys[0] = platform_get_resource(pdev, IORESOURCE_MEM, + DPAA_PORTAL_CE); + if (!addr_phys[0]) { + dev_err(dev, "Can't get %pOF property 'reg::CE'\n", node); + goto err_ioremap1; + } + + addr_phys[1] = platform_get_resource(pdev, IORESOURCE_MEM, + DPAA_PORTAL_CI); + if (!addr_phys[1]) { + dev_err(dev, "Can't get %pOF property 'reg::CI'\n", node); + goto err_ioremap1; + } + + pcfg->cpu = -1; + + irq = platform_get_irq(pdev, 0); + if (irq <= 0) + goto err_ioremap1; + pcfg->irq = irq; + + pcfg->addr_virt_ce = memremap(addr_phys[0]->start, + resource_size(addr_phys[0]), + QBMAN_MEMREMAP_ATTR); + if (!pcfg->addr_virt_ce) { + dev_err(dev, "memremap::CE failed\n"); + goto err_ioremap1; + } + + pcfg->addr_virt_ci = ioremap(addr_phys[1]->start, + resource_size(addr_phys[1])); + if (!pcfg->addr_virt_ci) { + dev_err(dev, "ioremap::CI failed\n"); + goto err_ioremap2; + } + + spin_lock(&bman_lock); + cpu = cpumask_first_zero(&portal_cpus); + if (cpu >= nr_cpu_ids) { + __bman_portals_probed = 1; + /* unassigned portal, skip init */ + spin_unlock(&bman_lock); + goto check_cleanup; + } + + cpumask_set_cpu(cpu, &portal_cpus); + spin_unlock(&bman_lock); + pcfg->cpu = cpu; + + if (!init_pcfg(pcfg)) { + dev_err(dev, "portal init failed\n"); + goto err_portal_init; + } + + /* clear irq affinity if assigned cpu is offline */ + if (!cpu_online(cpu)) + bman_offline_cpu(cpu); + +check_cleanup: + if (__bman_portals_probed == 1 && bman_requires_cleanup()) { + /* + * BMan wasn't reset prior to boot (Kexec for example) + * Empty all the buffer pools so they are in reset state + */ + for (i = 0; i < BM_POOL_MAX; i++) { + err = bm_shutdown_pool(i); + if (err) { + dev_err(dev, "Failed to shutdown bpool %d\n", + i); + goto err_portal_init; + } + } + bman_done_cleanup(); + } + + return 0; + +err_portal_init: + iounmap(pcfg->addr_virt_ci); +err_ioremap2: + memunmap(pcfg->addr_virt_ce); +err_ioremap1: + __bman_portals_probed = -1; + + return -ENXIO; +} + +static const struct of_device_id bman_portal_ids[] = { + { + .compatible = "fsl,bman-portal", + }, + {} +}; +MODULE_DEVICE_TABLE(of, bman_portal_ids); + +static struct platform_driver bman_portal_driver = { + .driver = { + .name = KBUILD_MODNAME, + .of_match_table = bman_portal_ids, + }, + .probe = bman_portal_probe, +}; + +static int __init bman_portal_driver_register(struct platform_driver *drv) +{ + int ret; + + ret = platform_driver_register(drv); + if (ret < 0) + return ret; + + ret = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, + "soc/qbman_portal:online", + bman_online_cpu, bman_offline_cpu); + if (ret < 0) { + pr_err("bman: failed to register hotplug callbacks.\n"); + platform_driver_unregister(drv); + return ret; + } + return 0; +} + +module_driver(bman_portal_driver, + bman_portal_driver_register, platform_driver_unregister); diff --git a/drivers/soc/fsl/qbman/bman_priv.h b/drivers/soc/fsl/qbman/bman_priv.h new file mode 100644 index 000000000..aa3981e04 --- /dev/null +++ b/drivers/soc/fsl/qbman/bman_priv.h @@ -0,0 +1,83 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + +#include "dpaa_sys.h" + +#include <soc/fsl/bman.h> + +/* Portal processing (interrupt) sources */ +#define BM_PIRQ_RCRI 0x00000002 /* RCR Ring (below threshold) */ + +/* Revision info (for errata and feature handling) */ +#define BMAN_REV10 0x0100 +#define BMAN_REV20 0x0200 +#define BMAN_REV21 0x0201 +extern u16 bman_ip_rev; /* 0 if uninitialised, otherwise BMAN_REVx */ + +extern struct gen_pool *bm_bpalloc; + +struct bm_portal_config { + /* Portal addresses */ + void *addr_virt_ce; + void __iomem *addr_virt_ci; + /* Allow these to be joined in lists */ + struct list_head list; + struct device *dev; + /* User-visible portal configuration settings */ + /* portal is affined to this cpu */ + int cpu; + /* portal interrupt line */ + int irq; +}; + +struct bman_portal *bman_create_affine_portal( + const struct bm_portal_config *config); +/* + * The below bman_p_***() variant might be called in a situation that the cpu + * which the portal affine to is not online yet. + * @bman_portal specifies which portal the API will use. + */ +int bman_p_irqsource_add(struct bman_portal *p, u32 bits); + +/* + * Used by all portal interrupt registers except 'inhibit' + * This mask contains all the "irqsource" bits visible to API users + */ +#define BM_PIRQ_VISIBLE BM_PIRQ_RCRI + +const struct bm_portal_config * +bman_get_bm_portal_config(const struct bman_portal *portal); + +int bman_requires_cleanup(void); +void bman_done_cleanup(void); + +int bm_shutdown_pool(u32 bpid); diff --git a/drivers/soc/fsl/qbman/bman_test.c b/drivers/soc/fsl/qbman/bman_test.c new file mode 100644 index 000000000..09b1c960b --- /dev/null +++ b/drivers/soc/fsl/qbman/bman_test.c @@ -0,0 +1,53 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "bman_test.h" + +MODULE_AUTHOR("Geoff Thorpe"); +MODULE_LICENSE("Dual BSD/GPL"); +MODULE_DESCRIPTION("BMan testing"); + +static int test_init(void) +{ +#ifdef CONFIG_FSL_BMAN_TEST_API + int loop = 1; + + while (loop--) + bman_test_api(); +#endif + return 0; +} + +static void test_exit(void) +{ +} + +module_init(test_init); +module_exit(test_exit); diff --git a/drivers/soc/fsl/qbman/bman_test.h b/drivers/soc/fsl/qbman/bman_test.h new file mode 100644 index 000000000..037ed342a --- /dev/null +++ b/drivers/soc/fsl/qbman/bman_test.h @@ -0,0 +1,35 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "bman_priv.h" + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + +void bman_test_api(void); diff --git a/drivers/soc/fsl/qbman/bman_test_api.c b/drivers/soc/fsl/qbman/bman_test_api.c new file mode 100644 index 000000000..6f6bdd154 --- /dev/null +++ b/drivers/soc/fsl/qbman/bman_test_api.c @@ -0,0 +1,151 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "bman_test.h" + +#define NUM_BUFS 93 +#define LOOPS 3 +#define BMAN_TOKEN_MASK 0x00FFFFFFFFFFLLU + +static struct bman_pool *pool; +static struct bm_buffer bufs_in[NUM_BUFS] ____cacheline_aligned; +static struct bm_buffer bufs_out[NUM_BUFS] ____cacheline_aligned; +static int bufs_received; + +static void bufs_init(void) +{ + int i; + + for (i = 0; i < NUM_BUFS; i++) + bm_buffer_set64(&bufs_in[i], 0xfedc01234567LLU * i); + bufs_received = 0; +} + +static inline int bufs_cmp(const struct bm_buffer *a, const struct bm_buffer *b) +{ + if (bman_ip_rev == BMAN_REV20 || bman_ip_rev == BMAN_REV21) { + + /* + * On SoCs with BMan revison 2.0, BMan only respects the 40 + * LS-bits of buffer addresses, masking off the upper 8-bits on + * release commands. The API provides for 48-bit addresses + * because some SoCs support all 48-bits. When generating + * garbage addresses for testing, we either need to zero the + * upper 8-bits when releasing to BMan (otherwise we'll be + * disappointed when the buffers we acquire back from BMan + * don't match), or we need to mask the upper 8-bits off when + * comparing. We do the latter. + */ + if ((bm_buffer_get64(a) & BMAN_TOKEN_MASK) < + (bm_buffer_get64(b) & BMAN_TOKEN_MASK)) + return -1; + if ((bm_buffer_get64(a) & BMAN_TOKEN_MASK) > + (bm_buffer_get64(b) & BMAN_TOKEN_MASK)) + return 1; + } else { + if (bm_buffer_get64(a) < bm_buffer_get64(b)) + return -1; + if (bm_buffer_get64(a) > bm_buffer_get64(b)) + return 1; + } + + return 0; +} + +static void bufs_confirm(void) +{ + int i, j; + + for (i = 0; i < NUM_BUFS; i++) { + int matches = 0; + + for (j = 0; j < NUM_BUFS; j++) + if (!bufs_cmp(&bufs_in[i], &bufs_out[j])) + matches++; + WARN_ON(matches != 1); + } +} + +/* test */ +void bman_test_api(void) +{ + int i, loops = LOOPS; + + bufs_init(); + + pr_info("%s(): Starting\n", __func__); + + pool = bman_new_pool(); + if (!pool) { + pr_crit("bman_new_pool() failed\n"); + goto failed; + } + + /* Release buffers */ +do_loop: + i = 0; + while (i < NUM_BUFS) { + int num = 8; + + if (i + num > NUM_BUFS) + num = NUM_BUFS - i; + if (bman_release(pool, bufs_in + i, num)) { + pr_crit("bman_release() failed\n"); + goto failed; + } + i += num; + } + + /* Acquire buffers */ + while (i > 0) { + int tmp, num = 8; + + if (num > i) + num = i; + tmp = bman_acquire(pool, bufs_out + i - num, num); + WARN_ON(tmp != num); + i -= num; + } + i = bman_acquire(pool, NULL, 1); + WARN_ON(i > 0); + + bufs_confirm(); + + if (--loops) + goto do_loop; + + /* Clean up */ + bman_free_pool(pool); + pr_info("%s(): Finished\n", __func__); + return; + +failed: + WARN_ON(1); +} diff --git a/drivers/soc/fsl/qbman/dpaa_sys.c b/drivers/soc/fsl/qbman/dpaa_sys.c new file mode 100644 index 000000000..9dd8bb571 --- /dev/null +++ b/drivers/soc/fsl/qbman/dpaa_sys.c @@ -0,0 +1,89 @@ +/* Copyright 2017 NXP Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of NXP Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY NXP Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL NXP Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include <linux/dma-mapping.h> +#include "dpaa_sys.h" + +/* + * Initialize a devices private memory region + */ +int qbman_init_private_mem(struct device *dev, int idx, dma_addr_t *addr, + size_t *size) +{ + struct device_node *mem_node; + struct reserved_mem *rmem; + struct property *prop; + int len, err; + __be32 *res_array; + + mem_node = of_parse_phandle(dev->of_node, "memory-region", idx); + if (!mem_node) { + dev_err(dev, "No memory-region found for index %d\n", idx); + return -ENODEV; + } + + rmem = of_reserved_mem_lookup(mem_node); + if (!rmem) { + dev_err(dev, "of_reserved_mem_lookup() returned NULL\n"); + return -ENODEV; + } + *addr = rmem->base; + *size = rmem->size; + + /* + * Check if the reg property exists - if not insert the node + * so upon kexec() the same memory region address will be preserved. + * This is needed because QBMan HW does not allow the base address/ + * size to be modified once set. + */ + prop = of_find_property(mem_node, "reg", &len); + if (!prop) { + prop = devm_kzalloc(dev, sizeof(*prop), GFP_KERNEL); + if (!prop) + return -ENOMEM; + prop->value = res_array = devm_kzalloc(dev, sizeof(__be32) * 4, + GFP_KERNEL); + if (!prop->value) + return -ENOMEM; + res_array[0] = cpu_to_be32(upper_32_bits(*addr)); + res_array[1] = cpu_to_be32(lower_32_bits(*addr)); + res_array[2] = cpu_to_be32(upper_32_bits(*size)); + res_array[3] = cpu_to_be32(lower_32_bits(*size)); + prop->length = sizeof(__be32) * 4; + prop->name = devm_kstrdup(dev, "reg", GFP_KERNEL); + if (!prop->name) + return -ENOMEM; + err = of_add_property(mem_node, prop); + if (err) + return err; + } + + return 0; +} diff --git a/drivers/soc/fsl/qbman/dpaa_sys.h b/drivers/soc/fsl/qbman/dpaa_sys.h new file mode 100644 index 000000000..ae8afa552 --- /dev/null +++ b/drivers/soc/fsl/qbman/dpaa_sys.h @@ -0,0 +1,134 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#ifndef __DPAA_SYS_H +#define __DPAA_SYS_H + +#include <linux/cpu.h> +#include <linux/slab.h> +#include <linux/module.h> +#include <linux/interrupt.h> +#include <linux/kthread.h> +#include <linux/sched/signal.h> +#include <linux/vmalloc.h> +#include <linux/platform_device.h> +#include <linux/of.h> +#include <linux/of_reserved_mem.h> +#include <linux/prefetch.h> +#include <linux/genalloc.h> +#include <asm/cacheflush.h> +#include <linux/io.h> +#include <linux/delay.h> + +/* For 2-element tables related to cache-inhibited and cache-enabled mappings */ +#define DPAA_PORTAL_CE 0 +#define DPAA_PORTAL_CI 1 + +static inline void dpaa_flush(void *p) +{ + /* + * Only PPC needs to flush the cache currently - on ARM the mapping + * is non cacheable + */ +#ifdef CONFIG_PPC + flush_dcache_range((unsigned long)p, (unsigned long)p+64); +#endif +} + +#define dpaa_invalidate(p) dpaa_flush(p) + +#define dpaa_zero(p) memset(p, 0, 64) + +static inline void dpaa_touch_ro(void *p) +{ +#if (L1_CACHE_BYTES == 32) + prefetch(p+32); +#endif + prefetch(p); +} + +/* Commonly used combo */ +static inline void dpaa_invalidate_touch_ro(void *p) +{ + dpaa_invalidate(p); + dpaa_touch_ro(p); +} + + +#ifdef CONFIG_FSL_DPAA_CHECKING +#define DPAA_ASSERT(x) WARN_ON(!(x)) +#else +#define DPAA_ASSERT(x) +#endif + +/* cyclic helper for rings */ +static inline u8 dpaa_cyc_diff(u8 ringsize, u8 first, u8 last) +{ + /* 'first' is included, 'last' is excluded */ + if (first <= last) + return last - first; + return ringsize + last - first; +} + +/* Offset applied to genalloc pools due to zero being an error return */ +#define DPAA_GENALLOC_OFF 0x80000000 + +/* Initialize the devices private memory region */ +int qbman_init_private_mem(struct device *dev, int idx, dma_addr_t *addr, + size_t *size); + +/* memremap() attributes for different platforms */ +#ifdef CONFIG_PPC +#define QBMAN_MEMREMAP_ATTR MEMREMAP_WB +#else +#define QBMAN_MEMREMAP_ATTR MEMREMAP_WC +#endif + +static inline int dpaa_set_portal_irq_affinity(struct device *dev, + int irq, int cpu) +{ + int ret = 0; + + if (!irq_can_set_affinity(irq)) { + dev_err(dev, "unable to set IRQ affinity\n"); + return -EINVAL; + } + + if (cpu == -1 || !cpu_online(cpu)) + cpu = cpumask_any(cpu_online_mask); + + ret = irq_set_affinity(irq, cpumask_of(cpu)); + if (ret) + dev_err(dev, "irq_set_affinity() on CPU %d failed\n", cpu); + + return ret; +} + +#endif /* __DPAA_SYS_H */ diff --git a/drivers/soc/fsl/qbman/qman.c b/drivers/soc/fsl/qbman/qman.c new file mode 100644 index 000000000..739e4eee6 --- /dev/null +++ b/drivers/soc/fsl/qbman/qman.c @@ -0,0 +1,3053 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "qman_priv.h" + +#define DQRR_MAXFILL 15 +#define EQCR_ITHRESH 4 /* if EQCR congests, interrupt threshold */ +#define IRQNAME "QMan portal %d" +#define MAX_IRQNAME 16 /* big enough for "QMan portal %d" */ +#define QMAN_POLL_LIMIT 32 +#define QMAN_PIRQ_DQRR_ITHRESH 12 +#define QMAN_DQRR_IT_MAX 15 +#define QMAN_ITP_MAX 0xFFF +#define QMAN_PIRQ_MR_ITHRESH 4 +#define QMAN_PIRQ_IPERIOD 100 + +/* Portal register assists */ + +#if defined(CONFIG_ARM) || defined(CONFIG_ARM64) +/* Cache-inhibited register offsets */ +#define QM_REG_EQCR_PI_CINH 0x3000 +#define QM_REG_EQCR_CI_CINH 0x3040 +#define QM_REG_EQCR_ITR 0x3080 +#define QM_REG_DQRR_PI_CINH 0x3100 +#define QM_REG_DQRR_CI_CINH 0x3140 +#define QM_REG_DQRR_ITR 0x3180 +#define QM_REG_DQRR_DCAP 0x31C0 +#define QM_REG_DQRR_SDQCR 0x3200 +#define QM_REG_DQRR_VDQCR 0x3240 +#define QM_REG_DQRR_PDQCR 0x3280 +#define QM_REG_MR_PI_CINH 0x3300 +#define QM_REG_MR_CI_CINH 0x3340 +#define QM_REG_MR_ITR 0x3380 +#define QM_REG_CFG 0x3500 +#define QM_REG_ISR 0x3600 +#define QM_REG_IER 0x3640 +#define QM_REG_ISDR 0x3680 +#define QM_REG_IIR 0x36C0 +#define QM_REG_ITPR 0x3740 + +/* Cache-enabled register offsets */ +#define QM_CL_EQCR 0x0000 +#define QM_CL_DQRR 0x1000 +#define QM_CL_MR 0x2000 +#define QM_CL_EQCR_PI_CENA 0x3000 +#define QM_CL_EQCR_CI_CENA 0x3040 +#define QM_CL_DQRR_PI_CENA 0x3100 +#define QM_CL_DQRR_CI_CENA 0x3140 +#define QM_CL_MR_PI_CENA 0x3300 +#define QM_CL_MR_CI_CENA 0x3340 +#define QM_CL_CR 0x3800 +#define QM_CL_RR0 0x3900 +#define QM_CL_RR1 0x3940 + +#else +/* Cache-inhibited register offsets */ +#define QM_REG_EQCR_PI_CINH 0x0000 +#define QM_REG_EQCR_CI_CINH 0x0004 +#define QM_REG_EQCR_ITR 0x0008 +#define QM_REG_DQRR_PI_CINH 0x0040 +#define QM_REG_DQRR_CI_CINH 0x0044 +#define QM_REG_DQRR_ITR 0x0048 +#define QM_REG_DQRR_DCAP 0x0050 +#define QM_REG_DQRR_SDQCR 0x0054 +#define QM_REG_DQRR_VDQCR 0x0058 +#define QM_REG_DQRR_PDQCR 0x005c +#define QM_REG_MR_PI_CINH 0x0080 +#define QM_REG_MR_CI_CINH 0x0084 +#define QM_REG_MR_ITR 0x0088 +#define QM_REG_CFG 0x0100 +#define QM_REG_ISR 0x0e00 +#define QM_REG_IER 0x0e04 +#define QM_REG_ISDR 0x0e08 +#define QM_REG_IIR 0x0e0c +#define QM_REG_ITPR 0x0e14 + +/* Cache-enabled register offsets */ +#define QM_CL_EQCR 0x0000 +#define QM_CL_DQRR 0x1000 +#define QM_CL_MR 0x2000 +#define QM_CL_EQCR_PI_CENA 0x3000 +#define QM_CL_EQCR_CI_CENA 0x3100 +#define QM_CL_DQRR_PI_CENA 0x3200 +#define QM_CL_DQRR_CI_CENA 0x3300 +#define QM_CL_MR_PI_CENA 0x3400 +#define QM_CL_MR_CI_CENA 0x3500 +#define QM_CL_CR 0x3800 +#define QM_CL_RR0 0x3900 +#define QM_CL_RR1 0x3940 +#endif + +/* + * BTW, the drivers (and h/w programming model) already obtain the required + * synchronisation for portal accesses and data-dependencies. Use of barrier()s + * or other order-preserving primitives simply degrade performance. Hence the + * use of the __raw_*() interfaces, which simply ensure that the compiler treats + * the portal registers as volatile + */ + +/* Cache-enabled ring access */ +#define qm_cl(base, idx) ((void *)base + ((idx) << 6)) + +/* + * Portal modes. + * Enum types; + * pmode == production mode + * cmode == consumption mode, + * dmode == h/w dequeue mode. + * Enum values use 3 letter codes. First letter matches the portal mode, + * remaining two letters indicate; + * ci == cache-inhibited portal register + * ce == cache-enabled portal register + * vb == in-band valid-bit (cache-enabled) + * dc == DCA (Discrete Consumption Acknowledgment), DQRR-only + * As for "enum qm_dqrr_dmode", it should be self-explanatory. + */ +enum qm_eqcr_pmode { /* matches QCSP_CFG::EPM */ + qm_eqcr_pci = 0, /* PI index, cache-inhibited */ + qm_eqcr_pce = 1, /* PI index, cache-enabled */ + qm_eqcr_pvb = 2 /* valid-bit */ +}; +enum qm_dqrr_dmode { /* matches QCSP_CFG::DP */ + qm_dqrr_dpush = 0, /* SDQCR + VDQCR */ + qm_dqrr_dpull = 1 /* PDQCR */ +}; +enum qm_dqrr_pmode { /* s/w-only */ + qm_dqrr_pci, /* reads DQRR_PI_CINH */ + qm_dqrr_pce, /* reads DQRR_PI_CENA */ + qm_dqrr_pvb /* reads valid-bit */ +}; +enum qm_dqrr_cmode { /* matches QCSP_CFG::DCM */ + qm_dqrr_cci = 0, /* CI index, cache-inhibited */ + qm_dqrr_cce = 1, /* CI index, cache-enabled */ + qm_dqrr_cdc = 2 /* Discrete Consumption Acknowledgment */ +}; +enum qm_mr_pmode { /* s/w-only */ + qm_mr_pci, /* reads MR_PI_CINH */ + qm_mr_pce, /* reads MR_PI_CENA */ + qm_mr_pvb /* reads valid-bit */ +}; +enum qm_mr_cmode { /* matches QCSP_CFG::MM */ + qm_mr_cci = 0, /* CI index, cache-inhibited */ + qm_mr_cce = 1 /* CI index, cache-enabled */ +}; + +/* --- Portal structures --- */ + +#define QM_EQCR_SIZE 8 +#define QM_DQRR_SIZE 16 +#define QM_MR_SIZE 8 + +/* "Enqueue Command" */ +struct qm_eqcr_entry { + u8 _ncw_verb; /* writes to this are non-coherent */ + u8 dca; + __be16 seqnum; + u8 __reserved[4]; + __be32 fqid; /* 24-bit */ + __be32 tag; + struct qm_fd fd; + u8 __reserved3[32]; +} __packed __aligned(8); +#define QM_EQCR_VERB_VBIT 0x80 +#define QM_EQCR_VERB_CMD_MASK 0x61 /* but only one value; */ +#define QM_EQCR_VERB_CMD_ENQUEUE 0x01 +#define QM_EQCR_SEQNUM_NESN 0x8000 /* Advance NESN */ +#define QM_EQCR_SEQNUM_NLIS 0x4000 /* More fragments to come */ +#define QM_EQCR_SEQNUM_SEQMASK 0x3fff /* sequence number goes here */ + +struct qm_eqcr { + struct qm_eqcr_entry *ring, *cursor; + u8 ci, available, ithresh, vbit; +#ifdef CONFIG_FSL_DPAA_CHECKING + u32 busy; + enum qm_eqcr_pmode pmode; +#endif +}; + +struct qm_dqrr { + const struct qm_dqrr_entry *ring, *cursor; + u8 pi, ci, fill, ithresh, vbit; +#ifdef CONFIG_FSL_DPAA_CHECKING + enum qm_dqrr_dmode dmode; + enum qm_dqrr_pmode pmode; + enum qm_dqrr_cmode cmode; +#endif +}; + +struct qm_mr { + union qm_mr_entry *ring, *cursor; + u8 pi, ci, fill, ithresh, vbit; +#ifdef CONFIG_FSL_DPAA_CHECKING + enum qm_mr_pmode pmode; + enum qm_mr_cmode cmode; +#endif +}; + +/* MC (Management Command) command */ +/* "FQ" command layout */ +struct qm_mcc_fq { + u8 _ncw_verb; + u8 __reserved1[3]; + __be32 fqid; /* 24-bit */ + u8 __reserved2[56]; +} __packed; + +/* "CGR" command layout */ +struct qm_mcc_cgr { + u8 _ncw_verb; + u8 __reserved1[30]; + u8 cgid; + u8 __reserved2[32]; +}; + +#define QM_MCC_VERB_VBIT 0x80 +#define QM_MCC_VERB_MASK 0x7f /* where the verb contains; */ +#define QM_MCC_VERB_INITFQ_PARKED 0x40 +#define QM_MCC_VERB_INITFQ_SCHED 0x41 +#define QM_MCC_VERB_QUERYFQ 0x44 +#define QM_MCC_VERB_QUERYFQ_NP 0x45 /* "non-programmable" fields */ +#define QM_MCC_VERB_QUERYWQ 0x46 +#define QM_MCC_VERB_QUERYWQ_DEDICATED 0x47 +#define QM_MCC_VERB_ALTER_SCHED 0x48 /* Schedule FQ */ +#define QM_MCC_VERB_ALTER_FE 0x49 /* Force Eligible FQ */ +#define QM_MCC_VERB_ALTER_RETIRE 0x4a /* Retire FQ */ +#define QM_MCC_VERB_ALTER_OOS 0x4b /* Take FQ out of service */ +#define QM_MCC_VERB_ALTER_FQXON 0x4d /* FQ XON */ +#define QM_MCC_VERB_ALTER_FQXOFF 0x4e /* FQ XOFF */ +#define QM_MCC_VERB_INITCGR 0x50 +#define QM_MCC_VERB_MODIFYCGR 0x51 +#define QM_MCC_VERB_CGRTESTWRITE 0x52 +#define QM_MCC_VERB_QUERYCGR 0x58 +#define QM_MCC_VERB_QUERYCONGESTION 0x59 +union qm_mc_command { + struct { + u8 _ncw_verb; /* writes to this are non-coherent */ + u8 __reserved[63]; + }; + struct qm_mcc_initfq initfq; + struct qm_mcc_initcgr initcgr; + struct qm_mcc_fq fq; + struct qm_mcc_cgr cgr; +}; + +/* MC (Management Command) result */ +/* "Query FQ" */ +struct qm_mcr_queryfq { + u8 verb; + u8 result; + u8 __reserved1[8]; + struct qm_fqd fqd; /* the FQD fields are here */ + u8 __reserved2[30]; +} __packed; + +/* "Alter FQ State Commands" */ +struct qm_mcr_alterfq { + u8 verb; + u8 result; + u8 fqs; /* Frame Queue Status */ + u8 __reserved1[61]; +}; +#define QM_MCR_VERB_RRID 0x80 +#define QM_MCR_VERB_MASK QM_MCC_VERB_MASK +#define QM_MCR_VERB_INITFQ_PARKED QM_MCC_VERB_INITFQ_PARKED +#define QM_MCR_VERB_INITFQ_SCHED QM_MCC_VERB_INITFQ_SCHED +#define QM_MCR_VERB_QUERYFQ QM_MCC_VERB_QUERYFQ +#define QM_MCR_VERB_QUERYFQ_NP QM_MCC_VERB_QUERYFQ_NP +#define QM_MCR_VERB_QUERYWQ QM_MCC_VERB_QUERYWQ +#define QM_MCR_VERB_QUERYWQ_DEDICATED QM_MCC_VERB_QUERYWQ_DEDICATED +#define QM_MCR_VERB_ALTER_SCHED QM_MCC_VERB_ALTER_SCHED +#define QM_MCR_VERB_ALTER_FE QM_MCC_VERB_ALTER_FE +#define QM_MCR_VERB_ALTER_RETIRE QM_MCC_VERB_ALTER_RETIRE +#define QM_MCR_VERB_ALTER_OOS QM_MCC_VERB_ALTER_OOS +#define QM_MCR_RESULT_NULL 0x00 +#define QM_MCR_RESULT_OK 0xf0 +#define QM_MCR_RESULT_ERR_FQID 0xf1 +#define QM_MCR_RESULT_ERR_FQSTATE 0xf2 +#define QM_MCR_RESULT_ERR_NOTEMPTY 0xf3 /* OOS fails if FQ is !empty */ +#define QM_MCR_RESULT_ERR_BADCHANNEL 0xf4 +#define QM_MCR_RESULT_PENDING 0xf8 +#define QM_MCR_RESULT_ERR_BADCOMMAND 0xff +#define QM_MCR_FQS_ORLPRESENT 0x02 /* ORL fragments to come */ +#define QM_MCR_FQS_NOTEMPTY 0x01 /* FQ has enqueued frames */ +#define QM_MCR_TIMEOUT 10000 /* us */ +union qm_mc_result { + struct { + u8 verb; + u8 result; + u8 __reserved1[62]; + }; + struct qm_mcr_queryfq queryfq; + struct qm_mcr_alterfq alterfq; + struct qm_mcr_querycgr querycgr; + struct qm_mcr_querycongestion querycongestion; + struct qm_mcr_querywq querywq; + struct qm_mcr_queryfq_np queryfq_np; +}; + +struct qm_mc { + union qm_mc_command *cr; + union qm_mc_result *rr; + u8 rridx, vbit; +#ifdef CONFIG_FSL_DPAA_CHECKING + enum { + /* Can be _mc_start()ed */ + qman_mc_idle, + /* Can be _mc_commit()ed or _mc_abort()ed */ + qman_mc_user, + /* Can only be _mc_retry()ed */ + qman_mc_hw + } state; +#endif +}; + +struct qm_addr { + void *ce; /* cache-enabled */ + __be32 *ce_be; /* same value as above but for direct access */ + void __iomem *ci; /* cache-inhibited */ +}; + +struct qm_portal { + /* + * In the non-CONFIG_FSL_DPAA_CHECKING case, the following stuff up to + * and including 'mc' fits within a cacheline (yay!). The 'config' part + * is setup-only, so isn't a cause for a concern. In other words, don't + * rearrange this structure on a whim, there be dragons ... + */ + struct qm_addr addr; + struct qm_eqcr eqcr; + struct qm_dqrr dqrr; + struct qm_mr mr; + struct qm_mc mc; +} ____cacheline_aligned; + +/* Cache-inhibited register access. */ +static inline u32 qm_in(struct qm_portal *p, u32 offset) +{ + return ioread32be(p->addr.ci + offset); +} + +static inline void qm_out(struct qm_portal *p, u32 offset, u32 val) +{ + iowrite32be(val, p->addr.ci + offset); +} + +/* Cache Enabled Portal Access */ +static inline void qm_cl_invalidate(struct qm_portal *p, u32 offset) +{ + dpaa_invalidate(p->addr.ce + offset); +} + +static inline void qm_cl_touch_ro(struct qm_portal *p, u32 offset) +{ + dpaa_touch_ro(p->addr.ce + offset); +} + +static inline u32 qm_ce_in(struct qm_portal *p, u32 offset) +{ + return be32_to_cpu(*(p->addr.ce_be + (offset/4))); +} + +/* --- EQCR API --- */ + +#define EQCR_SHIFT ilog2(sizeof(struct qm_eqcr_entry)) +#define EQCR_CARRY (uintptr_t)(QM_EQCR_SIZE << EQCR_SHIFT) + +/* Bit-wise logic to wrap a ring pointer by clearing the "carry bit" */ +static struct qm_eqcr_entry *eqcr_carryclear(struct qm_eqcr_entry *p) +{ + uintptr_t addr = (uintptr_t)p; + + addr &= ~EQCR_CARRY; + + return (struct qm_eqcr_entry *)addr; +} + +/* Bit-wise logic to convert a ring pointer to a ring index */ +static int eqcr_ptr2idx(struct qm_eqcr_entry *e) +{ + return ((uintptr_t)e >> EQCR_SHIFT) & (QM_EQCR_SIZE - 1); +} + +/* Increment the 'cursor' ring pointer, taking 'vbit' into account */ +static inline void eqcr_inc(struct qm_eqcr *eqcr) +{ + /* increment to the next EQCR pointer and handle overflow and 'vbit' */ + struct qm_eqcr_entry *partial = eqcr->cursor + 1; + + eqcr->cursor = eqcr_carryclear(partial); + if (partial != eqcr->cursor) + eqcr->vbit ^= QM_EQCR_VERB_VBIT; +} + +static inline int qm_eqcr_init(struct qm_portal *portal, + enum qm_eqcr_pmode pmode, + unsigned int eq_stash_thresh, + int eq_stash_prio) +{ + struct qm_eqcr *eqcr = &portal->eqcr; + u32 cfg; + u8 pi; + + eqcr->ring = portal->addr.ce + QM_CL_EQCR; + eqcr->ci = qm_in(portal, QM_REG_EQCR_CI_CINH) & (QM_EQCR_SIZE - 1); + qm_cl_invalidate(portal, QM_CL_EQCR_CI_CENA); + pi = qm_in(portal, QM_REG_EQCR_PI_CINH) & (QM_EQCR_SIZE - 1); + eqcr->cursor = eqcr->ring + pi; + eqcr->vbit = (qm_in(portal, QM_REG_EQCR_PI_CINH) & QM_EQCR_SIZE) ? + QM_EQCR_VERB_VBIT : 0; + eqcr->available = QM_EQCR_SIZE - 1 - + dpaa_cyc_diff(QM_EQCR_SIZE, eqcr->ci, pi); + eqcr->ithresh = qm_in(portal, QM_REG_EQCR_ITR); +#ifdef CONFIG_FSL_DPAA_CHECKING + eqcr->busy = 0; + eqcr->pmode = pmode; +#endif + cfg = (qm_in(portal, QM_REG_CFG) & 0x00ffffff) | + (eq_stash_thresh << 28) | /* QCSP_CFG: EST */ + (eq_stash_prio << 26) | /* QCSP_CFG: EP */ + ((pmode & 0x3) << 24); /* QCSP_CFG::EPM */ + qm_out(portal, QM_REG_CFG, cfg); + return 0; +} + +static inline void qm_eqcr_finish(struct qm_portal *portal) +{ + struct qm_eqcr *eqcr = &portal->eqcr; + u8 pi = qm_in(portal, QM_REG_EQCR_PI_CINH) & (QM_EQCR_SIZE - 1); + u8 ci = qm_in(portal, QM_REG_EQCR_CI_CINH) & (QM_EQCR_SIZE - 1); + + DPAA_ASSERT(!eqcr->busy); + if (pi != eqcr_ptr2idx(eqcr->cursor)) + pr_crit("losing uncommitted EQCR entries\n"); + if (ci != eqcr->ci) + pr_crit("missing existing EQCR completions\n"); + if (eqcr->ci != eqcr_ptr2idx(eqcr->cursor)) + pr_crit("EQCR destroyed unquiesced\n"); +} + +static inline struct qm_eqcr_entry *qm_eqcr_start_no_stash(struct qm_portal + *portal) +{ + struct qm_eqcr *eqcr = &portal->eqcr; + + DPAA_ASSERT(!eqcr->busy); + if (!eqcr->available) + return NULL; + +#ifdef CONFIG_FSL_DPAA_CHECKING + eqcr->busy = 1; +#endif + dpaa_zero(eqcr->cursor); + return eqcr->cursor; +} + +static inline struct qm_eqcr_entry *qm_eqcr_start_stash(struct qm_portal + *portal) +{ + struct qm_eqcr *eqcr = &portal->eqcr; + u8 diff, old_ci; + + DPAA_ASSERT(!eqcr->busy); + if (!eqcr->available) { + old_ci = eqcr->ci; + eqcr->ci = qm_ce_in(portal, QM_CL_EQCR_CI_CENA) & + (QM_EQCR_SIZE - 1); + diff = dpaa_cyc_diff(QM_EQCR_SIZE, old_ci, eqcr->ci); + eqcr->available += diff; + if (!diff) + return NULL; + } +#ifdef CONFIG_FSL_DPAA_CHECKING + eqcr->busy = 1; +#endif + dpaa_zero(eqcr->cursor); + return eqcr->cursor; +} + +static inline void eqcr_commit_checks(struct qm_eqcr *eqcr) +{ + DPAA_ASSERT(eqcr->busy); + DPAA_ASSERT(!(be32_to_cpu(eqcr->cursor->fqid) & ~QM_FQID_MASK)); + DPAA_ASSERT(eqcr->available >= 1); +} + +static inline void qm_eqcr_pvb_commit(struct qm_portal *portal, u8 myverb) +{ + struct qm_eqcr *eqcr = &portal->eqcr; + struct qm_eqcr_entry *eqcursor; + + eqcr_commit_checks(eqcr); + DPAA_ASSERT(eqcr->pmode == qm_eqcr_pvb); + dma_wmb(); + eqcursor = eqcr->cursor; + eqcursor->_ncw_verb = myverb | eqcr->vbit; + dpaa_flush(eqcursor); + eqcr_inc(eqcr); + eqcr->available--; +#ifdef CONFIG_FSL_DPAA_CHECKING + eqcr->busy = 0; +#endif +} + +static inline void qm_eqcr_cce_prefetch(struct qm_portal *portal) +{ + qm_cl_touch_ro(portal, QM_CL_EQCR_CI_CENA); +} + +static inline u8 qm_eqcr_cce_update(struct qm_portal *portal) +{ + struct qm_eqcr *eqcr = &portal->eqcr; + u8 diff, old_ci = eqcr->ci; + + eqcr->ci = qm_ce_in(portal, QM_CL_EQCR_CI_CENA) & (QM_EQCR_SIZE - 1); + qm_cl_invalidate(portal, QM_CL_EQCR_CI_CENA); + diff = dpaa_cyc_diff(QM_EQCR_SIZE, old_ci, eqcr->ci); + eqcr->available += diff; + return diff; +} + +static inline void qm_eqcr_set_ithresh(struct qm_portal *portal, u8 ithresh) +{ + struct qm_eqcr *eqcr = &portal->eqcr; + + eqcr->ithresh = ithresh; + qm_out(portal, QM_REG_EQCR_ITR, ithresh); +} + +static inline u8 qm_eqcr_get_avail(struct qm_portal *portal) +{ + struct qm_eqcr *eqcr = &portal->eqcr; + + return eqcr->available; +} + +static inline u8 qm_eqcr_get_fill(struct qm_portal *portal) +{ + struct qm_eqcr *eqcr = &portal->eqcr; + + return QM_EQCR_SIZE - 1 - eqcr->available; +} + +/* --- DQRR API --- */ + +#define DQRR_SHIFT ilog2(sizeof(struct qm_dqrr_entry)) +#define DQRR_CARRY (uintptr_t)(QM_DQRR_SIZE << DQRR_SHIFT) + +static const struct qm_dqrr_entry *dqrr_carryclear( + const struct qm_dqrr_entry *p) +{ + uintptr_t addr = (uintptr_t)p; + + addr &= ~DQRR_CARRY; + + return (const struct qm_dqrr_entry *)addr; +} + +static inline int dqrr_ptr2idx(const struct qm_dqrr_entry *e) +{ + return ((uintptr_t)e >> DQRR_SHIFT) & (QM_DQRR_SIZE - 1); +} + +static const struct qm_dqrr_entry *dqrr_inc(const struct qm_dqrr_entry *e) +{ + return dqrr_carryclear(e + 1); +} + +static inline void qm_dqrr_set_maxfill(struct qm_portal *portal, u8 mf) +{ + qm_out(portal, QM_REG_CFG, (qm_in(portal, QM_REG_CFG) & 0xff0fffff) | + ((mf & (QM_DQRR_SIZE - 1)) << 20)); +} + +static inline int qm_dqrr_init(struct qm_portal *portal, + const struct qm_portal_config *config, + enum qm_dqrr_dmode dmode, + enum qm_dqrr_pmode pmode, + enum qm_dqrr_cmode cmode, u8 max_fill) +{ + struct qm_dqrr *dqrr = &portal->dqrr; + u32 cfg; + + /* Make sure the DQRR will be idle when we enable */ + qm_out(portal, QM_REG_DQRR_SDQCR, 0); + qm_out(portal, QM_REG_DQRR_VDQCR, 0); + qm_out(portal, QM_REG_DQRR_PDQCR, 0); + dqrr->ring = portal->addr.ce + QM_CL_DQRR; + dqrr->pi = qm_in(portal, QM_REG_DQRR_PI_CINH) & (QM_DQRR_SIZE - 1); + dqrr->ci = qm_in(portal, QM_REG_DQRR_CI_CINH) & (QM_DQRR_SIZE - 1); + dqrr->cursor = dqrr->ring + dqrr->ci; + dqrr->fill = dpaa_cyc_diff(QM_DQRR_SIZE, dqrr->ci, dqrr->pi); + dqrr->vbit = (qm_in(portal, QM_REG_DQRR_PI_CINH) & QM_DQRR_SIZE) ? + QM_DQRR_VERB_VBIT : 0; + dqrr->ithresh = qm_in(portal, QM_REG_DQRR_ITR); +#ifdef CONFIG_FSL_DPAA_CHECKING + dqrr->dmode = dmode; + dqrr->pmode = pmode; + dqrr->cmode = cmode; +#endif + /* Invalidate every ring entry before beginning */ + for (cfg = 0; cfg < QM_DQRR_SIZE; cfg++) + dpaa_invalidate(qm_cl(dqrr->ring, cfg)); + cfg = (qm_in(portal, QM_REG_CFG) & 0xff000f00) | + ((max_fill & (QM_DQRR_SIZE - 1)) << 20) | /* DQRR_MF */ + ((dmode & 1) << 18) | /* DP */ + ((cmode & 3) << 16) | /* DCM */ + 0xa0 | /* RE+SE */ + (0 ? 0x40 : 0) | /* Ignore RP */ + (0 ? 