diff options
Diffstat (limited to 'arch/mips/sibyte/common/sb_tbprof.c')
| -rw-r--r-- | arch/mips/sibyte/common/sb_tbprof.c | 595 |
1 files changed, 595 insertions, 0 deletions
diff --git a/arch/mips/sibyte/common/sb_tbprof.c b/arch/mips/sibyte/common/sb_tbprof.c new file mode 100644 index 000000000..bc47681e8 --- /dev/null +++ b/arch/mips/sibyte/common/sb_tbprof.c @@ -0,0 +1,595 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * + * Copyright (C) 2001, 2002, 2003 Broadcom Corporation + * Copyright (C) 2007 Ralf Baechle <ralf@linux-mips.org> + * Copyright (C) 2007 MIPS Technologies, Inc. + * written by Ralf Baechle <ralf@linux-mips.org> + */ + +#undef DEBUG + +#include <linux/device.h> +#include <linux/module.h> +#include <linux/kernel.h> +#include <linux/types.h> +#include <linux/init.h> +#include <linux/interrupt.h> +#include <linux/sched.h> +#include <linux/vmalloc.h> +#include <linux/fs.h> +#include <linux/errno.h> +#include <linux/wait.h> +#include <asm/io.h> +#include <asm/sibyte/sb1250.h> + +#if defined(CONFIG_SIBYTE_BCM1x55) || defined(CONFIG_SIBYTE_BCM1x80) +#include <asm/sibyte/bcm1480_regs.h> +#include <asm/sibyte/bcm1480_scd.h> +#include <asm/sibyte/bcm1480_int.h> +#elif defined(CONFIG_SIBYTE_SB1250) || defined(CONFIG_SIBYTE_BCM112X) +#include <asm/sibyte/sb1250_regs.h> +#include <asm/sibyte/sb1250_scd.h> +#include <asm/sibyte/sb1250_int.h> +#else +#error invalid SiByte UART configuration +#endif + +#if defined(CONFIG_SIBYTE_BCM1x55) || defined(CONFIG_SIBYTE_BCM1x80) +#undef K_INT_TRACE_FREEZE +#define K_INT_TRACE_FREEZE K_BCM1480_INT_TRACE_FREEZE +#undef K_INT_PERF_CNT +#define K_INT_PERF_CNT K_BCM1480_INT_PERF_CNT +#endif + +#include <linux/uaccess.h> + +#define SBPROF_TB_MAJOR 240 + +typedef u64 tb_sample_t[6*256]; + +enum open_status { + SB_CLOSED, + SB_OPENING, + SB_OPEN +}; + +struct sbprof_tb { + wait_queue_head_t tb_sync; + wait_queue_head_t tb_read; + struct mutex lock; + enum open_status open; + tb_sample_t *sbprof_tbbuf; + int next_tb_sample; + + volatile int tb_enable; + volatile int tb_armed; + +}; + +static struct sbprof_tb sbp; + +#define MAX_SAMPLE_BYTES (24*1024*1024) +#define MAX_TBSAMPLE_BYTES (12*1024*1024) + +#define MAX_SAMPLES (MAX_SAMPLE_BYTES/sizeof(u_int32_t)) +#define TB_SAMPLE_SIZE (sizeof(tb_sample_t)) +#define MAX_TB_SAMPLES (MAX_TBSAMPLE_BYTES/TB_SAMPLE_SIZE) + +/* ioctls */ +#define SBPROF_ZBSTART _IOW('s', 0, int) +#define SBPROF_ZBSTOP _IOW('s', 1, int) +#define SBPROF_ZBWAITFULL _IOW('s', 2, int) + +/* + * Routines for using 40-bit SCD cycle counter + * + * Client responsible for either handling interrupts or making sure + * the cycles counter never saturates, e.g., by doing + * zclk_timer_init(0) at least every 2^40 - 1 ZCLKs. + */ + +/* + * Configures SCD counter 0 to count ZCLKs starting from val; + * Configures SCD counters1,2,3 to count nothing. + * Must not be called while gathering ZBbus profiles. + */ + +#define zclk_timer_init(val) \ + __asm__ __volatile__ (".set push;" \ + ".set mips64;" \ + "la $8, 0xb00204c0;" /* SCD perf_cnt_cfg */ \ + "sd %0, 0x10($8);" /* write val to counter0 */ \ + "sd %1, 0($8);" /* config counter0 for zclks*/ \ + ".set pop" \ + : /* no outputs */ \ + /* enable, counter0 */ \ + : /* inputs */ "r"(val), "r" ((1ULL << 33) | 1ULL) \ + : /* modifies */ "$8" ) + + +/* Reads SCD counter 0 and puts result in value + unsigned long long val; */ +#define zclk_get(val) \ + __asm__ __volatile__ (".set push;" \ + ".set mips64;" \ + "la $8, 0xb00204c0;" /* SCD perf_cnt_cfg */ \ + "ld %0, 0x10($8);" /* write val to counter0 */ \ + ".set pop" \ + : /* outputs */ "=r"(val) \ + : /* inputs */ \ + : /* modifies */ "$8" ) + +#define DEVNAME "sb_tbprof" + +#define TB_FULL (sbp.next_tb_sample == MAX_TB_SAMPLES) + +/* + * Support for ZBbus sampling using the trace buffer + * + * We use the SCD performance counter interrupt, caused by a Zclk counter + * overflow, to trigger the start of tracing. + * + * We set the trace buffer to sample everything and freeze on + * overflow. + * + * We map the interrupt for trace_buffer_freeze to handle it on CPU 0. + * + */ + +static u64 tb_period; + +static void arm_tb(void) +{ + u64 scdperfcnt; + u64 next = (1ULL << 40) - tb_period; + u64 tb_options = M_SCD_TRACE_CFG_FREEZE_FULL; + + /* + * Generate an SCD_PERFCNT interrupt in TB_PERIOD Zclks to + * trigger start of trace. XXX vary sampling period + */ + __raw_writeq(0, IOADDR(A_SCD_PERF_CNT_1)); + scdperfcnt = __raw_readq(IOADDR(A_SCD_PERF_CNT_CFG)); + + /* + * Unfortunately, in Pass 2 we must clear all counters to knock down + * a previous interrupt request. This means that bus profiling + * requires ALL of the SCD perf counters. + */ +#if defined(CONFIG_SIBYTE_BCM1x55) || defined(CONFIG_SIBYTE_BCM1x80) + __raw_writeq((scdperfcnt & ~M_SPC_CFG_SRC1) | + /* keep counters 0,2,3,4,5,6,7 as is */ + V_SPC_CFG_SRC1(1), /* counter 1 counts cycles */ + IOADDR(A_BCM1480_SCD_PERF_CNT_CFG0)); + __raw_writeq( + M_SPC_CFG_ENABLE | /* enable counting */ + M_SPC_CFG_CLEAR | /* clear all counters */ + V_SPC_CFG_SRC1(1), /* counter 1 counts cycles */ + IOADDR(A_BCM1480_SCD_PERF_CNT_CFG1)); +#else + __raw_writeq((scdperfcnt & ~M_SPC_CFG_SRC1) | + /* keep counters 0,2,3 as is */ + M_SPC_CFG_ENABLE | /* enable counting */ + M_SPC_CFG_CLEAR | /* clear all counters */ + V_SPC_CFG_SRC1(1), /* counter 1 counts cycles */ + IOADDR(A_SCD_PERF_CNT_CFG)); +#endif + __raw_writeq(next, IOADDR(A_SCD_PERF_CNT_1)); + /* Reset the trace buffer */ + __raw_writeq(M_SCD_TRACE_CFG_RESET, IOADDR(A_SCD_TRACE_CFG)); +#if 0 && defined(M_SCD_TRACE_CFG_FORCECNT) + /* XXXKW may want to expose control to the data-collector */ + tb_options |= M_SCD_TRACE_CFG_FORCECNT; +#endif + __raw_writeq(tb_options, IOADDR(A_SCD_TRACE_CFG)); + sbp.tb_armed = 1; +} + +static irqreturn_t sbprof_tb_intr(int irq, void *dev_id) +{ + int i; + + pr_debug(DEVNAME ": tb_intr\n"); + + if (sbp.next_tb_sample < MAX_TB_SAMPLES) { + /* XXX should use XKPHYS to make writes bypass L2 */ + u64 *p = sbp.sbprof_tbbuf[sbp.next_tb_sample++]; + /* Read out trace */ + __raw_writeq(M_SCD_TRACE_CFG_START_READ, + IOADDR(A_SCD_TRACE_CFG)); + __asm__ __volatile__ ("sync" : : : "memory"); + /* Loop runs backwards because bundles are read out in reverse order */ + for (i = 256 * 6; i > 0; i -= 6) { + /* Subscripts decrease to put bundle in the order */ + /* t0 lo, t0 hi, t1 lo, t1 hi, t2 lo, t2 hi */ + p[i - 1] = __raw_readq(IOADDR(A_SCD_TRACE_READ)); + /* read t2 hi */ + p[i - 2] = __raw_readq(IOADDR(A_SCD_TRACE_READ)); + /* read t2 lo */ + p[i - 