diff options
Diffstat (limited to 'drivers/accel/qaic')
| -rw-r--r-- | drivers/accel/qaic/Makefile | 3 | ||||
| -rw-r--r-- | drivers/accel/qaic/mhi_controller.c | 44 | ||||
| -rw-r--r-- | drivers/accel/qaic/mhi_controller.h | 2 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic.h | 21 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic_control.c | 7 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic_data.c | 147 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic_drv.c | 98 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic_timesync.c | 395 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic_timesync.h | 11 |
9 files changed, 598 insertions, 130 deletions
diff --git a/drivers/accel/qaic/Makefile b/drivers/accel/qaic/Makefile index 2418418f7a..3f7f6dfde7 100644 --- a/drivers/accel/qaic/Makefile +++ b/drivers/accel/qaic/Makefile @@ -9,4 +9,5 @@ qaic-y := \ mhi_controller.o \ qaic_control.o \ qaic_data.o \ - qaic_drv.o + qaic_drv.o \ + qaic_timesync.o diff --git a/drivers/accel/qaic/mhi_controller.c b/drivers/accel/qaic/mhi_controller.c index 1405623b03..cb77d048ed 100644 --- a/drivers/accel/qaic/mhi_controller.c +++ b/drivers/accel/qaic/mhi_controller.c @@ -348,7 +348,7 @@ static struct mhi_channel_config aic100_channels[] = { .local_elements = 0, .event_ring = 0, .dir = DMA_TO_DEVICE, - .ee_mask = MHI_CH_EE_SBL | MHI_CH_EE_AMSS, + .ee_mask = MHI_CH_EE_SBL, .pollcfg = 0, .doorbell = MHI_DB_BRST_DISABLE, .lpm_notify = false, @@ -364,7 +364,39 @@ static struct mhi_channel_config aic100_channels[] = { .local_elements = 0, .event_ring = 0, .dir = DMA_FROM_DEVICE, - .ee_mask = MHI_CH_EE_SBL | MHI_CH_EE_AMSS, + .ee_mask = MHI_CH_EE_SBL, + .pollcfg = 0, + .doorbell = MHI_DB_BRST_DISABLE, + .lpm_notify = false, + .offload_channel = false, + .doorbell_mode_switch = false, + .auto_queue = false, + .wake_capable = false, + }, + { + .name = "QAIC_TIMESYNC_PERIODIC", + .num = 22, + .num_elements = 32, + .local_elements = 0, + .event_ring = 0, + .dir = DMA_TO_DEVICE, + .ee_mask = MHI_CH_EE_AMSS, + .pollcfg = 0, + .doorbell = MHI_DB_BRST_DISABLE, + .lpm_notify = false, + .offload_channel = false, + .doorbell_mode_switch = false, + .auto_queue = false, + .wake_capable = false, + }, + { + .num = 23, + .name = "QAIC_TIMESYNC_PERIODIC", + .num_elements = 32, + .local_elements = 0, + .event_ring = 0, + .dir = DMA_FROM_DEVICE, + .ee_mask = MHI_CH_EE_AMSS, .pollcfg = 0, .doorbell = MHI_DB_BRST_DISABLE, .lpm_notify = false, @@ -450,7 +482,7 @@ static void mhi_status_cb(struct mhi_controller *mhi_cntrl, enum mhi_callback re pci_err(qdev->pdev, "Fatal error received from device. Attempting to recover\n"); /* this event occurs in non-atomic context */ if (reason == MHI_CB_SYS_ERROR) - qaic_dev_reset_clean_local_state(qdev, true); + qaic_dev_reset_clean_local_state(qdev); } static int mhi_reset_and_async_power_up(struct mhi_controller *mhi_cntrl) @@ -481,7 +513,7 @@ static int mhi_reset_and_async_power_up(struct mhi_controller *mhi_cntrl) } struct mhi_controller *qaic_mhi_register_controller(struct pci_dev *pci_dev, void __iomem *mhi_bar, - int mhi_irq) + int mhi_irq, bool shared_msi) { struct mhi_controller *mhi_cntrl; int ret; @@ -513,6 +545,10 @@ struct mhi_controller *qaic_mhi_register_controller(struct pci_dev *pci_dev, voi return ERR_PTR(-ENOMEM); mhi_cntrl->irq[0] = mhi_irq; + + if (shared_msi) /* MSI shared with data path, no IRQF_NO_SUSPEND */ + mhi_cntrl->irq_flags = IRQF_SHARED; + mhi_cntrl->fw_image = "qcom/aic100/sbl.bin"; /* use latest configured timeout */ diff --git a/drivers/accel/qaic/mhi_controller.h b/drivers/accel/qaic/mhi_controller.h index 2ae45d768e..500e7f4af2 100644 --- a/drivers/accel/qaic/mhi_controller.h +++ b/drivers/accel/qaic/mhi_controller.h @@ -8,7 +8,7 @@ #define MHICONTROLLERQAIC_H_ struct mhi_controller *qaic_mhi_register_controller(struct pci_dev *pci_dev, void __iomem *mhi_bar, - int mhi_irq); + int mhi_irq, bool shared_msi); void qaic_mhi_free_controller(struct mhi_controller *mhi_cntrl, bool link_up); void qaic_mhi_start_reset(struct mhi_controller *mhi_cntrl); void qaic_mhi_reset_done(struct mhi_controller *mhi_cntrl); diff --git a/drivers/accel/qaic/qaic.h b/drivers/accel/qaic/qaic.h index e3f4c30f3f..582836f953 100644 --- a/drivers/accel/qaic/qaic.h +++ b/drivers/accel/qaic/qaic.h @@ -31,6 +31,15 @@ #define to_drm(qddev) (&(qddev)->drm) #define to_accel_kdev(qddev) (to_drm(qddev)->accel->kdev) /* Return Linux device of accel node */ +enum __packed dev_states { + /* Device is offline or will be very soon */ + QAIC_OFFLINE, + /* Device is booting, not clear if it's in a usable state */ + QAIC_BOOT, + /* Device is fully operational */ + QAIC_ONLINE, +}; + extern bool datapath_polling; struct qaic_user { @@ -121,8 +130,10 @@ struct qaic_device { struct workqueue_struct *cntl_wq; /* Synchronizes all the users of device during cleanup */ struct srcu_struct dev_lock; - /* true: Device under reset; false: Device not under reset */ - bool in_reset; + /* Track the state of the device during resets */ + enum dev_states dev_state; + /* true: single MSI is used to operate device */ + bool single_msi; /* * true: A tx MHI transaction has failed and a rx buffer is still queued * in control device. Such a buffer is considered lost rx buffer @@ -137,6 +148,10 @@ struct qaic_device { u32 (*gen_crc)(void *msg); /* Validate the CRC of a control message */ bool (*valid_crc)(void *msg); + /* MHI "QAIC_TIMESYNC" channel device */ + struct mhi_device *qts_ch; + /* Work queue for tasks related to MHI "QAIC_TIMESYNC" channel */ + struct workqueue_struct *qts_wq; }; struct qaic_drm_device { @@ -268,7 +283,7 @@ void wakeup_dbc(struct qaic_device *qdev, u32 dbc_id); void release_dbc(struct qaic_device *qdev, u32 dbc_id); void wake_all_cntl(struct qaic_device *qdev); -void qaic_dev_reset_clean_local_state(struct qaic_device *qdev, bool exit_reset); +void qaic_dev_reset_clean_local_state(struct qaic_device *qdev); struct drm_gem_object *qaic_gem_prime_import(struct drm_device *dev, struct dma_buf *dma_buf); diff --git a/drivers/accel/qaic/qaic_control.c b/drivers/accel/qaic/qaic_control.c index 388abd4002..9e8a8cbadf 100644 --- a/drivers/accel/qaic/qaic_control.c +++ b/drivers/accel/qaic/qaic_control.c @@ -1022,7 +1022,8 @@ static void *msg_xfer(struct qaic_device *qdev, struct wrapper_list *wrappers, u int xfer_count = 0; int retry_count; - if (qdev->in_reset) { + /* Allow QAIC_BOOT state since we need to check control protocol version */ + if (qdev->dev_state == QAIC_OFFLINE) { mutex_unlock(&qdev->cntl_mutex); return ERR_PTR(-ENODEV); } @@ -1138,7 +1139,7 @@ static int abort_dma_cont(struct qaic_device *qdev, struct wrapper_list *wrapper if (!list_is_first(&wrapper->list, &wrappers->list)) kref_put(&wrapper->ref_count, free_wrapper); - wrapper = add_wrapper(wrappers, offsetof(struct wrapper_msg, trans) + sizeof(*out_trans)); + wrapper = add_wrapper(wrappers, sizeof(*wrapper)); if (!wrapper) return -ENOMEM; @@ -1306,7 +1307,7 @@ int qaic_manage_ioctl(struct drm_device *dev, void *data, struct drm_file *file_ qdev = usr->qddev->qdev; qdev_rcu_id = srcu_read_lock(&qdev->dev_lock); - if (qdev->in_reset) { + if (qdev->dev_state != QAIC_ONLINE) { srcu_read_unlock(&qdev->dev_lock, qdev_rcu_id); srcu_read_unlock(&usr->qddev_lock, usr_rcu_id); return -ENODEV; diff --git a/drivers/accel/qaic/qaic_data.c b/drivers/accel/qaic/qaic_data.c index d42f002bc0..03c9a793da 100644 --- a/drivers/accel/qaic/qaic_data.c +++ b/drivers/accel/qaic/qaic_data.c @@ -51,6 +51,7 @@ }) #define NUM_EVENTS 128 #define NUM_DELAYS 10 +#define fifo_at(base, offset) ((base) + (offset) * get_dbc_req_elem_size()) static unsigned int wait_exec_default_timeout_ms = 5000; /* 5 sec default */ module_param(wait_exec_default_timeout_ms, uint, 0600); @@ -451,7 +452,7 @@ static int create_sgt(struct qaic_device *qdev, struct sg_table **sgt_out, u64 s * later */ buf_extra = (PAGE_SIZE - size % PAGE_SIZE) % PAGE_SIZE; - max_order = min(MAX_ORDER - 1, get_order(size)); + max_order = min(MAX_PAGE_ORDER, get_order(size)); } else { /* allocate a single page for book keeping */ nr_pages = 1; @@ -689,7 +690,7 @@ int qaic_create_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *fi qdev = usr->qddev->qdev; qdev_rcu_id = srcu_read_lock(&qdev->dev_lock); - if (qdev->in_reset) { + if (qdev->dev_state != QAIC_ONLINE) { ret = -ENODEV; goto unlock_dev_srcu; } @@ -748,7 +749,7 @@ int qaic_mmap_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *file qdev = usr->qddev->qdev; qdev_rcu_id = srcu_read_lock(&qdev->dev_lock); - if (qdev->in_reset) { + if (qdev->dev_state != QAIC_ONLINE) { ret = -ENODEV; goto unlock_dev_srcu; } @@ -967,7 +968,7 @@ int qaic_attach_slice_bo_ioctl(struct drm_device *dev, void *data, struct drm_fi qdev = usr->qddev->qdev; qdev_rcu_id = srcu_read_lock(&qdev->dev_lock); - if (qdev->in_reset) { + if (qdev->dev_state != QAIC_ONLINE) { ret = -ENODEV; goto unlock_dev_srcu; } @@ -1056,6 +1057,16 @@ unlock_usr_srcu: return ret; } +static inline u32 fifo_space_avail(u32 head, u32 tail, u32 q_size) +{ + u32 avail = head - tail - 1; + + if (head <= tail) + avail += q_size; + + return avail; +} + static inline int copy_exec_reqs(struct qaic_device *qdev, struct bo_slice *slice, u32 dbc_id, u32 head, u32 *ptail) { @@ -1064,27 +1075,20 @@ static inline int copy_exec_reqs(struct qaic_device *qdev, struct bo_slice *slic u32 tail = *ptail; u32 avail; - avail = head - tail; - if (head <= tail) - avail += dbc->nelem; - - --avail; - + avail = fifo_space_avail(head, tail, dbc->nelem); if (avail < slice->nents) return -EAGAIN; if (tail + slice->nents > dbc->nelem) { avail = dbc->nelem - tail; avail = min_t(u32, avail, slice->nents); - memcpy(dbc->req_q_base + tail * get_dbc_req_elem_size(), reqs, - sizeof(*reqs) * avail); + memcpy(fifo_at(dbc->req_q_base, tail), reqs, sizeof(*reqs) * avail); reqs += avail; avail = slice->nents - avail; if (avail) memcpy(dbc->req_q_base, reqs, sizeof(*reqs) * avail); } else { - memcpy(dbc->req_q_base + tail * get_dbc_req_elem_size(), reqs, - sizeof(*reqs) * slice->nents); + memcpy(fifo_at(dbc->req_q_base, tail), reqs, sizeof(*reqs) * slice->nents); } *ptail = (tail + slice->nents) % dbc->nelem; @@ -1092,46 +1096,31 @@ static inline int copy_exec_reqs(struct qaic_device *qdev, struct bo_slice *slic return 