0x10 : 0); /* Ignore SP */ + qm_out(portal, QM_REG_CFG, cfg); + qm_dqrr_set_maxfill(portal, max_fill); + return 0; +} + +static inline void qm_dqrr_finish(struct qm_portal *portal) +{ +#ifdef CONFIG_FSL_DPAA_CHECKING + struct qm_dqrr *dqrr = &portal->dqrr; + + if (dqrr->cmode != qm_dqrr_cdc && + dqrr->ci != dqrr_ptr2idx(dqrr->cursor)) + pr_crit("Ignoring completed DQRR entries\n"); +#endif +} + +static inline const struct qm_dqrr_entry *qm_dqrr_current( + struct qm_portal *portal) +{ + struct qm_dqrr *dqrr = &portal->dqrr; + + if (!dqrr->fill) + return NULL; + return dqrr->cursor; +} + +static inline u8 qm_dqrr_next(struct qm_portal *portal) +{ + struct qm_dqrr *dqrr = &portal->dqrr; + + DPAA_ASSERT(dqrr->fill); + dqrr->cursor = dqrr_inc(dqrr->cursor); + return --dqrr->fill; +} + +static inline void qm_dqrr_pvb_update(struct qm_portal *portal) +{ + struct qm_dqrr *dqrr = &portal->dqrr; + struct qm_dqrr_entry *res = qm_cl(dqrr->ring, dqrr->pi); + + DPAA_ASSERT(dqrr->pmode == qm_dqrr_pvb); +#ifndef CONFIG_FSL_PAMU + /* + * If PAMU is not available we need to invalidate the cache. + * When PAMU is available the cache is updated by stash + */ + dpaa_invalidate_touch_ro(res); +#endif + if ((res->verb & QM_DQRR_VERB_VBIT) == dqrr->vbit) { + dqrr->pi = (dqrr->pi + 1) & (QM_DQRR_SIZE - 1); + if (!dqrr->pi) + dqrr->vbit ^= QM_DQRR_VERB_VBIT; + dqrr->fill++; + } +} + +static inline void qm_dqrr_cdc_consume_1ptr(struct qm_portal *portal, + const struct qm_dqrr_entry *dq, + int park) +{ + __maybe_unused struct qm_dqrr *dqrr = &portal->dqrr; + int idx = dqrr_ptr2idx(dq); + + DPAA_ASSERT(dqrr->cmode == qm_dqrr_cdc); + DPAA_ASSERT((dqrr->ring + idx) == dq); + DPAA_ASSERT(idx < QM_DQRR_SIZE); + qm_out(portal, QM_REG_DQRR_DCAP, (0 << 8) | /* DQRR_DCAP::S */ + ((park ? 1 : 0) << 6) | /* DQRR_DCAP::PK */ + idx); /* DQRR_DCAP::DCAP_CI */ +} + +static inline void qm_dqrr_cdc_consume_n(struct qm_portal *portal, u32 bitmask) +{ + __maybe_unused struct qm_dqrr *dqrr = &portal->dqrr; + + DPAA_ASSERT(dqrr->cmode == qm_dqrr_cdc); + qm_out(portal, QM_REG_DQRR_DCAP, (1 << 8) | /* DQRR_DCAP::S */ + (bitmask << 16)); /* DQRR_DCAP::DCAP_CI */ +} + +static inline void qm_dqrr_sdqcr_set(struct qm_portal *portal, u32 sdqcr) +{ + qm_out(portal, QM_REG_DQRR_SDQCR, sdqcr); +} + +static inline void qm_dqrr_vdqcr_set(struct qm_portal *portal, u32 vdqcr) +{ + qm_out(portal, QM_REG_DQRR_VDQCR, vdqcr); +} + +static inline int qm_dqrr_set_ithresh(struct qm_portal *portal, u8 ithresh) +{ + + if (ithresh > QMAN_DQRR_IT_MAX) + return -EINVAL; + + qm_out(portal, QM_REG_DQRR_ITR, ithresh); + + return 0; +} + +/* --- MR API --- */ + +#define MR_SHIFT ilog2(sizeof(union qm_mr_entry)) +#define MR_CARRY (uintptr_t)(QM_MR_SIZE << MR_SHIFT) + +static union qm_mr_entry *mr_carryclear(union qm_mr_entry *p) +{ + uintptr_t addr = (uintptr_t)p; + + addr &= ~MR_CARRY; + + return (union qm_mr_entry *)addr; +} + +static inline int mr_ptr2idx(const union qm_mr_entry *e) +{ + return ((uintptr_t)e >> MR_SHIFT) & (QM_MR_SIZE - 1); +} + +static inline union qm_mr_entry *mr_inc(union qm_mr_entry *e) +{ + return mr_carryclear(e + 1); +} + +static inline int qm_mr_init(struct qm_portal *portal, enum qm_mr_pmode pmode, + enum qm_mr_cmode cmode) +{ + struct qm_mr *mr = &portal->mr; + u32 cfg; + + mr->ring = portal->addr.ce + QM_CL_MR; + mr->pi = qm_in(portal, QM_REG_MR_PI_CINH) & (QM_MR_SIZE - 1); + mr->ci = qm_in(portal, QM_REG_MR_CI_CINH) & (QM_MR_SIZE - 1); + mr->cursor = mr->ring + mr->ci; + mr->fill = dpaa_cyc_diff(QM_MR_SIZE, mr->ci, mr->pi); + mr->vbit = (qm_in(portal, QM_REG_MR_PI_CINH) & QM_MR_SIZE) + ? QM_MR_VERB_VBIT : 0; + mr->ithresh = qm_in(portal, QM_REG_MR_ITR); +#ifdef CONFIG_FSL_DPAA_CHECKING + mr->pmode = pmode; + mr->cmode = cmode; +#endif + cfg = (qm_in(portal, QM_REG_CFG) & 0xfffff0ff) | + ((cmode & 1) << 8); /* QCSP_CFG:MM */ + qm_out(portal, QM_REG_CFG, cfg); + return 0; +} + +static inline void qm_mr_finish(struct qm_portal *portal) +{ + struct qm_mr *mr = &portal->mr; + + if (mr->ci != mr_ptr2idx(mr->cursor)) + pr_crit("Ignoring completed MR entries\n"); +} + +static inline const union qm_mr_entry *qm_mr_current(struct qm_portal *portal) +{ + struct qm_mr *mr = &portal->mr; + + if (!mr->fill) + return NULL; + return mr->cursor; +} + +static inline int qm_mr_next(struct qm_portal *portal) +{ + struct qm_mr *mr = &portal->mr; + + DPAA_ASSERT(mr->fill); + mr->cursor = mr_inc(mr->cursor); + return --mr->fill; +} + +static inline void qm_mr_pvb_update(struct qm_portal *portal) +{ + struct qm_mr *mr = &portal->mr; + union qm_mr_entry *res = qm_cl(mr->ring, mr->pi); + + DPAA_ASSERT(mr->pmode == qm_mr_pvb); + + if ((res->verb & QM_MR_VERB_VBIT) == mr->vbit) { + mr->pi = (mr->pi + 1) & (QM_MR_SIZE - 1); + if (!mr->pi) + mr->vbit ^= QM_MR_VERB_VBIT; + mr->fill++; + res = mr_inc(res); + } + dpaa_invalidate_touch_ro(res); +} + +static inline void qm_mr_cci_consume(struct qm_portal *portal, u8 num) +{ + struct qm_mr *mr = &portal->mr; + + DPAA_ASSERT(mr->cmode == qm_mr_cci); + mr->ci = (mr->ci + num) & (QM_MR_SIZE - 1); + qm_out(portal, QM_REG_MR_CI_CINH, mr->ci); +} + +static inline void qm_mr_cci_consume_to_current(struct qm_portal *portal) +{ + struct qm_mr *mr = &portal->mr; + + DPAA_ASSERT(mr->cmode == qm_mr_cci); + mr->ci = mr_ptr2idx(mr->cursor); + qm_out(portal, QM_REG_MR_CI_CINH, mr->ci); +} + +static inline void qm_mr_set_ithresh(struct qm_portal *portal, u8 ithresh) +{ + qm_out(portal, QM_REG_MR_ITR, ithresh); +} + +/* --- Management command API --- */ + +static inline int qm_mc_init(struct qm_portal *portal) +{ + u8 rr0, rr1; + struct qm_mc *mc = &portal->mc; + + mc->cr = portal->addr.ce + QM_CL_CR; + mc->rr = portal->addr.ce + QM_CL_RR0; + /* + * The expected valid bit polarity for the next CR command is 0 + * if RR1 contains a valid response, and is 1 if RR0 contains a + * valid response. If both RR contain all 0, this indicates either + * that no command has been executed since reset (in which case the + * expected valid bit polarity is 1) + */ + rr0 = mc->rr->verb; + rr1 = (mc->rr+1)->verb; + if ((rr0 == 0 && rr1 == 0) || rr0 != 0) + mc->rridx = 1; + else + mc->rridx = 0; + mc->vbit = mc->rridx ? QM_MCC_VERB_VBIT : 0; +#ifdef CONFIG_FSL_DPAA_CHECKING + mc->state = qman_mc_idle; +#endif + return 0; +} + +static inline void qm_mc_finish(struct qm_portal *portal) +{ +#ifdef CONFIG_FSL_DPAA_CHECKING + struct qm_mc *mc = &portal->mc; + + DPAA_ASSERT(mc->state == qman_mc_idle); + if (mc->state != qman_mc_idle) + pr_crit("Losing incomplete MC command\n"); +#endif +} + +static inline union qm_mc_command *qm_mc_start(struct qm_portal *portal) +{ + struct qm_mc *mc = &portal->mc; + + DPAA_ASSERT(mc->state == qman_mc_idle); +#ifdef CONFIG_FSL_DPAA_CHECKING + mc->state = qman_mc_user; +#endif + dpaa_zero(mc->cr); + return mc->cr; +} + +static inline void qm_mc_commit(struct qm_portal *portal, u8 myverb) +{ + struct qm_mc *mc = &portal->mc; + union qm_mc_result *rr = mc->rr + mc->rridx; + + DPAA_ASSERT(mc->state == qman_mc_user); + dma_wmb(); + mc->cr->_ncw_verb = myverb | mc->vbit; + dpaa_flush(mc->cr); + dpaa_invalidate_touch_ro(rr); +#ifdef CONFIG_FSL_DPAA_CHECKING + mc->state = qman_mc_hw; +#endif +} + +static inline union qm_mc_result *qm_mc_result(struct qm_portal *portal) +{ + struct qm_mc *mc = &portal->mc; + union qm_mc_result *rr = mc->rr + mc->rridx; + + DPAA_ASSERT(mc->state == qman_mc_hw); + /* + * The inactive response register's verb byte always returns zero until + * its command is submitted and completed. This includes the valid-bit, + * in case you were wondering... + */ + if (!rr->verb) { + dpaa_invalidate_touch_ro(rr); + return NULL; + } + mc->rridx ^= 1; + mc->vbit ^= QM_MCC_VERB_VBIT; +#ifdef CONFIG_FSL_DPAA_CHECKING + mc->state = qman_mc_idle; +#endif + return rr; +} + +static inline int qm_mc_result_timeout(struct qm_portal *portal, + union qm_mc_result **mcr) +{ + int timeout = QM_MCR_TIMEOUT; + + do { + *mcr = qm_mc_result(portal); + if (*mcr) + break; + udelay(1); + } while (--timeout); + + return timeout; +} + +static inline void fq_set(struct qman_fq *fq, u32 mask) +{ + fq->flags |= mask; +} + +static inline void fq_clear(struct qman_fq *fq, u32 mask) +{ + fq->flags &= ~mask; +} + +static inline int fq_isset(struct qman_fq *fq, u32 mask) +{ + return fq->flags & mask; +} + +static inline int fq_isclear(struct qman_fq *fq, u32 mask) +{ + return !(fq->flags & mask); +} + +struct qman_portal { + struct qm_portal p; + /* PORTAL_BITS_*** - dynamic, strictly internal */ + unsigned long bits; + /* interrupt sources processed by portal_isr(), configurable */ + unsigned long irq_sources; + u32 use_eqcr_ci_stashing; + /* only 1 volatile dequeue at a time */ + struct qman_fq *vdqcr_owned; + u32 sdqcr; + /* probing time config params for cpu-affine portals */ + const struct qm_portal_config *config; + /* 2-element array. cgrs[0] is mask, cgrs[1] is snapshot. */ + struct qman_cgrs *cgrs; + /* linked-list of CSCN handlers. */ + struct list_head cgr_cbs; + /* list lock */ + spinlock_t cgr_lock; + struct work_struct congestion_work; + struct work_struct mr_work; + char irqname[MAX_IRQNAME]; +}; + +static cpumask_t affine_mask; +static DEFINE_SPINLOCK(affine_mask_lock); +static u16 affine_channels[NR_CPUS]; +static DEFINE_PER_CPU(struct qman_portal, qman_affine_portal); +struct qman_portal *affine_portals[NR_CPUS]; + +static inline struct qman_portal *get_affine_portal(void) +{ + return &get_cpu_var(qman_affine_portal); +} + +static inline void put_affine_portal(void) +{ + put_cpu_var(qman_affine_portal); +} + + +static inline struct qman_portal *get_portal_for_channel(u16 channel) +{ + int i; + + for (i = 0; i < num_possible_cpus(); i++) { + if (affine_portals[i] && + affine_portals[i]->config->channel == channel) + return affine_portals[i]; + } + + return NULL; +} + +static struct workqueue_struct *qm_portal_wq; + +int qman_dqrr_set_ithresh(struct qman_portal *portal, u8 ithresh) +{ + int res; + + if (!portal) + return -EINVAL; + + res = qm_dqrr_set_ithresh(&portal->p, ithresh); + if (res) + return res; + + portal->p.dqrr.ithresh = ithresh; + + return 0; +} +EXPORT_SYMBOL(qman_dqrr_set_ithresh); + +void qman_dqrr_get_ithresh(struct qman_portal *portal, u8 *ithresh) +{ + if (portal && ithresh) + *ithresh = qm_in(&portal->p, QM_REG_DQRR_ITR); +} +EXPORT_SYMBOL(qman_dqrr_get_ithresh); + +void qman_portal_get_iperiod(struct qman_portal *portal, u32 *iperiod) +{ + if (portal && iperiod) + *iperiod = qm_in(&portal->p, QM_REG_ITPR); +} +EXPORT_SYMBOL(qman_portal_get_iperiod); + +int qman_portal_set_iperiod(struct qman_portal *portal, u32 iperiod) +{ + if (!portal || iperiod > QMAN_ITP_MAX) + return -EINVAL; + + qm_out(&portal->p, QM_REG_ITPR, iperiod); + + return 0; +} +EXPORT_SYMBOL(qman_portal_set_iperiod); + +int qman_wq_alloc(void) +{ + qm_portal_wq = alloc_workqueue("qman_portal_wq", 0, 1); + if (!qm_portal_wq) + return -ENOMEM; + return 0; +} + + +void qman_enable_irqs(void) +{ + int i; + + for (i = 0; i < num_possible_cpus(); i++) { + if (affine_portals[i]) { + qm_out(&affine_portals[i]->p, QM_REG_ISR, 0xffffffff); + qm_out(&affine_portals[i]->p, QM_REG_IIR, 0); + } + + } +} + +/* + * This is what everything can wait on, even if it migrates to a different cpu + * to the one whose affine portal it is waiting on. + */ +static DECLARE_WAIT_QUEUE_HEAD(affine_queue); + +static struct qman_fq **fq_table; +static u32 num_fqids; + +int qman_alloc_fq_table(u32 _num_fqids) +{ + num_fqids = _num_fqids; + + fq_table = vzalloc(array3_size(sizeof(struct qman_fq *), + num_fqids, 2)); + if (!fq_table) + return -ENOMEM; + + pr_debug("Allocated fq lookup table at %p, entry count %u\n", + fq_table, num_fqids * 2); + return 0; +} + +static struct qman_fq *idx_to_fq(u32 idx) +{ + struct qman_fq *fq; + +#ifdef CONFIG_FSL_DPAA_CHECKING + if (WARN_ON(idx >= num_fqids * 2)) + return NULL; +#endif + fq = fq_table[idx]; + DPAA_ASSERT(!fq || idx == fq->idx); + + return fq; +} + +/* + * Only returns full-service fq objects, not enqueue-only + * references (QMAN_FQ_FLAG_NO_MODIFY). + */ +static struct qman_fq *fqid_to_fq(u32 fqid) +{ + return idx_to_fq(fqid * 2); +} + +static struct qman_fq *tag_to_fq(u32 tag) +{ +#if BITS_PER_LONG == 64 + return idx_to_fq(tag); +#else + return (struct qman_fq *)tag; +#endif +} + +static u32 fq_to_tag(struct qman_fq *fq) +{ +#if BITS_PER_LONG == 64 + return fq->idx; +#else + return (u32)fq; +#endif +} + +static u32 __poll_portal_slow(struct qman_portal *p, u32 is); +static inline unsigned int __poll_portal_fast(struct qman_portal *p, + unsigned int poll_limit, bool sched_napi); +static void qm_congestion_task(struct work_struct *work); +static void qm_mr_process_task(struct work_struct *work); + +static irqreturn_t portal_isr(int irq, void *ptr) +{ + struct qman_portal *p = ptr; + u32 is = qm_in(&p->p, QM_REG_ISR) & p->irq_sources; + u32 clear = 0; + + if (unlikely(!is)) + return IRQ_NONE; + + /* DQRR-handling if it's interrupt-driven */ + if (is & QM_PIRQ_DQRI) { + __poll_portal_fast(p, QMAN_POLL_LIMIT, true); + clear = QM_DQAVAIL_MASK | QM_PIRQ_DQRI; + } + /* Handling of anything else that's interrupt-driven */ + clear |= __poll_portal_slow(p, is) & QM_PIRQ_SLOW; + qm_out(&p->p, QM_REG_ISR, clear); + return IRQ_HANDLED; +} + +static int drain_mr_fqrni(struct qm_portal *p) +{ + const union qm_mr_entry *msg; +loop: + qm_mr_pvb_update(p); + msg = qm_mr_current(p); + if (!msg) { + /* + * if MR was full and h/w had other FQRNI entries to produce, we + * need to allow it time to produce those entries once the + * existing entries are consumed. A worst-case situation + * (fully-loaded system) means h/w sequencers may have to do 3-4 + * other things before servicing the portal's MR pump, each of + * which (if slow) may take ~50 qman cycles (which is ~200 + * processor cycles). So rounding up and then multiplying this + * worst-case estimate by a factor of 10, just to be + * ultra-paranoid, goes as high as 10,000 cycles. NB, we consume + * one entry at a time, so h/w has an opportunity to produce new + * entries well before the ring has been fully consumed, so + * we're being *really* paranoid here. + */ + mdelay(1); + qm_mr_pvb_update(p); + msg = qm_mr_current(p); + if (!msg) + return 0; + } + if ((msg->verb & QM_MR_VERB_TYPE_MASK) != QM_MR_VERB_FQRNI) { + /* We aren't draining anything but FQRNIs */ + pr_err("Found verb 0x%x in MR\n", msg->verb); + return -1; + } + qm_mr_next(p); + qm_mr_cci_consume(p, 1); + goto loop; +} + +static int qman_create_portal(struct qman_portal *portal, + const struct qm_portal_config *c, + const struct qman_cgrs *cgrs) +{ + struct qm_portal *p; + int ret; + u32 isdr; + + p = &portal->p; + +#ifdef CONFIG_FSL_PAMU + /* PAMU is required for stashing */ + portal->use_eqcr_ci_stashing = ((qman_ip_rev >= QMAN_REV30) ? 1 : 0); +#else + portal->use_eqcr_ci_stashing = 0; +#endif + /* + * prep the low-level portal struct with the mapped addresses from the + * config, everything that follows depends on it and "config" is more + * for (de)reference + */ + p->addr.ce = c->addr_virt_ce; + p->addr.ce_be = c->addr_virt_ce; + p->addr.ci = c->addr_virt_ci; + /* + * If CI-stashing is used, the current defaults use a threshold of 3, + * and stash with high-than-DQRR priority. + */ + if (qm_eqcr_init(p, qm_eqcr_pvb, + portal->use_eqcr_ci_stashing ? 3 : 0, 1)) { + dev_err(c->dev, "EQCR initialisation failed\n"); + goto fail_eqcr; + } + if (qm_dqrr_init(p, c, qm_dqrr_dpush, qm_dqrr_pvb, + qm_dqrr_cdc, DQRR_MAXFILL)) { + dev_err(c->dev, "DQRR initialisation failed\n"); + goto fail_dqrr; + } + if (qm_mr_init(p, qm_mr_pvb, qm_mr_cci)) { + dev_err(c->dev, "MR initialisation failed\n"); + goto fail_mr; + } + if (qm_mc_init(p)) { + dev_err(c->dev, "MC initialisation failed\n"); + goto fail_mc; + } + /* static interrupt-gating controls */ + qm_dqrr_set_ithresh(p, QMAN_PIRQ_DQRR_ITHRESH); + qm_mr_set_ithresh(p, QMAN_PIRQ_MR_ITHRESH); + qm_out(p, QM_REG_ITPR, QMAN_PIRQ_IPERIOD); + portal->cgrs = kmalloc_array(2, sizeof(*cgrs), GFP_KERNEL); + if (!portal->cgrs) + goto fail_cgrs; + /* initial snapshot is no-depletion */ + qman_cgrs_init(&portal->cgrs[1]); + if (cgrs) + portal->cgrs[0] = *cgrs; + else + /* if the given mask is NULL, assume all CGRs can be seen */ + qman_cgrs_fill(&portal->cgrs[0]); + INIT_LIST_HEAD(&portal->cgr_cbs); + spin_lock_init(&portal->cgr_lock); + INIT_WORK(&portal->congestion_work, qm_congestion_task); + INIT_WORK(&portal->mr_work, qm_mr_process_task); + portal->bits = 0; + portal->sdqcr = QM_SDQCR_SOURCE_CHANNELS | QM_SDQCR_COUNT_UPTO3 | + QM_SDQCR_DEDICATED_PRECEDENCE | QM_SDQCR_TYPE_PRIO_QOS | + QM_SDQCR_TOKEN_SET(0xab) | QM_SDQCR_CHANNELS_DEDICATED; + isdr = 0xffffffff; + qm_out(p, QM_REG_ISDR, isdr); + portal->irq_sources = 0; + qm_out(p, QM_REG_IER, 0); + snprintf(portal->irqname, MAX_IRQNAME, IRQNAME, c->cpu); + qm_out(p, QM_REG_IIR, 1); + if (request_irq(c->irq, portal_isr, 0, portal->irqname, portal)) { + dev_err(c->dev, "request_irq() failed\n"); + goto fail_irq; + } + + if (dpaa_set_portal_irq_affinity(c->dev, c->irq, c->cpu)) + goto fail_affinity; + + /* Need EQCR to be empty before continuing */ + isdr &= ~QM_PIRQ_EQCI; + qm_out(p, QM_REG_ISDR, isdr); + ret = qm_eqcr_get_fill(p); + if (ret) { + dev_err(c->dev, "EQCR unclean\n"); + goto fail_eqcr_empty; + } + isdr &= ~(QM_PIRQ_DQRI | QM_PIRQ_MRI); + qm_out(p, QM_REG_ISDR, isdr); + if (qm_dqrr_current(p)) { + dev_dbg(c->dev, "DQRR unclean\n"); + qm_dqrr_cdc_consume_n(p, 0xffff); + } + if (qm_mr_current(p) && drain_mr_fqrni(p)) { + /* special handling, drain just in case it's a few FQRNIs */ + const union qm_mr_entry *e = qm_mr_current(p); + + dev_err(c->dev, "MR dirty, VB 0x%x, rc 0x%x, addr 0x%llx\n", + e->verb, e->ern.rc, qm_fd_addr_get64(&e->ern.fd)); + goto fail_dqrr_mr_empty; + } + /* Success */ + portal->config = c; + qm_out(p, QM_REG_ISR, 0xffffffff); + qm_out(p, QM_REG_ISDR, 0); + if (!qman_requires_cleanup()) + qm_out(p, QM_REG_IIR, 0); + /* Write a sane SDQCR */ + qm_dqrr_sdqcr_set(p, portal->sdqcr); + return 0; + +fail_dqrr_mr_empty: +fail_eqcr_empty: +fail_affinity: + free_irq(c->irq, portal); +fail_irq: + kfree(portal->cgrs); +fail_cgrs: + qm_mc_finish(p); +fail_mc: + qm_mr_finish(p); +fail_mr: + qm_dqrr_finish(p); +fail_dqrr: + qm_eqcr_finish(p); +fail_eqcr: + return -EIO; +} + +struct qman_portal *qman_create_affine_portal(const struct qm_portal_config *c, + const struct qman_cgrs *cgrs) +{ + struct qman_portal *portal; + int err; + + portal = &per_cpu(qman_affine_portal, c->cpu); + err = qman_create_portal(portal, c, cgrs); + if (err) + return NULL; + + spin_lock(&affine_mask_lock); + cpumask_set_cpu(c->cpu, &affine_mask); + affine_channels[c->cpu] = c->channel; + affine_portals[c->cpu] = portal; + spin_unlock(&affine_mask_lock); + + return portal; +} + +static void qman_destroy_portal(struct qman_portal *qm) +{ + const struct qm_portal_config *pcfg; + + /* Stop dequeues on the portal */ + qm_dqrr_sdqcr_set(&qm->p, 0); + + /* + * NB we do this to "quiesce" EQCR. If we add enqueue-completions or + * something related to QM_PIRQ_EQCI, this may need fixing. + * Also, due to the prefetching model used for CI updates in the enqueue + * path, this update will only invalidate the CI cacheline *after* + * working on it, so we need to call this twice to ensure a full update + * irrespective of where the enqueue processing was at when the teardown + * began. + */ + qm_eqcr_cce_update(&qm->p); + qm_eqcr_cce_update(&qm->p); + pcfg = qm->config; + + free_irq(pcfg->irq, qm); + + kfree(qm->cgrs); + qm_mc_finish(&qm->p); + qm_mr_finish(&qm->p); + qm_dqrr_finish(&qm->p); + qm_eqcr_finish(&qm->p); + + qm->config = NULL; +} + +const struct qm_portal_config *qman_destroy_affine_portal(void) +{ + struct qman_portal *qm = get_affine_portal(); + const struct qm_portal_config *pcfg; + int cpu; + + pcfg = qm->config; + cpu = pcfg->cpu; + + qman_destroy_portal(qm); + + spin_lock(&affine_mask_lock); + cpumask_clear_cpu(cpu, &affine_mask); + spin_unlock(&affine_mask_lock); + put_affine_portal(); + return pcfg; +} + +/* Inline helper to reduce nesting in __poll_portal_slow() */ +static inline void fq_state_change(struct qman_portal *p, struct qman_fq *fq, + const union qm_mr_entry *msg, u8 verb) +{ + switch (verb) { + case QM_MR_VERB_FQRL: + DPAA_ASSERT(fq_isset(fq, QMAN_FQ_STATE_ORL)); + fq_clear(fq, QMAN_FQ_STATE_ORL); + break; + case QM_MR_VERB_FQRN: + DPAA_ASSERT(fq->state == qman_fq_state_parked || + fq->state == qman_fq_state_sched); + DPAA_ASSERT(fq_isset(fq, QMAN_FQ_STATE_CHANGING)); + fq_clear(fq, QMAN_FQ_STATE_CHANGING); + if (msg->fq.fqs & QM_MR_FQS_NOTEMPTY) + fq_set(fq, QMAN_FQ_STATE_NE); + if (msg->fq.fqs & QM_MR_FQS_ORLPRESENT) + fq_set(fq, QMAN_FQ_STATE_ORL); + fq->state = qman_fq_state_retired; + break; + case QM_MR_VERB_FQPN: + DPAA_ASSERT(fq->state == qman_fq_state_sched); + DPAA_ASSERT(fq_isclear(fq, QMAN_FQ_STATE_CHANGING)); + fq->state = qman_fq_state_parked; + } +} + +static void qm_congestion_task(struct work_struct *work) +{ + struct qman_portal *p = container_of(work, struct qman_portal, + congestion_work); + struct qman_cgrs rr, c; + union qm_mc_result *mcr; + struct qman_cgr *cgr; + + spin_lock(&p->cgr_lock); + qm_mc_start(&p->p); + qm_mc_commit(&p->p, QM_MCC_VERB_QUERYCONGESTION); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + spin_unlock(&p->cgr_lock); + dev_crit(p->config->dev, "QUERYCONGESTION timeout\n"); + qman_p_irqsource_add(p, QM_PIRQ_CSCI); + return; + } + /* mask out the ones I'm not interested in */ + qman_cgrs_and(&rr, (struct qman_cgrs *)&mcr->querycongestion.state, + &p->cgrs[0]); + /* check previous snapshot for delta, enter/exit congestion */ + qman_cgrs_xor(&c, &rr, &p->cgrs[1]); + /* update snapshot */ + qman_cgrs_cp(&p->cgrs[1], &rr); + /* Invoke callback */ + list_for_each_entry(cgr, &p->cgr_cbs, node) + if (cgr->cb && qman_cgrs_get(&c, cgr->cgrid)) + cgr->cb(p, cgr, qman_cgrs_get(&rr, cgr->cgrid)); + spin_unlock(&p->cgr_lock); + qman_p_irqsource_add(p, QM_PIRQ_CSCI); +} + +static void qm_mr_process_task(struct work_struct *work) +{ + struct qman_portal *p = container_of(work, struct qman_portal, + mr_work); + const union qm_mr_entry *msg; + struct qman_fq *fq; + u8 verb, num = 0; + + preempt_disable(); + + while (1) { + qm_mr_pvb_update(&p->p); + msg = qm_mr_current(&p->p); + if (!msg) + break; + + verb = msg->verb & QM_MR_VERB_TYPE_MASK; + /* The message is a software ERN iff the 0x20 bit is clear */ + if (verb & 0x20) { + switch (verb) { + case QM_MR_VERB_FQRNI: + /* nada, we drop FQRNIs on the floor */ + break; + case QM_MR_VERB_FQRN: + case QM_MR_VERB_FQRL: + /* Lookup in the retirement table */ + fq = fqid_to_fq(qm_fqid_get(&msg->fq)); + if (WARN_ON(!fq)) + break; + fq_state_change(p, fq, msg, verb); + if (fq->cb.fqs) + fq->cb.fqs(p, fq, msg); + break; + case QM_MR_VERB_FQPN: + /* Parked */ + fq = tag_to_fq(be32_to_cpu(msg->fq.context_b)); + fq_state_change(p, fq, msg, verb); + if (fq->cb.fqs) + fq->cb.fqs(p, fq, msg); + break; + case QM_MR_VERB_DC_ERN: + /* DCP ERN */ + pr_crit_once("Leaking DCP ERNs!\n"); + break; + default: + pr_crit("Invalid MR verb 0x%02x\n", verb); + } + } else { + /* Its a software ERN */ + fq = tag_to_fq(be32_to_cpu(msg->ern.tag)); + fq->cb.ern(p, fq, msg); + } + num++; + qm_mr_next(&p->p); + } + + qm_mr_cci_consume(&p->p, num); + qman_p_irqsource_add(p, QM_PIRQ_MRI); + preempt_enable(); +} + +static u32 __poll_portal_slow(struct qman_portal *p, u32 is) +{ + if (is & QM_PIRQ_CSCI) { + qman_p_irqsource_remove(p, QM_PIRQ_CSCI); + queue_work_on(smp_processor_id(), qm_portal_wq, + &p->congestion_work); + } + + if (is & QM_PIRQ_EQRI) { + qm_eqcr_cce_update(&p->p); + qm_eqcr_set_ithresh(&p->p, 0); + wake_up(&affine_queue); + } + + if (is & QM_PIRQ_MRI) { + qman_p_irqsource_remove(p, QM_PIRQ_MRI); + queue_work_on(smp_processor_id(), qm_portal_wq, + &p->mr_work); + } + + return is; +} + +/* + * remove some slowish-path stuff from the "fast path" and make sure it isn't + * inlined. + */ +static noinline void clear_vdqcr(struct qman_portal *p, struct qman_fq *fq) +{ + p->vdqcr_owned = NULL; + fq_clear(fq, QMAN_FQ_STATE_VDQCR); + wake_up(&affine_queue); +} + +/* + * The only states that would conflict with other things if they ran at the + * same time on the same cpu are: + * + * (i) setting/clearing vdqcr_owned, and + * (ii) clearing the NE (Not Empty) flag. + * + * Both are safe. Because; + * + * (i) this clearing can only occur after qman_volatile_dequeue() has set the + * vdqcr_owned field (which it does before setting VDQCR), and + * qman_volatile_dequeue() blocks interrupts and preemption while this is + * done so that we can't interfere. + * (ii) the NE flag is only cleared after qman_retire_fq() has set it, and as + * with (i) that API prevents us from interfering until it's safe. + * + * The good thing is that qman_volatile_dequeue() and qman_retire_fq() run far + * less frequently (ie. per-FQ) than __poll_portal_fast() does, so the nett + * advantage comes from this function not having to "lock" anything at all. + * + * Note also that the callbacks are invoked at points which are safe against the + * above potential conflicts, but that this function itself is not re-entrant + * (this is because the function tracks one end of each FIFO in the portal and + * we do *not* want to lock that). So the consequence is that it is safe for + * user callbacks to call into any QMan API. + */ +static inline unsigned int __poll_portal_fast(struct qman_portal *p, + unsigned int poll_limit, bool sched_napi) +{ + const struct qm_dqrr_entry *dq; + struct qman_fq *fq; + enum qman_cb_dqrr_result res; + unsigned int limit = 0; + + do { + qm_dqrr_pvb_update(&p->p); + dq = qm_dqrr_current(&p->p); + if (!dq) + break; + + if (dq->stat & QM_DQRR_STAT_UNSCHEDULED) { + /* + * VDQCR: don't trust context_b as the FQ may have + * been configured for h/w consumption and we're + * draining it post-retirement. + */ + fq = p->vdqcr_owned; + /* + * We only set QMAN_FQ_STATE_NE when retiring, so we + * only need to check for clearing it when doing + * volatile dequeues. It's one less thing to check + * in the critical path (SDQCR). + */ + if (dq->stat & QM_DQRR_STAT_FQ_EMPTY) + fq_clear(fq, QMAN_FQ_STATE_NE); + /* + * This is duplicated from the SDQCR code, but we + * have stuff to do before *and* after this callback, + * and we don't want multiple if()s in the critical + * path (SDQCR). + */ + res = fq->cb.dqrr(p, fq, dq, sched_napi); + if (res == qman_cb_dqrr_stop) + break; + /* Check for VDQCR completion */ + if (dq->stat & QM_DQRR_STAT_DQCR_EXPIRED) + clear_vdqcr(p, fq); + } else { + /* SDQCR: context_b points to the FQ */ + fq = tag_to_fq(be32_to_cpu(dq->context_b)); + /* Now let the callback do its stuff */ + res = fq->cb.dqrr(p, fq, dq, sched_napi); + /* + * The callback can request that we exit without + * consuming this entry nor advancing; + */ + if (res == qman_cb_dqrr_stop) + break; + } + /* Interpret 'dq' from a driver perspective. */ + /* + * Parking isn't possible unless HELDACTIVE was set. NB, + * FORCEELIGIBLE implies HELDACTIVE, so we only need to + * check for HELDACTIVE to cover both. + */ + DPAA_ASSERT((dq->stat & QM_DQRR_STAT_FQ_HELDACTIVE) || + (res != qman_cb_dqrr_park)); + /* just means "skip it, I'll consume it myself later on" */ + if (res != qman_cb_dqrr_defer) + qm_dqrr_cdc_consume_1ptr(&p->p, dq, + res == qman_cb_dqrr_park); + /* Move forward */ + qm_dqrr_next(&p->p); + /* + * Entry processed and consumed, increment our counter. The + * callback can request that we exit after consuming the + * entry, and we also exit if we reach our processing limit, + * so loop back only if neither of these conditions is met. + */ + } while (++limit < poll_limit && res != qman_cb_dqrr_consume_stop); + + return limit; +} + +void qman_p_irqsource_add(struct qman_portal *p, u32 bits) +{ + unsigned long irqflags; + + local_irq_save(irqflags); + p->irq_sources |= bits & QM_PIRQ_VISIBLE; + qm_out(&p->p, QM_REG_IER, p->irq_sources); + local_irq_restore(irqflags); +} +EXPORT_SYMBOL(qman_p_irqsource_add); + +void qman_p_irqsource_remove(struct qman_portal *p, u32 bits) +{ + unsigned long irqflags; + u32 ier; + + /* + * Our interrupt handler only processes+clears status register bits that + * are in p->irq_sources. As we're trimming that mask, if one of them + * were to assert in the status register just before we remove it from + * the enable register, there would be an interrupt-storm when we + * release the IRQ lock. So we wait for the enable register update to + * take effect in h/w (by reading it back) and then clear all other bits + * in the status register. Ie. we clear them from ISR once it's certain + * IER won't allow them to reassert. + */ + local_irq_save(irqflags); + bits &= QM_PIRQ_VISIBLE; + p->irq_sources &= ~bits; + qm_out(&p->p, QM_REG_IER, p->irq_sources); + ier = qm_in(&p->p, QM_REG_IER); + /* + * Using "~ier" (rather than "bits" or "~p->irq_sources") creates a + * data-dependency, ie. to protect against re-ordering. + */ + qm_out(&p->p, QM_REG_ISR, ~ier); + local_irq_restore(irqflags); +} +EXPORT_SYMBOL(qman_p_irqsource_remove); + +const cpumask_t *qman_affine_cpus(void) +{ + return &affine_mask; +} +EXPORT_SYMBOL(qman_affine_cpus); + +u16 qman_affine_channel(int cpu) +{ + if (cpu < 0) { + struct qman_portal *portal = get_affine_portal(); + + cpu = portal->config->cpu; + put_affine_portal(); + } + WARN_ON(!cpumask_test_cpu(cpu, &affine_mask)); + return affine_channels[cpu]; +} +EXPORT_SYMBOL(qman_affine_channel); + +struct qman_portal *qman_get_affine_portal(int cpu) +{ + return affine_portals[cpu]; +} +EXPORT_SYMBOL(qman_get_affine_portal); + +int qman_start_using_portal(struct qman_portal *p, struct device *dev) +{ + return (!device_link_add(dev, p->config->dev, + DL_FLAG_AUTOREMOVE_CONSUMER)) ? -EINVAL : 0; +} +EXPORT_SYMBOL(qman_start_using_portal); + +int qman_p_poll_dqrr(struct qman_portal *p, unsigned int limit) +{ + return __poll_portal_fast(p, limit, false); +} +EXPORT_SYMBOL(qman_p_poll_dqrr); + +void qman_p_static_dequeue_add(struct qman_portal *p, u32 pools) +{ + unsigned long irqflags; + + local_irq_save(irqflags); + pools &= p->config->pools; + p->sdqcr |= pools; + qm_dqrr_sdqcr_set(&p->p, p->sdqcr); + local_irq_restore(irqflags); +} +EXPORT_SYMBOL(qman_p_static_dequeue_add); + +/* Frame queue API */ + +static const char *mcr_result_str(u8 result) +{ + switch (result) { + case QM_MCR_RESULT_NULL: + return "QM_MCR_RESULT_NULL"; + case QM_MCR_RESULT_OK: + return "QM_MCR_RESULT_OK"; + case QM_MCR_RESULT_ERR_FQID: + return "QM_MCR_RESULT_ERR_FQID"; + case QM_MCR_RESULT_ERR_FQSTATE: + return "QM_MCR_RESULT_ERR_FQSTATE"; + case QM_MCR_RESULT_ERR_NOTEMPTY: + return "QM_MCR_RESULT_ERR_NOTEMPTY"; + case QM_MCR_RESULT_PENDING: + return "QM_MCR_RESULT_PENDING"; + case QM_MCR_RESULT_ERR_BADCOMMAND: + return "QM_MCR_RESULT_ERR_BADCOMMAND"; + } + return "<unknown MCR result>"; +} + +int qman_create_fq(u32 fqid, u32 flags, struct qman_fq *fq) +{ + if (flags & QMAN_FQ_FLAG_DYNAMIC_FQID) { + int ret = qman_alloc_fqid(&fqid); + + if (ret) + return ret; + } + fq->fqid = fqid; + fq->flags = flags; + fq->state = qman_fq_state_oos; + fq->cgr_groupid = 0; + + /* A context_b of 0 is allegedly special, so don't use that fqid */ + if (fqid == 0 || fqid >= num_fqids) { + WARN(1, "bad fqid %d\n", fqid); + return -EINVAL; + } + + fq->idx = fqid * 2; + if (flags & QMAN_FQ_FLAG_NO_MODIFY) + fq->idx++; + + WARN_ON(fq_table[fq->idx]); + fq_table[fq->idx] = fq; + + return 0; +} +EXPORT_SYMBOL(qman_create_fq); + +void qman_destroy_fq(struct qman_fq *fq) +{ + /* + * We don't need to lock the FQ as it is a pre-condition that the FQ be + * quiesced. Instead, run some checks. + */ + switch (fq->state) { + case qman_fq_state_parked: + case qman_fq_state_oos: + if (fq_isset(fq, QMAN_FQ_FLAG_DYNAMIC_FQID)) + qman_release_fqid(fq->fqid); + + DPAA_ASSERT(fq_table[fq->idx]); + fq_table[fq->idx] = NULL; + return; + default: + break; + } + DPAA_ASSERT(NULL == "qman_free_fq() on unquiesced FQ!"); +} +EXPORT_SYMBOL(qman_destroy_fq); + +u32 qman_fq_fqid(struct qman_fq *fq) +{ + return fq->fqid; +} +EXPORT_SYMBOL(qman_fq_fqid); + +int qman_init_fq(struct qman_fq *fq, u32 flags, struct qm_mcc_initfq *opts) +{ + union qm_mc_command *mcc; + union qm_mc_result *mcr; + struct qman_portal *p; + u8 res, myverb; + int ret = 0; + + myverb = (flags & QMAN_INITFQ_FLAG_SCHED) + ? QM_MCC_VERB_INITFQ_SCHED : QM_MCC_VERB_INITFQ_PARKED; + + if (fq->state != qman_fq_state_oos && + fq->state != qman_fq_state_parked) + return -EINVAL; +#ifdef CONFIG_FSL_DPAA_CHECKING + if (fq_isset(fq, QMAN_FQ_FLAG_NO_MODIFY)) + return -EINVAL; +#endif + if (opts && (be16_to_cpu(opts->we_mask) & QM_INITFQ_WE_OAC)) { + /* And can't be set at the same time as TDTHRESH */ + if (be16_to_cpu(opts->we_mask) & QM_INITFQ_WE_TDTHRESH) + return -EINVAL; + } + /* Issue an INITFQ_[PARKED|SCHED] management command */ + p = get_affine_portal(); + if (fq_isset(fq, QMAN_FQ_STATE_CHANGING) || + (fq->state != qman_fq_state_oos && + fq->state != qman_fq_state_parked)) { + ret = -EBUSY; + goto out; + } + mcc = qm_mc_start(&p->p); + if (opts) + mcc->initfq = *opts; + qm_fqid_set(&mcc->fq, fq->fqid); + mcc->initfq.count = 0; + /* + * If the FQ does *not* have the TO_DCPORTAL flag, context_b is set as a + * demux pointer. Otherwise, the caller-provided value is allowed to + * stand, don't overwrite it. + */ + if (fq_isclear(fq, QMAN_FQ_FLAG_TO_DCPORTAL)) { + dma_addr_t phys_fq; + + mcc->initfq.we_mask |= cpu_to_be16(QM_INITFQ_WE_CONTEXTB); + mcc->initfq.fqd.context_b = cpu_to_be32(fq_to_tag(fq)); + /* + * and the physical address - NB, if the user wasn't trying to + * set CONTEXTA, clear the stashing settings. + */ + if (!