3] = __raw_readq(IOADDR(A_SCD_TRACE_READ)); + /* read t1 hi */ + p[i - 4] = __raw_readq(IOADDR(A_SCD_TRACE_READ)); + /* read t1 lo */ + p[i - 5] = __raw_readq(IOADDR(A_SCD_TRACE_READ)); + /* read t0 hi */ + p[i - 6] = __raw_readq(IOADDR(A_SCD_TRACE_READ)); + /* read t0 lo */ + } + if (!sbp.tb_enable) { + pr_debug(DEVNAME ": tb_intr shutdown\n"); + __raw_writeq(M_SCD_TRACE_CFG_RESET, + IOADDR(A_SCD_TRACE_CFG)); + sbp.tb_armed = 0; + wake_up_interruptible(&sbp.tb_sync); + } else { + /* knock down current interrupt and get another one later */ + arm_tb(); + } + } else { + /* No more trace buffer samples */ + pr_debug(DEVNAME ": tb_intr full\n"); + __raw_writeq(M_SCD_TRACE_CFG_RESET, IOADDR(A_SCD_TRACE_CFG)); + sbp.tb_armed = 0; + if (!sbp.tb_enable) + wake_up_interruptible(&sbp.tb_sync); + wake_up_interruptible(&sbp.tb_read); + } + return IRQ_HANDLED; +} + +static irqreturn_t sbprof_pc_intr(int irq, void *dev_id) +{ + printk(DEVNAME ": unexpected pc_intr"); + return IRQ_NONE; +} + +/* + * Requires: Already called zclk_timer_init with a value that won't + * saturate 40 bits. No subsequent use of SCD performance counters + * or trace buffer. + */ + +static int sbprof_zbprof_start(struct file *filp) +{ + u64 scdperfcnt; + int err; + + if (xchg(&sbp.tb_enable, 1)) + return -EBUSY; + + pr_debug(DEVNAME ": starting\n"); + + sbp.next_tb_sample = 0; + filp->f_pos = 0; + + err = request_irq(K_INT_TRACE_FREEZE, sbprof_tb_intr, 0, + DEVNAME " trace freeze", &sbp); + if (err) + return -EBUSY; + + /* Make sure there isn't a perf-cnt interrupt waiting */ + scdperfcnt = __raw_readq(IOADDR(A_SCD_PERF_CNT_CFG)); + /* Disable and clear counters, override SRC_1 */ + __raw_writeq((scdperfcnt & ~(M_SPC_CFG_SRC1 | M_SPC_CFG_ENABLE)) | + M_SPC_CFG_ENABLE | M_SPC_CFG_CLEAR | V_SPC_CFG_SRC1(1), + IOADDR(A_SCD_PERF_CNT_CFG)); + + /* + * We grab this interrupt to prevent others from trying to use + * it, even though we don't want to service the interrupts + * (they only feed into the trace-on-interrupt mechanism) + */ + if (request_irq(K_INT_PERF_CNT, sbprof_pc_intr, 0, DEVNAME " scd perfcnt", &sbp)) { + free_irq(K_INT_TRACE_FREEZE, &sbp); + return -EBUSY; + } + + /* + * I need the core to mask these, but the interrupt mapper to + * pass them through. I am exploiting my knowledge that + * cp0_status masks out IP[5]. krw + */ +#if defined(CONFIG_SIBYTE_BCM1x55) || defined(CONFIG_SIBYTE_BCM1x80) + __raw_writeq(K_BCM1480_INT_MAP_I3, + IOADDR(A_BCM1480_IMR_REGISTER(0, R_BCM1480_IMR_INTERRUPT_MAP_BASE_L) + + ((K_BCM1480_INT_PERF_CNT & 0x3f) << 3))); +#else + __raw_writeq(K_INT_MAP_I3, + IOADDR(A_IMR_REGISTER(0, R_IMR_INTERRUPT_MAP_BASE) + + (K_INT_PERF_CNT << 3))); +#endif + + /* Initialize address traps */ + __raw_writeq(0, IOADDR(A_ADDR_TRAP_UP_0)); + __raw_writeq(0, IOADDR(A_ADDR_TRAP_UP_1)); + __raw_writeq(0, IOADDR(A_ADDR_TRAP_UP_2)); + __raw_writeq(0, IOADDR(A_ADDR_TRAP_UP_3)); + + __raw_writeq(0, IOADDR(A_ADDR_TRAP_DOWN_0)); + __raw_writeq(0, IOADDR(A_ADDR_TRAP_DOWN_1)); + __raw_writeq(0, IOADDR(A_ADDR_TRAP_DOWN_2)); + __raw_writeq(0, IOADDR(A_ADDR_TRAP_DOWN_3)); + + __raw_writeq(0, IOADDR(A_ADDR_TRAP_CFG_0)); + __raw_writeq(0, IOADDR(A_ADDR_TRAP_CFG_1)); + __raw_writeq(0, IOADDR(A_ADDR_TRAP_CFG_2)); + __raw_writeq(0, IOADDR(A_ADDR_TRAP_CFG_3)); + + /* Initialize Trace