(be16_to_cpu(mcc->initfq.we_mask) & + QM_INITFQ_WE_CONTEXTA)) { + mcc->initfq.we_mask |= + cpu_to_be16(QM_INITFQ_WE_CONTEXTA); + memset(&mcc->initfq.fqd.context_a, 0, + sizeof(mcc->initfq.fqd.context_a)); + } else { + struct qman_portal *p = qman_dma_portal; + + phys_fq = dma_map_single(p->config->dev, fq, + sizeof(*fq), DMA_TO_DEVICE); + if (dma_mapping_error(p->config->dev, phys_fq)) { + dev_err(p->config->dev, "dma_mapping failed\n"); + ret = -EIO; + goto out; + } + + qm_fqd_stashing_set64(&mcc->initfq.fqd, phys_fq); + } + } + if (flags & QMAN_INITFQ_FLAG_LOCAL) { + int wq = 0; + + if (!(be16_to_cpu(mcc->initfq.we_mask) & + QM_INITFQ_WE_DESTWQ)) { + mcc->initfq.we_mask |= + cpu_to_be16(QM_INITFQ_WE_DESTWQ); + wq = 4; + } + qm_fqd_set_destwq(&mcc->initfq.fqd, p->config->channel, wq); + } + qm_mc_commit(&p->p, myverb); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + dev_err(p->config->dev, "MCR timeout\n"); + ret = -ETIMEDOUT; + goto out; + } + + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == myverb); + res = mcr->result; + if (res != QM_MCR_RESULT_OK) { + ret = -EIO; + goto out; + } + if (opts) { + if (be16_to_cpu(opts->we_mask) & QM_INITFQ_WE_FQCTRL) { + if (be16_to_cpu(opts->fqd.fq_ctrl) & QM_FQCTRL_CGE) + fq_set(fq, QMAN_FQ_STATE_CGR_EN); + else + fq_clear(fq, QMAN_FQ_STATE_CGR_EN); + } + if (be16_to_cpu(opts->we_mask) & QM_INITFQ_WE_CGID) + fq->cgr_groupid = opts->fqd.cgid; + } + fq->state = (flags & QMAN_INITFQ_FLAG_SCHED) ? + qman_fq_state_sched : qman_fq_state_parked; + +out: + put_affine_portal(); + return ret; +} +EXPORT_SYMBOL(qman_init_fq); + +int qman_schedule_fq(struct qman_fq *fq) +{ + union qm_mc_command *mcc; + union qm_mc_result *mcr; + struct qman_portal *p; + int ret = 0; + + if (fq->state != qman_fq_state_parked) + return -EINVAL; +#ifdef CONFIG_FSL_DPAA_CHECKING + if (fq_isset(fq, QMAN_FQ_FLAG_NO_MODIFY)) + return -EINVAL; +#endif + /* Issue a ALTERFQ_SCHED management command */ + p = get_affine_portal(); + if (fq_isset(fq, QMAN_FQ_STATE_CHANGING) || + fq->state != qman_fq_state_parked) { + ret = -EBUSY; + goto out; + } + mcc = qm_mc_start(&p->p); + qm_fqid_set(&mcc->fq, fq->fqid); + qm_mc_commit(&p->p, QM_MCC_VERB_ALTER_SCHED); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + dev_err(p->config->dev, "ALTER_SCHED timeout\n"); + ret = -ETIMEDOUT; + goto out; + } + + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == QM_MCR_VERB_ALTER_SCHED); + if (mcr->result != QM_MCR_RESULT_OK) { + ret = -EIO; + goto out; + } + fq->state = qman_fq_state_sched; +out: + put_affine_portal(); + return ret; +} +EXPORT_SYMBOL(qman_schedule_fq); + +int qman_retire_fq(struct qman_fq *fq, u32 *flags) +{ + union qm_mc_command *mcc; + union qm_mc_result *mcr; + struct qman_portal *p; + int ret; + u8 res; + + if (fq->state != qman_fq_state_parked && + fq->state != qman_fq_state_sched) + return -EINVAL; +#ifdef CONFIG_FSL_DPAA_CHECKING + if (fq_isset(fq, QMAN_FQ_FLAG_NO_MODIFY)) + return -EINVAL; +#endif + p = get_affine_portal(); + if (fq_isset(fq, QMAN_FQ_STATE_CHANGING) || + fq->state == qman_fq_state_retired || + fq->state == qman_fq_state_oos) { + ret = -EBUSY; + goto out; + } + mcc = qm_mc_start(&p->p); + qm_fqid_set(&mcc->fq, fq->fqid); + qm_mc_commit(&p->p, QM_MCC_VERB_ALTER_RETIRE); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + dev_crit(p->config->dev, "ALTER_RETIRE timeout\n"); + ret = -ETIMEDOUT; + goto out; + } + + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == QM_MCR_VERB_ALTER_RETIRE); + res = mcr->result; + /* + * "Elegant" would be to treat OK/PENDING the same way; set CHANGING, + * and defer the flags until FQRNI or FQRN (respectively) show up. But + * "Friendly" is to process OK immediately, and not set CHANGING. We do + * friendly, otherwise the caller doesn't necessarily have a fully + * "retired" FQ on return even if the retirement was immediate. However + * this does mean some code duplication between here and + * fq_state_change(). + */ + if (res == QM_MCR_RESULT_OK) { + ret = 0; + /* Process 'fq' right away, we'll ignore FQRNI */ + if (mcr->alterfq.fqs & QM_MCR_FQS_NOTEMPTY) + fq_set(fq, QMAN_FQ_STATE_NE); + if (mcr->alterfq.fqs & QM_MCR_FQS_ORLPRESENT) + fq_set(fq, QMAN_FQ_STATE_ORL); + if (flags) + *flags = fq->flags; + fq->state = qman_fq_state_retired; + if (fq->cb.fqs) { + /* + * Another issue with supporting "immediate" retirement + * is that we're forced to drop FQRNIs, because by the + * time they're seen it may already be "too late" (the + * fq may have been OOS'd and free()'d already). But if + * the upper layer wants a callback whether it's + * immediate or not, we have to fake a "MR" entry to + * look like an FQRNI... + */ + union qm_mr_entry msg; + + msg.verb = QM_MR_VERB_FQRNI; + msg.fq.fqs = mcr->alterfq.fqs; + qm_fqid_set(&msg.fq, fq->fqid); + msg.fq.context_b = cpu_to_be32(fq_to_tag(fq)); + fq->cb.fqs(p, fq, &msg); + } + } else if (res == QM_MCR_RESULT_PENDING) { + ret = 1; + fq_set(fq, QMAN_FQ_STATE_CHANGING); + } else { + ret = -EIO; + } +out: + put_affine_portal(); + return ret; +} +EXPORT_SYMBOL(qman_retire_fq); + +int qman_oos_fq(struct qman_fq *fq) +{ + union qm_mc_command *mcc; + union qm_mc_result *mcr; + struct qman_portal *p; + int ret = 0; + + if (fq->state != qman_fq_state_retired) + return -EINVAL; +#ifdef CONFIG_FSL_DPAA_CHECKING + if (fq_isset(fq, QMAN_FQ_FLAG_NO_MODIFY)) + return -EINVAL; +#endif + p = get_affine_portal(); + if (fq_isset(fq, QMAN_FQ_STATE_BLOCKOOS) || + fq->state != qman_fq_state_retired) { + ret = -EBUSY; + goto out; + } + mcc = qm_mc_start(&p->p); + qm_fqid_set(&mcc->fq, fq->fqid); + qm_mc_commit(&p->p, QM_MCC_VERB_ALTER_OOS); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + ret = -ETIMEDOUT; + goto out; + } + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == QM_MCR_VERB_ALTER_OOS); + if (mcr->result != QM_MCR_RESULT_OK) { + ret = -EIO; + goto out; + } + fq->state = qman_fq_state_oos; +out: + put_affine_portal(); + return ret; +} +EXPORT_SYMBOL(qman_oos_fq); + +int qman_query_fq(struct qman_fq *fq, struct qm_fqd *fqd) +{ + union qm_mc_command *mcc; + union qm_mc_result *mcr; + struct qman_portal *p = get_affine_portal(); + int ret = 0; + + mcc = qm_mc_start(&p->p); + qm_fqid_set(&mcc->fq, fq->fqid); + qm_mc_commit(&p->p, QM_MCC_VERB_QUERYFQ); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + ret = -ETIMEDOUT; + goto out; + } + + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == QM_MCR_VERB_QUERYFQ); + if (mcr->result == QM_MCR_RESULT_OK) + *fqd = mcr->queryfq.fqd; + else + ret = -EIO; +out: + put_affine_portal(); + return ret; +} + +int qman_query_fq_np(struct qman_fq *fq, struct qm_mcr_queryfq_np *np) +{ + union qm_mc_command *mcc; + union qm_mc_result *mcr; + struct qman_portal *p = get_affine_portal(); + int ret = 0; + + mcc = qm_mc_start(&p->p); + qm_fqid_set(&mcc->fq, fq->fqid); + qm_mc_commit(&p->p, QM_MCC_VERB_QUERYFQ_NP); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + ret = -ETIMEDOUT; + goto out; + } + + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == QM_MCR_VERB_QUERYFQ_NP); + if (mcr->result == QM_MCR_RESULT_OK) + *np = mcr->queryfq_np; + else if (mcr->result == QM_MCR_RESULT_ERR_FQID) + ret = -ERANGE; + else + ret = -EIO; +out: + put_affine_portal(); + return ret; +} +EXPORT_SYMBOL(qman_query_fq_np); + +static int qman_query_cgr(struct qman_cgr *cgr, + struct qm_mcr_querycgr *cgrd) +{ + union qm_mc_command *mcc; + union qm_mc_result *mcr; + struct qman_portal *p = get_affine_portal(); + int ret = 0; + + mcc = qm_mc_start(&p->p); + mcc->cgr.cgid = cgr->cgrid; + qm_mc_commit(&p->p, QM_MCC_VERB_QUERYCGR); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + ret = -ETIMEDOUT; + goto out; + } + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == QM_MCC_VERB_QUERYCGR); + if (mcr->result == QM_MCR_RESULT_OK) + *cgrd = mcr->querycgr; + else { + dev_err(p->config->dev, "QUERY_CGR failed: %s\n", + mcr_result_str(mcr->result)); + ret = -EIO; + } +out: + put_affine_portal(); + return ret; +} + +int qman_query_cgr_congested(struct qman_cgr *cgr, bool *result) +{ + struct qm_mcr_querycgr query_cgr; + int err; + + err = qman_query_cgr(cgr, &query_cgr); + if (err) + return err; + + *result = !!query_cgr.cgr.cs; + return 0; +} +EXPORT_SYMBOL(qman_query_cgr_congested); + +/* internal function used as a wait_event() expression */ +static int set_p_vdqcr(struct qman_portal *p, struct qman_fq *fq, u32 vdqcr) +{ + unsigned long irqflags; + int ret = -EBUSY; + + local_irq_save(irqflags); + if (p->vdqcr_owned) + goto out; + if (fq_isset(fq, QMAN_FQ_STATE_VDQCR)) + goto out; + + fq_set(fq, QMAN_FQ_STATE_VDQCR); + p->vdqcr_owned = fq; + qm_dqrr_vdqcr_set(&p->p, vdqcr); + ret = 0; +out: + local_irq_restore(irqflags); + return ret; +} + +static int set_vdqcr(struct qman_portal **p, struct qman_fq *fq, u32 vdqcr) +{ + int ret; + + *p = get_affine_portal(); + ret = set_p_vdqcr(*p, fq, vdqcr); + put_affine_portal(); + return ret; +} + +static int wait_vdqcr_start(struct qman_portal **p, struct qman_fq *fq, + u32 vdqcr, u32 flags) +{ + int ret = 0; + + if (flags & QMAN_VOLATILE_FLAG_WAIT_INT) + ret = wait_event_interruptible(affine_queue, + !set_vdqcr(p, fq, vdqcr)); + else + wait_event(affine_queue, !set_vdqcr(p, fq, vdqcr)); + return ret; +} + +int qman_volatile_dequeue(struct qman_fq *fq, u32 flags, u32 vdqcr) +{ + struct qman_portal *p; + int ret; + + if (fq->state != qman_fq_state_parked && + fq->state != qman_fq_state_retired) + return -EINVAL; + if (vdqcr & QM_VDQCR_FQID_MASK) + return -EINVAL; + if (fq_isset(fq, QMAN_FQ_STATE_VDQCR)) + return -EBUSY; + vdqcr = (vdqcr & ~QM_VDQCR_FQID_MASK) | fq->fqid; + if (flags & QMAN_VOLATILE_FLAG_WAIT) + ret = wait_vdqcr_start(&p, fq, vdqcr, flags); + else + ret = set_vdqcr(&p, fq, vdqcr); + if (ret) + return ret; + /* VDQCR is set */ + if (flags & QMAN_VOLATILE_FLAG_FINISH) { + if (flags & QMAN_VOLATILE_FLAG_WAIT_INT) + /* + * NB: don't propagate any error - the caller wouldn't + * know whether the VDQCR was issued or not. A signal + * could arrive after returning anyway, so the caller + * can check signal_pending() if that's an issue. + */ + wait_event_interruptible(affine_queue, + !fq_isset(fq, QMAN_FQ_STATE_VDQCR)); + else + wait_event(affine_queue, + !fq_isset(fq, QMAN_FQ_STATE_VDQCR)); + } + return 0; +} +EXPORT_SYMBOL(qman_volatile_dequeue); + +static void update_eqcr_ci(struct qman_portal *p, u8 avail) +{ + if (avail) + qm_eqcr_cce_prefetch(&p->p); + else + qm_eqcr_cce_update(&p->p); +} + +int qman_enqueue(struct qman_fq *fq, const struct qm_fd *fd) +{ + struct qman_portal *p; + struct qm_eqcr_entry *eq; + unsigned long irqflags; + u8 avail; + + p = get_affine_portal(); + local_irq_save(irqflags); + + if (p->use_eqcr_ci_stashing) { + /* + * The stashing case is easy, only update if we need to in + * order to try and liberate ring entries. + */ + eq = qm_eqcr_start_stash(&p->p); + } else { + /* + * The non-stashing case is harder, need to prefetch ahead of + * time. + */ + avail = qm_eqcr_get_avail(&p->p); + if (avail < 2) + update_eqcr_ci(p, avail); + eq = qm_eqcr_start_no_stash(&p->p); + } + + if (unlikely(!eq)) + goto out; + + qm_fqid_set(eq, fq->fqid); + eq->tag = cpu_to_be32(fq_to_tag(fq)); + eq->fd = *fd; + + qm_eqcr_pvb_commit(&p->p, QM_EQCR_VERB_CMD_ENQUEUE); +out: + local_irq_restore(irqflags); + put_affine_portal(); + return 0; +} +EXPORT_SYMBOL(qman_enqueue); + +static int qm_modify_cgr(struct qman_cgr *cgr, u32 flags, + struct qm_mcc_initcgr *opts) +{ + union qm_mc_command *mcc; + union qm_mc_result *mcr; + struct qman_portal *p = get_affine_portal(); + u8 verb = QM_MCC_VERB_MODIFYCGR; + int ret = 0; + + mcc = qm_mc_start(&p->p); + if (opts) + mcc->initcgr = *opts; + mcc->initcgr.cgid = cgr->cgrid; + if (flags & QMAN_CGR_FLAG_USE_INIT) + verb = QM_MCC_VERB_INITCGR; + qm_mc_commit(&p->p, verb); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + ret = -ETIMEDOUT; + goto out; + } + + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == verb); + if (mcr->result != QM_MCR_RESULT_OK) + ret = -EIO; + +out: + put_affine_portal(); + return ret; +} + +#define PORTAL_IDX(n) (n->config->channel - QM_CHANNEL_SWPORTAL0) + +/* congestion state change notification target update control */ +static void qm_cgr_cscn_targ_set(struct __qm_mc_cgr *cgr, int pi, u32 val) +{ + if (qman_ip_rev >= QMAN_REV30) + cgr->cscn_targ_upd_ctrl = cpu_to_be16(pi | + QM_CGR_TARG_UDP_CTRL_WRITE_BIT); + else + cgr->cscn_targ = cpu_to_be32(val | QM_CGR_TARG_PORTAL(pi)); +} + +static void qm_cgr_cscn_targ_clear(struct __qm_mc_cgr *cgr, int pi, u32 val) +{ + if (qman_ip_rev >= QMAN_REV30) + cgr->cscn_targ_upd_ctrl = cpu_to_be16(pi); + else + cgr->cscn_targ = cpu_to_be32(val & ~QM_CGR_TARG_PORTAL(pi)); +} + +static u8 qman_cgr_cpus[CGR_NUM]; + +void qman_init_cgr_all(void) +{ + struct qman_cgr cgr; + int err_cnt = 0; + + for (cgr.cgrid = 0; cgr.cgrid < CGR_NUM; cgr.cgrid++) { + if (qm_modify_cgr(&cgr, QMAN_CGR_FLAG_USE_INIT, NULL)) + err_cnt++; + } + + if (err_cnt) + pr_err("Warning: %d error%s while initialising CGR h/w\n", + err_cnt, (err_cnt > 1) ? "s" : ""); +} + +int qman_create_cgr(struct qman_cgr *cgr, u32 flags, + struct qm_mcc_initcgr *opts) +{ + struct qm_mcr_querycgr cgr_state; + int ret; + struct qman_portal *p; + + /* + * We have to check that the provided CGRID is within the limits of the + * data-structures, for obvious reasons. However we'll let h/w take + * care of determining whether it's within the limits of what exists on + * the SoC. + */ + if (cgr->cgrid >= CGR_NUM) + return -EINVAL; + + preempt_disable(); + p = get_affine_portal(); + qman_cgr_cpus[cgr->cgrid] = smp_processor_id(); + preempt_enable(); + + cgr->chan = p->config->channel; + spin_lock(&p->cgr_lock); + + if (opts) { + struct qm_mcc_initcgr local_opts = *opts; + + ret = qman_query_cgr(cgr, &cgr_state); + if (ret) + goto out; + + qm_cgr_cscn_targ_set(&local_opts.cgr, PORTAL_IDX(p), + be32_to_cpu(cgr_state.cgr.cscn_targ)); + local_opts.we_mask |= cpu_to_be16(QM_CGR_WE_CSCN_TARG); + + /* send init if flags indicate so */ + if (flags & QMAN_CGR_FLAG_USE_INIT) + ret = qm_modify_cgr(cgr, QMAN_CGR_FLAG_USE_INIT, + &local_opts); + else + ret = qm_modify_cgr(cgr, 0, &local_opts); + if (ret) + goto out; + } + + list_add(&cgr->node, &p->cgr_cbs); + + /* Determine if newly added object requires its callback to be called */ + ret = qman_query_cgr(cgr, &cgr_state); + if (ret) { + /* we can't go back, so proceed and return success */ + dev_err(p->config->dev, "CGR HW state partially modified\n"); + ret = 0; + goto out; + } + if (cgr->cb && cgr_state.cgr.cscn_en && + qman_cgrs_get(&p->cgrs[1], cgr->cgrid)) + cgr->cb(p, cgr, 1); +out: + spin_unlock(&p->cgr_lock); + put_affine_portal(); + return ret; +} +EXPORT_SYMBOL(qman_create_cgr); + +static struct qman_portal *qman_cgr_get_affine_portal(struct qman_cgr *cgr) +{ + struct qman_portal *p = get_affine_portal(); + + if (cgr->chan != p->config->channel) { + /* attempt to delete from other portal than creator */ + dev_err(p->config->dev, "CGR not owned by current portal"); + dev_dbg(p->config->dev, " create 0x%x, delete 0x%x\n", + cgr->chan, p->config->channel); + put_affine_portal(); + return NULL; + } + + return p; +} + +int qman_delete_cgr(struct qman_cgr *cgr) +{ + unsigned long irqflags; + struct qm_mcr_querycgr cgr_state; + struct qm_mcc_initcgr local_opts; + int ret = 0; + struct qman_cgr *i; + struct qman_portal *p = qman_cgr_get_affine_portal(cgr); + + if (!p) + return -EINVAL; + + memset(&local_opts, 0, sizeof(struct qm_mcc_initcgr)); + spin_lock_irqsave(&p->cgr_lock, irqflags); + list_del(&cgr->node); + /* + * If there are no other CGR objects for this CGRID in the list, + * update CSCN_TARG accordingly + */ + list_for_each_entry(i, &p->cgr_cbs, node) + if (i->cgrid == cgr->cgrid && i->cb) + goto release_lock; + ret = qman_query_cgr(cgr, &cgr_state); + if (ret) { + /* add back to the list */ + list_add(&cgr->node, &p->cgr_cbs); + goto release_lock; + } + + local_opts.we_mask = cpu_to_be16(QM_CGR_WE_CSCN_TARG); + qm_cgr_cscn_targ_clear(&local_opts.cgr, PORTAL_IDX(p), + be32_to_cpu(cgr_state.cgr.cscn_targ)); + + ret = qm_modify_cgr(cgr, 0, &local_opts); + if (ret) + /* add back to the list */ + list_add(&cgr->node, &p->cgr_cbs); +release_lock: + spin_unlock_irqrestore(&p->cgr_lock, irqflags); + put_affine_portal(); + return ret; +} +EXPORT_SYMBOL(qman_delete_cgr); + +struct cgr_comp { + struct qman_cgr *cgr; + struct completion completion; +}; + +static void qman_delete_cgr_smp_call(void *p) +{ + qman_delete_cgr((struct qman_cgr *)p); +} + +void qman_delete_cgr_safe(struct qman_cgr *cgr) +{ + preempt_disable(); + if (qman_cgr_cpus[cgr->cgrid] != smp_processor_id()) { + smp_call_function_single(qman_cgr_cpus[cgr->cgrid], + qman_delete_cgr_smp_call, cgr, true); + preempt_enable(); + return; + } + + qman_delete_cgr(cgr); + preempt_enable(); +} +EXPORT_SYMBOL(qman_delete_cgr_safe); + +static int qman_update_cgr(struct qman_cgr *cgr, struct qm_mcc_initcgr *opts) +{ + int ret; + unsigned long irqflags; + struct qman_portal *p = qman_cgr_get_affine_portal(cgr); + + if (!p) + return -EINVAL; + + spin_lock_irqsave(&p->cgr_lock, irqflags); + ret = qm_modify_cgr(cgr, 0, opts); + spin_unlock_irqrestore(&p->cgr_lock, irqflags); + put_affine_portal(); + return ret; +} + +struct update_cgr_params { + struct qman_cgr *cgr; + struct qm_mcc_initcgr *opts; + int ret; +}; + +static void qman_update_cgr_smp_call(void *p) +{ + struct update_cgr_params *params = p; + + params->ret = qman_update_cgr(params->cgr, params->opts); +} + +int qman_update_cgr_safe(struct qman_cgr *cgr, struct qm_mcc_initcgr *opts) +{ + struct update_cgr_params params = { + .cgr = cgr, + .opts = opts, + }; + + preempt_disable(); + if (qman_cgr_cpus[cgr->cgrid] != smp_processor_id()) + smp_call_function_single(qman_cgr_cpus[cgr->cgrid], + qman_update_cgr_smp_call, ¶ms, + true); + else + params.ret = qman_update_cgr(cgr, opts); + preempt_enable(); + return params.ret; +} +EXPORT_SYMBOL(qman_update_cgr_safe); + +/* Cleanup FQs */ + +static int _qm_mr_consume_and_match_verb(struct qm_portal *p, int v) +{ + const union qm_mr_entry *msg; + int found = 0; + + qm_mr_pvb_update(p); + msg = qm_mr_current(p); + while (msg) { + if ((msg->verb & QM_MR_VERB_TYPE_MASK) == v) + found = 1; + qm_mr_next(p); + qm_mr_cci_consume_to_current(p); + qm_mr_pvb_update(p); + msg = qm_mr_current(p); + } + return found; +} + +static int _qm_dqrr_consume_and_match(struct qm_portal *p, u32 fqid, int s, + bool wait) +{ + const struct qm_dqrr_entry *dqrr; + int found = 0; + + do { + qm_dqrr_pvb_update(p); + dqrr = qm_dqrr_current(p); + if (!dqrr) + cpu_relax(); + } while (wait && !dqrr); + + while (dqrr) { + if (qm_fqid_get(dqrr) == fqid && (dqrr->stat & s)) + found = 1; + qm_dqrr_cdc_consume_1ptr(p, dqrr, 0); + qm_dqrr_pvb_update(p); + qm_dqrr_next(p); + dqrr = qm_dqrr_current(p); + } + return found; +} + +#define qm_mr_drain(p, V) \ + _qm_mr_consume_and_match_verb(p, QM_MR_VERB_##V) + +#define qm_dqrr_drain(p, f, S) \ + _qm_dqrr_consume_and_match(p, f, QM_DQRR_STAT_##S, false) + +#define qm_dqrr_drain_wait(p, f, S) \ + _qm_dqrr_consume_and_match(p, f, QM_DQRR_STAT_##S, true) + +#define qm_dqrr_drain_nomatch(p) \ + _qm_dqrr_consume_and_match(p, 0, 0, false) + +int qman_shutdown_fq(u32 fqid) +{ + struct qman_portal *p, *channel_portal; + struct device *dev; + union qm_mc_command *mcc; + union qm_mc_result *mcr; + int orl_empty, drain = 0, ret = 0; + u32 channel, res; + u8 state; + + p = get_affine_portal(); + dev = p->config->dev; + /* Determine the state of the FQID */ + mcc = qm_mc_start(&p->p); + qm_fqid_set(&mcc->fq, fqid); + qm_mc_commit(&p->p, QM_MCC_VERB_QUERYFQ_NP); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + dev_err(dev, "QUERYFQ_NP timeout\n"); + ret = -ETIMEDOUT; + goto out; + } + + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == QM_MCR_VERB_QUERYFQ_NP); + state = mcr->queryfq_np.state & QM_MCR_NP_STATE_MASK; + if (state == QM_MCR_NP_STATE_OOS) + goto out; /* Already OOS, no need to do anymore checks */ + + /* Query which channel the FQ is using */ + mcc = qm_mc_start(&p->p); + qm_fqid_set(&mcc->fq, fqid); + qm_mc_commit(&p->p, QM_MCC_VERB_QUERYFQ); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + dev_err(dev, "QUERYFQ timeout\n"); + ret = -ETIMEDOUT; + goto out; + } + + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == QM_MCR_VERB_QUERYFQ); + /* Need to store these since the MCR gets reused */ + channel = qm_fqd_get_chan(&mcr->queryfq.fqd); + qm_fqd_get_wq(&mcr->queryfq.fqd); + + if (channel < qm_channel_pool1) { + channel_portal = get_portal_for_channel(channel); + if (channel_portal == NULL) { + dev_err(dev, "Can't find portal for dedicated channel 0x%x\n", + channel); + ret = -EIO; + goto out; + } + } else + channel_portal = p; + + switch (state) { + case QM_MCR_NP_STATE_TEN_SCHED: + case QM_MCR_NP_STATE_TRU_SCHED: + case QM_MCR_NP_STATE_ACTIVE: + case QM_MCR_NP_STATE_PARKED: + orl_empty = 0; + mcc = qm_mc_start(&channel_portal->p); + qm_fqid_set(&mcc->fq, fqid); + qm_mc_commit(&channel_portal->p, QM_MCC_VERB_ALTER_RETIRE); + if (!qm_mc_result_timeout(&channel_portal->p, &mcr)) { + dev_err(dev, "ALTER_RETIRE timeout\n"); + ret = -ETIMEDOUT; + goto out; + } + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == + QM_MCR_VERB_ALTER_RETIRE); + res = mcr->result; /* Make a copy as we reuse MCR below */ + + if (res == QM_MCR_RESULT_OK) + drain_mr_fqrni(&channel_portal->p); + + if (res == QM_MCR_RESULT_PENDING) { + /* + * Need to wait for the FQRN in the message ring, which + * will only occur once the FQ has been drained. In + * order for the FQ to drain the portal needs to be set + * to dequeue from the channel the FQ is scheduled on + */ + int found_fqrn = 0; + + /* Flag that we need to drain FQ */ + drain = 1; + + if (channel >= qm_channel_pool1 && + channel < qm_channel_pool1 + 15) { + /* Pool channel, enable the bit in the portal */ + } else if (channel < qm_channel_pool1) { + /* Dedicated channel */ + } else { + dev_err(dev, "Can't recover FQ 0x%x, ch: 0x%x", + fqid, channel); + ret = -EBUSY; + goto out; + } + /* Set the sdqcr to drain this channel */ + if (channel < qm_channel_pool1) + qm_dqrr_sdqcr_set(&channel_portal->p, + QM_SDQCR_TYPE_ACTIVE | + QM_SDQCR_CHANNELS_DEDICATED); + else + qm_dqrr_sdqcr_set(&channel_portal->p, + QM_SDQCR_TYPE_ACTIVE | + QM_SDQCR_CHANNELS_POOL_CONV + (channel)); + do { + /* Keep draining DQRR while checking the MR*/ + qm_dqrr_drain_nomatch(&channel_portal->p); + /* Process message ring too */ + found_fqrn = qm_mr_drain(&channel_portal->p, + FQRN); + cpu_relax(); + } while (!found_fqrn); + /* Restore SDQCR */ + qm_dqrr_sdqcr_set(&channel_portal->p, + channel_portal->sdqcr); + + } + if (res != QM_MCR_RESULT_OK && + res != QM_MCR_RESULT_PENDING) { + dev_err(dev, "retire_fq failed: FQ 0x%x, res=0x%x\n", + fqid, res); + ret = -EIO; + goto out; + } + if (!(mcr->alterfq.fqs & QM_MCR_FQS_ORLPRESENT)) { + /* + * ORL had no entries, no need to wait until the + * ERNs come in + */ + orl_empty = 1; + } + /* + * Retirement succeeded, check to see if FQ needs + * to be drained + */ + if (drain || mcr->alterfq.fqs & QM_MCR_FQS_NOTEMPTY) { + /* FQ is Not Empty, drain using volatile DQ commands */ + do { + u32 vdqcr = fqid | QM_VDQCR_NUMFRAMES_SET(3); + + qm_dqrr_vdqcr_set(&p->p, vdqcr); + /* + * Wait for a dequeue and process the dequeues, + * making sure to empty the ring completely + */ + } while (!qm_dqrr_drain_wait(&p->p, fqid, FQ_EMPTY)); + } + + while (!orl_empty) { + /* Wait for the ORL to have been completely drained */ + orl_empty = qm_mr_drain(&p->p, FQRL); + cpu_relax(); + } + mcc = qm_mc_start(&p->p); + qm_fqid_set(&mcc->fq, fqid); + qm_mc_commit(&p->p, QM_MCC_VERB_ALTER_OOS); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + ret = -ETIMEDOUT; + goto out; + } + + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == + QM_MCR_VERB_ALTER_OOS); + if (mcr->result != QM_MCR_RESULT_OK) { + dev_err(dev, "OOS after drain fail: FQ 0x%x (0x%x)\n", + fqid, mcr->result); + ret = -EIO; + goto out; + } + break; + + case QM_MCR_NP_STATE_RETIRED: + /* Send OOS Command */ + mcc = qm_mc_start(&p->p); + qm_fqid_set(&mcc->fq, fqid); + qm_mc_commit(&p->p, QM_MCC_VERB_ALTER_OOS); + if (!qm_mc_result_timeout(&p->p, &mcr)) { + ret = -ETIMEDOUT; + goto out; + } + + DPAA_ASSERT((mcr->verb & QM_MCR_VERB_MASK) == + QM_MCR_VERB_ALTER_OOS); + if (mcr->result != QM_MCR_RESULT_OK) { + dev_err(dev, "OOS fail: FQ 0x%x (0x%x)\n", + fqid, mcr->result); + ret = -EIO; + goto out; + } + break; + + case QM_MCR_NP_STATE_OOS: + /* Done */ + break; + + default: + ret = -EIO; + } + +out: + put_affine_portal(); + return ret; +} + +const struct qm_portal_config *qman_get_qm_portal_config( + struct qman_portal *portal) +{ + return portal->config; +} +EXPORT_SYMBOL(qman_get_qm_portal_config); + +struct gen_pool *qm_fqalloc; /* FQID allocator */ +struct gen_pool *qm_qpalloc; /* pool-channel allocator */ +struct gen_pool *qm_cgralloc; /* CGR ID allocator */ + +static int qman_alloc_range(struct gen_pool *p, u32 *result, u32 cnt) +{ + unsigned long addr; + + if (!p) + return -ENODEV; + + addr = gen_pool_alloc(p, cnt); + if (!addr) + return -ENOMEM; + + *result = addr & ~DPAA_GENALLOC_OFF; + + return 0; +} + +int qman_alloc_fqid_range(u32 *result, u32 count) +{ + return qman_alloc_range(qm_fqalloc, result, count); +} +EXPORT_SYMBOL(qman_alloc_fqid_range); + +int qman_alloc_pool_range(u32 *result, u32 count) +{ + return qman_alloc_range(qm_qpalloc, result, count); +} +EXPORT_SYMBOL(qman_alloc_pool_range); + +int qman_alloc_cgrid_range(u32 *result, u32 count) +{ + return qman_alloc_range(qm_cgralloc, result, count); +} +EXPORT_SYMBOL(qman_alloc_cgrid_range); + +int qman_release_fqid(u32 fqid) +{ + int ret = qman_shutdown_fq(fqid); + + if (ret) { + pr_debug("FQID %d leaked\n", fqid); + return ret; + } + + gen_pool_free(qm_fqalloc, fqid | DPAA_GENALLOC_OFF, 1); + return 0; +} +EXPORT_SYMBOL(qman_release_fqid); + +static int qpool_cleanup(u32 qp) +{ + /* + * We query all FQDs starting from + * FQID 1 until we get an "invalid FQID" error, looking for non-OOS FQDs + * whose destination channel is the pool-channel being released. + * When a non-OOS FQD is found we attempt to clean it up + */ + struct qman_fq fq = { + .fqid = QM_FQID_RANGE_START + }; + int err; + + do { + struct qm_mcr_queryfq_np np; + + err = qman_query_fq_np(&fq, &np); + if (err == -ERANGE) + /* FQID range exceeded, found no problems */ + return 0; + else if (WARN_ON(err)) + return err; + + if ((np.state & QM_MCR_NP_STATE_MASK) != QM_MCR_NP_STATE_OOS) { + struct qm_fqd fqd; + + err = qman_query_fq(&fq, &fqd); + if (WARN_ON(err)) + return err; + if (qm_fqd_get_chan(&fqd) == qp) { + /* The channel is the FQ's target, clean it */ + err = qman_shutdown_fq(fq.fqid); + if (err) + /* + * Couldn't shut down the FQ + * so the pool must be leaked + */ + return err; + } + } + /* Move to the next FQID */ + fq.fqid++; + } while (1); +} + +int qman_release_pool(u32 qp) +{ + int ret; + + ret = qpool_cleanup(qp); + if (ret) { + pr_debug("CHID %d leaked\n", qp); + return ret; + } + + gen_pool_free(qm_qpalloc, qp | DPAA_GENALLOC_OFF, 1); + return 0; +} +EXPORT_SYMBOL(qman_release_pool); + +static int cgr_cleanup(u32 cgrid) +{ + /* + * query all FQDs starting from FQID 1 until we get an "invalid FQID" + * error, looking for non-OOS FQDs whose CGR is the CGR being released + */ + struct qman_fq fq = { + .fqid = QM_FQID_RANGE_START + }; + int err; + + do { + struct qm_mcr_queryfq_np np; + + err = qman_query_fq_np(&fq, &np); + if (err == -ERANGE) + /* FQID range exceeded, found no problems */ + return 0; + else if (WARN_ON(err)) + return err; + + if ((np.state & QM_MCR_NP_STATE_MASK) != QM_MCR_NP_STATE_OOS) { + struct qm_fqd fqd; + + err = qman_query_fq(&fq, &fqd); + if (WARN_ON(err)) + return err; + if (be16_to_cpu(fqd.fq_ctrl) & QM_FQCTRL_CGE && + fqd.cgid == cgrid) { + pr_err("CRGID 0x%x is being used by FQID 0x%x, CGR will be leaked\n", + cgrid, fq.fqid); + return -EIO; + } + } + /* Move to the next FQID */ + fq.fqid++; + } while (1); +} + +int qman_release_cgrid(u32 cgrid) +{ + int ret; + + ret = cgr_cleanup(cgrid); + if (ret) { + pr_debug("CGRID %d leaked\n", cgrid); + return ret; + } + + gen_pool_free(qm_cgralloc, cgrid | DPAA_GENALLOC_OFF, 1); + return 0; +} +EXPORT_SYMBOL(qman_release_cgrid); diff --git a/drivers/soc/fsl/qbman/qman_ccsr.c b/drivers/soc/fsl/qbman/qman_ccsr.c new file mode 100644 index 000000000..157659fd0 --- /dev/null +++ b/drivers/soc/fsl/qbman/qman_ccsr.c @@ -0,0 +1,917 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "qman_priv.h" + +u16 qman_ip_rev; +EXPORT_SYMBOL(qman_ip_rev); +u16 qm_channel_pool1 = QMAN_CHANNEL_POOL1; +EXPORT_SYMBOL(qm_channel_pool1); +u16 qm_channel_caam = QMAN_CHANNEL_CAAM; +EXPORT_SYMBOL(qm_channel_caam); + +/* Register offsets */ +#define REG_QCSP_LIO_CFG(n) (0x0000 + ((n) * 0x10)) +#define REG_QCSP_IO_CFG(n) (0x0004 + ((n) * 0x10)) +#define REG_QCSP_DD_CFG(n) (0x000c + ((n) * 0x10)) +#define REG_DD_CFG 0x0200 +#define REG_DCP_CFG(n) (0x0300 + ((n) * 0x10)) +#define REG_DCP_DD_CFG(n) (0x0304 + ((n) * 0x10)) +#define REG_DCP_DLM_AVG(n) (0x030c + ((n) * 0x10)) +#define REG_PFDR_FPC 0x0400 +#define REG_PFDR_FP_HEAD 0x0404 +#define REG_PFDR_FP_TAIL 0x0408 +#define REG_PFDR_FP_LWIT 0x0410 +#define REG_PFDR_CFG 0x0414 +#define REG_SFDR_CFG 0x0500 +#define REG_SFDR_IN_USE 0x0504 +#define REG_WQ_CS_CFG(n) (0x0600 + ((n) * 0x04)) +#define REG_WQ_DEF_ENC_WQID 0x0630 +#define REG_WQ_SC_DD_CFG(n) (0x640 + ((n) * 0x04)) +#define REG_WQ_PC_DD_CFG(n) (0x680 + ((n) * 0x04)) +#define REG_WQ_DC0_DD_CFG(n) (0x6c0 + ((n) * 0x04)) +#define REG_WQ_DC1_DD_CFG(n) (0x700 + ((n) * 0x04)) +#define REG_WQ_DCn_DD_CFG(n) (0x6c0 + ((n) * 0x40)) /* n=2,3 */ +#define REG_CM_CFG 0x0800 +#define REG_ECSR 0x0a00 +#define REG_ECIR 0x0a04 +#define REG_EADR 0x0a08 +#define REG_ECIR2 0x0a0c +#define REG_EDATA(n) (0x0a10 + ((n) * 0x04)) +#define REG_SBEC(n) (0x0a80 + ((n) * 0x04)) +#define REG_MCR 0x0b00 +#define REG_MCP(n) (0x0b04 + ((n) * 0x04)) +#define REG_MISC_CFG 0x0be0 +#define REG_HID_CFG 0x0bf0 +#define REG_IDLE_STAT 0x0bf4 +#define REG_IP_REV_1 0x0bf8 +#define REG_IP_REV_2 0x0bfc +#define REG_FQD_BARE 0x0c00 +#define REG_PFDR_BARE 0x0c20 +#define REG_offset_BAR 0x0004 /* relative to REG_[FQD|PFDR]_BARE */ +#define REG_offset_AR 0x0010 /* relative to REG_[FQD|PFDR]_BARE */ +#define REG_QCSP_BARE 0x0c80 +#define REG_QCSP_BAR 0x0c84 +#define REG_CI_SCHED_CFG 0x0d00 +#define REG_SRCIDR 0x0d04 +#define REG_LIODNR 0x0d08 +#define REG_CI_RLM_AVG 0x0d14 +#define REG_ERR_ISR 0x0e00 +#define REG_ERR_IER 0x0e04 +#define REG_REV3_QCSP_LIO_CFG(n) (0x1000 + ((n) * 0x10)) +#define REG_REV3_QCSP_IO_CFG(n) (0x1004 + ((n) * 0x10)) +#define REG_REV3_QCSP_DD_CFG(n) (0x100c + ((n) * 0x10)) + +/* Assists for QMAN_MCR */ +#define MCR_INIT_PFDR 0x01000000 +#define MCR_get_rslt(v) (u8)((v) >> 24) +#define MCR_rslt_idle(r) (!(r) || ((r) >= 0xf0)) +#define MCR_rslt_ok(r) ((r) == 0xf0) +#define MCR_rslt_eaccess(r) ((r) == 0xf8) +#define MCR_rslt_inval(r) ((r) == 0xff) + +/* + * Corenet initiator settings. Stash request queues are 4-deep to match cores + * ability to snarf. Stash priority is 3, other priorities are 2. + */ +#define QM_CI_SCHED_CFG_SRCCIV 4 +#define QM_CI_SCHED_CFG_SRQ_W 3 +#define QM_CI_SCHED_CFG_RW_W 2 +#define QM_CI_SCHED_CFG_BMAN_W 2 +/* write SRCCIV enable */ +#define QM_CI_SCHED_CFG_SRCCIV_EN BIT(31) + +/* Follows WQ_CS_CFG0-5 */ +enum qm_wq_class { + qm_wq_portal = 0, + qm_wq_pool = 1, + qm_wq_fman0 = 2, + qm_wq_fman1 = 3, + qm_wq_caam = 4, + qm_wq_pme = 5, + qm_wq_first = qm_wq_portal, + qm_wq_last = qm_wq_pme +}; + +/* Follows FQD_[BARE|BAR|AR] and PFDR_[BARE|BAR|AR] */ +enum qm_memory { + qm_memory_fqd, + qm_memory_pfdr +}; + +/* Used by all error interrupt registers except 'inhibit' */ +#define QM_EIRQ_CIDE 0x20000000 /* Corenet Initiator Data Error */ +#define QM_EIRQ_CTDE 0x10000000 /* Corenet Target Data Error */ +#define QM_EIRQ_CITT 0x08000000 /* Corenet Invalid Target Transaction */ +#define QM_EIRQ_PLWI 0x04000000 /* PFDR Low Watermark */ +#define QM_EIRQ_MBEI 0x02000000 /* Multi-bit ECC Error */ +#define QM_EIRQ_SBEI 0x01000000 /* Single-bit ECC Error */ +#define QM_EIRQ_PEBI 0x00800000 /* PFDR Enqueues Blocked Interrupt */ +#define QM_EIRQ_IFSI 0x00020000 /* Invalid FQ Flow Control State */ +#define QM_EIRQ_ICVI 0x00010000 /* Invalid Command Verb */ +#define QM_EIRQ_IDDI 0x00000800 /* Invalid Dequeue (Direct-connect) */ +#define QM_EIRQ_IDFI 0x00000400 /* Invalid Dequeue FQ */ +#define QM_EIRQ_IDSI 0x00000200 /* Invalid Dequeue Source */ +#define QM_EIRQ_IDQI 0x00000100 /* Invalid Dequeue Queue */ +#define QM_EIRQ_IECE 0x00000010 /* Invalid Enqueue Configuration */ +#define QM_EIRQ_IEOI 0x00000008 /* Invalid Enqueue Overflow */ +#define QM_EIRQ_IESI 0x00000004 /* Invalid Enqueue State */ +#define QM_EIRQ_IECI 0x00000002 /* Invalid Enqueue Channel */ +#define QM_EIRQ_IEQI 0x00000001 /* Invalid Enqueue Queue */ + +/* QMAN_ECIR valid error bit */ +#define PORTAL_ECSR_ERR (QM_EIRQ_IEQI | QM_EIRQ_IESI | QM_EIRQ_IEOI | \ + QM_EIRQ_IDQI | QM_EIRQ_IDSI | QM_EIRQ_IDFI | \ + QM_EIRQ_IDDI | QM_EIRQ_ICVI | QM_EIRQ_IFSI) +#define FQID_ECSR_ERR (QM_EIRQ_IEQI | QM_EIRQ_IECI | QM_EIRQ_IESI | \ + QM_EIRQ_IEOI | QM_EIRQ_IDQI | QM_EIRQ_IDFI | \ + QM_EIRQ_IFSI) + +struct qm_ecir { + u32 info; /* res[30-31], ptyp[29], pnum[24-28], fqid[0-23] */ +}; + +static bool qm_ecir_is_dcp(const struct qm_ecir *p) +{ + return p->info & BIT(29); +} + +static int qm_ecir_get_pnum(const struct qm_ecir *p) +{ + return (p->info >> 24) & 0x1f; +} + +static int qm_ecir_get_fqid(const struct qm_ecir *p) +{ + return p->info & (BIT(24) - 1); +} + +struct qm_ecir2 { + u32 info; /* ptyp[31], res[10-30], pnum[0-9] */ +}; + +static bool qm_ecir2_is_dcp(const struct qm_ecir2 *p) +{ + return p->info & BIT(31); +} + +static int qm_ecir2_get_pnum(const struct qm_ecir2 *p) +{ + return p->info & (BIT(10) - 1); +} + +struct qm_eadr { + u32 info; /* memid[24-27], eadr[0-11] */ + /* v3: memid[24-28], eadr[0-15] */ +}; + +static int qm_eadr_get_memid(const struct qm_eadr *p) +{ + return (p->info >> 24) & 0xf; +} + +static int qm_eadr_get_eadr(const struct qm_eadr *p) +{ + return p->info & (BIT(12) - 1); +} + +static int qm_eadr_v3_get_memid(const struct qm_eadr *p) +{ + return (p->info >> 24) & 0x1f; +} + +static int qm_eadr_v3_get_eadr(const struct qm_eadr *p) +{ + return p->info & (BIT(16) - 1); +} + +struct qman_hwerr_txt { + u32 mask; + const char *txt; +}; + + +static const struct qman_hwerr_txt qman_hwerr_txts[] = { + { QM_EIRQ_CIDE, "Corenet Initiator Data Error" }, + { QM_EIRQ_CTDE, "Corenet Target Data Error" }, + { QM_EIRQ_CITT, "Corenet Invalid Target Transaction" }, + { QM_EIRQ_PLWI, "PFDR Low Watermark" }, + { QM_EIRQ_MBEI, "Multi-bit ECC Error" }, + { QM_EIRQ_SBEI, "Single-bit ECC Error" }, + { QM_EIRQ_PEBI, "PFDR Enqueues Blocked Interrupt" }, + { QM_EIRQ_ICVI, "Invalid Command Verb" }, + { QM_EIRQ_IFSI, "Invalid Flow Control State" }, + { QM_EIRQ_IDDI, "Invalid Dequeue (Direct-connect)" }, + { QM_EIRQ_IDFI, "Invalid Dequeue FQ" }, + { QM_EIRQ_IDSI, "Invalid Dequeue Source" }, + { QM_EIRQ_IDQI, "Invalid Dequeue Queue" }, + { QM_EIRQ_IECE, "Invalid Enqueue Configuration" }, + { QM_EIRQ_IEOI, "Invalid Enqueue Overflow" }, + { QM_EIRQ_IESI, "Invalid Enqueue State" }, + { QM_EIRQ_IECI, "Invalid Enqueue Channel" }, + { QM_EIRQ_IEQI, "Invalid Enqueue Queue" }, +}; + +struct qman_error_info_mdata { + u16 addr_mask; + u16 bits; + const char *txt; +}; + +static const struct qman_error_info_mdata error_mdata[] = { + { 0x01FF, 24, "FQD cache tag memory 0" }, + { 0x01FF, 24, "FQD cache tag memory 1" }, + { 0x01FF, 24, "FQD cache tag memory 2" }, + { 0x01FF, 24, "FQD cache tag memory 3" }, + { 0x0FFF, 512, "FQD cache memory" }, + { 0x07FF, 128, "SFDR memory" }, + { 0x01FF, 72, "WQ context memory" }, + { 0x00FF, 240, "CGR memory" }, + { 0x00FF, 302, "Internal Order Restoration List memory" }, + { 0x01FF, 256, "SW portal ring memory" }, +}; + +#define QMAN_ERRS_TO_DISABLE (QM_EIRQ_PLWI | QM_EIRQ_PEBI) + +/* + * TODO: unimplemented registers + * + * Keeping a list here of QMan registers I have not yet covered; + * QCSP_DD_IHRSR, QCSP_DD_IHRFR, QCSP_DD_HASR, + * DCP_DD_IHRSR, DCP_DD_IHRFR, DCP_DD_HASR, CM_CFG, + * QMAN_EECC, QMAN_SBET, QMAN_EINJ, QMAN_SBEC0-12 + */ + +/* Pointer to the start of the QMan's CCSR space */ +static u32 __iomem *qm_ccsr_start; +/* A SDQCR mask comprising all the available/visible pool channels */ +static u32 qm_pools_sdqcr; +static int __qman_probed; +static int __qman_requires_cleanup; + +static inline u32 qm_ccsr_in(u32 offset) +{ + return ioread32be(qm_ccsr_start + offset/4); +} + +static inline void qm_ccsr_out(u32 offset, u32 val) +{ + iowrite32be(val, qm_ccsr_start + offset/4); +} + +u32 qm_get_pools_sdqcr(void) +{ + return qm_pools_sdqcr; +} + +enum qm_dc_portal { + qm_dc_portal_fman0 = 0, + qm_dc_portal_fman1 = 1 +}; + +static void qm_set_dc(enum qm_dc_portal portal, int ed, u8 sernd) +{ + DPAA_ASSERT(!ed || portal == qm_dc_portal_fman0 || + portal == qm_dc_portal_fman1); + if ((qman_ip_rev & 0xFF00) >= QMAN_REV30) + qm_ccsr_out(REG_DCP_CFG(portal), + (ed ? 0x1000 : 0) | (sernd & 0x3ff)); + else + qm_ccsr_out(REG_DCP_CFG(portal), + (ed ? 0x100 : 0) | (sernd & 0x1f)); +} + +static void qm_set_wq_scheduling(enum qm_wq_class wq_class, + u8 cs_elev, u8 csw2, u8 csw3, u8 csw4, + u8 csw5, u8 csw6, u8 csw7) +{ + qm_ccsr_out(REG_WQ_CS_CFG(wq_class), ((cs_elev & 0xff) << 24) | + ((csw2 & 0x7) << 20) | ((csw3 & 0x7) << 16) | + ((csw4 & 0x7) << 12) | ((csw5 & 0x7) << 8) | + ((csw6 & 0x7) << 4) | (csw7 & 0x7)); +} + +static void qm_set_hid(void) +{ + qm_ccsr_out(REG_HID_CFG, 0); +} + +static void qm_set_corenet_initiator(void) +{ + qm_ccsr_out(REG_CI_SCHED_CFG, QM_CI_SCHED_CFG_SRCCIV_EN | + (QM_CI_SCHED_CFG_SRCCIV << 24) | + (QM_CI_SCHED_CFG_SRQ_W << 8) | + (QM_CI_SCHED_CFG_RW_W << 4) | + QM_CI_SCHED_CFG_BMAN_W); +} + +static void qm_get_version(u16 *id, u8 *major, u8 *minor) +{ + u32 v = qm_ccsr_in(REG_IP_REV_1); + *id = (v >> 16); + *major = (v >> 8) & 0xff; + *minor = v & 0xff; +} + +#define PFDR_AR_EN BIT(31) +static int qm_set_memory(enum qm_memory memory, u64 ba, u32 size) +{ + void *ptr; + u32 offset = (memory == qm_memory_fqd) ? REG_FQD_BARE : REG_PFDR_BARE; + u32 exp = ilog2(size); + u32 bar, bare; + + /* choke if size isn't within range */ + DPAA_ASSERT((size >= 4096) && (size <= 1024*1024*1024) && + is_power_of_2(size)); + /* choke if 'ba' has lower-alignment than 'size' */ + DPAA_ASSERT(!(ba & (size - 1))); + + /* Check to see if QMan has already been initialized */ + bar = qm_ccsr_in(offset + REG_offset_BAR); + if (bar) { + /* Maker sure ba == what was programmed) */ + bare = qm_ccsr_in(offset); + if (bare != upper_32_bits(ba) || bar != lower_32_bits(ba)) { + pr_err("Attempted to reinitialize QMan with different BAR, got 0x%llx read BARE=0x%x BAR=0x%x\n", + ba, bare, bar); + return -ENOMEM; + } + __qman_requires_cleanup = 1; + /* Return 1 to indicate memory was previously programmed */ + return 1; + } + /* Need to temporarily map the area to make sure it is zeroed */ + ptr = memremap(ba, size, MEMREMAP_WB); + if (!ptr) { + pr_crit("memremap() of QMan private memory failed\n"); + return -ENOMEM; + } + memset(ptr, 0, size); + +#ifdef CONFIG_PPC + /* + * PPC doesn't appear to flush the cache on memunmap() but the + * cache must be flushed since QMan does non coherent accesses + * to this memory + */ + flush_dcache_range((unsigned long) ptr, (unsigned long) ptr+size); +#endif + memunmap(ptr); + + qm_ccsr_out(offset, upper_32_bits(ba)); + qm_ccsr_out(offset + REG_offset_BAR, lower_32_bits(ba)); + qm_ccsr_out(offset + REG_offset_AR, PFDR_AR_EN | (exp - 1)); + return 0; +} + +static void qm_set_pfdr_threshold(u32 th, u8 k) +{ + qm_ccsr_out(REG_PFDR_FP_LWIT, th & 0xffffff); + qm_ccsr_out(REG_PFDR_CFG, k); +} + +static void qm_set_sfdr_threshold(u16 th) +{ + qm_ccsr_out(REG_SFDR_CFG, th & 0x3ff); +} + +static int qm_init_pfdr(struct device *dev, u32 pfdr_start, u32 num) +{ + u8 rslt = MCR_get_rslt(qm_ccsr_in(REG_MCR)); + + DPAA_ASSERT(pfdr_start && !(pfdr_start & 7) && !(num & 7) && num); + /* Make sure the command interface is 'idle' */ + if (!MCR_rslt_idle(rslt)) { + dev_crit(dev, "QMAN_MCR isn't idle"); + WARN_ON(1); + } + + /* Write the MCR command params then the verb */ + qm_ccsr_out(REG_MCP(0), pfdr_start); + /* + * TODO: remove this - it's a workaround for a model bug that is + * corrected in more recent versions. We use the workaround until + * everyone has upgraded. + */ + qm_ccsr_out(REG_MCP(1), pfdr_start + num - 16); + dma_wmb(); + qm_ccsr_out(REG_MCR, MCR_INIT_PFDR); + /* Poll for the result */ + do { + rslt = MCR_get_rslt(qm_ccsr_in(REG_MCR)); + } while (!MCR_rslt_idle(rslt)); + if (MCR_rslt_ok(rslt)) + return 0; + if (MCR_rslt_eaccess(rslt)) + return -EACCES; + if (MCR_rslt_inval(rslt)) + return -EINVAL; + dev_crit(dev, "Unexpected result from MCR_INIT_PFDR: %02x\n", rslt); + return -ENODEV; +} + +/* + * QMan needs two global memory areas initialized at boot time: + * 1) FQD: Frame Queue Descriptors used to manage frame queues + * 2) PFDR: Packed Frame Queue Descriptor Records used to store frames + * Both areas are reserved using the device tree reserved memory framework + * and the addresses and sizes are initialized when the QMan device is probed + */ +static dma_addr_t fqd_a, pfdr_a; +static size_t fqd_sz, pfdr_sz; + +#ifdef CONFIG_PPC +/* + * Support for PPC Device Tree backward compatibility when compatible + * string is set to fsl-qman-fqd and fsl-qman-pfdr + */ +static int zero_priv_mem(phys_addr_t addr, size_t sz) +{ + /* map as cacheable, non-guarded */ + void __iomem *tmpp = ioremap_cache(addr, sz); + + if (!tmpp) + return -ENOMEM; + + memset_io(tmpp, 0, sz); + flush_dcache_range((unsigned long)tmpp, + (unsigned long)tmpp + sz); + iounmap(tmpp); + + return 0; +} + +static int qman_fqd(struct reserved_mem *rmem) +{ + fqd_a = rmem->base; + fqd_sz = rmem->size; + + WARN_ON(!(fqd_a && fqd_sz)); + return 0; +} +RESERVEDMEM_OF_DECLARE(qman_fqd, "fsl,qman-fqd", qman_fqd); + +static int qman_pfdr(struct reserved_mem *rmem) +{ + pfdr_a = rmem->base; + pfdr_sz = rmem->size; + + WARN_ON(!(pfdr_a && pfdr_sz)); + + return 0; +} +RESERVEDMEM_OF_DECLARE(qman_pfdr, "fsl,qman-pfdr", qman_pfdr); + +#endif + +unsigned int qm_get_fqid_maxcnt(void) +{ + return fqd_sz / 64; +} + +static void log_edata_bits(struct device *dev, u32 bit_count) +{ + u32 i, j, mask = 0xffffffff; + + dev_warn(dev, "ErrInt, EDATA:\n"); + i = bit_count / 32; + if (bit_count % 32) { + i++; + mask = ~(mask << bit_count % 32); + } + j = 16 - i; + dev_warn(dev, " 0x%08x\n", qm_ccsr_in(REG_EDATA(j)) & mask); + j++; + for (; j < 16; j++) + dev_warn(dev, " 0x%08x\n", qm_ccsr_in(REG_EDATA(j))); +} + +static void log_additional_error_info(struct device *dev, u32 isr_val, + u32 ecsr_val) +{ + struct qm_ecir ecir_val; + struct qm_eadr eadr_val; + int memid; + + ecir_val.info = qm_ccsr_in(REG_ECIR); + /* Is portal info valid */ + if ((qman_ip_rev & 0xFF00) >= QMAN_REV30) { + struct qm_ecir2 ecir2_val; + + ecir2_val.info = qm_ccsr_in(REG_ECIR2); + if (ecsr_val & PORTAL_ECSR_ERR) { + dev_warn(dev, "ErrInt: %s id %d\n", + qm_ecir2_is_dcp(&ecir2_val) ? "DCP" : "SWP", + qm_ecir2_get_pnum(&ecir2_val)); + } + if (ecsr_val & (FQID_ECSR_ERR | QM_EIRQ_IECE)) + dev_warn(dev, "ErrInt: ecir.fqid 0x%x\n", + qm_ecir_get_fqid(&ecir_val)); + + if (ecsr_val & (QM_EIRQ_SBEI|QM_EIRQ_MBEI)) { + eadr_val.info = qm_ccsr_in(REG_EADR); + memid = qm_eadr_v3_get_memid(&eadr_val); + dev_warn(dev, "ErrInt: EADR Memory: %s, 0x%x\n", + error_mdata[memid].txt, + error_mdata[memid].addr_mask + & qm_eadr_v3_get_eadr(&eadr_val)); + log_edata_bits(dev, error_mdata[memid].bits); + } + } else { + if (ecsr_val & PORTAL_ECSR_ERR) { + dev_warn(dev, "ErrInt: %s id %d\n", + qm_ecir_is_dcp(&ecir_val) ? "DCP" : "SWP", + qm_ecir_get_pnum(&ecir_val)); + } + if (ecsr_val & FQID_ECSR_ERR) + dev_warn(dev, "ErrInt: ecir.fqid 0x%x\n", + qm_ecir_get_fqid(&ecir_val)); + + if (ecsr_val & (QM_EIRQ_SBEI|QM_EIRQ_MBEI)) { + eadr_val.info = qm_ccsr_in(REG_EADR); + memid = qm_eadr_get_memid(&eadr_val); + dev_warn(dev, "ErrInt: EADR Memory: %s, 0x%x\n", + error_mdata[memid].txt, + error_mdata[memid].addr_mask + & qm_eadr_get_eadr(&eadr_val)); + log_edata_bits(dev, error_mdata[memid].bits); + } + } +} + +static irqreturn_t qman_isr(int irq, void *ptr) +{ + u32 isr_val, ier_val, ecsr_val, isr_mask, i; + struct device *dev = ptr; + + ier_val = qm_ccsr_in(REG_ERR_IER); + isr_val = qm_ccsr_in(REG_ERR_ISR); + ecsr_val = qm_ccsr_in(REG_ECSR); + isr_mask = isr_val & ier_val; + + if (!isr_mask) + return IRQ_NONE; + + for (i = 0; i < ARRAY_SIZE(qman_hwerr_txts); i++) { + if (qman_hwerr_txts[i].mask & isr_mask) { + dev_err_ratelimited(dev, "ErrInt: %s\n", + qman_hwerr_txts[i].txt); + if (qman_hwerr_txts[i].mask & ecsr_val) { + log_additional_error_info(dev, isr_mask, + ecsr_val); + /* Re-arm error capture registers */ + qm_ccsr_out(REG_ECSR, ecsr_val); + } + if (qman_hwerr_txts[i].mask & QMAN_ERRS_TO_DISABLE) { + dev_dbg(dev, "Disabling error 0x%x\n", + qman_hwerr_txts[i].mask); + ier_val &= ~qman_hwerr_txts[i].mask; + qm_ccsr_out(REG_ERR_IER, ier_val); + } + } + } + qm_ccsr_out(REG_ERR_ISR, isr_val); + + return IRQ_HANDLED; +} + +static int qman_init_ccsr(struct device *dev) +{ + int i, err; + + /* FQD memory */ + err = qm_set_memory(qm_memory_fqd, fqd_a, fqd_sz); + if (err < 0) + return err; + /* PFDR memory */ + err = qm_set_memory(qm_memory_pfdr, pfdr_a, pfdr_sz); + if (err < 0) + return err; + /* Only initialize PFDRs if the QMan was not initialized before */ + if (err == 0) { + err = qm_init_pfdr(dev, 8, pfdr_sz / 64 - 8); + if (err) + return err; + } + /* thresholds */ + qm_set_pfdr_threshold(512, 64); + qm_set_sfdr_threshold(128); + /* clear stale PEBI bit from interrupt status register */ + qm_ccsr_out(REG_ERR_ISR, QM_EIRQ_PEBI); + /* corenet initiator settings */ + qm_set_corenet_initiator(); + /* HID settings */ + qm_set_hid(); + /* Set scheduling weights to defaults */ + for (i = qm_wq_first; i <= qm_wq_last; i++) + qm_set_wq_scheduling(i, 0, 0, 0, 0, 0, 0, 0); + /* We are not prepared to accept ERNs for hardware enqueues */ + qm_set_dc(qm_dc_portal_fman0, 1, 0); + qm_set_dc(qm_dc_portal_fman1, 1, 0); + return 0; +} + +#define LIO_CFG_LIODN_MASK 0x0fff0000 +void __qman_liodn_fixup(u16 channel) +{ + static int done; + static u32 liodn_offset; + u32 before, after; + int idx = channel - QM_CHANNEL_SWPORTAL0; + + if ((qman_ip_rev & 0xFF00) >= QMAN_REV30) + before = qm_ccsr_in(REG_REV3_QCSP_LIO_CFG(idx)); + else + before = qm_ccsr_in(REG_QCSP_LIO_CFG(idx)); + if (!done) { + liodn_offset = before & LIO_CFG_LIODN_MASK; + done = 1; + return; + } + after = (before & (~LIO_CFG_LIODN_MASK)) | liodn_offset; + if ((qman_ip_rev & 0xFF00) >= QMAN_REV30) + qm_ccsr_out(REG_REV3_QCSP_LIO_CFG(idx), after); + else + qm_ccsr_out(REG_QCSP_LIO_CFG(idx), after); +} + +#define IO_CFG_SDEST_MASK 0x00ff0000 +void qman_set_sdest(u16 channel, unsigned int cpu_idx) +{ + int idx = channel - QM_CHANNEL_SWPORTAL0; + u32 before, after; + + if ((qman_ip_rev & 0xFF00) >= QMAN_REV30) { + before = qm_ccsr_in(REG_REV3_QCSP_IO_CFG(idx)); + /* Each pair of vcpu share the same SRQ(SDEST) */ + cpu_idx /= 2; + after = (before & (~IO_CFG_SDEST_MASK)) | (cpu_idx << 16); + qm_ccsr_out(REG_REV3_QCSP_IO_CFG(idx), after); + } else { + before = qm_ccsr_in(REG_QCSP_IO_CFG(idx)); + after = (before & (~IO_CFG_SDEST_MASK)) | (cpu_idx << 16); + qm_ccsr_out(REG_QCSP_IO_CFG(idx), after); + } +} + +static int qman_resource_init(struct device *dev) +{ + int pool_chan_num, cgrid_num; + int ret, i; + + switch (qman_ip_rev >> 8) { + case 1: + pool_chan_num = 15; + cgrid_num = 256; + break; + case 2: + pool_chan_num = 3; + cgrid_num = 64; + break; + case 3: + pool_chan_num = 15; + cgrid_num = 256; + break; + default: + return -ENODEV; + } + + ret = gen_pool_add(qm_qpalloc, qm_channel_pool1 | DPAA_GENALLOC_OFF, + pool_chan_num, -1); + if (ret) { + dev_err(dev, "Failed to seed pool channels (%d)\n", ret); + return ret; + } + + ret = gen_pool_add(qm_cgralloc, DPAA_GENALLOC_OFF, cgrid_num, -1); + if (ret) { + dev_err(dev, "Failed to seed CGRID range (%d)\n", ret); + return ret; + } + + /* parse pool channels into the SDQCR mask */ + for (i = 0; i < cgrid_num; i++) + qm_pools_sdqcr |= QM_SDQCR_CHANNELS_POOL_CONV(i); + + ret = gen_pool_add(qm_fqalloc, QM_FQID_RANGE_START | DPAA_GENALLOC_OFF, + qm_get_fqid_maxcnt() - QM_FQID_RANGE_START, -1); + if (ret) { + dev_err(dev, "Failed to seed FQID range (%d)\n", ret); + return ret; + } + + return 0; +} + +int qman_is_probed(void) +{ + return __qman_probed; +} +EXPORT_SYMBOL_GPL(qman_is_probed); + +int qman_requires_cleanup(void) +{ + return __qman_requires_cleanup; +} + +void qman_done_cleanup(void) +{ + qman_enable_irqs(); + __qman_requires_cleanup = 0; +} + + +static int fsl_qman_probe(struct platform_device *pdev) +{ + struct device *dev = &pdev->dev; + struct device_node *node = dev->of_node; + struct resource *res; + int ret, err_irq; + u16 id; + u8 major, minor; + + __qman_probed = -1; + + res = platform_get_resource(pdev, IORESOURCE_MEM, 0); + if (!res) { + dev_err(dev, "Can't get %pOF property 'IORESOURCE_MEM'\n", + node); + return -ENXIO; + } + qm_ccsr_start = devm_ioremap(dev, res->start, resource_size(res)); + if (!qm_ccsr_start) + return -ENXIO; + + qm_get_version(&id, &major, &minor); + if (major == 1 && minor == 0) { + dev_err(dev, "Rev1.0 on P4080 rev1 is not supported!\n"); + return -ENODEV; + } else if (major == 1 && minor == 1) + qman_ip_rev = QMAN_REV11; + else if (major == 1 && minor == 2) + qman_ip_rev = QMAN_REV12; + else if (major == 2 && minor == 0) + qman_ip_rev = QMAN_REV20; + else if (major == 3 && minor == 0) + qman_ip_rev = QMAN_REV30; + else if (major == 3 && minor == 1) + qman_ip_rev = QMAN_REV31; + else if (major == 3 && minor == 2) + qman_ip_rev = QMAN_REV32; + else { + dev_err(dev, "Unknown QMan version\n"); + return -ENODEV; + } + + if ((qman_ip_rev & 0xff00) >= QMAN_REV30) { + qm_channel_pool1 = QMAN_CHANNEL_POOL1_REV3; + qm_channel_caam = QMAN_CHANNEL_CAAM_REV3; + } + + if (fqd_a) { +#ifdef CONFIG_PPC + /* + * For PPC backward DT compatibility + * FQD memory MUST be zero'd by software + */ + zero_priv_mem(fqd_a, fqd_sz); +#else + WARN(1, "Unexpected architecture using non shared-dma-mem reservations"); +#endif + } else { + /* + * Order of memory regions is assumed as FQD followed by PFDR + * in order to ensure allocations from the correct regions the + * driver initializes then allocates each piece in order + */ + ret = qbman_init_private_mem(dev, 0, &fqd_a, &fqd_sz); + if (ret) { + dev_err(dev, "qbman_init_private_mem() for FQD failed 0x%x\n", + ret); + return -ENODEV; + } + } + dev_dbg(dev, "Allocated FQD 0x%llx 0x%zx\n", fqd_a, fqd_sz); + + if (!pfdr_a) { + /* Setup PFDR memory */ + ret = qbman_init_private_mem(dev, 1, &pfdr_a, &pfdr_sz); + if (ret) { + dev_err(dev, "qbman_init_private_mem() for PFDR failed 0x%x\n", + ret); + return -ENODEV; + } + } + dev_dbg(dev, "Allocated PFDR 0x%llx 0x%zx\n", pfdr_a, pfdr_sz); + + ret = qman_init_ccsr(dev); + if (ret) { + dev_err(dev, "CCSR setup failed\n"); + return ret; + } + + err_irq = platform_get_irq(pdev, 0); + if (err_irq <= 0) { + dev_info(dev, "Can't get %pOF property 'interrupts'\n", + node); + return -ENODEV; + } + ret = devm_request_irq(dev, err_irq, qman_isr, IRQF_SHARED, "qman-err", + dev); + if (ret) { + dev_err(dev, "devm_request_irq() failed %d for '%pOF'\n", + ret, node); + return ret; + } + + /* + * Write-to-clear any stale bits, (eg. starvation being asserted prior + * to resource allocation during driver init). + */ + qm_ccsr_out(REG_ERR_ISR, 0xffffffff); + /* Enable Error Interrupts */ + qm_ccsr_out(REG_ERR_IER, 0xffffffff); + + qm_fqalloc = devm_gen_pool_create(dev, 0, -1, "qman-fqalloc"); + if (IS_ERR(qm_fqalloc)) { + ret = PTR_ERR(qm_fqalloc); + dev_err(dev, "qman-fqalloc pool init failed (%d)\n", ret); + return ret; + } + + qm_qpalloc = devm_gen_pool_create(dev, 0, -1, "qman-qpalloc"); + if (IS_ERR(qm_qpalloc)) { + ret = PTR_ERR(qm_qpalloc); + dev_err(dev, "qman-qpalloc pool init failed (%d)\n", ret); + return ret; + } + + qm_cgralloc = devm_gen_pool_create(dev, 0, -1, "qman-cgralloc"); + if (IS_ERR(qm_cgralloc)) { + ret = PTR_ERR(qm_cgralloc); + dev_err(dev, "qman-cgralloc pool init failed (%d)\n", ret); + return ret; + } + + ret = qman_resource_init(dev); + if (ret) + return ret; + + ret = qman_alloc_fq_table(qm_get_fqid_maxcnt()); + if (ret) + return ret; + + ret = qman_wq_alloc(); + if (ret) + return ret; + + __qman_probed = 1; + + return 0; +} + +static const struct of_device_id fsl_qman_ids[] = { + { + .compatible = "fsl,qman", + }, + {} +}; + +static struct platform_driver fsl_qman_driver = { + .driver = { + .name = KBUILD_MODNAME, + .of_match_table = fsl_qman_ids, + .suppress_bind_attrs = true, + }, + .probe = fsl_qman_probe, +}; + +builtin_platform_driver(fsl_qman_driver); diff --git a/drivers/soc/fsl/qbman/qman_portal.c b/drivers/soc/fsl/qbman/qman_portal.c new file mode 100644 index 000000000..e23b60618 --- /dev/null +++ b/drivers/soc/fsl/qbman/qman_portal.c @@ -0,0 +1,342 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "qman_priv.h" + +struct qman_portal *qman_dma_portal; +EXPORT_SYMBOL(qman_dma_portal); + +/* Enable portal interupts (as opposed to polling mode) */ +#define CONFIG_FSL_DPA_PIRQ_SLOW 1 +#define CONFIG_FSL_DPA_PIRQ_FAST 1 + +static struct cpumask portal_cpus; +static int __qman_portals_probed; +/* protect qman global registers and global data shared among portals */ +static DEFINE_SPINLOCK(qman_lock); + +static void portal_set_cpu(struct qm_portal_config *pcfg, int cpu) +{ +#ifdef CONFIG_FSL_PAMU + struct device *dev = pcfg->dev; + int ret; + + pcfg->iommu_domain = iommu_domain_alloc(&platform_bus_type); + if (!pcfg->iommu_domain) { + dev_err(dev, "%s(): iommu_domain_alloc() failed", __func__); + goto no_iommu; + } + ret = fsl_pamu_configure_l1_stash(pcfg->iommu_domain, cpu); + if (ret < 0) { + dev_err(dev, "%s(): fsl_pamu_configure_l1_stash() = %d", + __func__, ret); + goto out_domain_free; + } + ret = iommu_attach_device(pcfg->iommu_domain, dev); + if (ret < 0) { + dev_err(dev, "%s(): iommu_device_attach() = %d", __func__, + ret); + goto out_domain_free; + } + +no_iommu: +#endif + qman_set_sdest(pcfg->channel, cpu); + + return; + +#ifdef CONFIG_FSL_PAMU +out_domain_free: + iommu_domain_free(pcfg->iommu_domain); + pcfg->iommu_domain = NULL; +#endif +} + +static struct qman_portal *init_pcfg(struct qm_portal_config *pcfg) +{ + struct qman_portal *p; + u32 irq_sources = 0; + + /* We need the same LIODN offset for all portals */ + qman_liodn_fixup(pcfg->channel); + + pcfg->iommu_domain = NULL; + portal_set_cpu(pcfg, pcfg->cpu); + + p = qman_create_affine_portal(pcfg, NULL); + if (!p) { + dev_crit(pcfg->dev, "%s: Portal failure on cpu %d\n", + __func__, pcfg->cpu); + return NULL; + } + + /* Determine what should be interrupt-vs-poll driven */ +#ifdef CONFIG_FSL_DPA_PIRQ_SLOW + irq_sources |= QM_PIRQ_EQCI | QM_PIRQ_EQRI | QM_PIRQ_MRI | + QM_PIRQ_CSCI; +#endif +#ifdef CONFIG_FSL_DPA_PIRQ_FAST + irq_sources |= QM_PIRQ_DQRI; +#endif + qman_p_irqsource_add(p, irq_sources); + + spin_lock(&qman_lock); + if (cpumask_equal(&portal_cpus, cpu_possible_mask)) { + /* all assigned portals are initialized now */ + qman_init_cgr_all(); + } + + if (!qman_dma_portal) + qman_dma_portal = p; + + spin_unlock(&qman_lock); + + dev_info(pcfg->dev, "Portal initialised, cpu %d\n", pcfg->cpu); + + return p; +} + +static void qman_portal_update_sdest(const struct qm_portal_config *pcfg, + unsigned int cpu) +{ +#ifdef CONFIG_FSL_PAMU /* TODO */ + if (pcfg->iommu_domain) { + if (fsl_pamu_configure_l1_stash(pcfg->iommu_domain, cpu) < 0) { + dev_err(pcfg->dev, + "Failed to update pamu stash setting\n"); + return; + } + } +#endif + qman_set_sdest(pcfg->channel, cpu); +} + +static int qman_offline_cpu(unsigned int cpu) +{ + struct qman_portal *p; + const struct qm_portal_config *pcfg; + + p = affine_portals[cpu]; + if (p) { + pcfg = qman_get_qm_portal_config(p); + if (pcfg) { + /* select any other online CPU */ + cpu = cpumask_any_but(cpu_online_mask, cpu); + irq_set_affinity(pcfg->irq, cpumask_of(cpu)); + qman_portal_update_sdest(pcfg, cpu); + } + } + return 0; +} + +static int qman_online_cpu(unsigned int cpu) +{ + struct qman_portal *p; + const struct qm_portal_config *pcfg; + + p = affine_portals[cpu]; + if (p) { + pcfg = qman_get_qm_portal_config(p); + if (pcfg) { + irq_set_affinity(pcfg->irq, cpumask_of(cpu)); + qman_portal_update_sdest(pcfg, cpu); + } + } + return 0; +} + +int qman_portals_probed(void) +{ + return __qman_portals_probed; +} +EXPORT_SYMBOL_GPL(qman_portals_probed); + +static int qman_portal_probe(struct platform_device *pdev) +{ + struct device *dev = &pdev->dev; + struct device_node *node = dev->of_node; + struct qm_portal_config *pcfg; + struct resource *addr_phys[2]; + int irq, cpu, err, i; + u32 val; + + err = qman_is_probed(); + if (!err) + return -EPROBE_DEFER; + if (err < 0) { + dev_err(&pdev->dev, "failing probe due to qman probe error\n"); + return -ENODEV; + } + + pcfg = devm_kmalloc(dev, sizeof(*pcfg), GFP_KERNEL); + if (!pcfg) { + __qman_portals_probed = -1; + return -ENOMEM; + } + + pcfg->dev = dev; + + addr_phys[0] = platform_get_resource(pdev, IORESOURCE_MEM, + DPAA_PORTAL_CE); + if (!addr_phys[0]) { + dev_err(dev, "Can't get %pOF property 'reg::CE'\n", node); + goto err_ioremap1; + } + + addr_phys[1] = platform_get_resource(pdev, IORESOURCE_MEM, + DPAA_PORTAL_CI); + if (!addr_phys[1]) { + dev_err(dev, "Can't get %pOF property 'reg::CI'\n", node); + goto err_ioremap1; + } + + err = of_property_read_u32(node, "cell-index", &val); + if (err) { + dev_err(dev, "Can't get %pOF property 'cell-index'\n", node); + __qman_portals_probed = -1; + return err; + } + pcfg->channel = val; + pcfg->cpu = -1; + irq = platform_get_irq(pdev, 0); + if (irq <= 0) + goto err_ioremap1; + pcfg->irq = irq; + + pcfg->addr_virt_ce = memremap(addr_phys[0]->start, + resource_size(addr_phys[0]), + QBMAN_MEMREMAP_ATTR); + if (!pcfg->addr_virt_ce) { + dev_err(dev, "memremap::CE failed\n"); + goto err_ioremap1; + } + + pcfg->addr_virt_ci = ioremap(addr_phys[1]->start, + resource_size(addr_phys[1])); + if (!pcfg->addr_virt_ci) { + dev_err(dev, "ioremap::CI failed\n"); + goto err_ioremap2; + } + + pcfg->pools = qm_get_pools_sdqcr(); + + spin_lock(&qman_lock); + cpu = cpumask_first_zero(&portal_cpus); + if (cpu >= nr_cpu_ids) { + __qman_portals_probed = 1; + /* unassigned portal, skip init */ + spin_unlock(&qman_lock); + goto check_cleanup; + } + + cpumask_set_cpu(cpu, &portal_cpus); + spin_unlock(&qman_lock); + pcfg->cpu = cpu; + + if (dma_set_mask(dev, DMA_BIT_MASK(40))) { + dev_err(dev, "dma_set_mask() failed\n"); + goto err_portal_init; + } + + if (!init_pcfg(pcfg)) { + dev_err(dev, "portal init failed\n"); + goto err_portal_init; + } + + /* clear irq affinity if assigned cpu is offline */ + if (!cpu_online(cpu)) + qman_offline_cpu(cpu); + +check_cleanup: + if (__qman_portals_probed == 1 && qman_requires_cleanup()) { + /* + * QMan wasn't reset prior to boot (Kexec for example) + * Empty all the frame queues so they are in reset state + */ + for (i = 0; i < qm_get_fqid_maxcnt(); i++) { + err = qman_shutdown_fq(i); + if (err) { + dev_err(dev, "Failed to shutdown frame queue %d\n", + i); + goto err_portal_init; + } + } + qman_done_cleanup(); + } + + return 0; + +err_portal_init: + iounmap(pcfg->addr_virt_ci); +err_ioremap2: + memunmap(pcfg->addr_virt_ce); +err_ioremap1: + __qman_portals_probed = -1; + + return -ENXIO; +} + +static const struct of_device_id qman_portal_ids[] = { + { + .compatible = "fsl,qman-portal", + }, + {} +}; +MODULE_DEVICE_TABLE(of, qman_portal_ids); + +static struct platform_driver qman_portal_driver = { + .driver = { + .name = KBUILD_MODNAME, + .of_match_table = qman_portal_ids, + }, + .probe = qman_portal_probe, +}; + +static int __init qman_portal_driver_register(struct platform_driver *drv) +{ + int ret; + + ret = platform_driver_register(drv); + if (ret < 0) + return ret; + + ret = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, + "soc/qman_portal:online", + qman_online_cpu, qman_offline_cpu); + if (ret < 0) { + pr_err("qman: failed to register hotplug callbacks.