Event 0-7 */ + /* when interrupt */ + __raw_writeq(M_SCD_TREVT_INTERRUPT, IOADDR(A_SCD_TRACE_EVENT_0)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_1)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_2)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_3)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_4)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_5)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_6)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_EVENT_7)); + + /* Initialize Trace Sequence 0-7 */ + /* Start on event 0 (interrupt) */ + __raw_writeq(V_SCD_TRSEQ_FUNC_START | 0x0fff, + IOADDR(A_SCD_TRACE_SEQUENCE_0)); + /* dsamp when d used | asamp when a used */ + __raw_writeq(M_SCD_TRSEQ_ASAMPLE | M_SCD_TRSEQ_DSAMPLE | + K_SCD_TRSEQ_TRIGGER_ALL, + IOADDR(A_SCD_TRACE_SEQUENCE_1)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_SEQUENCE_2)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_SEQUENCE_3)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_SEQUENCE_4)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_SEQUENCE_5)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_SEQUENCE_6)); + __raw_writeq(0, IOADDR(A_SCD_TRACE_SEQUENCE_7)); + + /* Now indicate the PERF_CNT interrupt as a trace-relevant interrupt */ +#if defined(CONFIG_SIBYTE_BCM1x55) || defined(CONFIG_SIBYTE_BCM1x80) + __raw_writeq(1ULL << (K_BCM1480_INT_PERF_CNT & 0x3f), + IOADDR(A_BCM1480_IMR_REGISTER(0, R_BCM1480_IMR_INTERRUPT_TRACE_L))); +#else + __raw_writeq(1ULL << K_INT_PERF_CNT, + IOADDR(A_IMR_REGISTER(0, R_IMR_INTERRUPT_TRACE))); +#endif + arm_tb(); + + pr_debug(DEVNAME ": done starting\n"); + + return 0; +} + +static int sbprof_zbprof_stop(void) +{ + int err = 0; + + pr_debug(DEVNAME ": stopping\n"); + + if (sbp.tb_enable) { + /* + * XXXKW there is a window here where the intr handler may run, + * see the disable, and do the wake_up before this sleep + * happens. + */ + pr_debug(DEVNAME ": wait for disarm\n"); + err = wait_event_interruptible(sbp.tb_sync, !sbp.tb_armed); + pr_debug(DEVNAME ": disarm complete, stat %d\n", err); + + if (err) + return err; + + sbp.tb_enable = 0; + free_irq(K_INT_TRACE_FREEZE, &sbp); + free_irq(K_INT_PERF_CNT, &sbp); + } + + pr_debug(DEVNAME ": done stopping\n"); + + return err; +} + +static int sbprof_tb_open(struct inode *inode, struct file *filp) +{ + int minor; + + minor = iminor(inode); + if (minor != 0) + return -ENODEV; + + if (xchg(&sbp.open, SB_OPENING) != SB_CLOSED) + return -EBUSY; + + memset(&sbp, 0, sizeof(struct sbprof_tb)); + sbp.sbprof_tbbuf = vzalloc(MAX_TBSAMPLE_BYTES); + if (!sbp.sbprof_tbbuf) { + sbp.open = SB_CLOSED; + wmb(); + return -ENOMEM; + } + + init_waitqueue_head(&sbp.tb_sync); + init_waitqueue_head(&sbp.tb_read); + mutex_init(&sbp.lock); + + sbp.open = SB_OPEN; + wmb(); + + return 0; +} + +static int sbprof_tb_release(struct inode *inode, struct file *filp) +{ + int minor; + + minor = iminor(inode); + if (minor != 0 || sbp.open != SB_CLOSED) + return -ENODEV; + + mutex_lock(&sbp.lock); + + if (sbp.tb_armed || sbp.tb_enable) + sbprof_zbprof_stop(); + + vfree(sbp.sbprof_tbbuf); + sbp.open = SB_CLOSED; + wmb(); + + mutex_unlock(&sbp.lock); + + return 0; +} + +static ssize_t sbprof_tb_read(struct file *filp, char __user *buf, + size_t size, loff_t *offp) +{ + int cur_sample, sample_off, cur_count, sample_left; + char *src; + int count = 0; + char __user *dest = buf; + long cur_off = *offp; + + if (!access_ok(buf, size)) + return -EFAULT; + + mutex_lock(&sbp.lock); + + count = 0; + cur_sample = cur_off / TB_SAMPLE_SIZE; + sample_off = cur_off % TB_SAMPLE_SIZE; + sample_left = TB_SAMPLE_SIZE - sample_off; + + while (size && (cur_sample < sbp.next_tb_sample)) { + int err; + + cur_count = size < sample_left ? size : sample_left; + src = (char *)(((long)sbp.sbprof_tbbuf[cur_sample])+sample_off); + err = __copy_to_user(dest, src, cur_count); + if (err) { + *offp = cur_off + cur_count - err; + mutex_unlock(&sbp.lock); + return err; + } + pr_debug(DEVNAME ": read from sample %d, %d bytes\n", + cur_sample, cur_count); + size -= cur_count; + sample_left -= cur_count; + if (!sample_left) { + cur_sample++; + sample_off = 0; + sample_left = TB_SAMPLE_SIZE; + } else { + sample_off += cur_count; + } + cur_off += cur_count; + dest += cur_count; + count += cur_count; + } + *offp = cur_off; + mutex_unlock(&sbp.lock); + + return count; +} + +static long sbprof_tb_ioctl(struct file *filp, + unsigned int command, + unsigned long arg) +{ + int err = 0; + + switch (command) { + case SBPROF_ZBSTART: + mutex_lock(&sbp.lock); + err = sbprof_zbprof_start(filp); + mutex_unlock(&sbp.lock); + break; + + case SBPROF_ZBSTOP: + mutex_lock(&sbp.lock); + err = sbprof_zbprof_stop(); + mutex_unlock(&sbp.lock); + break; + + case SBPROF_ZBWAITFULL: { + err = wait_event_interruptible(sbp.tb_read, TB_FULL); + if (err) + break; + + err = put_user(TB_FULL, (int __user *) arg); + break; + } + + default: + err = -EINVAL; + break; + } + + return err; +} + +static const struct file_operations sbprof_tb_fops = { + .owner = THIS_MODULE, + .open = sbprof_tb_open, + .release = sbprof_tb_release, + .read = sbprof_tb_read, + .unlocked_ioctl = sbprof_tb_ioctl, + .compat_ioctl = sbprof_tb_ioctl, + .mmap = NULL, + .llseek = default_llseek, +}; + +static struct class *tb_class; +static struct device *tb_dev; + +static int __init sbprof_tb_init(void) +{ + struct device *dev; + struct class *tbc; + int err; + + if (register_chrdev(SBPROF_TB_MAJOR, DEVNAME, &sbprof_tb_fops)) { + printk(KERN_WARNING DEVNAME ": initialization failed (dev %d)\n", + SBPROF_TB_MAJOR); + return -EIO; + } + + tbc = class_create(THIS_MODULE, "sb_tracebuffer"); + if (IS_ERR(tbc)) { + err = PTR_ERR(tbc); + goto out_chrdev; + } + + tb_class = tbc; + + dev = device_create(tbc, NULL, MKDEV(SBPROF_TB_MAJOR, 0), NULL, "tb"); + if (IS_ERR(dev)) { + err = PTR_ERR(dev); + goto out_class; + } + tb_dev = dev; + + sbp.open = SB_CLOSED; + wmb(); + tb_period = zbbus_mhz * 10000LL; + pr_info(DEVNAME ": initialized - tb_period = %lld\n", + (long long) tb_period); + return 0; + +out_class: + class_destroy(tb_class); +out_chrdev: + unregister_chrdev(SBPROF_TB_MAJOR, DEVNAME); + + return err; +} + +static void __exit sbprof_tb_cleanup(void) +{ + device_destroy(tb_class, MKDEV(SBPROF_TB_MAJOR, 0)); + unregister_chrdev(SBPROF_TB_MAJOR, DEVNAME); + class_destroy(tb_class); +} + +module_init(sbprof_tb_init); +module_exit(sbprof_tb_cleanup); + +MODULE_ALIAS_CHARDEV_MAJOR(SBPROF_TB_MAJOR); +MODULE_AUTHOR("Ralf Baechle <ralf@linux-mips.org>"); +MODULE_LICENSE("GPL"); |