\n"); + platform_driver_unregister(drv); + return ret; + } + return 0; +} + +module_driver(qman_portal_driver, + qman_portal_driver_register, platform_driver_unregister); diff --git a/drivers/soc/fsl/qbman/qman_priv.h b/drivers/soc/fsl/qbman/qman_priv.h new file mode 100644 index 000000000..fd1cf543f --- /dev/null +++ b/drivers/soc/fsl/qbman/qman_priv.h @@ -0,0 +1,282 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "dpaa_sys.h" + +#include <soc/fsl/qman.h> +#include <linux/dma-mapping.h> +#include <linux/iommu.h> + +#if defined(CONFIG_FSL_PAMU) +#include <asm/fsl_pamu_stash.h> +#endif + +struct qm_mcr_querywq { + u8 verb; + u8 result; + u16 channel_wq; /* ignores wq (3 lsbits): _res[0-2] */ + u8 __reserved[28]; + u32 wq_len[8]; +} __packed; + +static inline u16 qm_mcr_querywq_get_chan(const struct qm_mcr_querywq *wq) +{ + return wq->channel_wq >> 3; +} + +struct __qm_mcr_querycongestion { + u32 state[8]; +}; + +/* "Query Congestion Group State" */ +struct qm_mcr_querycongestion { + u8 verb; + u8 result; + u8 __reserved[30]; + /* Access this struct using qman_cgrs_get() */ + struct __qm_mcr_querycongestion state; +} __packed; + +/* "Query CGR" */ +struct qm_mcr_querycgr { + u8 verb; + u8 result; + u16 __reserved1; + struct __qm_mc_cgr cgr; /* CGR fields */ + u8 __reserved2[6]; + u8 i_bcnt_hi; /* high 8-bits of 40-bit "Instant" */ + __be32 i_bcnt_lo; /* low 32-bits of 40-bit */ + u8 __reserved3[3]; + u8 a_bcnt_hi; /* high 8-bits of 40-bit "Average" */ + __be32 a_bcnt_lo; /* low 32-bits of 40-bit */ + __be32 cscn_targ_swp[4]; +} __packed; + +static inline u64 qm_mcr_querycgr_i_get64(const struct qm_mcr_querycgr *q) +{ + return ((u64)q->i_bcnt_hi << 32) | be32_to_cpu(q->i_bcnt_lo); +} +static inline u64 qm_mcr_querycgr_a_get64(const struct qm_mcr_querycgr *q) +{ + return ((u64)q->a_bcnt_hi << 32) | be32_to_cpu(q->a_bcnt_lo); +} + +/* Congestion Groups */ + +/* + * This wrapper represents a bit-array for the state of the 256 QMan congestion + * groups. Is also used as a *mask* for congestion groups, eg. so we ignore + * those that don't concern us. We harness the structure and accessor details + * already used in the management command to query congestion groups. + */ +#define CGR_BITS_PER_WORD 5 +#define CGR_WORD(x) ((x) >> CGR_BITS_PER_WORD) +#define CGR_BIT(x) (BIT(31) >> ((x) & 0x1f)) +#define CGR_NUM (sizeof(struct __qm_mcr_querycongestion) << 3) + +struct qman_cgrs { + struct __qm_mcr_querycongestion q; +}; + +static inline void qman_cgrs_init(struct qman_cgrs *c) +{ + memset(c, 0, sizeof(*c)); +} + +static inline void qman_cgrs_fill(struct qman_cgrs *c) +{ + memset(c, 0xff, sizeof(*c)); +} + +static inline int qman_cgrs_get(struct qman_cgrs *c, u8 cgr) +{ + return c->q.state[CGR_WORD(cgr)] & CGR_BIT(cgr); +} + +static inline void qman_cgrs_cp(struct qman_cgrs *dest, + const struct qman_cgrs *src) +{ + *dest = *src; +} + +static inline void qman_cgrs_and(struct qman_cgrs *dest, + const struct qman_cgrs *a, const struct qman_cgrs *b) +{ + int ret; + u32 *_d = dest->q.state; + const u32 *_a = a->q.state; + const u32 *_b = b->q.state; + + for (ret = 0; ret < 8; ret++) + *_d++ = *_a++ & *_b++; +} + +static inline void qman_cgrs_xor(struct qman_cgrs *dest, + const struct qman_cgrs *a, const struct qman_cgrs *b) +{ + int ret; + u32 *_d = dest->q.state; + const u32 *_a = a->q.state; + const u32 *_b = b->q.state; + + for (ret = 0; ret < 8; ret++) + *_d++ = *_a++ ^ *_b++; +} + +void qman_init_cgr_all(void); + +struct qm_portal_config { + /* Portal addresses */ + void *addr_virt_ce; + void __iomem *addr_virt_ci; + struct device *dev; + struct iommu_domain *iommu_domain; + /* Allow these to be joined in lists */ + struct list_head list; + /* User-visible portal configuration settings */ + /* portal is affined to this cpu */ + int cpu; + /* portal interrupt line */ + int irq; + /* + * the portal's dedicated channel id, used initialising + * frame queues to target this portal when scheduled + */ + u16 channel; + /* + * mask of pool channels this portal has dequeue access to + * (using QM_SDQCR_CHANNELS_POOL(n) for the bitmask) + */ + u32 pools; +}; + +/* Revision info (for errata and feature handling) */ +#define QMAN_REV11 0x0101 +#define QMAN_REV12 0x0102 +#define QMAN_REV20 0x0200 +#define QMAN_REV30 0x0300 +#define QMAN_REV31 0x0301 +#define QMAN_REV32 0x0302 +extern u16 qman_ip_rev; /* 0 if uninitialised, otherwise QMAN_REVx */ + +#define QM_FQID_RANGE_START 1 /* FQID 0 reserved for internal use */ +extern struct gen_pool *qm_fqalloc; /* FQID allocator */ +extern struct gen_pool *qm_qpalloc; /* pool-channel allocator */ +extern struct gen_pool *qm_cgralloc; /* CGR ID allocator */ +u32 qm_get_pools_sdqcr(void); + +int qman_wq_alloc(void); +#ifdef CONFIG_FSL_PAMU +#define qman_liodn_fixup __qman_liodn_fixup +#else +static inline void qman_liodn_fixup(u16 channel) +{ +} +#endif +void __qman_liodn_fixup(u16 channel); +void qman_set_sdest(u16 channel, unsigned int cpu_idx); + +struct qman_portal *qman_create_affine_portal( + const struct qm_portal_config *config, + const struct qman_cgrs *cgrs); +const struct qm_portal_config *qman_destroy_affine_portal(void); + +/* + * qman_query_fq - Queries FQD fields (via h/w query command) + * @fq: the frame queue object to be queried + * @fqd: storage for the queried FQD fields + */ +int qman_query_fq(struct qman_fq *fq, struct qm_fqd *fqd); + +int qman_alloc_fq_table(u32 num_fqids); + +/* QMan s/w corenet portal, low-level i/face */ + +/* + * For qm_dqrr_sdqcr_set(); Choose one SOURCE. Choose one COUNT. Choose one + * dequeue TYPE. Choose TOKEN (8-bit). + * If SOURCE == CHANNELS, + * Choose CHANNELS_DEDICATED and/or CHANNELS_POOL(n). + * You can choose DEDICATED_PRECEDENCE if the portal channel should have + * priority. + * If SOURCE == SPECIFICWQ, + * Either select the work-queue ID with SPECIFICWQ_WQ(), or select the + * channel (SPECIFICWQ_DEDICATED or SPECIFICWQ_POOL()) and specify the + * work-queue priority (0-7) with SPECIFICWQ_WQ() - either way, you get the + * same value. + */ +#define QM_SDQCR_SOURCE_CHANNELS 0x0 +#define QM_SDQCR_SOURCE_SPECIFICWQ 0x40000000 +#define QM_SDQCR_COUNT_EXACT1 0x0 +#define QM_SDQCR_COUNT_UPTO3 0x20000000 +#define QM_SDQCR_DEDICATED_PRECEDENCE 0x10000000 +#define QM_SDQCR_TYPE_MASK 0x03000000 +#define QM_SDQCR_TYPE_NULL 0x0 +#define QM_SDQCR_TYPE_PRIO_QOS 0x01000000 +#define QM_SDQCR_TYPE_ACTIVE_QOS 0x02000000 +#define QM_SDQCR_TYPE_ACTIVE 0x03000000 +#define QM_SDQCR_TOKEN_MASK 0x00ff0000 +#define QM_SDQCR_TOKEN_SET(v) (((v) & 0xff) << 16) +#define QM_SDQCR_TOKEN_GET(v) (((v) >> 16) & 0xff) +#define QM_SDQCR_CHANNELS_DEDICATED 0x00008000 +#define QM_SDQCR_SPECIFICWQ_MASK 0x000000f7 +#define QM_SDQCR_SPECIFICWQ_DEDICATED 0x00000000 +#define QM_SDQCR_SPECIFICWQ_POOL(n) ((n) << 4) +#define QM_SDQCR_SPECIFICWQ_WQ(n) (n) + +/* For qm_dqrr_vdqcr_set(): use FQID(n) to fill in the frame queue ID */ +#define QM_VDQCR_FQID_MASK 0x00ffffff +#define QM_VDQCR_FQID(n) ((n) & QM_VDQCR_FQID_MASK) + +/* + * Used by all portal interrupt registers except 'inhibit' + * Channels with frame availability + */ +#define QM_PIRQ_DQAVAIL 0x0000ffff + +/* The DQAVAIL interrupt fields break down into these bits; */ +#define QM_DQAVAIL_PORTAL 0x8000 /* Portal channel */ +#define QM_DQAVAIL_POOL(n) (0x8000 >> (n)) /* Pool channel, n==[1..15] */ +#define QM_DQAVAIL_MASK 0xffff +/* This mask contains all the "irqsource" bits visible to API users */ +#define QM_PIRQ_VISIBLE (QM_PIRQ_SLOW | QM_PIRQ_DQRI) + +extern struct qman_portal *affine_portals[NR_CPUS]; +extern struct qman_portal *qman_dma_portal; +const struct qm_portal_config *qman_get_qm_portal_config( + struct qman_portal *portal); + +unsigned int qm_get_fqid_maxcnt(void); + +int qman_shutdown_fq(u32 fqid); + +int qman_requires_cleanup(void); +void qman_done_cleanup(void); +void qman_enable_irqs(void); diff --git a/drivers/soc/fsl/qbman/qman_test.c b/drivers/soc/fsl/qbman/qman_test.c new file mode 100644 index 000000000..18f7f0202 --- /dev/null +++ b/drivers/soc/fsl/qbman/qman_test.c @@ -0,0 +1,62 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "qman_test.h" + +MODULE_AUTHOR("Geoff Thorpe"); +MODULE_LICENSE("Dual BSD/GPL"); +MODULE_DESCRIPTION("QMan testing"); + +static int test_init(void) +{ + int loop = 1; + int err = 0; + + while (loop--) { +#ifdef CONFIG_FSL_QMAN_TEST_STASH + err = qman_test_stash(); + if (err) + break; +#endif +#ifdef CONFIG_FSL_QMAN_TEST_API + err = qman_test_api(); + if (err) + break; +#endif + } + return err; +} + +static void test_exit(void) +{ +} + +module_init(test_init); +module_exit(test_exit); diff --git a/drivers/soc/fsl/qbman/qman_test.h b/drivers/soc/fsl/qbman/qman_test.h new file mode 100644 index 000000000..41bdbc48c --- /dev/null +++ b/drivers/soc/fsl/qbman/qman_test.h @@ -0,0 +1,34 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "qman_priv.h" + +int qman_test_stash(void); +int qman_test_api(void); diff --git a/drivers/soc/fsl/qbman/qman_test_api.c b/drivers/soc/fsl/qbman/qman_test_api.c new file mode 100644 index 000000000..28fbddc3c --- /dev/null +++ b/drivers/soc/fsl/qbman/qman_test_api.c @@ -0,0 +1,247 @@ +/* Copyright 2008 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "qman_test.h" + +#define CGR_ID 27 +#define POOL_ID 2 +#define FQ_FLAGS QMAN_FQ_FLAG_DYNAMIC_FQID +#define NUM_ENQUEUES 10 +#define NUM_PARTIAL 4 +#define PORTAL_SDQCR (QM_SDQCR_SOURCE_CHANNELS | \ + QM_SDQCR_TYPE_PRIO_QOS | \ + QM_SDQCR_TOKEN_SET(0x98) | \ + QM_SDQCR_CHANNELS_DEDICATED | \ + QM_SDQCR_CHANNELS_POOL(POOL_ID)) +#define PORTAL_OPAQUE ((void *)0xf00dbeef) +#define VDQCR_FLAGS (QMAN_VOLATILE_FLAG_WAIT | QMAN_VOLATILE_FLAG_FINISH) + +static enum qman_cb_dqrr_result cb_dqrr(struct qman_portal *, + struct qman_fq *, + const struct qm_dqrr_entry *, + bool sched_napi); +static void cb_ern(struct qman_portal *, struct qman_fq *, + const union qm_mr_entry *); +static void cb_fqs(struct qman_portal *, struct qman_fq *, + const union qm_mr_entry *); + +static struct qm_fd fd, fd_dq; +static struct qman_fq fq_base = { + .cb.dqrr = cb_dqrr, + .cb.ern = cb_ern, + .cb.fqs = cb_fqs +}; +static DECLARE_WAIT_QUEUE_HEAD(waitqueue); +static int retire_complete, sdqcr_complete; + +/* Helpers for initialising and "incrementing" a frame descriptor */ +static void fd_init(struct qm_fd *fd) +{ + qm_fd_addr_set64(fd, 0xabdeadbeefLLU); + qm_fd_set_contig_big(fd, 0x0000ffff); + fd->cmd = cpu_to_be32(0xfeedf00d); +} + +static void fd_inc(struct qm_fd *fd) +{ + u64 t = qm_fd_addr_get64(fd); + int z = t >> 40; + unsigned int len, off; + enum qm_fd_format fmt; + + t <<= 1; + if (z) + t |= 1; + qm_fd_addr_set64(fd, t); + + fmt = qm_fd_get_format(fd); + off = qm_fd_get_offset(fd); + len = qm_fd_get_length(fd); + len--; + qm_fd_set_param(fd, fmt, off, len); + + be32_add_cpu(&fd->cmd, 1); +} + +/* The only part of the 'fd' we can't memcmp() is the ppid */ +static bool fd_neq(const struct qm_fd *a, const struct qm_fd *b) +{ + bool neq = qm_fd_addr_get64(a) != qm_fd_addr_get64(b); + + neq |= qm_fd_get_format(a) != qm_fd_get_format(b); + neq |= a->cfg != b->cfg; + neq |= a->cmd != b->cmd; + + return neq; +} + +/* test */ +static int do_enqueues(struct qman_fq *fq) +{ + unsigned int loop; + int err = 0; + + for (loop = 0; loop < NUM_ENQUEUES; loop++) { + if (qman_enqueue(fq, &fd)) { + pr_crit("qman_enqueue() failed\n"); + err = -EIO; + } + fd_inc(&fd); + } + + return err; +} + +int qman_test_api(void) +{ + unsigned int flags, frmcnt; + int err; + struct qman_fq *fq = &fq_base; + + pr_info("%s(): Starting\n", __func__); + fd_init(&fd); + fd_init(&fd_dq); + + /* Initialise (parked) FQ */ + err = qman_create_fq(0, FQ_FLAGS, fq); + if (err) { + pr_crit("qman_create_fq() failed\n"); + goto failed; + } + err = qman_init_fq(fq, QMAN_INITFQ_FLAG_LOCAL, NULL); + if (err) { + pr_crit("qman_init_fq() failed\n"); + goto failed; + } + /* Do enqueues + VDQCR, twice. (Parked FQ) */ + err = do_enqueues(fq); + if (err) + goto failed; + pr_info("VDQCR (till-empty);\n"); + frmcnt = QM_VDQCR_NUMFRAMES_TILLEMPTY; + err = qman_volatile_dequeue(fq, VDQCR_FLAGS, frmcnt); + if (err) { + pr_crit("qman_volatile_dequeue() failed\n"); + goto failed; + } + err = do_enqueues(fq); + if (err) + goto failed; + pr_info("VDQCR (%d of %d);\n", NUM_PARTIAL, NUM_ENQUEUES); + frmcnt = QM_VDQCR_NUMFRAMES_SET(NUM_PARTIAL); + err = qman_volatile_dequeue(fq, VDQCR_FLAGS, frmcnt); + if (err) { + pr_crit("qman_volatile_dequeue() failed\n"); + goto failed; + } + pr_info("VDQCR (%d of %d);\n", NUM_ENQUEUES - NUM_PARTIAL, + NUM_ENQUEUES); + frmcnt = QM_VDQCR_NUMFRAMES_SET(NUM_ENQUEUES - NUM_PARTIAL); + err = qman_volatile_dequeue(fq, VDQCR_FLAGS, frmcnt); + if (err) { + pr_err("qman_volatile_dequeue() failed\n"); + goto failed; + } + + err = do_enqueues(fq); + if (err) + goto failed; + pr_info("scheduled dequeue (till-empty)\n"); + err = qman_schedule_fq(fq); + if (err) { + pr_crit("qman_schedule_fq() failed\n"); + goto failed; + } + wait_event(waitqueue, sdqcr_complete); + + /* Retire and OOS the FQ */ + err = qman_retire_fq(fq, &flags); + if (err < 0) { + pr_crit("qman_retire_fq() failed\n"); + goto failed; + } + wait_event(waitqueue, retire_complete); + if (flags & QMAN_FQ_STATE_BLOCKOOS) { + err = -EIO; + pr_crit("leaking frames\n"); + goto failed; + } + err = qman_oos_fq(fq); + if (err) { + pr_crit("qman_oos_fq() failed\n"); + goto failed; + } + qman_destroy_fq(fq); + pr_info("%s(): Finished\n", __func__); + return 0; + +failed: + WARN_ON(1); + return err; +} + +static enum qman_cb_dqrr_result cb_dqrr(struct qman_portal *p, + struct qman_fq *fq, + const struct qm_dqrr_entry *dq, + bool sched_napi) +{ + if (WARN_ON(fd_neq(&fd_dq, &dq->fd))) { + pr_err("BADNESS: dequeued frame doesn't match;\n"); + return qman_cb_dqrr_consume; + } + fd_inc(&fd_dq); + if (!(dq->stat & QM_DQRR_STAT_UNSCHEDULED) && !fd_neq(&fd_dq, &fd)) { + sdqcr_complete = 1; + wake_up(&waitqueue); + } + return qman_cb_dqrr_consume; +} + +static void cb_ern(struct qman_portal *p, struct qman_fq *fq, + const union qm_mr_entry *msg) +{ + pr_crit("cb_ern() unimplemented"); + WARN_ON(1); +} + +static void cb_fqs(struct qman_portal *p, struct qman_fq *fq, + const union qm_mr_entry *msg) +{ + u8 verb = (msg->verb & QM_MR_VERB_TYPE_MASK); + + if ((verb != QM_MR_VERB_FQRN) && (verb != QM_MR_VERB_FQRNI)) { + pr_crit("unexpected FQS message"); + WARN_ON(1); + return; + } + pr_info("Retirement message received\n"); + retire_complete = 1; + wake_up(&waitqueue); +} diff --git a/drivers/soc/fsl/qbman/qman_test_stash.c b/drivers/soc/fsl/qbman/qman_test_stash.c new file mode 100644 index 000000000..b7e8e5ec8 --- /dev/null +++ b/drivers/soc/fsl/qbman/qman_test_stash.c @@ -0,0 +1,629 @@ +/* Copyright 2009 - 2016 Freescale Semiconductor, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Freescale Semiconductor nor the + * names of its contributors may be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * ALTERNATIVELY, this software may be distributed under the terms of the + * GNU General Public License ("GPL") as published by the Free Software + * Foundation, either version 2 of that License or (at your option) any + * later version. + * + * THIS SOFTWARE IS PROVIDED BY Freescale Semiconductor ``AS IS'' AND ANY + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + * DISCLAIMED. IN NO EVENT SHALL Freescale Semiconductor BE LIABLE FOR ANY + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "qman_test.h" + +#include <linux/dma-mapping.h> +#include <linux/delay.h> + +/* + * Algorithm: + * + * Each cpu will have HP_PER_CPU "handlers" set up, each of which incorporates + * an rx/tx pair of FQ objects (both of which are stashed on dequeue). The + * organisation of FQIDs is such that the HP_PER_CPU*NUM_CPUS handlers will + * shuttle a "hot potato" frame around them such that every forwarding action + * moves it from one cpu to another. (The use of more than one handler per cpu + * is to allow enough handlers/FQs to truly test the significance of caching - + * ie. when cache-expiries are occurring.) + * + * The "hot potato" frame content will be HP_NUM_WORDS*4 bytes in size, and the + * first and last words of the frame data will undergo a transformation step on + * each forwarding action. To achieve this, each handler will be assigned a + * 32-bit "mixer", that is produced using a 32-bit LFSR. When a frame is + * received by a handler, the mixer of the expected sender is XOR'd into all + * words of the entire frame, which is then validated against the original + * values. Then, before forwarding, the entire frame is XOR'd with the mixer of + * the current handler. Apart from validating that the frame is taking the + * expected path, this also provides some quasi-realistic overheads to each + * forwarding action - dereferencing *all* the frame data, computation, and + * conditional branching. There is a "special" handler designated to act as the + * instigator of the test by creating an enqueuing the "hot potato" frame, and + * to determine when the test has completed by counting HP_LOOPS iterations. + * + * Init phases: + * + * 1. prepare each cpu's 'hp_cpu' struct using on_each_cpu(,,1) and link them + * into 'hp_cpu_list'. Specifically, set processor_id, allocate HP_PER_CPU + * handlers and link-list them (but do no other handler setup). + * + * 2. scan over 'hp_cpu_list' HP_PER_CPU times, the first time sets each + * hp_cpu's 'iterator' to point to its first handler. With each loop, + * allocate rx/tx FQIDs and mixer values to the hp_cpu's iterator handler + * and advance the iterator for the next loop. This includes a final fixup, + * which connects the last handler to the first (and which is why phase 2 + * and 3 are separate). + * + * 3. scan over 'hp_cpu_list' HP_PER_CPU times, the first time sets each + * hp_cpu's 'iterator' to point to its first handler. With each loop, + * initialise FQ objects and advance the iterator for the next loop. + * Moreover, do this initialisation on the cpu it applies to so that Rx FQ + * initialisation targets the correct cpu. + */ + +/* + * helper to run something on all cpus (can't use on_each_cpu(), as that invokes + * the fn from irq context, which is too restrictive). + */ +struct bstrap { + int (*fn)(void); + atomic_t started; +}; +static int bstrap_fn(void *bs) +{ + struct bstrap *bstrap = bs; + int err; + + atomic_inc(&bstrap->started); + err = bstrap->fn(); + if (err) + return err; + while (!kthread_should_stop()) + msleep(20); + return 0; +} +static int on_all_cpus(int (*fn)(void)) +{ + int cpu; + + for_each_cpu(cpu, cpu_online_mask) { + struct bstrap bstrap = { + .fn = fn, + .started = ATOMIC_INIT(0) + }; + struct task_struct *k = kthread_create(bstrap_fn, &bstrap, + "hotpotato%d", cpu); + int ret; + + if (IS_ERR(k)) + return -ENOMEM; + kthread_bind(k, cpu); + wake_up_process(k); + /* + * If we call kthread_stop() before the "wake up" has had an + * effect, then the thread may exit with -EINTR without ever + * running the function. So poll until it's started before + * requesting it to stop. + */ + while (!atomic_read(&bstrap.started)) + msleep(20); + ret = kthread_stop(k); + if (ret) + return ret; + } + return 0; +} + +struct hp_handler { + + /* The following data is stashed when 'rx' is dequeued; */ + /* -------------- */ + /* The Rx FQ, dequeues of which will stash the entire hp_handler */ + struct qman_fq rx; + /* The Tx FQ we should forward to */ + struct qman_fq tx; + /* The value we XOR post-dequeue, prior to validating */ + u32 rx_mixer; + /* The value we XOR pre-enqueue, after validating */ + u32 tx_mixer; + /* what the hotpotato address should be on dequeue */ + dma_addr_t addr; + u32 *frame_ptr; + + /* The following data isn't (necessarily) stashed on dequeue; */ + /* -------------- */ + u32 fqid_rx, fqid_tx; + /* list node for linking us into 'hp_cpu' */ + struct list_head node; + /* Just to check ... */ + unsigned int processor_id; +} ____cacheline_aligned; + +struct hp_cpu { + /* identify the cpu we run on; */ + unsigned int processor_id; + /* root node for the per-cpu list of handlers */ + struct list_head handlers; + /* list node for linking us into 'hp_cpu_list' */ + struct list_head node; + /* + * when repeatedly scanning 'hp_list', each time linking the n'th + * handlers together, this is used as per-cpu iterator state + */ + struct hp_handler *iterator; +}; + +/* Each cpu has one of these */ +static DEFINE_PER_CPU(struct hp_cpu, hp_cpus); + +/* links together the hp_cpu structs, in first-come first-serve order. */ +static LIST_HEAD(hp_cpu_list); +static DEFINE_SPINLOCK(hp_lock); + +static unsigned int hp_cpu_list_length; + +/* the "special" handler, that starts and terminates the test. */ +static struct hp_handler *special_handler; +static int loop_counter; + +/* handlers are allocated out of this, so they're properly aligned. */ +static struct kmem_cache *hp_handler_slab; + +/* this is the frame data */ +static void *__frame_ptr; +static u32 *frame_ptr; +static dma_addr_t frame_dma; + +/* needed for dma_map*() */ +static const struct qm_portal_config *pcfg; + +/* the main function waits on this */ +static DECLARE_WAIT_QUEUE_HEAD(queue); + +#define HP_PER_CPU 2 +#define HP_LOOPS 8 +/* 80 bytes, like a small ethernet frame, and bleeds into a second cacheline */ +#define HP_NUM_WORDS 80 +/* First word of the LFSR-based frame data */ +#define HP_FIRST_WORD 0xabbaf00d + +static inline u32 do_lfsr(u32 prev) +{ + return (prev >> 1) ^ (-(prev & 1u) & 0xd0000001u); +} + +static int allocate_frame_data(void) +{ + u32 lfsr = HP_FIRST_WORD; + int loop; + + if (!qman_dma_portal) { + pr_crit("portal not available\n"); + return -EIO; + } + + pcfg = qman_get_qm_portal_config(qman_dma_portal); + + __frame_ptr = kmalloc(4 * HP_NUM_WORDS, GFP_KERNEL); + if (!__frame_ptr) + return -ENOMEM; + + frame_ptr = PTR_ALIGN(__frame_ptr, 64); + for (loop = 0; loop < HP_NUM_WORDS; loop++) { + frame_ptr[loop] = lfsr; + lfsr = do_lfsr(lfsr); + } + + frame_dma = dma_map_single(pcfg->dev, frame_ptr, 4 * HP_NUM_WORDS, + DMA_BIDIRECTIONAL); + if (dma_mapping_error(pcfg->dev, frame_dma)) { + pr_crit("dma mapping failure\n"); + kfree(__frame_ptr); + return -EIO; + } + + return 0; +} + +static void deallocate_frame_data(void) +{ + dma_unmap_single(pcfg->dev, frame_dma, 4 * HP_NUM_WORDS, + DMA_BIDIRECTIONAL); + kfree(__frame_ptr); +} + +static inline int process_frame_data(struct hp_handler *handler, + const struct qm_fd *fd) +{ + u32 *p = handler->frame_ptr; + u32 lfsr = HP_FIRST_WORD; + int loop; + + if (qm_fd_addr_get64(fd) != handler->addr) { + pr_crit("bad frame address, [%llX != %llX]\n", + qm_fd_addr_get64(fd), handler->addr); + return -EIO; + } + for (loop = 0; loop < HP_NUM_WORDS; loop++, p++) { + *p ^= handler->rx_mixer; + if (*p != lfsr) { + pr_crit("corrupt frame data"); + return -EIO; + } + *p ^= handler->tx_mixer; + lfsr = do_lfsr(lfsr); + } + return 0; +} + +static enum qman_cb_dqrr_result normal_dqrr(struct qman_portal *portal, + struct qman_fq *fq, + const struct qm_dqrr_entry *dqrr, + bool sched_napi) +{ + struct hp_handler *handler = (struct hp_handler *)fq; + + if (process_frame_data(handler, &dqrr->fd)) { + WARN_ON(1); + goto skip; + } + if (qman_enqueue(&handler->tx, &dqrr->fd)) { + pr_crit("qman_enqueue() failed"); + WARN_ON(1); + } +skip: + return qman_cb_dqrr_consume; +} + +static enum qman_cb_dqrr_result special_dqrr(struct qman_portal *portal, + struct qman_fq *fq, + const struct qm_dqrr_entry *dqrr, + bool sched_napi) +{ + struct hp_handler *handler = (struct hp_handler *)fq; + + process_frame_data(handler, &dqrr->fd); + if (++loop_counter < HP_LOOPS) { + if (qman_enqueue(&handler->tx, &dqrr->fd)) { + pr_crit("qman_enqueue() failed"); + WARN_ON(1); + goto skip; + } + } else { + pr_info("Received final (%dth) frame\n", loop_counter); + wake_up(&queue); + } +skip: + return qman_cb_dqrr_consume; +} + +static int create_per_cpu_handlers(void) +{ + struct hp_handler *handler; + int loop; + struct hp_cpu *hp_cpu = this_cpu_ptr(&hp_cpus); + + hp_cpu->processor_id = smp_processor_id(); + spin_lock(&hp_lock); + list_add_tail(&hp_cpu->node, &hp_cpu_list); + hp_cpu_list_length++; + spin_unlock(&hp_lock); + INIT_LIST_HEAD(&hp_cpu->handlers); + for (loop = 0; loop < HP_PER_CPU; loop++) { + handler = kmem_cache_alloc(hp_handler_slab, GFP_KERNEL); + if (!handler) { + pr_crit("kmem_cache_alloc() failed"); + WARN_ON(1); + return -EIO; + } + handler->processor_id = hp_cpu->processor_id; + handler->addr = frame_dma; + handler->frame_ptr = frame_ptr; + list_add_tail(&handler->node, &hp_cpu->handlers); + } + return 0; +} + +static int destroy_per_cpu_handlers(void) +{ + struct list_head *loop, *tmp; + struct hp_cpu *hp_cpu = this_cpu_ptr(&hp_cpus); + + spin_lock(&hp_lock); + list_del(&hp_cpu->node); + spin_unlock(&hp_lock); + list_for_each_safe(loop, tmp, &hp_cpu->handlers) { + u32 flags = 0; + struct hp_handler *handler = list_entry(loop, struct hp_handler, + node); + if (qman_retire_fq(&handler->rx, &flags) || + (flags & QMAN_FQ_STATE_BLOCKOOS)) { + pr_crit("qman_retire_fq(rx) failed, flags: %x", flags); + WARN_ON(1); + return -EIO; + } + if (qman_oos_fq(&handler->rx)) { + pr_crit("qman_oos_fq(rx) failed"); + WARN_ON(1); + return -EIO; + } + qman_destroy_fq(&handler->rx); + qman_destroy_fq(&handler->tx); + qman_release_fqid(handler->fqid_rx); + list_del(&handler->node); + kmem_cache_free(hp_handler_slab, handler); + } + return 0; +} + +static inline u8 num_cachelines(u32 offset) +{ + u8 res = (offset + (L1_CACHE_BYTES - 1)) + / (L1_CACHE_BYTES); + if (res > 3) + return 3; + return res; +} +#define STASH_DATA_CL \ + num_cachelines(HP_NUM_WORDS * 4) +#define STASH_CTX_CL \ + num_cachelines(offsetof(struct hp_handler, fqid_rx)) + +static int init_handler(void *h) +{ + struct qm_mcc_initfq opts; + struct hp_handler *handler = h; + int err; + + if (handler->processor_id != smp_processor_id()) { + err = -EIO; + goto failed; + } + /* Set up rx */ + memset(&handler->rx, 0, sizeof(handler->rx)); + if (handler == special_handler) + handler->rx.cb.dqrr = special_dqrr; + else + handler->rx.cb.dqrr = normal_dqrr; + err = qman_create_fq(handler->fqid_rx, 0, &handler->rx); + if (err) { + pr_crit("qman_create_fq(rx) failed"); + goto failed; + } + memset(&opts, 0, sizeof(opts)); + opts.we_mask = cpu_to_be16(QM_INITFQ_WE_FQCTRL | + QM_INITFQ_WE_CONTEXTA); + opts.fqd.fq_ctrl = cpu_to_be16(QM_FQCTRL_CTXASTASHING); + qm_fqd_set_stashing(&opts.fqd, 0, STASH_DATA_CL, STASH_CTX_CL); + err = qman_init_fq(&handler->rx, QMAN_INITFQ_FLAG_SCHED | + QMAN_INITFQ_FLAG_LOCAL, &opts); + if (err) { + pr_crit("qman_init_fq(rx) failed"); + goto failed; + } + /* Set up tx */ + memset(&handler->tx, 0, sizeof(handler->tx)); + err = qman_create_fq(handler->fqid_tx, QMAN_FQ_FLAG_NO_MODIFY, + &handler->tx); + if (err) { + pr_crit("qman_create_fq(tx) failed"); + goto failed; + } + + return 0; +failed: + return err; +} + +static void init_handler_cb(void *h) +{ + if (init_handler(h)) + WARN_ON(1); +} + +static int init_phase2(void) +{ + int loop; + u32 fqid = 0; + u32 lfsr = 0xdeadbeef; + struct hp_cpu *hp_cpu; + struct hp_handler *handler; + + for (loop = 0; loop < HP_PER_CPU; loop++) { + list_for_each_entry(hp_cpu, &hp_cpu_list, node) { + int err; + + if (!loop) + hp_cpu->iterator = list_first_entry( + &hp_cpu->handlers, + struct hp_handler, node); + else + hp_cpu->iterator = list_entry( + hp_cpu->iterator->node.next, + struct hp_handler, node); + /* Rx FQID is the previous handler's Tx FQID */ + hp_cpu->iterator->fqid_rx = fqid; + /* Allocate new FQID for Tx */ + err = qman_alloc_fqid(&fqid); + if (err) { + pr_crit("qman_alloc_fqid() failed"); + return err; + } + hp_cpu->iterator->fqid_tx = fqid; + /* Rx mixer is the previous handler's Tx mixer */ + hp_cpu->iterator->rx_mixer = lfsr; + /* Get new mixer for Tx */ + lfsr = do_lfsr(lfsr); + hp_cpu->iterator->tx_mixer = lfsr; + } + } + /* Fix up the first handler (fqid_rx==0, rx_mixer=0xdeadbeef) */ + hp_cpu = list_first_entry(&hp_cpu_list, struct hp_cpu, node); + handler = list_first_entry(&hp_cpu->handlers, struct hp_handler, node); + if (handler->fqid_rx != 0 || handler->rx_mixer != 0xdeadbeef) + return 1; + handler->fqid_rx = fqid; + handler->rx_mixer = lfsr; + /* and tag it as our "special" handler */ + special_handler = handler; + return 0; +} + +static int init_phase3(void) +{ + int loop, err; + struct hp_cpu *hp_cpu; + + for (loop = 0; loop < HP_PER_CPU; loop++) { + list_for_each_entry(hp_cpu, &hp_cpu_list, node) { + if (!loop) + hp_cpu->iterator = list_first_entry( + &hp_cpu->handlers, + struct hp_handler, node); + else + hp_cpu->iterator = list_entry( + hp_cpu->iterator->node.next, + struct hp_handler, node); + preempt_disable(); + if (hp_cpu->processor_id == smp_processor_id()) { + err = init_handler(hp_cpu->iterator); + if (err) + return err; + } else { + smp_call_function_single(hp_cpu->processor_id, + init_handler_cb, hp_cpu->iterator, 1); + } + preempt_enable(); + } + } + return 0; +} + +static int send_first_frame(void *ignore) +{ + u32 *p = special_handler->frame_ptr; + u32 lfsr = HP_FIRST_WORD; + int loop, err; + struct qm_fd fd; + + if (special_handler->processor_id != smp_processor_id()) { + err = -EIO; + goto failed; + } + memset(&fd, 0, sizeof(fd)); + qm_fd_addr_set64(&fd, special_handler->addr); + qm_fd_set_contig_big(&fd, HP_NUM_WORDS * 4); + for (loop = 0; loop < HP_NUM_WORDS; loop++, p++) { + if (*p != lfsr) { + err = -EIO; + pr_crit("corrupt frame data"); + goto failed; + } + *p ^= special_handler->tx_mixer; + lfsr = do_lfsr(lfsr); + } + pr_info("Sending first frame\n"); + err = qman_enqueue(&special_handler->tx, &fd); + if (err) { + pr_crit("qman_enqueue() failed"); + goto failed; + } + + return 0; +failed: + return err; +} + +static void send_first_frame_cb(void *ignore) +{ + if (send_first_frame(NULL)) + WARN_ON(1); +} + +int qman_test_stash(void) +{ + int err; + + if (cpumask_weight(cpu_online_mask) < 2) { + pr_info("%s(): skip - only 1 CPU\n", __func__); + return 0; + } + + pr_info("%s(): Starting\n", __func__); + + hp_cpu_list_length = 0; + loop_counter = 0; + hp_handler_slab = kmem_cache_create("hp_handler_slab", + sizeof(struct hp_handler), L1_CACHE_BYTES, + SLAB_HWCACHE_ALIGN, NULL); + if (!hp_handler_slab) { + err = -EIO; + pr_crit("kmem_cache_create() failed"); + goto failed; + } + + err = allocate_frame_data(); + if (err) + goto failed; + + /* Init phase 1 */ + pr_info("Creating %d handlers per cpu...\n", HP_PER_CPU); + if (on_all_cpus(create_per_cpu_handlers)) { + err = -EIO; + pr_crit("on_each_cpu() failed"); + goto failed; + } + pr_info("Number of cpus: %d, total of %d handlers\n", + hp_cpu_list_length, hp_cpu_list_length * HP_PER_CPU); + + err = init_phase2(); + if (err) + goto failed; + + err = init_phase3(); + if (err) + goto failed; + + preempt_disable(); + if (special_handler->processor_id == smp_processor_id()) { + err = send_first_frame(NULL); + if (err) + goto failed; + } else { + smp_call_function_single(special_handler->processor_id, + send_first_frame_cb, NULL, 1); + } + preempt_enable(); + + wait_event(queue, loop_counter == HP_LOOPS); + deallocate_frame_data(); + if (on_all_cpus(destroy_per_cpu_handlers)) { + err = -EIO; + pr_crit("on_each_cpu() failed"); + goto failed; + } + kmem_cache_destroy(hp_handler_slab); + pr_info("%s(): Finished\n", __func__); + + return 0; +failed: + WARN_ON(1); + return err; +} diff --git a/drivers/soc/fsl/qe/Kconfig b/drivers/soc/fsl/qe/Kconfig new file mode 100644 index 000000000..7afa796db --- /dev/null +++ b/drivers/soc/fsl/qe/Kconfig @@ -0,0 +1,45 @@ +# SPDX-License-Identifier: GPL-2.0-only +# +# QE Communication options +# + +config QUICC_ENGINE + bool "QUICC Engine (QE) framework support" + depends on OF && HAS_IOMEM + depends on PPC || ARM || ARM64 || COMPILE_TEST + select GENERIC_ALLOCATOR + select CRC32 + help + The QUICC Engine (QE) is a new generation of communications + coprocessors on Freescale embedded CPUs (akin to CPM in older chips). + Selecting this option means that you wish to build a kernel + for a machine with a QE coprocessor. + +config UCC_SLOW + bool + default y if SERIAL_QE + help + This option provides qe_lib support to UCC slow + protocols: UART, BISYNC, QMC + +config UCC_FAST + bool + default y if UCC_GETH || QE_TDM + help + This option provides qe_lib support to UCC fast + protocols: HDLC, Ethernet, ATM, transparent + +config UCC + bool + default y if UCC_FAST || UCC_SLOW + +config QE_TDM + bool + default y if FSL_UCC_HDLC + +config QE_USB + bool + depends on QUICC_ENGINE + default y if USB_FSL_QE + help + QE USB Controller support diff --git a/drivers/soc/fsl/qe/Makefile b/drivers/soc/fsl/qe/Makefile new file mode 100644 index 000000000..55a555304 --- /dev/null +++ b/drivers/soc/fsl/qe/Makefile @@ -0,0 +1,12 @@ +# SPDX-License-Identifier: GPL-2.0 +# +# Makefile for the linux ppc-specific parts of QE +# +obj-$(CONFIG_QUICC_ENGINE)+= qe.o qe_common.o qe_ic.o qe_io.o +obj-$(CONFIG_CPM) += qe_common.o +obj-$(CONFIG_UCC) += ucc.o +obj-$(CONFIG_UCC_SLOW) += ucc_slow.o +obj-$(CONFIG_UCC_FAST) += ucc_fast.o +obj-$(CONFIG_QE_TDM) += qe_tdm.o +obj-$(CONFIG_QE_USB) += usb.o +obj-$(CONFIG_QE_GPIO) += gpio.o diff --git a/drivers/soc/fsl/qe/gpio.c b/drivers/soc/fsl/qe/gpio.c new file mode 100644 index 000000000..99f7de43c --- /dev/null +++ b/drivers/soc/fsl/qe/gpio.c @@ -0,0 +1,342 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * QUICC Engine GPIOs + * + * Copyright (c) MontaVista Software, Inc. 2008. + * + * Author: Anton Vorontsov <avorontsov@ru.mvista.com> + */ + +#include <linux/kernel.h> +#include <linux/init.h> +#include <linux/spinlock.h> +#include <linux/err.h> +#include <linux/io.h> +#include <linux/of.h> +#include <linux/of_gpio.h> +#include <linux/gpio/driver.h> +/* FIXME: needed for gpio_to_chip() get rid of this */ +#include <linux/gpio.h> +#include <linux/slab.h> +#include <linux/export.h> +#include <soc/fsl/qe/qe.h> + +struct qe_gpio_chip { + struct of_mm_gpio_chip mm_gc; + spinlock_t lock; + + unsigned long pin_flags[QE_PIO_PINS]; +#define QE_PIN_REQUESTED 0 + + /* shadowed data register to clear/set bits safely */ + u32 cpdata; + + /* saved_regs used to restore dedicated functions */ + struct qe_pio_regs saved_regs; +}; + +static void qe_gpio_save_regs(struct of_mm_gpio_chip *mm_gc) +{ + struct qe_gpio_chip *qe_gc = + container_of(mm_gc, struct qe_gpio_chip, mm_gc); + struct qe_pio_regs __iomem *regs = mm_gc->regs; + + qe_gc->cpdata = ioread32be(®s->cpdata); + qe_gc->saved_regs.cpdata = qe_gc->cpdata; + qe_gc->saved_regs.cpdir1 = ioread32be(®s->cpdir1); + qe_gc->saved_regs.cpdir2 = ioread32be(®s->cpdir2); + qe_gc->saved_regs.cppar1 = ioread32be(®s->cppar1); + qe_gc->saved_regs.cppar2 = ioread32be(®s->cppar2); + qe_gc->saved_regs.cpodr = ioread32be(®s->cpodr); +} + +static int qe_gpio_get(struct gpio_chip *gc, unsigned int gpio) +{ + struct of_mm_gpio_chip *mm_gc = to_of_mm_gpio_chip(gc); + struct qe_pio_regs __iomem *regs = mm_gc->regs; + u32 pin_mask = 1 << (QE_PIO_PINS - 1 - gpio); + + return !!(ioread32be(®s->cpdata) & pin_mask); +} + +static void qe_gpio_set(struct gpio_chip *gc, unsigned int gpio, int val) +{ + struct of_mm_gpio_chip *mm_gc = to_of_mm_gpio_chip(gc); + struct qe_gpio_chip *qe_gc = gpiochip_get_data(gc); + struct qe_pio_regs __iomem *regs = mm_gc->regs; + unsigned long flags; + u32 pin_mask = 1 << (QE_PIO_PINS - 1 - gpio); + + spin_lock_irqsave(&qe_gc->lock, flags); + + if (val) + qe_gc->cpdata |= pin_mask; + else + qe_gc->cpdata &= ~pin_mask; + + iowrite32be(qe_gc->cpdata, ®s->cpdata); + + spin_unlock_irqrestore(&qe_gc->lock, flags); +} + +static void qe_gpio_set_multiple(struct gpio_chip *gc, + unsigned long *mask, unsigned long *bits) +{ + struct of_mm_gpio_chip *mm_gc = to_of_mm_gpio_chip(gc); + struct qe_gpio_chip *qe_gc = gpiochip_get_data(gc); + struct qe_pio_regs __iomem *regs = mm_gc->regs; + unsigned long flags; + int i; + + spin_lock_irqsave(&qe_gc->lock, flags); + + for (i = 0; i < gc->ngpio; i++) { + if (*mask == 0) + break; + if (__test_and_clear_bit(i, mask)) { + if (test_bit(i, bits)) + qe_gc->cpdata |= (1U << (QE_PIO_PINS - 1 - i)); + else + qe_gc->cpdata &= ~(1U << (QE_PIO_PINS - 1 - i)); + } + } + + iowrite32be(qe_gc->cpdata, ®s->cpdata); + + spin_unlock_irqrestore(&qe_gc->lock, flags); +} + +static int qe_gpio_dir_in(struct gpio_chip *gc, unsigned int gpio) +{ + struct of_mm_gpio_chip *mm_gc = to_of_mm_gpio_chip(gc); + struct qe_gpio_chip *qe_gc = gpiochip_get_data(gc); + unsigned long flags; + + spin_lock_irqsave(&qe_gc->lock, flags); + + __par_io_config_pin(mm_gc->regs, gpio, QE_PIO_DIR_IN, 0, 0, 0); + + spin_unlock_irqrestore(&qe_gc->lock, flags); + + return 0; +} + +static int qe_gpio_dir_out(struct gpio_chip *gc, unsigned int gpio, int val) +{ + struct of_mm_gpio_chip *mm_gc = to_of_mm_gpio_chip(gc); + struct qe_gpio_chip *qe_gc = gpiochip_get_data(gc); + unsigned long flags; + + qe_gpio_set(gc, gpio, val); + + spin_lock_irqsave(&qe_gc->lock, flags); + + __par_io_config_pin(mm_gc->regs, gpio, QE_PIO_DIR_OUT, 0, 0, 0); + + spin_unlock_irqrestore(&qe_gc->lock, flags); + + return 0; +} + +struct qe_pin { + /* + * The qe_gpio_chip name is unfortunate, we should change that to + * something like qe_pio_controller. Someday. + */ + struct qe_gpio_chip *controller; + int num; +}; + +/** + * qe_pin_request - Request a QE pin + * @np: device node to get a pin from + * @index: index of a pin in the device tree + * Context: non-atomic + * + * This function return qe_pin so that you could use it with the rest of + * the QE Pin Multiplexing API. + */ +struct qe_pin *qe_pin_request(struct device_node *np, int index) +{ + struct qe_pin *qe_pin; + struct gpio_chip *gc; + struct qe_gpio_chip *qe_gc; + int err; + unsigned long flags; + + qe_pin = kzalloc(sizeof(*qe_pin), GFP_KERNEL); + if (!qe_pin) { + pr_debug("%s: can't allocate memory\n", __func__); + return ERR_PTR(-ENOMEM); + } + + err = of_get_gpio(np, index); + if (err < 0) + goto err0; + gc = gpio_to_chip(err); + if (WARN_ON(!gc)) { + err = -ENODEV; + goto err0; + } + + if (!of_device_is_compatible(gc->of_node, "fsl,mpc8323-qe-pario-bank")) { + pr_debug("%s: tried to get a non-qe pin\n", __func__); + err = -EINVAL; + goto err0; + } + + qe_gc = gpiochip_get_data(gc); + + spin_lock_irqsave(&qe_gc->lock, flags); + + err -= gc->base; + if (test_and_set_bit(QE_PIN_REQUESTED, &qe_gc->pin_flags[err]) == 0) { + qe_pin->controller = qe_gc; + qe_pin->num = err; + err = 0; + } else { + err = -EBUSY; + } + + spin_unlock_irqrestore(&qe_gc->lock, flags); + + if (!err) + return qe_pin; +err0: + kfree(qe_pin); + pr_debug("%s failed with status %d\n", __func__, err); + return ERR_PTR(err); +} +EXPORT_SYMBOL(qe_pin_request); + +/** + * qe_pin_free - Free a pin + * @qe_pin: pointer to the qe_pin structure + * Context: any + * + * This function frees the qe_pin structure and makes a pin available + * for further qe_pin_request() calls. + */ +void qe_pin_free(struct qe_pin *qe_pin) +{ + struct qe_gpio_chip *qe_gc = qe_pin->controller; + unsigned long flags; + const int pin = qe_pin->num; + + spin_lock_irqsave(&qe_gc->lock, flags); + test_and_clear_bit(QE_PIN_REQUESTED, &qe_gc->pin_flags[pin]); + spin_unlock_irqrestore(&qe_gc->lock, flags); + + kfree(qe_pin); +} +EXPORT_SYMBOL(qe_pin_free); + +/** + * qe_pin_set_dedicated - Revert a pin to a dedicated peripheral function mode + * @qe_pin: pointer to the qe_pin structure + * Context: any + * + * This function resets a pin to a dedicated peripheral function that + * has been set up by the firmware. + */ +void qe_pin_set_dedicated(struct qe_pin *qe_pin) +{ + struct qe_gpio_chip *qe_gc = qe_pin->controller; + struct qe_pio_regs __iomem *regs = qe_gc->mm_gc.regs; + struct qe_pio_regs *sregs = &qe_gc->saved_regs; + int pin = qe_pin->num; + u32 mask1 = 1 << (QE_PIO_PINS - (pin + 1)); + u32 mask2 = 0x3 << (QE_PIO_PINS - (pin % (QE_PIO_PINS / 2) + 1) * 2); + bool second_reg = pin > (QE_PIO_PINS / 2) - 1; + unsigned long flags; + + spin_lock_irqsave(&qe_gc->lock, flags); + + if (second_reg) { + qe_clrsetbits_be32(®s->cpdir2, mask2, + sregs->cpdir2 & mask2); + qe_clrsetbits_be32(®s->cppar2, mask2, + sregs->cppar2 & mask2); + } else { + qe_clrsetbits_be32(®s->cpdir1, mask2, + sregs->cpdir1 & mask2); + qe_clrsetbits_be32(®s->cppar1, mask2, + sregs->cppar1 & mask2); + } + + if (sregs->cpdata & mask1) + qe_gc->cpdata |= mask1; + else + qe_gc->cpdata &= ~mask1; + + iowrite32be(qe_gc->cpdata, ®s->cpdata); + qe_clrsetbits_be32(®s->cpodr, mask1, sregs->cpodr & mask1); + + spin_unlock_irqrestore(&qe_gc->lock, flags); +} +EXPORT_SYMBOL(qe_pin_set_dedicated); + +/** + * qe_pin_set_gpio - Set a pin to the GPIO mode + * @qe_pin: pointer to the qe_pin structure + * Context: any + * + * This function sets a pin to the GPIO mode. + */ +void qe_pin_set_gpio(struct qe_pin *qe_pin) +{ + struct qe_gpio_chip *qe_gc = qe_pin->controller; + struct qe_pio_regs __iomem *regs = qe_gc->mm_gc.regs; + unsigned long flags; + + spin_lock_irqsave(&qe_gc->lock, flags); + + /* Let's make it input by default, GPIO API is able to change that. */ + __par_io_config_pin(regs, qe_pin->num, QE_PIO_DIR_IN, 0, 0, 0); + + spin_unlock_irqrestore(&qe_gc->lock, flags); +} +EXPORT_SYMBOL(qe_pin_set_gpio); + +static int __init qe_add_gpiochips(void) +{ + struct device_node *np; + + for_each_compatible_node(np, NULL, "fsl,mpc8323-qe-pario-bank") { + int ret; + struct qe_gpio_chip *qe_gc; + struct of_mm_gpio_chip *mm_gc; + struct gpio_chip *gc; + + qe_gc = kzalloc(sizeof(*qe_gc), GFP_KERNEL); + if (!qe_gc) { + ret = -ENOMEM; + goto err; + } + + spin_lock_init(&qe_gc->lock); + + mm_gc = &qe_gc->mm_gc; + gc = &mm_gc->gc; + + mm_gc->save_regs = qe_gpio_save_regs; + gc->ngpio = QE_PIO_PINS; + gc->direction_input = qe_gpio_dir_in; + gc->direction_output = qe_gpio_dir_out; + gc->get = qe_gpio_get; + gc->set = qe_gpio_set; + gc->set_multiple = qe_gpio_set_multiple; + + ret = of_mm_gpiochip_add_data(np, mm_gc, qe_gc); + if (ret) + goto err; + continue; +err: + pr_err("%pOF: registration failed with status %d\n", + np, ret); + kfree(qe_gc); + /* try others anyway */ + } + return 0; +} +arch_initcall(qe_add_gpiochips); diff --git a/drivers/soc/fsl/qe/qe.c b/drivers/soc/fsl/qe/qe.c new file mode 100644 index 000000000..b3c226eb5 --- /dev/null +++ b/drivers/soc/fsl/qe/qe.c @@ -0,0 +1,681 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Copyright (C) 2006-2010 Freescale Semiconductor, Inc. All rights reserved. + * + * Authors: Shlomi Gridish <gridish@freescale.com> + * Li Yang <leoli@freescale.com> + * Based on cpm2_common.c from Dan Malek (dmalek@jlc.net) + * + * Description: + * General Purpose functions for the global management of the + * QUICC Engine (QE). + */ +#include <linux/bitmap.h> +#include <linux/errno.h> +#include <linux/sched.h> +#include <linux/kernel.h> +#include <linux/param.h> +#include <linux/string.h> +#include <linux/spinlock.h> +#include <linux/mm.h> +#include <linux/interrupt.h> +#include <linux/module.h> +#include <linux/delay.h> +#include <linux/ioport.h> +#include <linux/iopoll.h> +#include <linux/crc32.h> +#include <linux/mod_devicetable.h> +#include <linux/of_platform.h> +#include <soc/fsl/qe/immap_qe.h> +#include <soc/fsl/qe/qe.h> + +static void qe_snums_init(void); +static int qe_sdma_init(void); + +static DEFINE_SPINLOCK(qe_lock); +DEFINE_SPINLOCK(cmxgcr_lock); +EXPORT_SYMBOL(cmxgcr_lock); + +/* We allocate this here because it is used almost exclusively for + * the communication processor devices. + */ +struct qe_immap __iomem *qe_immr; +EXPORT_SYMBOL(qe_immr); + +static u8 snums[QE_NUM_OF_SNUM]; /* Dynamically allocated SNUMs */ +static DECLARE_BITMAP(snum_state, QE_NUM_OF_SNUM); +static unsigned int qe_num_of_snum; + +static phys_addr_t qebase = -1; + +static struct device_node *qe_get_device_node(void) +{ + struct device_node *qe; + + /* + * Newer device trees have an "fsl,qe" compatible property for the QE + * node, but we still need to support older device trees. + */ + qe = of_find_compatible_node(NULL, NULL, "fsl,qe"); + if (qe) + return qe; + return of_find_node_by_type(NULL, "qe"); +} + +static phys_addr_t get_qe_base(void) +{ + struct device_node *qe; + int ret; + struct resource res; + + if (qebase != -1) + return qebase; + + qe = qe_get_device_node(); + if (!qe) + return qebase; + + ret = of_address_to_resource(qe, 0, &res); + if (!ret) + qebase = res.start; + of_node_put(qe); + + return qebase; +} + +void qe_reset(void) +{ + if (qe_immr == NULL) + qe_immr = ioremap(get_qe_base(), QE_IMMAP_SIZE); + + qe_snums_init(); + + qe_issue_cmd(QE_RESET, QE_CR_SUBBLOCK_INVALID, + QE_CR_PROTOCOL_UNSPECIFIED, 0); + + /* Reclaim the MURAM memory for our use. */ + qe_muram_init(); + + if (qe_sdma_init()) + panic("sdma init failed!"); +} + +int qe_issue_cmd(u32 cmd, u32 device, u8 mcn_protocol, u32 cmd_input) +{ + unsigned long flags; + u8 mcn_shift = 0, dev_shift = 0; + u32 val; + int ret; + + spin_lock_irqsave(&qe_lock, flags); + if (cmd == QE_RESET) { + iowrite32be((u32)(cmd | QE_CR_FLG), &qe_immr->cp.cecr); + } else { + if (cmd == QE_ASSIGN_PAGE) { + /* Here device is the SNUM, not sub-block */ + dev_shift = QE_CR_SNUM_SHIFT; + } else if (cmd == QE_ASSIGN_RISC) { + /* Here device is the SNUM, and mcnProtocol is + * e_QeCmdRiscAssignment value */ + dev_shift = QE_CR_SNUM_SHIFT; + mcn_shift = QE_CR_MCN_RISC_ASSIGN_SHIFT; + } else { + if (device == QE_CR_SUBBLOCK_USB) + mcn_shift = QE_CR_MCN_USB_SHIFT; + else + mcn_shift = QE_CR_MCN_NORMAL_SHIFT; + } + + iowrite32be(cmd_input, &qe_immr->cp.cecdr); + iowrite32be((cmd | QE_CR_FLG | ((u32)device << dev_shift) | (u32)mcn_protocol << mcn_shift), + &qe_immr->cp.cecr); + } + + /* wait for the QE_CR_FLG to clear */ + ret = readx_poll_timeout_atomic(ioread32be, &qe_immr->cp.cecr, val, + (val & QE_CR_FLG) == 0, 0, 100); + /* On timeout, ret is -ETIMEDOUT, otherwise it will be 0. */ + spin_unlock_irqrestore(&qe_lock, flags); + + return ret == 0; +} +EXPORT_SYMBOL(qe_issue_cmd); + +/* Set a baud rate generator. This needs lots of work. There are + * 16 BRGs, which can be connected to the QE channels or output + * as clocks. The BRGs are in two different block of internal + * memory mapped space. + * The BRG clock is the QE clock divided by 2. + * It was set up long ago during the initial boot phase and is + * given to us. + * Baud rate clocks are zero-based in the driver code (as that maps + * to port numbers). Documentation uses 1-based numbering. + */ +static unsigned int brg_clk = 0; + +#define CLK_GRAN (1000) +#define CLK_GRAN_LIMIT (5) + +unsigned int qe_get_brg_clk(void) +{ + struct device_node *qe; + u32 brg; + unsigned int mod; + + if (brg_clk) + return brg_clk; + + qe = qe_get_device_node(); + if (!qe) + return brg_clk; + + if (!of_property_read_u32(qe, "brg-frequency", &brg)) + brg_clk = brg; + + of_node_put(qe); + + /* round this if near to a multiple of CLK_GRAN */ + mod = brg_clk % CLK_GRAN; + if (mod) { + if (mod < CLK_GRAN_LIMIT) + brg_clk -= mod; + else if (mod > (CLK_GRAN - CLK_GRAN_LIMIT)) + brg_clk += CLK_GRAN - mod; + } + + return brg_clk; +} +EXPORT_SYMBOL(qe_get_brg_clk); + +#define PVR_VER_836x 0x8083 +#define PVR_VER_832x 0x8084 + +static bool qe_general4_errata(void) +{ +#ifdef CONFIG_PPC32 + return pvr_version_is(PVR_VER_836x) || pvr_version_is(PVR_VER_832x); +#endif + return false; +} + +/* Program the BRG to the given sampling rate and multiplier + * + * @brg: the BRG, QE_BRG1 - QE_BRG16 + * @rate: the desired sampling rate + * @multiplier: corresponds to the value programmed in GUMR_L[RDCR] or + * GUMR_L[TDCR]. E.g., if this BRG is the RX clock, and GUMR_L[RDCR]=01, + * then 'multiplier' should be 8. + */ +int qe_setbrg(enum qe_clock brg, unsigned int rate, unsigned int multiplier) +{ + u32 divisor, tempval; + u32 div16 = 0; + + if ((brg < QE_BRG1) || (brg > QE_BRG16)) + return -EINVAL; + + divisor = qe_get_brg_clk() / (rate * multiplier); + + if (divisor > QE_BRGC_DIVISOR_MAX + 1) { + div16 = QE_BRGC_DIV16; + divisor /= 16; + } + + /* Errata QE_General4, which affects some MPC832x and MPC836x SOCs, says + that the BRG divisor must be even if you're not using divide-by-16 + mode. */ + if (qe_general4_errata()) + if (!div16 && (divisor & 1) && (divisor > 3)) + divisor++; + + tempval = ((divisor - 1) << QE_BRGC_DIVISOR_SHIFT) | + QE_BRGC_ENABLE | div16; + + iowrite32be(tempval, &qe_immr->brg.brgc[brg - QE_BRG1]); + + return 0; +} +EXPORT_SYMBOL(qe_setbrg); + +/* Convert a string to a QE clock source enum + * + * This function takes a string, typically from a property in the device + * tree, and returns the corresponding "enum qe_clock" value. +*/ +enum qe_clock qe_clock_source(const char *source) +{ + unsigned int i; + + if (strcasecmp(source, "none") == 0) + return QE_CLK_NONE; + + if (strcmp(source, "tsync_pin") == 0) + return QE_TSYNC_PIN; + + if (strcmp(source, "rsync_pin") == 0) + return QE_RSYNC_PIN; + + if (strncasecmp(source, "brg", 3) == 0) { + i = simple_strtoul(source + 3, NULL, 10); + if ((i >= 1) && (i <= 16)) + return (QE_BRG1 - 1) + i; + else + return QE_CLK_DUMMY; + } + + if (strncasecmp(source, "clk", 3) == 0) { + i = simple_strtoul(source + 3, NULL, 10); + if ((i >= 1) && (i <= 24)) + return (QE_CLK1 - 1) + i; + else + return QE_CLK_DUMMY; + } + + return QE_CLK_DUMMY; +} +EXPORT_SYMBOL(qe_clock_source); + +/* Initialize SNUMs (thread serial numbers) according to + * QE Module Control chapter, SNUM table + */ +static void qe_snums_init(void) +{ + static const u8 snum_init_76[] = { + 0x04, 0x05, 0x0C, 0x0D, 0x14, 0x15, 0x1C, 0x1D, + 0x24, 0x25, 0x2C, 0x2D, 0x34, 0x35, 0x88, 0x89, + 0x98, 0x99, 0xA8, 0xA9, 0xB8, 0xB9, 0xC8, 0xC9, + 0xD8, 0xD9, 0xE8, 0xE9, 0x44, 0x45, 0x4C, 0x4D, + 0x54, 0x55, 0x5C, 0x5D, 0x64, 0x65, 0x6C, 0x6D, + 0x74, 0x75, 0x7C, 0x7D, 0x84, 0x85, 0x8C, 0x8D, + 0x94, 0x95, 0x9C, 0x9D, 0xA4, 0xA5, 0xAC, 0xAD, + 0xB4, 0xB5, 0xBC, 0xBD, 0xC4, 0xC5, 0xCC, 0xCD, + 0xD4, 0xD5, 0xDC, 0xDD, 0xE4, 0xE5, 0xEC, 0xED, + 0xF4, 0xF5, 0xFC, 0xFD, + }; + static const u8 snum_init_46[] = { + 0x04, 0x05, 0x0C, 0x0D, 0x14, 0x15, 0x1C, 0x1D, + 0x24, 0x25, 0x2C, 0x2D, 0x34, 0x35, 0x88, 0x89, + 0x98, 0x99, 0xA8, 0xA9, 0xB8, 0xB9, 0xC8, 0xC9, + 0xD8, 0xD9, 0xE8, 0xE9, 0x08, 0x09, 0x18, 0x19, + 0x28, 0x29, 0x38, 0x39, 0x48, 0x49, 0x58, 0x59, + 0x68, 0x69, 0x78, 0x79, 0x80, 0x81, + }; + struct device_node *qe; + const u8 *snum_init; + int i; + + bitmap_zero(snum_state, QE_NUM_OF_SNUM); + qe_num_of_snum = 28; /* The default number of snum for threads is 28 */ + qe = qe_get_device_node(); + if (qe) { + i = of_property_read_variable_u8_array(qe, "fsl,qe-snums", + snums, 1, QE_NUM_OF_SNUM); + if (i > 0) { + of_node_put(qe); + qe_num_of_snum = i; + return; + } + /* + * Fall back to legacy binding of using the value of + * fsl,qe-num-snums to choose one of the static arrays + * above. + */ + of_property_read_u32(qe, "fsl,qe-num-snums", &qe_num_of_snum); + of_node_put(qe); + } + + if (qe_num_of_snum == 76) { + snum_init = snum_init_76; + } else if (qe_num_of_snum == 28 || qe_num_of_snum == 46) { + snum_init = snum_init_46; + } else { + pr_err("QE: unsupported value of fsl,qe-num-snums: %u\n", qe_num_of_snum); + return; + } + memcpy(snums, snum_init, qe_num_of_snum); +} + +int qe_get_snum(void) +{ + unsigned long flags; + int snum = -EBUSY; + int i; + + spin_lock_irqsave(&qe_lock, flags); + i = find_first_zero_bit(snum_state, qe_num_of_snum); + if (i < qe_num_of_snum) { + set_bit(i, snum_state); + snum = snums[i]; + } + spin_unlock_irqrestore(&qe_lock, flags); + + return snum; +} +EXPORT_SYMBOL(qe_get_snum); + +void qe_put_snum(u8 snum) +{ + const u8 *p = memchr(snums, snum, qe_num_of_snum); + + if (p) + clear_bit(p - snums, snum_state); +} +EXPORT_SYMBOL(qe_put_snum); + +static int qe_sdma_init(void) +{ + struct sdma __iomem *sdma = &qe_immr->sdma; + static s32 sdma_buf_offset = -ENOMEM; + + /* allocate 2 internal temporary buffers (512 bytes size each) for + * the SDMA */ + if (sdma_buf_offset < 0) { + sdma_buf_offset = qe_muram_alloc(512 * 2, 4096); + if (sdma_buf_offset < 0) + return -ENOMEM; + } + + iowrite32be((u32)sdma_buf_offset & QE_SDEBCR_BA_MASK, + &sdma->sdebcr); + iowrite32be((QE_SDMR_GLB_1_MSK | (0x1 << QE_SDMR_CEN_SHIFT)), + &sdma->sdmr); + + return 0; +} + +/* The maximum number of RISCs we support */ +#define MAX_QE_RISC 4 + +/* Firmware information stored here for qe_get_firmware_info() */ +static struct qe_firmware_info qe_firmware_info; + +/* + * Set to 1 if QE firmware has been uploaded, and therefore + * qe_firmware_info contains valid data. + */ +static int qe_firmware_uploaded; + +/* + * Upload a QE microcode + * + * This function is a worker function for qe_upload_firmware(). It does + * the actual uploading of the microcode. + */ +static void qe_upload_microcode(const void *base, + const struct qe_microcode *ucode) +{ + const __be32 *code = base + be32_to_cpu(ucode->code_offset); + unsigned int i; + + if (ucode->major || ucode->minor || ucode->revision) + printk(KERN_INFO "qe-firmware: " + "uploading microcode '%s' version %u.%u.%u\n", + ucode->id, ucode->major, ucode->minor, ucode->revision); + else + printk(KERN_INFO "qe-firmware: " + "uploading microcode '%s'\n", ucode->id); + + /* Use auto-increment */ + iowrite32be(be32_to_cpu(ucode->iram_offset) | QE_IRAM_IADD_AIE | QE_IRAM_IADD_BADDR, + &qe_immr->iram.iadd); + + for (i = 0; i < be32_to_cpu(ucode->count); i++) + iowrite32be(be32_to_cpu(code[i]), &qe_immr->iram.idata); + + /* Set I-RAM Ready Register */ + iowrite32be(QE_IRAM_READY, &qe_immr->iram.iready); +} + +/* + * Upload a microcode to the I-RAM at a specific address. + * + * See Documentation/powerpc/qe_firmware.rst for information on QE microcode + * uploading. + * + * Currently, only version 1 is supported, so the 'version' field must be + * set to 1. + * + * The SOC model and revision are not validated, they are only displayed for + * informational purposes. + * + * 'calc_size' is the calculated size, in bytes, of the firmware structure and + * all of the microcode structures, minus the CRC. + * + * 'length' is the size that the structure says it is, including the CRC. + */ +int qe_upload_firmware(const struct qe_firmware *firmware) +{ + unsigned int i; + unsigned int j; + u32 crc; + size_t calc_size; + size_t length; + const struct qe_header *hdr; + + if (!firmware) { + printk(KERN_ERR "qe-firmware: invalid pointer\n"); + return -EINVAL; + } + + hdr = &firmware->header; + length = be32_to_cpu(hdr->length); + + /* Check the magic */ + if ((hdr->magic[0] != 'Q') || (hdr->magic[1] != 'E') || + (hdr->magic[2] != 'F')) { + printk(KERN_ERR "qe-firmware: not a microcode\n"); + return -EPERM; + } + + /* Check the version */ + if (hdr->version != 1) { + printk(KERN_ERR "qe-firmware: unsupported version\n"); + return -EPERM; + } + + /* Validate some of the fields */ + if ((firmware->count < 1) || (firmware->count > MAX_QE_RISC)) { + printk(KERN_ERR "qe-firmware: invalid data\n"); + return -EINVAL; + } + + /* Validate the length and check if there's a CRC */ + calc_size = struct_size(firmware, microcode, firmware->count); + + for (i = 0; i < firmware->count; i++) + /* + * For situations where the second RISC uses the same microcode + * as the first, the 'code_offset' and 'count' fields will be + * zero, so it's okay to add those. + */ + calc_size += sizeof(__be32) * + be32_to_cpu(firmware->microcode[i].count); + + /* Validate the length */ + if (length != calc_size + sizeof(__be32)) { + printk(KERN_ERR "qe-firmware: invalid length\n"); + return -EPERM; + } + + /* Validate the CRC */ + crc = be32_to_cpu(*(__be32 *)((void *)firmware + calc_size)); + if (crc != crc32(0, firmware, calc_size)) { + printk(KERN_ERR "qe-firmware: firmware CRC is invalid\n"); + return -EIO; + } + + /* + * If the microcode calls for it, split the I-RAM. + */ + if (!firmware->split) + qe_setbits_be16(&qe_immr->cp.cercr, QE_CP_CERCR_CIR); + + if (firmware->soc.model) + printk(KERN_INFO + "qe-firmware: firmware '%s' for %u V%u.%u\n", + firmware->id, be16_to_cpu(firmware->soc.model), + firmware->soc.major, firmware->soc.minor); + else + printk(KERN_INFO "qe-firmware: firmware '%s'\n", + firmware->id); + + /* + * The QE only supports one microcode per RISC, so clear out all the + * saved microcode information and put in the new. + */ + memset(&qe_firmware_info, 0, sizeof(qe_firmware_info)); + strlcpy(qe_firmware_info.id, firmware->id, sizeof(qe_firmware_info.id)); + qe_firmware_info.extended_modes = be64_to_cpu(firmware->extended_modes); + memcpy(qe_firmware_info.vtraps, firmware->vtraps, + sizeof(firmware->vtraps)); + + /* Loop through each microcode. */ + for (i = 0; i < firmware->count; i++) { + const struct qe_microcode *ucode = &firmware->microcode[i]; + + /* Upload a microcode if it's present */ + if (ucode->code_offset) + qe_upload_microcode(firmware, ucode); + + /* Program the traps for this processor */ + for (j = 0; j < 16; j++) { + u32 trap = be32_to_cpu(ucode->traps[j]); + + if (trap) + iowrite32be(trap, + &qe_immr->rsp[i].tibcr[j]); + } + + /* Enable traps */ + iowrite32be(be32_to_cpu(ucode->eccr), + &qe_immr->rsp[i].eccr); + } + + qe_firmware_uploaded = 1; + + return 0; +} +EXPORT_SYMBOL(qe_upload_firmware); + +/* + * Get info on the currently-loaded firmware + * + * This function also checks the device tree to see if the boot loader has + * uploaded a firmware already. + */ +struct qe_firmware_info *qe_get_firmware_info(void) +{ + static int initialized; + struct device_node *qe; + struct device_node *fw = NULL; + const char *sprop; + + /* + * If we haven't checked yet, and a driver hasn't uploaded a firmware + * yet, then check the device tree for information. + */ + if (qe_firmware_uploaded) + return &qe_firmware_info; + + if (initialized) + return NULL; + + initialized = 1; + + qe = qe_get_device_node(); + if (!qe) + return NULL; + + /* Find the 'firmware' child node */ + fw = of_get_child_by_name(qe, "firmware"); + of_node_put(qe); + + /* Did we find the 'firmware' node? */ + if (!fw) + return NULL; + + qe_firmware_uploaded = 1; + + /* Copy the data into qe_firmware_info*/ + sprop = of_get_property(fw, "id", NULL); + if (sprop) + strlcpy(qe_firmware_info.id, sprop, + sizeof(qe_firmware_info.id)); + + of_property_read_u64(fw, "extended-modes", + &qe_firmware_info.extended_modes); + + of_property_read_u32_array(fw, "virtual-traps", qe_firmware_info.vtraps, + ARRAY_SIZE(qe_firmware_info.vtraps)); + + of_node_put(fw); + + return &qe_firmware_info; +} +EXPORT_SYMBOL(qe_get_firmware_info); + +unsigned int qe_get_num_of_risc(void) +{ + struct device_node *qe; + unsigned int num_of_risc = 0; + + qe = qe_get_device_node(); + if (!qe) + return num_of_risc; + + of_property_read_u32(qe, "fsl,qe-num-riscs", &num_of_risc); + + of_node_put(qe); + + return num_of_risc; +} +EXPORT_SYMBOL(qe_get_num_of_risc); + +unsigned int qe_get_num_of_snums(void) +{ + return qe_num_of_snum; +} +EXPORT_SYMBOL(qe_get_num_of_snums); + +static int __init qe_init(void) +{ + struct device_node *np; + + np = of_find_compatible_node(NULL, NULL, "fsl,qe"); + if (!np) + return -ENODEV; + qe_reset(); + of_node_put(np); + return 0; +} +subsys_initcall(qe_init); + +#if defined(CONFIG_SUSPEND) && defined(CONFIG_PPC_85xx) +static int qe_resume(struct platform_device *ofdev) +{ + if (!qe_alive_during_sleep()) + qe_reset(); + return 0; +} + +static int qe_probe(struct platform_device *ofdev) +{ + return 0; +} + +static const struct of_device_id qe_ids[] = { + { .compatible = "fsl,qe", }, + { }, +}; + +static struct platform_driver qe_driver = { + .driver = { + .name = "fsl-qe", + .of_match_table = qe_ids, + }, + .probe = qe_probe, + .resume = qe_resume, +}; + +builtin_platform_driver(qe_driver); +#endif /* defined(CONFIG_SUSPEND) && defined(CONFIG_PPC_85xx) */ diff --git a/drivers/soc/fsl/qe/qe_common.c b/drivers/soc/fsl/qe/qe_common.c new file mode 100644 index 000000000..a0cb8e746 --- /dev/null +++ b/drivers/soc/fsl/qe/qe_common.c @@ -0,0 +1,251 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Common CPM code + * + * Author: Scott Wood <scottwood@freescale.com> + * + * Copyright 2007-2008,2010 Freescale Semiconductor, Inc. + * + * Some parts derived from commproc.c/cpm2_common.c, which is: + * Copyright (c) 1997 Dan error_act (dmalek@jlc.net) + * Copyright (c) 1999-2001 Dan Malek <dan@embeddedalley.com> + * Copyright (c) 2000 MontaVista Software, Inc (source@mvista.com) + * 2006 (c) MontaVista Software, Inc. + * Vitaly Bordug <vbordug@ru.mvista.com> + */ +#include <linux/genalloc.h> +#include <linux/init.h> +#include <linux/list.h> +#include <linux/of_device.h> +#include <linux/spinlock.h> +#include <linux/export.h> +#include <linux/of.h> +#include <linux/of_address.h> +#include <linux/slab.h> +#include <linux/io.h> +#include <soc/fsl/qe/qe.h> + +static struct gen_pool *muram_pool; +static DEFINE_SPINLOCK(cpm_muram_lock); +static void __iomem *muram_vbase; +static phys_addr_t muram_pbase; + +struct muram_block { + struct list_head head; + s32 start; + int size; +}; + +static LIST_HEAD(muram_block_list); + +/* max address size we deal with */ +#define OF_MAX_ADDR_CELLS 4 +#define GENPOOL_OFFSET (4096 * 8) + +int cpm_muram_init(void) +{ + struct device_node *np; + struct resource r; + __be32 zero[OF_MAX_ADDR_CELLS] = {}; + resource_size_t max = 0; + int i = 0; + int ret = 0; + + if (muram_pbase) + return 0; + + np = of_find_compatible_node(NULL, NULL, "fsl,cpm-muram-data"); + if (!np) { + /* try legacy bindings */ + np = of_find_node_by_name(NULL, "data-only"); + if (!np) { + pr_err("Cannot find CPM muram data node"); + ret = -ENODEV; + goto out_muram; + } + } + + muram_pool = gen_pool_create(0, -1); + if (!muram_pool) { + pr_err("Cannot allocate memory pool for CPM/QE muram"); + ret = -ENOMEM; + goto out_muram; + } + muram_pbase = of_translate_address(np, zero); + if (muram_pbase == (phys_addr_t)OF_BAD_ADDR) { + pr_err("Cannot translate zero through CPM muram node"); + ret = -ENODEV; + goto out_pool; + } + + while (of_address_to_resource(np, i++, &r) == 0) { + if (r.end > max) + max = r.end; + ret = gen_pool_add(muram_pool, r.start - muram_pbase + + GENPOOL_OFFSET, resource_size(&r), -1); + if (ret) { + pr_err("QE: couldn't add muram to pool!\n"); + goto out_pool; + } + } + + muram_vbase = ioremap(muram_pbase, max - muram_pbase + 1); + if (!muram_vbase) { + pr_err("Cannot map QE muram"); + ret = -ENOMEM; + goto out_pool; + } + goto out_muram; +out_pool: + gen_pool_destroy(muram_pool); +out_muram: + of_node_put(np); + return ret; +} + +/* + * cpm_muram_alloc_common - cpm_muram_alloc common code + * @size: number of bytes to allocate + * @algo: algorithm for alloc. + * @data: data for genalloc's algorithm. + * + * This function returns a non-negative offset into the muram area, or + * a negative errno on failure. + */ +static s32 cpm_muram_alloc_common(unsigned long size, + genpool_algo_t algo, void *data) +{ + struct muram_block *entry; + s32 start; + + entry = kmalloc(sizeof(*entry), GFP_ATOMIC); + if (!entry) + return -ENOMEM; + start = gen_pool_alloc_algo(muram_pool, size, algo, data); + if (!start) { + kfree(entry); + return -ENOMEM; + } + start = start - GENPOOL_OFFSET; + memset_io(cpm_muram_addr(start), 0, size); + entry->start = start; + entry->size = size; + list_add(&entry->head, &muram_block_list); + + return start; +} + +/* + * cpm_muram_alloc - allocate the requested size worth of multi-user ram + * @size: number of bytes to allocate + * @align: requested alignment, in bytes + * + * This function returns a non-negative offset into the muram area, or + * a negative errno on failure. + * Use cpm_dpram_addr() to get the virtual address of the area. + * Use cpm_muram_free() to free the allocation. + */ +s32 cpm_muram_alloc(unsigned long size, unsigned long align) +{ + s32 start; + unsigned long flags; + struct genpool_data_align muram_pool_data; + + spin_lock_irqsave(&cpm_muram_lock, flags); + muram_pool_data.align = align; + start = cpm_muram_alloc_common(size, gen_pool_first_fit_align, + &muram_pool_data); + spin_unlock_irqrestore(&cpm_muram_lock, flags); + return start; +} +EXPORT_SYMBOL(cpm_muram_alloc); + +/** + * cpm_muram_free - free a chunk of multi-user ram + * @offset: The beginning of the chunk as returned by cpm_muram_alloc(). + */ +void cpm_muram_free(s32 offset) +{ + unsigned long flags; + int size; + struct muram_block *tmp; + + if (offset < 0) + return; + + size = 0; + spin_lock_irqsave(&cpm_muram_lock, flags); + list_for_each_entry(tmp, &muram_block_list, head) { + if (tmp->start == offset) { + size = tmp->size; + list_del(&tmp->head); + kfree(tmp); + break; + } + } + gen_pool_free(muram_pool, offset + GENPOOL_OFFSET, size); + spin_unlock_irqrestore(&cpm_muram_lock, flags); +} +EXPORT_SYMBOL(cpm_muram_free); + +/* + * cpm_muram_alloc_fixed - reserve a specific region of multi-user ram + * @offset: offset of allocation start address + * @size: number of bytes to allocate + * This function returns @offset if the area was available, a negative + * errno otherwise. + * Use cpm_dpram_addr() to get the virtual address of the area. + * Use cpm_muram_free() to free the allocation. + */ +s32 cpm_muram_alloc_fixed(unsigned long offset, unsigned long size) +{ + s32 start; + unsigned long flags; + struct genpool_data_fixed muram_pool_data_fixed; + + spin_lock_irqsave(&cpm_muram_lock, flags); + muram_pool_data_fixed.offset = offset + GENPOOL_OFFSET; + start = cpm_muram_alloc_common(size, gen_pool_fixed_alloc, + &muram_pool_data_fixed); + spin_unlock_irqrestore(&cpm_muram_lock, flags); + return start; +} +EXPORT_SYMBOL(cpm_muram_alloc_fixed); + +/** + * cpm_muram_addr - turn a muram offset into a virtual address + * @offset: muram offset to convert + */ +void __iomem *cpm_muram_addr(unsigned long offset) +{ + return muram_vbase + offset; +} +EXPORT_SYMBOL(cpm_muram_addr); + +unsigned long cpm_muram_offset(const void __iomem *addr) +{ + return addr - muram_vbase; +} +EXPORT_SYMBOL(cpm_muram_offset); + +/** + * cpm_muram_dma - turn a muram virtual address into a DMA address + * @addr: virtual address from cpm_muram_addr() to convert + */ +dma_addr_t cpm_muram_dma(void __iomem *addr) +{ + return muram_pbase + (addr - muram_vbase); +} +EXPORT_SYMBOL(cpm_muram_dma); + +/* + * As cpm_muram_free, but takes the virtual address rather than the + * muram offset. + */ +void cpm_muram_free_addr(const void __iomem *addr) +{ + if (!addr) + return; + cpm_muram_free(cpm_muram_offset(addr)); +} +EXPORT_SYMBOL(cpm_muram_free_addr); diff --git a/drivers/soc/fsl/qe/qe_ic.c b/drivers/soc/fsl/qe/qe_ic.c new file mode 100644 index 000000000..bbae3d39c --- /dev/null +++ b/drivers/soc/fsl/qe/qe_ic.c @@ -0,0 +1,487 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * arch/powerpc/sysdev/qe_lib/qe_ic.c + * + * Copyright (C) 2006 Freescale Semiconductor, Inc. All rights reserved. + * + * Author: Li Yang <leoli@freescale.com> + * Based on code from Shlomi Gridish <gridish@freescale.com> + * + * QUICC ENGINE Interrupt Controller + */ + +#include <linux/of_irq.h> +#include <linux/of_address.h> +#include <linux/kernel.h> +#include <linux/init.h> +#include <linux/errno.h> +#include <linux/irq.h> +#include <linux/reboot.h> +#include <linux/slab.h> +#include <linux/stddef.h> +#include <linux/sched.h> +#include <linux/signal.h> +#include <linux/device.h> +#include <linux/spinlock.h> +#include <linux/platform_device.h> +#include <asm/irq.h> +#include <asm/io.h> +#include <soc/fsl/qe/qe.h> + +#define NR_QE_IC_INTS 64 + +/* QE IC registers offset */ +#define QEIC_CICR 0x00 +#define QEIC_CIVEC 0x04 +#define QEIC_CIPXCC 0x10 +#define QEIC_CIPYCC 0x14 +#define QEIC_CIPWCC 0x18 +#define QEIC_CIPZCC 0x1c +#define QEIC_CIMR 0x20 +#define QEIC_CRIMR 0x24 +#define QEIC_CIPRTA 0x30 +#define QEIC_CIPRTB 0x34 +#define QEIC_CHIVEC 0x60 + +struct qe_ic { + /* Control registers offset */ + __be32 __iomem *regs; + + /* The remapper for this QEIC */ + struct irq_domain *irqhost; + + /* The "linux" controller struct */ + struct irq_chip hc_irq; + + /* VIRQ numbers of QE high/low irqs */ + int virq_high; + int virq_low; +}; + +/* + * QE interrupt controller internal structure + */ +struct qe_ic_info { + /* Location of this source at the QIMR register */ + u32 mask; + + /* Mask register offset */ + u32 mask_reg; + + /* + * For grouped interrupts sources - the interrupt code as + * appears at the group priority register + */ + u8 pri_code; + + /* Group priority register offset */ + u32 pri_reg; +}; + +static DEFINE_RAW_SPINLOCK(qe_ic_lock); + +static struct qe_ic_info qe_ic_info[] = { + [1] = { + .mask = 0x00008000, + .mask_reg = QEIC_CIMR, + .pri_code = 0, + .pri_reg = QEIC_CIPWCC, + }, + [2] = { + .mask = 0x00004000, + .mask_reg = QEIC_CIMR, + .pri_code = 1, + .pri_reg = QEIC_CIPWCC, + }, + [3] = { + .mask = 0x00002000, + .mask_reg = QEIC_CIMR, + .pri_code = 2, + .pri_reg = QEIC_CIPWCC, + }, + [10] = { + .mask = 0x00000040, + .mask_reg = QEIC_CIMR, + .pri_code = 1, + .pri_reg = QEIC_CIPZCC, + }, + [11] = { + .mask = 0x00000020, + .mask_reg = QEIC_CIMR, + .pri_code = 2, + .pri_reg = QEIC_CIPZCC, + }, + [12] = { + .mask = 0x00000010, + .mask_reg = QEIC_CIMR, + .pri_code = 3, + .pri_reg = QEIC_CIPZCC, + }, + [13] = { + .mask = 0x00000008, + .mask_reg = QEIC_CIMR, + .pri_code = 4, + .pri_reg = QEIC_CIPZCC, + }, + [14] = { + .mask = 0x00000004, + .mask_reg = QEIC_CIMR, + .pri_code = 5, + .pri_reg = QEIC_CIPZCC, + }, + [15] = { + .mask = 0x00000002, + .mask_reg = QEIC_CIMR, + .pri_code = 6, + .pri_reg = QEIC_CIPZCC, + }, + [20] = { + .mask = 0x10000000, + .mask_reg = QEIC_CRIMR, + .pri_code = 3, + .pri_reg = QEIC_CIPRTA, + }, + [25] = { + .mask = 0x00800000, + .mask_reg = QEIC_CRIMR, + .pri_code = 0, + .pri_reg = QEIC_CIPRTB, + }, + [26] = { + .mask = 0x00400000, + .mask_reg = QEIC_CRIMR, + .pri_code = 1, + .pri_reg = QEIC_CIPRTB, + }, + [27] = { + .mask = 0x00200000, + .mask_reg = QEIC_CRIMR, + .pri_code = 2, + .pri_reg = QEIC_CIPRTB, + }, + [28] = { + .mask = 0x00100000, + .mask_reg = QEIC_CRIMR, + .pri_code = 3, + .pri_reg = QEIC_CIPRTB, + }, + [32] = { + .mask = 0x80000000, + .mask_reg = QEIC_CIMR, + .pri_code = 0, + .pri_reg = QEIC_CIPXCC, + }, + [33] = { + .mask = 0x40000000, + .mask_reg = QEIC_CIMR, + .pri_code = 1, + .pri_reg = QEIC_CIPXCC, + }, + [34] = { + .mask = 0x20000000, + .mask_reg = QEIC_CIMR, + .pri_code = 2, + .pri_reg = QEIC_CIPXCC, + }, + [35] = { + .mask = 0x10000000, + .mask_reg = QEIC_CIMR, + .pri_code = 3, + .pri_reg = QEIC_CIPXCC, + }, + [36] = { + .mask = 0x08000000, + .mask_reg = QEIC_CIMR, + .pri_code = 4, + .pri_reg = QEIC_CIPXCC, + }, + [40] = { + .mask = 0x00800000, + .mask_reg = QEIC_CIMR, + .pri_code = 0, + .pri_reg = QEIC_CIPYCC, + }, + [41] = { + .mask = 0x00400000, + .mask_reg = QEIC_CIMR, + .pri_code = 1, + .pri_reg = QEIC_CIPYCC, + }, + [42] = { + .mask = 0x00200000, + .mask_reg = QEIC_CIMR, + .pri_code = 2, + .pri_reg = QEIC_CIPYCC, + }, + [43] = { + .mask = 0x00100000, + .mask_reg = QEIC_CIMR, + .pri_code = 3, + .pri_reg = QEIC_CIPYCC, + }, +}; + +static inline u32 qe_ic_read(__be32 __iomem *base, unsigned int reg) +{ + return ioread32be(base + (reg >> 2)); +} + +static inline void qe_ic_write(__be32 __iomem *base, unsigned int reg, + u32 value) +{ + iowrite32be(value, base + (reg >> 2)); +} + +static inline struct qe_ic *qe_ic_from_irq(unsigned int virq) +{ + return irq_get_chip_data(virq); +} + +static inline struct qe_ic *qe_ic_from_irq_data(struct irq_data *d) +{ + return irq_data_get_irq_chip_data(d); +} + +static void qe_ic_unmask_irq(struct irq_data *d) +{ + struct qe_ic *qe_ic = qe_ic_from_irq_data(d); + unsigned int src = irqd_to_hwirq(d); + unsigned long flags; + u32 temp; + + raw_spin_lock_irqsave(&qe_ic_lock, flags); + + temp = qe_ic_read(qe_ic->regs, qe_ic_info[src].mask_reg); + qe_ic_write(qe_ic->regs, qe_ic_info[src].mask_reg, + temp | qe_ic_info[src].mask); + + raw_spin_unlock_irqrestore(&qe_ic_lock, flags); +} + +static void qe_ic_mask_irq(struct irq_data *d) +{ + struct qe_ic *qe_ic = qe_ic_from_irq_data(d); + unsigned int src = irqd_to_hwirq(d); + unsigned long flags; + u32 temp; + + raw_spin_lock_irqsave(&qe_ic_lock, flags); + + temp = qe_ic_read(qe_ic->regs, qe_ic_info[src].mask_reg); + qe_ic_write(qe_ic->regs, qe_ic_info[src].mask_reg, + temp & ~qe_ic_info[src].mask); + + /* Flush the above write before enabling interrupts; otherwise, + * spurious interrupts will sometimes happen. To be 100% sure + * that the write has reached the device before interrupts are + * enabled, the mask register would have to be read back; however, + * this is not required for correctness, only to avoid wasting + * time on a large number of spurious interrupts. In testing, + * a sync reduced the observed spurious interrupts to zero. + */ + mb(); + + raw_spin_unlock_irqrestore(&qe_ic_lock, flags); +} + +static struct irq_chip qe_ic_irq_chip = { + .name = "QEIC", + .irq_unmask = qe_ic_unmask_irq, + .irq_mask = qe_ic_mask_irq, + .irq_mask_ack = qe_ic_mask_irq, +}; + +static int qe_ic_host_match(struct irq_domain *h, struct device_node *node, + enum irq_domain_bus_token bus_token) +{ + /* Exact match, unless qe_ic node is NULL */ + struct device_node *of_node = irq_domain_get_of_node(h); + return of_node == NULL || of_node == node; +} + +static int qe_ic_host_map(struct irq_domain *h, unsigned int virq, + irq_hw_number_t hw) +{ + struct qe_ic *qe_ic = h->host_data; + struct irq_chip *chip; + + if (hw >= ARRAY_SIZE(qe_ic_info)) { + pr_err("%s: Invalid hw irq number for QEIC\n", __func__); + return -EINVAL; + } + + if (qe_ic_info[hw].mask == 0) { + printk(KERN_ERR "Can't map reserved IRQ\n"); + return -EINVAL; + } + /* Default chip */ + chip = &qe_ic->hc_irq; + + irq_set_chip_data(virq, qe_ic); + irq_set_status_flags(virq, IRQ_LEVEL); + + irq_set_chip_and_handler(virq, chip, handle_level_irq); + + return 0; +} + +static const struct irq_domain_ops qe_ic_host_ops = { + .match = qe_ic_host_match, + .map = qe_ic_host_map, + .xlate = irq_domain_xlate_onetwocell, +}; + +/* Return an interrupt vector or 0 if no interrupt is pending. */ +static unsigned int qe_ic_get_low_irq(struct qe_ic *qe_ic) +{ + int irq; + + BUG_ON(qe_ic == NULL); + + /* get the interrupt source vector. */ + irq = qe_ic_read(qe_ic->regs, QEIC_CIVEC) >> 26; + + if (irq == 0) + return 0; + + return irq_linear_revmap(qe_ic->irqhost, irq); +} + +/* Return an interrupt vector or 0 if no interrupt is pending. */ +static unsigned int qe_ic_get_high_irq(struct qe_ic *qe_ic) +{ + int irq; + + BUG_ON(qe_ic == NULL); + + /* get the interrupt source vector. */ + irq = qe_ic_read(qe_ic->regs, QEIC_CHIVEC) >> 26; + + if (irq == 0) + return 0; + + return irq_linear_revmap(qe_ic->irqhost, irq); +} + +static void qe_ic_cascade_low(struct irq_desc *desc) +{ + struct qe_ic *qe_ic = irq_desc_get_handler_data(desc); + unsigned int cascade_irq = qe_ic_get_low_irq(qe_ic); + struct irq_chip *chip = irq_desc_get_chip(desc); + + if (cascade_irq != 0) + generic_handle_irq(cascade_irq); + + if (chip->irq_eoi) + chip->irq_eoi(&desc->irq_data); +} + +static void qe_ic_cascade_high(struct irq_desc *desc) +{ + struct qe_ic *qe_ic = irq_desc_get_handler_data(desc); + unsigned int cascade_irq = qe_ic_get_high_irq(qe_ic); + struct irq_chip *chip = irq_desc_get_chip(desc); + + if (cascade_irq != 0) + generic_handle_irq(cascade_irq); + + if (chip->irq_eoi) + chip->irq_eoi(&desc->irq_data); +} + +static void qe_ic_cascade_muxed_mpic(struct irq_desc *desc) +{ + struct qe_ic *qe_ic = irq_desc_get_handler_data(desc); + unsigned int cascade_irq; + struct irq_chip *chip = irq_desc_get_chip(desc); + + cascade_irq = qe_ic_get_high_irq(qe_ic); + if (cascade_irq == 0) + cascade_irq = qe_ic_get_low_irq(qe_ic); + + if (cascade_irq != 0) + generic_handle_irq(cascade_irq); + + chip->irq_eoi(&desc->irq_data); +} + +static int qe_ic_init(struct platform_device *pdev) +{ + struct device *dev = &pdev->dev; + void (*low_handler)(struct irq_desc *desc); + void (*high_handler)(struct irq_desc *desc); + struct qe_ic *qe_ic; + struct resource *res; + struct device_node *node = pdev->dev.of_node; + + res = platform_get_resource(pdev, IORESOURCE_MEM, 0); + if (res == NULL) { + dev_err(dev, "no memory resource defined\n"); + return -ENODEV; + } + + qe_ic = devm_kzalloc(dev, sizeof(*qe_ic), GFP_KERNEL); + if (qe_ic == NULL) + return -ENOMEM; + + qe_ic->regs = devm_ioremap(dev, res->start, resource_size(res)); + if (qe_ic->regs == NULL) { + dev_err(dev, "failed to ioremap() registers\n"); + return -ENODEV; + } + + qe_ic->hc_irq = qe_ic_irq_chip; + + qe_ic->virq_high = platform_get_irq(pdev, 0); + qe_ic->virq_low = platform_get_irq(pdev, 1); + + if (qe_ic->virq_low <= 0) + return -ENODEV; + + if (qe_ic->virq_high > 0 && qe_ic->virq_high != qe_ic->virq_low) { + low_handler = qe_ic_cascade_low; + high_handler = qe_ic_cascade_high; + } else { + low_handler = qe_ic_cascade_muxed_mpic; + high_handler = NULL; + } + + qe_ic->irqhost = irq_domain_add_linear(node, NR_QE_IC_INTS, + &qe_ic_host_ops, qe_ic); + if (qe_ic->irqhost == NULL) { + dev_err(dev, "failed to add irq domain\n"); + return -ENODEV; + } + + qe_ic_write(qe_ic->regs, QEIC_CICR, 0); + + irq_set_handler_data(qe_ic->virq_low, qe_ic); + irq_set_chained_handler(qe_ic->virq_low, low_handler); + + if (high_handler) { + irq_set_handler_data(qe_ic->virq_high, qe_ic); + irq_set_chained_handler(qe_ic->virq_high, high_handler); + } + return 0; +} +static const struct of_device_id qe_ic_ids[] = { + { .compatible = "fsl,qe-ic"}, + { .type = "qeic"}, + {}, +}; + +static struct platform_driver qe_ic_driver = +{ + .driver = { + .name = "qe-ic", + .of_match_table = qe_ic_ids, + }, + .probe = qe_ic_init, +}; + +static int __init qe_ic_of_init(void) +{ + platform_driver_register(&qe_ic_driver); + return 0; +} +subsys_initcall(qe_ic_of_init); diff --git a/drivers/soc/fsl/qe/qe_io.c b/drivers/soc/fsl/qe/qe_io.c new file mode 100644 index 000000000..a5e2d0e5a --- /dev/null +++ b/drivers/soc/fsl/qe/qe_io.c @@ -0,0 +1,186 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * arch/powerpc/sysdev/qe_lib/qe_io.c + * + * QE Parallel I/O ports configuration routines + * + * Copyright 2006 Freescale Semiconductor, Inc. All rights reserved. + * + * Author: Li Yang <LeoLi@freescale.com> + * Based on code from Shlomi Gridish <gridish@freescale.com> + */ + +#include <linux/stddef.h> +#include <linux/kernel.h> +#include <linux/errno.h> +#include <linux/module.h> +#include <linux/ioport.h> + +#include <asm/io.h> +#include <soc/fsl/qe/qe.h> + +#undef DEBUG + +static struct qe_pio_regs __iomem *par_io; +static int num_par_io_ports = 0; + +int par_io_init(struct device_node *np) +{ + struct resource res; + int ret; + u32 num_ports; + + /* Map Parallel I/O ports registers */ + ret = of_address_to_resource(np, 0, &res); + if (ret) + return ret; + par_io = ioremap(res.start, resource_size(&res)); + if (!par_io) + return -ENOMEM; + + if (!of_property_read_u32(np, "num-ports", &num_ports)) + num_par_io_ports = num_ports; + + return 0; +} + +void __par_io_config_pin(struct qe_pio_regs __iomem *par_io, u8 pin, int dir, + int open_drain, int assignment, int has_irq) +{ + u32 pin_mask1bit; + u32 pin_mask2bits; + u32 new_mask2bits; + u32 tmp_val; + + /* calculate pin location for single and 2 bits information */ + pin_mask1bit = (u32) (1 << (QE_PIO_PINS - (pin + 1))); + + /* Set open drain, if required */ + tmp_val = ioread32be(&par_io->cpodr); + if (open_drain) + iowrite32be(pin_mask1bit | tmp_val, &par_io->cpodr); + else + iowrite32be(~pin_mask1bit & tmp_val, &par_io->cpodr); + + /* define direction */ + tmp_val = (pin > (QE_PIO_PINS / 2) - 1) ? + ioread32be(&par_io->cpdir2) : + ioread32be(&par_io->cpdir1); + + /* get all bits mask for 2 bit per port */ + pin_mask2bits = (u32) (0x3 << (QE_PIO_PINS - + (pin % (QE_PIO_PINS / 2) + 1) * 2)); + + /* Get the final mask we need for the right definition */ + new_mask2bits = (u32) (dir << (QE_PIO_PINS - + (pin % (QE_PIO_PINS / 2) + 1) * 2)); + + /* clear and set 2 bits mask */ + if (pin > (QE_PIO_PINS / 2) - 1) { + iowrite32be(~pin_mask2bits & tmp_val, &par_io->cpdir2); + tmp_val &= ~pin_mask2bits; + iowrite32be(new_mask2bits | tmp_val, &par_io->cpdir2); + } else { + iowrite32be(~pin_mask2bits & tmp_val, &par_io->cpdir1); + tmp_val &= ~pin_mask2bits; + iowrite32be(new_mask2bits | tmp_val, &par_io->cpdir1); + } + /* define pin assignment */ + tmp_val = (pin > (QE_PIO_PINS / 2) - 1) ? + ioread32be(&par_io->cppar2) : + ioread32be(&par_io->cppar1); + + new_mask2bits = (u32) (assignment << (QE_PIO_PINS - + (pin % (QE_PIO_PINS / 2) + 1) * 2)); + /* clear and set 2 bits mask */ + if (pin > (QE_PIO_PINS / 2) - 1) { + iowrite32be(~pin_mask2bits & tmp_val, &par_io->cppar2); + tmp_val &= ~pin_mask2bits; + iowrite32be(new_mask2bits | tmp_val, &par_io->cppar2); + } else { + iowrite32be(~pin_mask2bits & tmp_val, &par_io->cppar1); + tmp_val &= ~pin_mask2bits; + iowrite32be(new_mask2bits | tmp_val, &par_io->cppar1); + } +} +EXPORT_SYMBOL(__par_io_config_pin); + +int par_io_config_pin(u8 port, u8 pin, int dir, int open_drain, + int assignment, int has_irq) +{ + if (!par_io || port >= num_par_io_ports) + return -EINVAL; + + __par_io_config_pin(&par_io[port], pin, dir, open_drain, assignment, + has_irq); + return 0; +} +EXPORT_SYMBOL(par_io_config_pin); + +int par_io_data_set(u8 port, u8 pin, u8 val) +{ + u32 pin_mask, tmp_val; + + if (port >= num_par_io_ports) + return -EINVAL; + if (pin >= QE_PIO_PINS) + return -EINVAL; + /* calculate pin location */ + pin_mask = (u32) (1 << (QE_PIO_PINS - 1 - pin)); + + tmp_val = ioread32be(&par_io[port].cpdata); + + if (val == 0) /* clear */ + iowrite32be(~pin_mask & tmp_val, &par_io[port].cpdata); + else /* set */ + iowrite32be(pin_mask | tmp_val, &par_io[port].cpdata); + + return 0; +} +EXPORT_SYMBOL(par_io_data_set); + +int par_io_of_config(struct device_node *np) +{ + struct device_node *pio; + int pio_map_len; + const __be32 *pio_map; + + if (par_io == NULL) { + printk(KERN_ERR "par_io not initialized\n"); + return -1; + } + + pio = of_parse_phandle(np, "pio-handle", 0); + if (pio == NULL) { + printk(KERN_ERR "pio-handle not available\n"); + return -1; + } + + pio_map = of_get_property(pio, "pio-map", &pio_map_len); + if (pio_map == NULL) { + printk(KERN_ERR "pio-map is not set!\n"); + return -1; + } + pio_map_len /= sizeof(unsigned int); + if ((pio_map_len % 6) != 0) { + printk(KERN_ERR "pio-map format wrong!\n"); + return -1; + } + + while (pio_map_len > 0) { + u8 port = be32_to_cpu(pio_map[0]); + u8 pin = be32_to_cpu(pio_map[1]); + int dir = be32_to_cpu(pio_map[2]); + int open_drain = be32_to_cpu(pio_map[3]); + int assignment = be32_to_cpu(pio_map[4]); + int has_irq = be32_to_cpu(pio_map[5]); + + par_io_config_pin(port, pin, dir, open_drain, + assignment, has_irq); + pio_map += 6; + pio_map_len -= 6; + } + of_node_put(pio); + return 0; +} +EXPORT_SYMBOL(par_io_of_config); diff --git a/drivers/soc/fsl/qe/qe_tdm.c b/drivers/soc/fsl/qe/qe_tdm.c new file mode 100644 index 000000000..7d7d78d3e --- /dev/null +++ b/drivers/soc/fsl/qe/qe_tdm.c @@ -0,0 +1,219 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Copyright (C) 2015 Freescale Semiconductor, Inc. All rights reserved. + * + * Authors: Zhao Qiang <qiang.zhao@nxp.com> + * + * Description: + * QE TDM API Set - TDM specific routines implementations. + */ +#include <linux/io.h> +#include <linux/kernel.h> +#include <linux/of_address.h> +#include <linux/of_irq.h> +#include <linux/of_platform.h> +#include <soc/fsl/qe/qe_tdm.h> + +static int set_tdm_framer(const char *tdm_framer_type) +{ + if (strcmp(tdm_framer_type, "e1") == 0) + return TDM_FRAMER_E1; + else if (strcmp(tdm_framer_type, "t1") == 0) + return TDM_FRAMER_T1; + else + return -EINVAL; +} + +static void set_si_param(struct ucc_tdm *utdm, struct ucc_tdm_info *ut_info) +{ + struct si_mode_info *si_info = &ut_info->si_info; + + if (utdm->tdm_mode == TDM_INTERNAL_LOOPBACK) { + si_info->simr_crt = 1; + si_info->simr_rfsd = 0; + } +} + +int ucc_of_parse_tdm(struct device_node *np, struct ucc_tdm *utdm, + struct ucc_tdm_info *ut_info) +{ + const char *sprop; + int ret = 0; + u32 val; + + sprop = of_get_property(np, "fsl,rx-sync-clock", NULL); + if (sprop) { + ut_info->uf_info.rx_sync = qe_clock_source(sprop); + if ((ut_info->uf_info.rx_sync < QE_CLK_NONE) || + (ut_info->uf_info.rx_sync > QE_RSYNC_PIN)) { + pr_err("QE-TDM: Invalid rx-sync-clock property\n"); + return -EINVAL; + } + } else { + pr_err("QE-TDM: Invalid rx-sync-clock property\n"); + return -EINVAL; + } + + sprop = of_get_property(np, "fsl,tx-sync-clock", NULL); + if (sprop) { + ut_info->uf_info.tx_sync = qe_clock_source(sprop); + if ((ut_info->uf_info.tx_sync < QE_CLK_NONE) || + (ut_info->uf_info.tx_sync > QE_TSYNC_PIN)) { + pr_err("QE-TDM: Invalid tx-sync-clock property\n"); + return -EINVAL; + } + } else { + pr_err("QE-TDM: Invalid tx-sync-clock property\n"); + return -EINVAL; + } + + ret = of_property_read_u32_index(np, "fsl,tx-timeslot-mask", 0, &val); + if (ret) { + pr_err("QE-TDM: Invalid tx-timeslot-mask property\n"); + return -EINVAL; + } + utdm->tx_ts_mask = val; + + ret = of_property_read_u32_index(np, "fsl,rx-timeslot-mask", 0, &val); + if (ret) { + ret = -EINVAL; + pr_err("QE-TDM: Invalid rx-timeslot-mask property\n"); + return ret; + } + utdm->rx_ts_mask = val; + + ret = of_property_read_u32_index(np, "fsl,tdm-id", 0, &val); + if (ret) { + ret = -EINVAL; + pr_err("QE-TDM: No fsl,tdm-id property for this UCC\n"); + return ret; + } + utdm->tdm_port = val; + ut_info->uf_info.tdm_num = utdm->tdm_port; + + if (of_property_read_bool(np, "fsl,tdm-internal-loopback")) + utdm->tdm_mode = TDM_INTERNAL_LOOPBACK; + else + utdm->tdm_mode = TDM_NORMAL; + + sprop = of_get_property(np, "fsl,tdm-framer-type", NULL); + if (!sprop) { + ret = -EINVAL; + pr_err("QE-TDM: No tdm-framer-type property for UCC\n"); + return ret; + } + ret = set_tdm_framer(sprop); + if (ret < 0) + return -EINVAL; + utdm->tdm_framer_type = ret; + + ret = of_property_read_u32_index(np, "fsl,siram-entry-id", 0, &val); + if (ret) { + ret = -EINVAL; + pr_err("QE-TDM: No siram entry id for UCC\n"); + return ret; + } + utdm->siram_entry_id = val; + + set_si_param(utdm, ut_info); + return ret; +} +EXPORT_SYMBOL(ucc_of_parse_tdm); + +void ucc_tdm_init(struct ucc_tdm *utdm, struct ucc_tdm_info *ut_info) +{ + struct si1 __iomem *si_regs; + u16 __iomem *siram; + u16 siram_entry_valid; + u16 siram_entry_closed; + u16 ucc_num; + u8 csel; + u16 sixmr; + u16 tdm_port; + u32 siram_entry_id; + u32 mask; + int i; + + si_regs = utdm->si_regs; + siram = utdm->siram; + ucc_num = ut_info->uf_info.ucc_num; + tdm_port = utdm->tdm_port; + siram_entry_id = utdm->siram_entry_id; + + if (utdm->tdm_framer_type == TDM_FRAMER_T1) + utdm->num_of_ts = 24; + if (utdm->tdm_framer_type == TDM_FRAMER_E1) + utdm->num_of_ts = 32; + + /* set siram table */ + csel = (ucc_num < 4) ? ucc_num + 9 : ucc_num - 3; + + siram_entry_valid = SIR_CSEL(csel) | SIR_BYTE | SIR_CNT(0); + siram_entry_closed = SIR_IDLE | SIR_BYTE | SIR_CNT(0); + + for (i = 0; i < utdm->num_of_ts; i++) { + mask = 0x01 << i; + + if (utdm->tx_ts_mask & mask) + iowrite16be(siram_entry_valid, + &siram[siram_entry_id * 32 + i]); + else + iowrite16be(siram_entry_closed, + &siram[siram_entry_id * 32 + i]); + + if (utdm->rx_ts_mask & mask) + iowrite16be(siram_entry_valid, + &siram[siram_entry_id * 32 + 0x200 + i]); + else + iowrite16be(siram_entry_closed, + &siram[siram_entry_id * 32 + 0x200 + i]); + } + + qe_setbits_be16(&siram[(siram_entry_id * 32) + (utdm->num_of_ts - 1)], + SIR_LAST); + qe_setbits_be16(&siram[(siram_entry_id * 32) + 0x200 + (utdm->num_of_ts - 1)], + SIR_LAST); + + /* Set SIxMR register */ + sixmr = SIMR_SAD(siram_entry_id); + + sixmr &= ~SIMR_SDM_MASK; + + if (utdm->tdm_mode == TDM_INTERNAL_LOOPBACK) + sixmr |= SIMR_SDM_INTERNAL_LOOPBACK; + else + sixmr |= SIMR_SDM_NORMAL; + + sixmr |= SIMR_RFSD(ut_info->si_info.simr_rfsd) | + SIMR_TFSD(ut_info->si_info.simr_tfsd); + + if (ut_info->si_info.simr_crt) + sixmr |= SIMR_CRT; + if (ut_info->si_info.simr_sl) + sixmr |= SIMR_SL; + if (ut_info->si_info.simr_ce) + sixmr |= SIMR_CE; + if (ut_info->si_info.simr_fe) + sixmr |= SIMR_FE; + if (ut_info->si_info.simr_gm) + sixmr |= SIMR_GM; + + switch (tdm_port) { + case 0: + iowrite16be(sixmr, &si_regs->sixmr1[0]); + break; + case 1: + iowrite16be(sixmr, &si_regs->sixmr1[1]); + break; + case 2: + iowrite16be(sixmr, &si_regs->sixmr1[2]); + break; + case 3: + iowrite16be(sixmr, &si_regs->sixmr1[3]); + break; + default: + pr_err("QE-TDM: can not find tdm sixmr reg\n"); + break; + } +} +EXPORT_SYMBOL(ucc_tdm_init); diff --git a/drivers/soc/fsl/qe/ucc.c b/drivers/soc/fsl/qe/ucc.c new file mode 100644 index 000000000..21dbcd787 --- /dev/null +++ b/drivers/soc/fsl/qe/ucc.c @@ -0,0 +1,657 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * arch/powerpc/sysdev/qe_lib/ucc.c + * + * QE UCC API Set - UCC specific routines implementations. + * + * Copyright (C) 2006 Freescale Semiconductor, Inc. All rights reserved. + * + * Authors: Shlomi Gridish <gridish@freescale.com> + * Li Yang <leoli@freescale.com> + */ +#include <linux/kernel.h> +#include <linux/errno.h> +#include <linux/stddef.h> +#include <linux/spinlock.h> +#include <linux/export.h> + +#include <asm/io.h> +#include <soc/fsl/qe/immap_qe.h> +#include <soc/fsl/qe/qe.h> +#include <soc/fsl/qe/ucc.h> + +#define UCC_TDM_NUM 8 +#define RX_SYNC_SHIFT_BASE 30 +#define TX_SYNC_SHIFT_BASE 14 +#define RX_CLK_SHIFT_BASE 28 +#define TX_CLK_SHIFT_BASE 12 + +int ucc_set_qe_mux_mii_mng(unsigned int ucc_num) +{ + unsigned long flags; + + if (ucc_num > UCC_MAX_NUM - 1) + return -EINVAL; + + spin_lock_irqsave(&cmxgcr_lock, flags); + qe_clrsetbits_be32(&qe_immr->qmx.cmxgcr, QE_CMXGCR_MII_ENET_MNG, + ucc_num << QE_CMXGCR_MII_ENET_MNG_SHIFT); + spin_unlock_irqrestore(&cmxgcr_lock, flags); + + return 0; +} +EXPORT_SYMBOL(ucc_set_qe_mux_mii_mng); + +/* Configure the UCC to either Slow or Fast. + * + * A given UCC can be figured to support either "slow" devices (e.g. UART) + * or "fast" devices (e.g. Ethernet). + * + * 'ucc_num' is the UCC number, from 0 - 7. + * + * This function also sets the UCC_GUEMR_SET_RESERVED3 bit because that bit + * must always be set to 1. + */ +int ucc_set_type(unsigned int ucc_num, enum ucc_speed_type speed) +{ + u8 __iomem *guemr; + + /* The GUEMR register is at the same location for both slow and fast + devices, so we just use uccX.slow.guemr. */ + switch (ucc_num) { + case 0: guemr = &qe_immr->ucc1.slow.guemr; + break; + case 1: guemr = &qe_immr->ucc2.slow.guemr; + break; + case 2: guemr = &qe_immr->ucc3.slow.guemr; + break; + case 3: guemr = &qe_immr->ucc4.slow.guemr; + break; + case 4: guemr = &qe_immr->ucc5.slow.guemr; + break; + case 5: guemr = &qe_immr->ucc6.slow.guemr; + break; + case 6: guemr = &qe_immr->ucc7.slow.guemr; + break; + case 7: guemr = &qe_immr->ucc8.slow.guemr; + break; + default: + return -EINVAL; + } + + qe_clrsetbits_8(guemr, UCC_GUEMR_MODE_MASK, + UCC_GUEMR_SET_RESERVED3 | speed); + + return 0; +} + +static void get_cmxucr_reg(unsigned int ucc_num, __be32 __iomem **cmxucr, + unsigned int *reg_num, unsigned int *shift) +{ + unsigned int cmx = ((ucc_num & 1) << 1) + (ucc_num > 3); + + *reg_num = cmx + 1; + *cmxucr = &qe_immr->qmx.cmxucr[cmx]; + *shift = 16 - 8 * (ucc_num & 2); +} + +int ucc_mux_set_grant_tsa_bkpt(unsigned int ucc_num, int set, u32 mask) +{ + __be32 __iomem *cmxucr; + unsigned int reg_num; + unsigned int shift; + + /* check if the UCC number is in range. */ + if (ucc_num > UCC_MAX_NUM - 1) + return -EINVAL; + + get_cmxucr_reg(ucc_num, &cmxucr, ®_num, &shift); + + if (set) + qe_setbits_be32(cmxucr, mask << shift); + else + qe_clrbits_be32(cmxucr, mask << shift); + + return 0; +} + +int ucc_set_qe_mux_rxtx(unsigned int ucc_num, enum qe_clock clock, + enum comm_dir mode) +{ + __be32 __iomem *cmxucr; + unsigned int reg_num; + unsigned int shift; + u32 clock_bits = 0; + + /* check if the UCC number is in range. */ + if (ucc_num > UCC_MAX_NUM - 1) + return -EINVAL; + + /* The communications direction must be RX or TX */ + if (!((mode == COMM_DIR_RX) || (mode == COMM_DIR_TX))) + return -EINVAL; + + get_cmxucr_reg(ucc_num, &cmxucr, ®_num, &shift); + + switch (reg_num) { + case 1: + switch (clock) { + case QE_BRG1: clock_bits = 1; break; + case QE_BRG2: clock_bits = 2; break; + case QE_BRG7: clock_bits = 3; break; + case QE_BRG8: clock_bits = 4; break; + case QE_CLK9: clock_bits = 5; break; + case QE_CLK10: clock_bits = 6; break; + case QE_CLK11: clock_bits = 7; break; + case QE_CLK12: clock_bits = 8; break; + case QE_CLK15: clock_bits = 9; break; + case QE_CLK16: clock_bits = 10; break; + default: break; + } + break; + case 2: + switch (clock) { + case QE_BRG5: clock_bits = 1; break; + case QE_BRG6: clock_bits = 2; break; + case QE_BRG7: clock_bits = 3; break; + case QE_BRG8: clock_bits = 4; break; + case QE_CLK13: clock_bits = 5; break; + case QE_CLK14: clock_bits = 6; break; + case QE_CLK19: clock_bits = 7; break; + case QE_CLK20: clock_bits = 8; break; + case QE_CLK15: clock_bits = 9; break; + case QE_CLK16: clock_bits = 10; break; + default: break; + } + break; + case 3: + switch (clock) { + case QE_BRG9: clock_bits = 1; break; + case QE_BRG10: clock_bits = 2; break; + case QE_BRG15: clock_bits = 3; break; + case QE_BRG16: clock_bits = 4; break; + case QE_CLK3: clock_bits = 5; break; + case QE_CLK4: clock_bits = 6; break; + case QE_CLK17: clock_bits = 7; break; + case QE_CLK18: clock_bits = 8; break; + case QE_CLK7: clock_bits = 9; break; + case QE_CLK8: clock_bits = 10; break; + case QE_CLK16: clock_bits = 11; break; + default: break; + } + break; + case 4: + switch (clock) { + case QE_BRG13: clock_bits = 1; break; + case QE_BRG14: clock_bits = 2; break; + case QE_BRG15: clock_bits = 3; break; + case QE_BRG16: clock_bits = 4; break; + case QE_CLK5: clock_bits = 5; break; + case QE_CLK6: clock_bits = 6; break; + case QE_CLK21: clock_bits = 7; break; + case QE_CLK22: clock_bits = 8; break; + case QE_CLK7: clock_bits = 9; break; + case QE_CLK8: clock_bits = 10; break; + case QE_CLK16: clock_bits = 11; break; + default: break; + } + break; + default: break; + } + + /* Check for invalid combination of clock and UCC number */ + if (!clock_bits) + return -ENOENT; + + if (mode == COMM_DIR_RX) + shift += 4; + + qe_clrsetbits_be32(cmxucr, QE_CMXUCR_TX_CLK_SRC_MASK << shift, + clock_bits << shift); + + return 0; +} + +static int ucc_get_tdm_common_clk(u32 tdm_num, enum qe_clock clock) +{ + int clock_bits = -EINVAL; + + /* + * for TDM[0, 1, 2, 3], TX and RX use common + * clock source BRG3,4 and CLK1,2 + * for TDM[4, 5, 6, 7], TX and RX use common + * clock source BRG12,13 and CLK23,24 + */ + switch (tdm_num) { + case 0: + case 1: + case 2: + case 3: + switch (clock) { + case QE_BRG3: + clock_bits = 1; + break; + case QE_BRG4: + clock_bits = 2; + break; + case QE_CLK1: + clock_bits = 4; + break; + case QE_CLK2: + clock_bits = 5; + break; + default: + break; + } + break; + case 4: + case 5: + case 6: + case 7: + switch (clock) { + case QE_BRG12: + clock_bits = 1; + break; + case QE_BRG13: + clock_bits = 2; + break; + case QE_CLK23: + clock_bits = 4; + break; + case QE_CLK24: + clock_bits = 5; + break; + default: + break; + } + break; + default: + break; + } + + return clock_bits; +} + +static int ucc_get_tdm_rx_clk(u32 tdm_num, enum qe_clock clock) +{ + int clock_bits = -EINVAL; + + switch (tdm_num) { + case 0: + switch (clock) { + case QE_CLK3: + clock_bits = 6; + break; + case QE_CLK8: + clock_bits = 7; + break; + default: + break; + } + break; + case 1: + switch (clock) { + case QE_CLK5: + clock_bits = 6; + break; + case QE_CLK10: + clock_bits = 7; + break; + default: + break; + } + break; + case 2: + switch (clock) { + case QE_CLK7: + clock_bits = 6; + break; + case QE_CLK12: + clock_bits = 7; + break; + default: + break; + } + break; + case 3: + switch (clock) { + case QE_CLK9: + clock_bits = 6; + break; + case QE_CLK14: + clock_bits = 7; + break; + default: + break; + } + break; + case 4: + switch (clock) { + case QE_CLK11: + clock_bits = 6; + break; + case QE_CLK16: + clock_bits = 7; + break; + default: + break; + } + break; + case 5: + switch (clock) { + case QE_CLK13: + clock_bits = 6; + break; + case QE_CLK18: + clock_bits = 7; + break; + default: + break; + } + break; + case 6: + switch (clock) { + case QE_CLK15: + clock_bits = 6; + break; + case QE_CLK20: + clock_bits = 7; + break; + default: + break; + } + break; + case 7: + switch (clock) { + case QE_CLK17: + clock_bits = 6; + break; + case QE_CLK22: + clock_bits = 7; + break; + default: + break; + } + break; + } + + return clock_bits; +} + +static int ucc_get_tdm_tx_clk(u32 tdm_num, enum qe_clock clock) +{ + int clock_bits = -EINVAL; + + switch (tdm_num) { + case 0: + switch (clock) { + case QE_CLK4: + clock_bits = 6; + break; + case QE_CLK9: + clock_bits = 7; + break; + default: + break; + } + break; + case 1: + switch (clock) { + case QE_CLK6: + clock_bits = 6; + break; + case QE_CLK11: + clock_bits = 7; + break; + default: + break; + } + break; + case 2: + switch (clock) { + case QE_CLK8: + clock_bits = 6; + break; + case QE_CLK13: + clock_bits = 7; + break; + default: + break; + } + break; + case 3: + switch (clock) { + case QE_CLK10: + clock_bits = 6; + break; + case QE_CLK15: + clock_bits = 7; + break; + default: + break; + } + break; + case 4: + switch (clock) { + case QE_CLK12: + clock_bits = 6; + break; + case QE_CLK17: + clock_bits = 7; + break; + default: + break; + } + break; + case 5: + switch (clock) { + case QE_CLK14: + clock_bits = 6; + break; + case QE_CLK19: + clock_bits = 7; + break; + default: + break; + } + break; + case 6: + switch (clock) { + case QE_CLK16: + clock_bits = 6; + break; + case QE_CLK21: + clock_bits = 7; + break; + default: + break; + } + break; + case 7: + switch (clock) { + case QE_CLK18: + clock_bits = 6; + break; + case QE_CLK3: + clock_bits = 7; + break; + default: + break; + } + break; + } + + return clock_bits; +} + +/* tdm_num: TDM A-H port num is 0-7 */ +static int ucc_get_tdm_rxtx_clk(enum comm_dir mode, u32 tdm_num, + enum qe_clock clock) +{ + int clock_bits; + + clock_bits = ucc_get_tdm_common_clk(tdm_num, clock); + if (clock_bits > 0) + return clock_bits; + if (mode == COMM_DIR_RX) + clock_bits = ucc_get_tdm_rx_clk(tdm_num, clock); + if (mode == COMM_DIR_TX) + clock_bits = ucc_get_tdm_tx_clk(tdm_num, clock); + return clock_bits; +} + +static u32 ucc_get_tdm_clk_shift(enum comm_dir mode, u32 tdm_num) +{ + u32 shift; + + shift = (mode == COMM_DIR_RX) ? RX_CLK_SHIFT_BASE : TX_CLK_SHIFT_BASE; + if (tdm_num < 4) + shift -= tdm_num * 4; + else + shift -= (tdm_num - 4) * 4; + + return shift; +} + +int ucc_set_tdm_rxtx_clk(u32 tdm_num, enum qe_clock clock, + enum comm_dir mode) +{ + int clock_bits; + u32 shift; + struct qe_mux __iomem *qe_mux_reg; + __be32 __iomem *cmxs1cr; + + qe_mux_reg = &qe_immr->qmx; + + if (tdm_num > 7) + return -EINVAL; + + /* The communications direction must be RX or TX */ + if (mode != COMM_DIR_RX && mode != COMM_DIR_TX) + return -EINVAL; + + clock_bits = ucc_get_tdm_rxtx_clk(mode, tdm_num, clock); + if (clock_bits < 0) + return -EINVAL; + + shift = ucc_get_tdm_clk_shift(mode, tdm_num); + + cmxs1cr = (tdm_num < 4) ? &qe_mux_reg->cmxsi1cr_l : + &qe_mux_reg->cmxsi1cr_h; + + qe_clrsetbits_be32(cmxs1cr, QE_CMXUCR_TX_CLK_SRC_MASK << shift, + clock_bits << shift); + + return 0; +} + +static int ucc_get_tdm_sync_source(u32 tdm_num, enum qe_clock clock, + enum comm_dir mode) +{ + int source = -EINVAL; + + if (mode == COMM_DIR_RX && clock == QE_RSYNC_PIN) { + source = 0; + return source; + } + if (mode == COMM_DIR_TX && clock == QE_TSYNC_PIN) { + source = 0; + return source; + } + + switch (tdm_num) { + case 0: + case 1: + switch (clock) { + case QE_BRG9: + source = 1; + break; + case QE_BRG10: + source = 2; + break; + default: + break; + } + break; + case 2: + case 3: + switch (clock) { + case QE_BRG9: + source = 1; + break; + case QE_BRG11: + source = 2; + break; + default: + break; + } + break; + case 4: + case 5: + switch (clock) { + case QE_BRG13: + source = 1; + break; + case QE_BRG14: + source = 2; + break; + default: + break; + } + break; + case 6: + case 7: + switch (clock) { + case QE_BRG13: + source = 1; + break; + case QE_BRG15: + source = 2; + break; + default: + break; + } + break; + } + + return source; +} + +static u32 ucc_get_tdm_sync_shift(enum comm_dir mode, u32 tdm_num) +{ + u32 shift; + + shift = (mode == COMM_DIR_RX) ? RX_SYNC_SHIFT_BASE : TX_SYNC_SHIFT_BASE; + shift -= tdm_num * 2; + + return shift; +} + +int ucc_set_tdm_rxtx_sync(u32 tdm_num, enum qe_clock clock, + enum comm_dir mode) +{ + int source; + u32 shift; + struct qe_mux __iomem *qe_mux_reg; + + qe_mux_reg = &qe_immr->qmx; + + if (tdm_num >= UCC_TDM_NUM) + return -EINVAL; + + /* The communications direction must be RX or TX */ + if (mode != COMM_DIR_RX && mode != COMM_DIR_TX) + return -EINVAL; + + source = ucc_get_tdm_sync_source(tdm_num, clock, mode); + if (source < 0) + return -EINVAL; + + shift = ucc_get_tdm_sync_shift(mode, tdm_num); + + qe_clrsetbits_be32(&qe_mux_reg->cmxsi1syr, + QE_CMXUCR_TX_CLK_SRC_MASK << shift, + source << shift); + + return 0; +} diff --git a/drivers/soc/fsl/qe/ucc_fast.c b/drivers/soc/fsl/qe/ucc_fast.c new file mode 100644 index 000000000..53d8aafc9 --- /dev/null +++ b/drivers/soc/fsl/qe/ucc_fast.c @@ -0,0 +1,395 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Copyright (C) 2006 Freescale Semiconductor, Inc. All rights reserved. + * + * Authors: Shlomi Gridish <gridish@freescale.com> + * Li Yang <leoli@freescale.com> + * + * Description: + * QE UCC Fast API Set - UCC Fast specific routines implementations. + */ +#include <linux/kernel.h> +#include <linux/errno.h> +#include <linux/slab.h> +#include <linux/stddef.h> +#include <linux/interrupt.h> +#include <linux/err.h> +#include <linux/export.h> + +#include <asm/io.h> +#include <soc/fsl/qe/immap_qe.h> +#include <soc/fsl/qe/qe.h> + +#include <soc/fsl/qe/ucc.h> +#include <soc/fsl/qe/ucc_fast.h> + +void ucc_fast_dump_regs(struct ucc_fast_private * uccf) +{ + printk(KERN_INFO "UCC%u Fast registers:\n", uccf->uf_info->ucc_num); + printk(KERN_INFO "Base address: 0x%p\n", uccf->uf_regs); + + printk(KERN_INFO "gumr : addr=0x%p, val=0x%08x\n", + &uccf->uf_regs->gumr, ioread32be(&uccf->uf_regs->gumr)); + printk(KERN_INFO "upsmr : addr=0x%p, val=0x%08x\n", + &uccf->uf_regs->upsmr, ioread32be(&uccf->uf_regs->upsmr)); + printk(KERN_INFO "utodr : addr=0x%p, val=0x%04x\n", + &uccf->uf_regs->utodr, ioread16be(&uccf->uf_regs->utodr)); + printk(KERN_INFO "udsr : addr=0x%p, val=0x%04x\n", + &uccf->uf_regs->udsr, ioread16be(&uccf->uf_regs->udsr)); + printk(KERN_INFO "ucce : addr=0x%p, val=0x%08x\n", + &uccf->uf_regs->ucce, ioread32be(&uccf->uf_regs->ucce)); + printk(KERN_INFO "uccm : addr=0x%p, val=0x%08x\n", + &uccf->uf_regs->uccm, ioread32be(&uccf->uf_regs->uccm)); + printk(KERN_INFO "uccs : addr=0x%p, val=0x%02x\n", + &uccf->uf_regs->uccs, ioread8(&uccf->uf_regs->uccs)); + printk(KERN_INFO "urfb : addr=0x%p, val=0x%08x\n", + &uccf->uf_regs->urfb, ioread32be(&uccf->uf_regs->urfb)); + printk(KERN_INFO "urfs : addr=0x%p, val=0x%04x\n", + &uccf->uf_regs->urfs, ioread16be(&uccf->uf_regs->urfs)); + printk(KERN_INFO "urfet : addr=0x%p, val=0x%04x\n", + &uccf->uf_regs->urfet, ioread16be(&uccf->uf_regs->urfet)); + printk(KERN_INFO "urfset: addr=0x%p, val=0x%04x\n", + &uccf->uf_regs->urfset, + ioread16be(&uccf->uf_regs->urfset)); + printk(KERN_INFO "utfb : addr=0x%p, val=0x%08x\n", + &uccf->uf_regs->utfb, ioread32be(&uccf->uf_regs->utfb)); + printk(KERN_INFO "utfs : addr=0x%p, val=0x%04x\n", + &uccf->uf_regs->utfs, ioread16be(&uccf->uf_regs->utfs)); + printk(KERN_INFO "utfet : addr=0x%p, val=0x%04x\n", + &uccf->uf_regs->utfet, ioread16be(&uccf->uf_regs->utfet)); + printk(KERN_INFO "utftt : addr=0x%p, val=0x%04x\n", + &uccf->uf_regs->utftt, ioread16be(&uccf->uf_regs->utftt)); + printk(KERN_INFO "utpt : addr=0x%p, val=0x%04x\n", + &uccf->uf_regs->utpt, ioread16be(&uccf->uf_regs->utpt)); + printk(KERN_INFO "urtry : addr=0x%p, val=0x%08x\n", + &uccf->uf_regs->urtry, ioread32be(&uccf->uf_regs->urtry)); + printk(KERN_INFO "guemr : addr=0x%p, val=0x%02x\n", + &uccf->uf_regs->guemr, ioread8(&uccf->uf_regs->guemr)); +} +EXPORT_SYMBOL(ucc_fast_dump_regs); + +u32 ucc_fast_get_qe_cr_subblock(int uccf_num) +{ + switch (uccf_num) { + case 0: return QE_CR_SUBBLOCK_UCCFAST1; + case 1: return QE_CR_SUBBLOCK_UCCFAST2; + case 2: return QE_CR_SUBBLOCK_UCCFAST3; + case 3: return QE_CR_SUBBLOCK_UCCFAST4; + case 4: return QE_CR_SUBBLOCK_UCCFAST5; + case 5: return QE_CR_SUBBLOCK_UCCFAST6; + case 6: return QE_CR_SUBBLOCK_UCCFAST7; + case 7: return QE_CR_SUBBLOCK_UCCFAST8; + default: return QE_CR_SUBBLOCK_INVALID; + } +} +EXPORT_SYMBOL(ucc_fast_get_qe_cr_subblock); + +void ucc_fast_transmit_on_demand(struct ucc_fast_private * uccf) +{ + iowrite16be(UCC_FAST_TOD, &uccf->uf_regs->utodr); +} +EXPORT_SYMBOL(ucc_fast_transmit_on_demand); + +void ucc_fast_enable(struct ucc_fast_private * uccf, enum comm_dir mode) +{ + struct ucc_fast __iomem *uf_regs; + u32 gumr; + + uf_regs = uccf->uf_regs; + + /* Enable reception and/or transmission on this UCC. */ + gumr = ioread32be(&uf_regs->gumr); + if (mode & COMM_DIR_TX) { + gumr |= UCC_FAST_GUMR_ENT; + uccf->enabled_tx = 1; + } + if (mode & COMM_DIR_RX) { + gumr |= UCC_FAST_GUMR_ENR; + uccf->enabled_rx = 1; + } + iowrite32be(gumr, &uf_regs->gumr); +} +EXPORT_SYMBOL(ucc_fast_enable); + +void ucc_fast_disable(struct ucc_fast_private * uccf, enum comm_dir mode) +{ + struct ucc_fast __iomem *uf_regs; + u32 gumr; + + uf_regs = uccf->uf_regs; + + /* Disable reception and/or transmission on this UCC. */ + gumr = ioread32be(&uf_regs->gumr); + if (mode & COMM_DIR_TX) { + gumr &= ~UCC_FAST_GUMR_ENT; + uccf->enabled_tx = 0; + } + if (mode & COMM_DIR_RX) { + gumr &= ~UCC_FAST_GUMR_ENR; + uccf->enabled_rx = 0; + } + iowrite32be(gumr, &uf_regs->gumr); +} +EXPORT_SYMBOL(ucc_fast_disable); + +int ucc_fast_init(struct ucc_fast_info * uf_info, struct ucc_fast_private ** uccf_ret) +{ + struct ucc_fast_private *uccf; + struct ucc_fast __iomem *uf_regs; + u32 gumr; + int ret; + + if (!uf_info) + return -EINVAL; + + /* check if the UCC port number is in range. */ + if ((uf_info->ucc_num < 0) || (uf_info->ucc_num > UCC_MAX_NUM - 1)) { + printk(KERN_ERR "%s: illegal UCC number\n", __func__); + return -EINVAL; + } + + /* Check that 'max_rx_buf_length' is properly aligned (4). */ + if (uf_info->max_rx_buf_length & (UCC_FAST_MRBLR_ALIGNMENT - 1)) { + printk(KERN_ERR "%s: max_rx_buf_length not aligned\n", + __func__); + return -EINVAL; + } + + /* Validate Virtual Fifo register values */ + if (uf_info->urfs < UCC_FAST_URFS_MIN_VAL) { + printk(KERN_ERR "%s: urfs is too small\n", __func__); + return -EINVAL; + } + + if (uf_info->urfs & (UCC_FAST_VIRT_FIFO_REGS_ALIGNMENT - 1)) { + printk(KERN_ERR "%s: urfs is not aligned\n", __func__); + return -EINVAL; + } + + if (uf_info->urfet & (UCC_FAST_VIRT_FIFO_REGS_ALIGNMENT - 1)) { + printk(KERN_ERR "%s: urfet is not aligned.\n", __func__); + return -EINVAL; + } + + if (uf_info->urfset & (UCC_FAST_VIRT_FIFO_REGS_ALIGNMENT - 1)) { + printk(KERN_ERR "%s: urfset is not aligned\n", __func__); + return -EINVAL; + } + + if (uf_info->utfs & (UCC_FAST_VIRT_FIFO_REGS_ALIGNMENT - 1)) { + printk(KERN_ERR "%s: utfs is not aligned\n", __func__); + return -EINVAL; + } + + if (uf_info->utfet & (UCC_FAST_VIRT_FIFO_REGS_ALIGNMENT - 1)) { + printk(KERN_ERR "%s: utfet is not aligned\n", __func__); + return -EINVAL; + } + + if (uf_info->utftt & (UCC_FAST_VIRT_FIFO_REGS_ALIGNMENT - 1)) { + printk(KERN_ERR "%s: utftt is not aligned\n", __func__); + return -EINVAL; + } + + uccf = kzalloc(sizeof(struct ucc_fast_private), GFP_KERNEL); + if (!uccf) { + printk(KERN_ERR "%s: Cannot allocate private data\n", + __func__); + return -ENOMEM; + } + uccf->ucc_fast_tx_virtual_fifo_base_offset = -1; + uccf->ucc_fast_rx_virtual_fifo_base_offset = -1; + + /* Fill fast UCC structure */ + uccf->uf_info = uf_info; + /* Set the PHY base address */ + uccf->uf_regs = ioremap(uf_info->regs, sizeof(struct ucc_fast)); + if (uccf->uf_regs == NULL) { + printk(KERN_ERR "%s: Cannot map UCC registers\n", __func__); + kfree(uccf); + return -ENOMEM; + } + + uccf->enabled_tx = 0; + uccf->enabled_rx = 0; + uccf->stopped_tx = 0; + uccf->stopped_rx = 0; + uf_regs = uccf->uf_regs; + uccf->p_ucce = &uf_regs->ucce; + uccf->p_uccm = &uf_regs->uccm; +#ifdef CONFIG_UGETH_TX_ON_DEMAND + uccf->p_utodr = &uf_regs->utodr; +#endif +#ifdef STATISTICS + uccf->tx_frames = 0; + uccf->rx_frames = 0; + uccf->rx_discarded = 0; +#endif /* STATISTICS */ + + /* Set UCC to fast type */ + ret = ucc_set_type(uf_info->ucc_num, UCC_SPEED_TYPE_FAST); + if (ret) { + printk(KERN_ERR "%s: cannot set UCC type\n", __func__); + ucc_fast_free(uccf); + return ret; + } + + uccf->mrblr = uf_info->max_rx_buf_length; + + /* Set GUMR */ + /* For more details see the hardware spec. */ + gumr = uf_info->ttx_trx; + if (uf_info->tci) + gumr |= UCC_FAST_GUMR_TCI; + if (uf_info->cdp) + gumr |= UCC_FAST_GUMR_CDP; + if (uf_info->ctsp) + gumr |= UCC_FAST_GUMR_CTSP; + if (uf_info->cds) + gumr |= UCC_FAST_GUMR_CDS; + if (uf_info->ctss) + gumr |= UCC_FAST_GUMR_CTSS; + if (uf_info->txsy) + gumr |= UCC_FAST_GUMR_TXSY; + if (uf_info->rsyn) + gumr |= UCC_FAST_GUMR_RSYN; + gumr |= uf_info->synl; + if (uf_info->rtsm) + gumr |= UCC_FAST_GUMR_RTSM; + gumr |= uf_info->renc; + if (uf_info->revd) + gumr |= UCC_FAST_GUMR_REVD; + gumr |= uf_info->tenc; + gumr |= uf_info->tcrc; + gumr |= uf_info->mode; + iowrite32be(gumr, &uf_regs->gumr); + + /* Allocate memory for Tx Virtual Fifo */ + uccf->ucc_fast_tx_virtual_fifo_base_offset = + qe_muram_alloc(uf_info->utfs, UCC_FAST_VIRT_FIFO_REGS_ALIGNMENT); + if (uccf->ucc_fast_tx_virtual_fifo_base_offset < 0) { + printk(KERN_ERR "%s: cannot allocate MURAM for TX FIFO\n", + __func__); + ucc_fast_free(uccf); + return -ENOMEM; + } + + /* Allocate memory for Rx Virtual Fifo */ + uccf->ucc_fast_rx_virtual_fifo_base_offset = + qe_muram_alloc(uf_info->urfs + + UCC_FAST_RECEIVE_VIRTUAL_FIFO_SIZE_FUDGE_FACTOR, + UCC_FAST_VIRT_FIFO_REGS_ALIGNMENT); + if (uccf->ucc_fast_rx_virtual_fifo_base_offset < 0) { + printk(KERN_ERR "%s: cannot allocate MURAM for RX FIFO\n", + __func__); + ucc_fast_free(uccf); + return -ENOMEM; + } + + /* Set Virtual Fifo registers */ + iowrite16be(uf_info->urfs, &uf_regs->urfs); + iowrite16be(uf_info->urfet, &uf_regs->urfet); + iowrite16be(uf_info->urfset, &uf_regs->urfset); + iowrite16be(uf_info->utfs, &uf_regs->utfs); + iowrite16be(uf_info->utfet, &uf_regs->utfet); + iowrite16be(uf_info->utftt, &uf_regs->utftt); + /* utfb, urfb are offsets from MURAM base */ + iowrite32be(uccf->ucc_fast_tx_virtual_fifo_base_offset, + &uf_regs->utfb); + iowrite32be(uccf->ucc_fast_rx_virtual_fifo_base_offset, + &uf_regs->urfb); + + /* Mux clocking */ + /* Grant Support */ + ucc_set_qe_mux_grant(uf_info->ucc_num, uf_info->grant_support); + /* Breakpoint Support */ + ucc_set_qe_mux_bkpt(uf_info->ucc_num, uf_info->brkpt_support); + /* Set Tsa or NMSI mode. */ + ucc_set_qe_mux_tsa(uf_info->ucc_num, uf_info->tsa); + /* If NMSI (not Tsa), set Tx and Rx clock. */ + if (!uf_info->tsa) { + /* Rx clock routing */ + if ((uf_info->rx_clock != QE_CLK_NONE) && + ucc_set_qe_mux_rxtx(uf_info->ucc_num, uf_info->rx_clock, + COMM_DIR_RX)) { + printk(KERN_ERR "%s: illegal value for RX clock\n", + __func__); + ucc_fast_free(uccf); + return -EINVAL; + } + /* Tx clock routing */ + if ((uf_info->tx_clock != QE_CLK_NONE) && + ucc_set_qe_mux_rxtx(uf_info->ucc_num, uf_info->tx_clock, + COMM_DIR_TX)) { + printk(KERN_ERR "%s: illegal value for TX clock\n", + __func__); + ucc_fast_free(uccf); + return -EINVAL; + } + } else { + /* tdm Rx clock routing */ + if ((uf_info->rx_clock != QE_CLK_NONE) && + ucc_set_tdm_rxtx_clk(uf_info->tdm_num, uf_info->rx_clock, + COMM_DIR_RX)) { + pr_err("%s: illegal value for RX clock", __func__); + ucc_fast_free(uccf); + return -EINVAL; + } + + /* tdm Tx clock routing */ + if ((uf_info->tx_clock != QE_CLK_NONE) && + ucc_set_tdm_rxtx_clk(uf_info->tdm_num, uf_info->tx_clock, + COMM_DIR_TX)) { + pr_err("%s: illegal value for TX clock", __func__); + ucc_fast_free(uccf); + return -EINVAL; + } + + /* tdm Rx sync clock routing */ + if ((uf_info->rx_sync != QE_CLK_NONE) && + ucc_set_tdm_rxtx_sync(uf_info->tdm_num, uf_info->rx_sync, + COMM_DIR_RX)) { + pr_err("%s: illegal value for RX clock", __func__); + ucc_fast_free(uccf); + return -EINVAL; + } + + /* tdm Tx sync clock routing */ + if ((uf_info->tx_sync != QE_CLK_NONE) && + ucc_set_tdm_rxtx_sync(uf_info->tdm_num, uf_info->tx_sync, + COMM_DIR_TX)) { + pr_err("%s: illegal value for TX clock", __func__); + ucc_fast_free(uccf); + return -EINVAL; + } + } + + /* Set interrupt mask register at UCC level. */ + iowrite32be(uf_info->uccm_mask, &uf_regs->uccm); + + /* First, clear anything pending at UCC level, + * otherwise, old garbage may come through + * as soon as the dam is opened. */ + + /* Writing '1' clears */ + iowrite32be(0xffffffff, &uf_regs->ucce); + + *uccf_ret = uccf; + return 0; +} +EXPORT_SYMBOL(ucc_fast_init); + +void ucc_fast_free(struct ucc_fast_private * uccf) +{ + if (!uccf) + return; + + qe_muram_free(uccf->ucc_fast_tx_virtual_fifo_base_offset); + qe_muram_free(uccf->ucc_fast_rx_virtual_fifo_base_offset); + + if (uccf->uf_regs) + iounmap(uccf->uf_regs); + + kfree(uccf); +} +EXPORT_SYMBOL(ucc_fast_free); diff --git a/drivers/soc/fsl/qe/ucc_slow.c b/drivers/soc/fsl/qe/ucc_slow.c new file mode 100644 index 000000000..d5ac1ac0e --- /dev/null +++ b/drivers/soc/fsl/qe/ucc_slow.c @@ -0,0 +1,359 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Copyright (C) 2006 Freescale Semiconductor, Inc. All rights reserved. + * + * Authors: Shlomi Gridish <gridish@freescale.com> + * Li Yang <leoli@freescale.com> + * + * Description: + * QE UCC Slow API Set - UCC Slow specific routines implementations. + */ +#include <linux/kernel.h> +#include <linux/errno.h> +#include <linux/slab.h> +#include <linux/stddef.h> +#include <linux/interrupt.h> +#include <linux/err.h> +#include <linux/export.h> + +#include <asm/io.h> +#include <soc/fsl/qe/immap_qe.h> +#include <soc/fsl/qe/qe.h> + +#include <soc/fsl/qe/ucc.h> +#include <soc/fsl/qe/ucc_slow.h> + +u32 ucc_slow_get_qe_cr_subblock(int uccs_num) +{ + switch (uccs_num) { + case 0: return QE_CR_SUBBLOCK_UCCSLOW1; + case 1: return QE_CR_SUBBLOCK_UCCSLOW2; + case 2: return QE_CR_SUBBLOCK_UCCSLOW3; + case 3: return QE_CR_SUBBLOCK_UCCSLOW4; + case 4: return QE_CR_SUBBLOCK_UCCSLOW5; + case 5: return QE_CR_SUBBLOCK_UCCSLOW6; + case 6: return QE_CR_SUBBLOCK_UCCSLOW7; + case 7: return QE_CR_SUBBLOCK_UCCSLOW8; + default: return QE_CR_SUBBLOCK_INVALID; + } +} +EXPORT_SYMBOL(ucc_slow_get_qe_cr_subblock); + +void ucc_slow_graceful_stop_tx(struct ucc_slow_private * uccs) +{ + struct ucc_slow_info *us_info = uccs->us_info; + u32 id; + + id = ucc_slow_get_qe_cr_subblock(us_info->ucc_num); + qe_issue_cmd(QE_GRACEFUL_STOP_TX, id, + QE_CR_PROTOCOL_UNSPECIFIED, 0); +} +EXPORT_SYMBOL(ucc_slow_graceful_stop_tx); + +void ucc_slow_stop_tx(struct ucc_slow_private * uccs) +{ + struct ucc_slow_info *us_info = uccs->us_info; + u32 id; + + id = ucc_slow_get_qe_cr_subblock(us_info->ucc_num); + qe_issue_cmd(QE_STOP_TX, id, QE_CR_PROTOCOL_UNSPECIFIED, 0); +} +EXPORT_SYMBOL(ucc_slow_stop_tx); + +void ucc_slow_restart_tx(struct ucc_slow_private * uccs) +{ + struct ucc_slow_info *us_info = uccs->us_info; + u32 id; + + id = ucc_slow_get_qe_cr_subblock(us_info->ucc_num); + qe_issue_cmd(QE_RESTART_TX, id, QE_CR_PROTOCOL_UNSPECIFIED, 0); +} +EXPORT_SYMBOL(ucc_slow_restart_tx); + +void ucc_slow_enable(struct ucc_slow_private * uccs, enum comm_dir mode) +{ + struct ucc_slow __iomem *us_regs; + u32 gumr_l; + + us_regs = uccs->us_regs; + + /* Enable reception and/or transmission on this UCC. */ + gumr_l = ioread32be(&us_regs->gumr_l); + if (mode & COMM_DIR_TX) { + gumr_l |= UCC_SLOW_GUMR_L_ENT; + uccs->enabled_tx = 1; + } + if (mode & COMM_DIR_RX) { + gumr_l |= UCC_SLOW_GUMR_L_ENR; + uccs->enabled_rx = 1; + } + iowrite32be(gumr_l, &us_regs->gumr_l); +} +EXPORT_SYMBOL(ucc_slow_enable); + +void ucc_slow_disable(struct ucc_slow_private * uccs, enum comm_dir mode) +{ + struct ucc_slow __iomem *us_regs; + u32 gumr_l; + + us_regs = uccs->us_regs; + + /* Disable reception and/or transmission on this UCC. */ + gumr_l = ioread32be(&us_regs->gumr_l); + if (mode & COMM_DIR_TX) { + gumr_l &= ~UCC_SLOW_GUMR_L_ENT; + uccs->enabled_tx = 0; + } + if (mode & COMM_DIR_RX) { + gumr_l &= ~UCC_SLOW_GUMR_L_ENR; + uccs->enabled_rx = 0; + } + iowrite32be(gumr_l, &us_regs->gumr_l); +} +EXPORT_SYMBOL(ucc_slow_disable); + +/* Initialize the UCC for Slow operations + * + * The caller should initialize the following us_info + */ +int ucc_slow_init(struct ucc_slow_info * us_info, struct ucc_slow_private ** uccs_ret) +{ + struct ucc_slow_private *uccs; + u32 i; + struct ucc_slow __iomem *us_regs; + u32 gumr; + struct qe_bd __iomem *bd; + u32 id; + u32 command; + int ret = 0; + + if (!us_info) + return -EINVAL; + + /* check if the UCC port number is in range. */ + if ((us_info->ucc_num < 0) || (us_info->ucc_num > UCC_MAX_NUM - 1)) { + printk(KERN_ERR "%s: illegal UCC number\n", __func__); + return -EINVAL; + } + + /* + * Set mrblr + * Check that 'max_rx_buf_length' is properly aligned (4), unless + * rfw is 1, meaning that QE accepts one byte at a time, unlike normal + * case when QE accepts 32 bits at a time. + */ + if ((!us_info->rfw) && + (us_info->max_rx_buf_length & (UCC_SLOW_MRBLR_ALIGNMENT - 1))) { + printk(KERN_ERR "max_rx_buf_length not aligned.\n"); + return -EINVAL; + } + + uccs = kzalloc(sizeof(struct ucc_slow_private), GFP_KERNEL); + if (!uccs) { + printk(KERN_ERR "%s: Cannot allocate private data\n", + __func__); + return -ENOMEM; + } + uccs->rx_base_offset = -1; + uccs->tx_base_offset = -1; + uccs->us_pram_offset = -1; + + /* Fill slow UCC structure */ + uccs->us_info = us_info; + /* Set the PHY base address */ + uccs->us_regs = ioremap(us_info->regs, sizeof(struct ucc_slow)); + if (uccs->us_regs == NULL) { + printk(KERN_ERR "%s: Cannot map UCC registers\n", __func__); + kfree(uccs); + return -ENOMEM; + } + + us_regs = uccs->us_regs; + uccs->p_ucce = &us_regs->ucce; + uccs->p_uccm = &us_regs->uccm; + + /* Get PRAM base */ + uccs->us_pram_offset = + qe_muram_alloc(UCC_SLOW_PRAM_SIZE, ALIGNMENT_OF_UCC_SLOW_PRAM); + if (uccs->us_pram_offset < 0) { + printk(KERN_ERR "%s: cannot allocate MURAM for PRAM", __func__); + ucc_slow_free(uccs); + return -ENOMEM; + } + id = ucc_slow_get_qe_cr_subblock(us_info->ucc_num); + qe_issue_cmd(QE_ASSIGN_PAGE_TO_DEVICE, id, us_info->protocol, + uccs->us_pram_offset); + + uccs->us_pram = qe_muram_addr(uccs->us_pram_offset); + + /* Set UCC to slow type */ + ret = ucc_set_type(us_info->ucc_num, UCC_SPEED_TYPE_SLOW); + if (ret) { + printk(KERN_ERR "%s: cannot set UCC type", __func__); + ucc_slow_free(uccs); + return ret; + } + + iowrite16be(us_info->max_rx_buf_length, &uccs->us_pram->mrblr); + + INIT_LIST_HEAD(&uccs->confQ); + + /* Allocate BDs. */ + uccs->rx_base_offset = + qe_muram_alloc(us_info->rx_bd_ring_len * sizeof(struct qe_bd), + QE_ALIGNMENT_OF_BD); + if (uccs->rx_base_offset < 0) { + printk(KERN_ERR "%s: cannot allocate %u RX BDs\n", __func__, + us_info->rx_bd_ring_len); + ucc_slow_free(uccs); + return -ENOMEM; + } + + uccs->tx_base_offset = + qe_muram_alloc(us_info->tx_bd_ring_len * sizeof(struct qe_bd), + QE_ALIGNMENT_OF_BD); + if (uccs->tx_base_offset < 0) { + printk(KERN_ERR "%s: cannot allocate TX BDs", __func__); + ucc_slow_free(uccs); + return -ENOMEM; + } + + /* Init Tx bds */ + bd = uccs->confBd = uccs->tx_bd = qe_muram_addr(uccs->tx_base_offset); + for (i = 0; i < us_info->tx_bd_ring_len - 1; i++) { + /* clear bd buffer */ + iowrite32be(0, &bd->buf); + /* set bd status and length */ + iowrite32be(0, (u32 __iomem *)bd); + bd++; + } + /* for last BD set Wrap bit */ + iowrite32be(0, &bd->buf); + iowrite32be(T_W, (u32 __iomem *)bd); + + /* Init Rx bds */ + bd = uccs->rx_bd = qe_muram_addr(uccs->rx_base_offset); + for (i = 0; i < us_info->rx_bd_ring_len - 1; i++) { + /* set bd status and length */ + iowrite32be(0, (u32 __iomem *)bd); + /* clear bd buffer */ + iowrite32be(0, &bd->buf); + bd++; + } + /* for last BD set Wrap bit */ + iowrite32be(R_W, (u32 __iomem *)bd); + iowrite32be(0, &bd->buf); + + /* Set GUMR (For more details see the hardware spec.). */ + /* gumr_h */ + gumr = us_info->tcrc; + if (us_info->cdp) + gumr |= UCC_SLOW_GUMR_H_CDP; + if (us_info->ctsp) + gumr |= UCC_SLOW_GUMR_H_CTSP; + if (us_info->cds) + gumr |= UCC_SLOW_GUMR_H_CDS; + if (us_info->ctss) + gumr |= UCC_SLOW_GUMR_H_CTSS; + if (us_info->tfl) + gumr |= UCC_SLOW_GUMR_H_TFL; + if (us_info->rfw) + gumr |= UCC_SLOW_GUMR_H_RFW; + if (us_info->txsy) + gumr |= UCC_SLOW_GUMR_H_TXSY; + if (us_info->rtsm) + gumr |= UCC_SLOW_GUMR_H_RTSM; + iowrite32be(gumr, &us_regs->gumr_h); + + /* gumr_l */ + gumr = (u32)us_info->tdcr | (u32)us_info->rdcr | (u32)us_info->tenc | + (u32)us_info->renc | (u32)us_info->diag | (u32)us_info->mode; + if (us_info->tci) + gumr |= UCC_SLOW_GUMR_L_TCI; + if (us_info->rinv) + gumr |= UCC_SLOW_GUMR_L_RINV; + if (us_info->tinv) + gumr |= UCC_SLOW_GUMR_L_TINV; + if (us_info->tend) + gumr |= UCC_SLOW_GUMR_L_TEND; + iowrite32be(gumr, &us_regs->gumr_l); + + /* Function code registers */ + + /* if the data is in cachable memory, the 'global' */ + /* in the function code should be set. */ + iowrite8(UCC_BMR_BO_BE, &uccs->us_pram->tbmr); + iowrite8(UCC_BMR_BO_BE, &uccs->us_pram->rbmr); + + /* rbase, tbase are offsets from MURAM base */ + iowrite16be(uccs->rx_base_offset, &uccs->us_pram->rbase); + iowrite16be(uccs->tx_base_offset, &uccs->us_pram->tbase); + + /* Mux clocking */ + /* Grant Support */ + ucc_set_qe_mux_grant(us_info->ucc_num, us_info->grant_support); + /* Breakpoint Support */ + ucc_set_qe_mux_bkpt(us_info->ucc_num, us_info->brkpt_support); + /* Set Tsa or NMSI mode. */ + ucc_set_qe_mux_tsa(us_info->ucc_num, us_info->tsa); + /* If NMSI (not Tsa), set Tx and Rx clock. */ + if (!us_info->tsa) { + /* Rx clock routing */ + if (ucc_set_qe_mux_rxtx(us_info->ucc_num, us_info->rx_clock, + COMM_DIR_RX)) { + printk(KERN_ERR "%s: illegal value for RX clock\n", + __func__); + ucc_slow_free(uccs); + return -EINVAL; + } + /* Tx clock routing */ + if (ucc_set_qe_mux_rxtx(us_info->ucc_num, us_info->tx_clock, + COMM_DIR_TX)) { + printk(KERN_ERR "%s: illegal value for TX clock\n", + __func__); + ucc_slow_free(uccs); + return -EINVAL; + } + } + + /* Set interrupt mask register at UCC level. */ + iowrite16be(us_info->uccm_mask, &us_regs->uccm); + + /* First, clear anything pending at UCC level, + * otherwise, old garbage may come through + * as soon as the dam is opened. */ + + /* Writing '1' clears */ + iowrite16be(0xffff, &us_regs->ucce); + + /* Issue QE Init command */ + if (us_info->init_tx && us_info->init_rx) + command = QE_INIT_TX_RX; + else if (us_info->init_tx) + command = QE_INIT_TX; + else + command = QE_INIT_RX; /* We know at least one is TRUE */ + + qe_issue_cmd(command, id, us_info->protocol, 0); + + *uccs_ret = uccs; + return 0; +} +EXPORT_SYMBOL(ucc_slow_init); + +void ucc_slow_free(struct ucc_slow_private * uccs) +{ + if (!uccs) + return; + + qe_muram_free(uccs->rx_base_offset); + qe_muram_free(uccs->tx_base_offset); + qe_muram_free(uccs->us_pram_offset); + + if (uccs->us_regs) + iounmap(uccs->us_regs); + + kfree(uccs); +} +EXPORT_SYMBOL(ucc_slow_free); + diff --git a/drivers/soc/fsl/qe/usb.c b/drivers/soc/fsl/qe/usb.c new file mode 100644 index 000000000..890f236ea --- /dev/null +++ b/drivers/soc/fsl/qe/usb.c @@ -0,0 +1,52 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * QE USB routines + * + * Copyright 2006 Freescale Semiconductor, Inc. + * Shlomi Gridish <gridish@freescale.com> + * Jerry Huang <Chang-Ming.Huang@freescale.com> + * Copyright (c) MontaVista Software, Inc. 2008. + * Anton Vorontsov <avorontsov@ru.mvista.com> + */ + +#include <linux/kernel.h> +#include <linux/errno.h> +#include <linux/export.h> +#include <linux/io.h> +#include <soc/fsl/qe/immap_qe.h> +#include <soc/fsl/qe/qe.h> + +int qe_usb_clock_set(enum qe_clock clk, int rate) +{ + struct qe_mux __iomem *mux = &qe_immr->qmx; + unsigned long flags; + u32 val; + + switch (clk) { + case QE_CLK3: val = QE_CMXGCR_USBCS_CLK3; break; + case QE_CLK5: val = QE_CMXGCR_USBCS_CLK5; break; + case QE_CLK7: val = QE_CMXGCR_USBCS_CLK7; break; + case QE_CLK9: val = QE_CMXGCR_USBCS_CLK9; break; + case QE_CLK13: val = QE_CMXGCR_USBCS_CLK13; break; + case QE_CLK17: val = QE_CMXGCR_USBCS_CLK17; break; + case QE_CLK19: val = QE_CMXGCR_USBCS_CLK19; break; + case QE_CLK21: val = QE_CMXGCR_USBCS_CLK21; break; + case QE_BRG9: val = QE_CMXGCR_USBCS_BRG9; break; + case QE_BRG10: val = QE_CMXGCR_USBCS_BRG10; break; + default: + pr_err("%s: requested unknown clock %d\n", __func__, clk); + return -EINVAL; + } + + if (qe_clock_is_brg(clk)) + qe_setbrg(clk, rate, 1); + + spin_lock_irqsave(&cmxgcr_lock, flags); + + qe_clrsetbits_be32(&mux->cmxgcr, QE_CMXGCR_USBCS, val); + + spin_unlock_irqrestore(&cmxgcr_lock, flags); + + return 0; +} +EXPORT_SYMBOL(qe_usb_clock_set); diff --git a/drivers/soc/fsl/rcpm.c b/drivers/soc/fsl/rcpm.c new file mode 100644 index 000000000..3d0cae30c --- /dev/null +++ b/drivers/soc/fsl/rcpm.c @@ -0,0 +1,199 @@ +// SPDX-License-Identifier: GPL-2.0 +// +// rcpm.c - Freescale QorIQ RCPM driver +// +// Copyright 2019-2020 NXP +// +// Author: Ran Wang <ran.wang_1@nxp.com> + +#include <linux/init.h> +#include <linux/module.h> +#include <linux/platform_device.h> +#include <linux/of_address.h> +#include <linux/slab.h> +#include <linux/suspend.h> +#include <linux/kernel.h> +#include <linux/acpi.h> + +#define RCPM_WAKEUP_CELL_MAX_SIZE 7 + +struct rcpm { + unsigned int wakeup_cells; + void __iomem *ippdexpcr_base; + bool little_endian; +}; + +#define SCFG_SPARECR8 0x051c + +static void copy_ippdexpcr1_setting(u32 val) +{ + struct device_node *np; + void __iomem *regs; + u32 reg_val; + + np = of_find_compatible_node(NULL, NULL, "fsl,ls1021a-scfg"); + if (!np) + return; + + regs = of_iomap(np, 0); + if (!regs) + return; + + reg_val = ioread32be(regs + SCFG_SPARECR8); + iowrite32be(val | reg_val, regs + SCFG_SPARECR8); + + iounmap(regs); +} + +/** + * rcpm_pm_prepare - performs device-level tasks associated with power + * management, such as programming related to the wakeup source control. + * @dev: Device to handle. + * + */ +static int rcpm_pm_prepare(struct device *dev) +{ + int i, ret, idx; + void __iomem *base; + struct wakeup_source *ws; + struct rcpm *rcpm; + struct device_node *np = dev->of_node; + u32 value[RCPM_WAKEUP_CELL_MAX_SIZE + 1]; + u32 setting[RCPM_WAKEUP_CELL_MAX_SIZE] = {0}; + + rcpm = dev_get_drvdata(dev); + if (!rcpm) + return -EINVAL; + + base = rcpm->ippdexpcr_base; + idx = wakeup_sources_read_lock(); + + /* Begin with first registered wakeup source */ + for_each_wakeup_source(ws) { + + /* skip object which is not attached to device */ + if (!ws->dev || !ws->dev->parent) + continue; + + ret = device_property_read_u32_array(ws->dev->parent, + "fsl,rcpm-wakeup", value, + rcpm->wakeup_cells + 1); + + if (ret) + continue; + + /* + * For DT mode, would handle devices with "fsl,rcpm-wakeup" + * pointing to the current RCPM node. + * + * For ACPI mode, currently we assume there is only one + * RCPM controller existing. + */ + if (is_of_node(dev->fwnode)) + if (np->phandle != value[0]) + continue; + + /* Property "#fsl,rcpm-wakeup-cells" of rcpm node defines the + * number of IPPDEXPCR register cells, and "fsl,rcpm-wakeup" + * of wakeup source IP contains an integer array: <phandle to + * RCPM node, IPPDEXPCR0 setting, IPPDEXPCR1 setting, + * IPPDEXPCR2 setting, etc>. + * + * So we will go thought them to collect setting data. + */ + for (i = 0; i < rcpm->wakeup_cells; i++) + setting[i] |= value[i + 1]; + } + + wakeup_sources_read_unlock(idx); + + /* Program all IPPDEXPCRn once */ + for (i = 0; i < rcpm->wakeup_cells; i++) { + u32 tmp = setting[i]; + void __iomem *address = base + i * 4; + + if (!tmp) + continue; + + /* We can only OR related bits */ + if (rcpm->little_endian) { + tmp |= ioread32(address); + iowrite32(tmp, address); + } else { + tmp |= ioread32be(address); + iowrite32be(tmp, address); + } + /* + * Workaround of errata A-008646 on SoC LS1021A: + * There is a bug of register ippdexpcr1. + * Reading configuration register RCPM_IPPDEXPCR1 + * always return zero. So save ippdexpcr1's value + * to register SCFG_SPARECR8.And the value of + * ippdexpcr1 will be read from SCFG_SPARECR8. + */ + if (dev_of_node(dev) && (i == 1)) + if (of_device_is_compatible(np, "fsl,ls1021a-rcpm")) + copy_ippdexpcr1_setting(tmp); + } + + return 0; +} + +static const struct dev_pm_ops rcpm_pm_ops = { + .prepare = rcpm_pm_prepare, +}; + +static int rcpm_probe(struct platform_device *pdev) +{ + struct device *dev = &pdev->dev; + struct rcpm *rcpm; + int ret; + + rcpm = devm_kzalloc(dev, sizeof(*rcpm), GFP_KERNEL); + if (!rcpm) + return -ENOMEM; + + rcpm->ippdexpcr_base = devm_platform_ioremap_resource(pdev, 0); + if (IS_ERR(rcpm->ippdexpcr_base)) { + ret = PTR_ERR(rcpm->ippdexpcr_base); + return ret; + } + + rcpm->little_endian = device_property_read_bool( + &pdev->dev, "little-endian"); + + ret = device_property_read_u32(&pdev->dev, + "#fsl,rcpm-wakeup-cells", &rcpm->wakeup_cells); + if (ret) + return ret; + + dev_set_drvdata(&pdev->dev, rcpm); + + return 0; +} + +static const struct of_device_id rcpm_of_match[] = { + { .compatible = "fsl,qoriq-rcpm-2.1+", }, + {} +}; +MODULE_DEVICE_TABLE(of, rcpm_of_match); + +#ifdef CONFIG_ACPI +static const struct acpi_device_id rcpm_acpi_ids[] = { + {"NXP0015",}, + { } +}; +MODULE_DEVICE_TABLE(acpi, rcpm_acpi_ids); +#endif + +static struct platform_driver rcpm_driver = { + .driver = { + .name = "rcpm", + .of_match_table = rcpm_of_match, + .acpi_match_table = ACPI_PTR(rcpm_acpi_ids), + .pm = &rcpm_pm_ops, + }, + .probe = rcpm_probe, +}; + +module_platform_driver(rcpm_driver); |