0; } -/* - * Based on the value of resize we may only need to transmit first_n - * entries and the last entry, with last_bytes to send from the last entry. - * Note that first_n could be 0. - */ static inline int copy_partial_exec_reqs(struct qaic_device *qdev, struct bo_slice *slice, - u64 resize, u32 dbc_id, u32 head, u32 *ptail) + u64 resize, struct dma_bridge_chan *dbc, u32 head, + u32 *ptail) { - struct dma_bridge_chan *dbc = &qdev->dbc[dbc_id]; struct dbc_req *reqs = slice->reqs; struct dbc_req *last_req; u32 tail = *ptail; - u64 total_bytes; u64 last_bytes; u32 first_n; u32 avail; - int ret; - int i; - - avail = head - tail; - if (head <= tail) - avail += dbc->nelem; - --avail; + avail = fifo_space_avail(head, tail, dbc->nelem); - total_bytes = 0; - for (i = 0; i < slice->nents; i++) { - total_bytes += le32_to_cpu(reqs[i].len); - if (total_bytes >= resize) + /* + * After this for loop is complete, first_n represents the index + * of the last DMA request of this slice that needs to be + * transferred after resizing and last_bytes represents DMA size + * of that request. + */ + last_bytes = resize; + for (first_n = 0; first_n < slice->nents; first_n++) + if (last_bytes > le32_to_cpu(reqs[first_n].len)) + last_bytes -= le32_to_cpu(reqs[first_n].len); + else break; - } - - if (total_bytes < resize) { - /* User space should have used the full buffer path. */ - ret = -EINVAL; - return ret; - } - - first_n = i; - last_bytes = i ? resize + le32_to_cpu(reqs[i].len) - total_bytes : resize; if (avail < (first_n + 1)) return -EAGAIN; @@ -1140,22 +1129,21 @@ static inline int copy_partial_exec_reqs(struct qaic_device *qdev, struct bo_sli if (tail + first_n > dbc->nelem) { avail = dbc->nelem - tail; avail = min_t(u32, avail, first_n); - memcpy(dbc->req_q_base + tail * get_dbc_req_elem_size(), reqs, - sizeof(*reqs) * avail); + memcpy(fifo_at(dbc->req_q_base, tail), reqs, sizeof(*reqs) * avail); last_req = reqs + avail; avail = first_n - avail; if (avail) memcpy(dbc->req_q_base, last_req, sizeof(*reqs) * avail); } else { - memcpy(dbc->req_q_base + tail * get_dbc_req_elem_size(), reqs, - sizeof(*reqs) * first_n); + memcpy(fifo_at(dbc->req_q_base, tail), reqs, sizeof(*reqs) * first_n); } } - /* Copy over the last entry. Here we need to adjust len to the left over + /* + * Copy over the last entry. Here we need to adjust len to the left over * size, and set src and dst to the entry it is copied to. */ - last_req = dbc->req_q_base + (tail + first_n) % dbc->nelem * get_dbc_req_elem_size(); + last_req = fifo_at(dbc->req_q_base, (tail + first_n) % dbc->nelem); memcpy(last_req, reqs + slice->nents - 1, sizeof(*reqs)); /* @@ -1166,6 +1154,9 @@ static inline int copy_partial_exec_reqs(struct qaic_device *qdev, struct bo_sli last_req->len = cpu_to_le32((u32)last_bytes); last_req->src_addr = reqs[first_n].src_addr; last_req->dest_addr = reqs[first_n].dest_addr; + if (!last_bytes) + /* Disable DMA transfer */ + last_req->cmd = GENMASK(7, 2) & reqs[first_n].cmd; *ptail = (tail + first_n + 1) % dbc->nelem; @@ -1225,26 +1216,17 @@ static int send_bo_list_to_device(struct qaic_device *qdev, struct drm_file *fil bo->req_id = dbc->next_req_id++; list_for_each_entry(slice, &bo->slices, slice) { - /* - * If this slice does not fall under the given - * resize then skip this slice and continue the loop - */ - if (is_partial && pexec[i].resize && pexec[i].resize <= slice->offset) - continue; - for (j = 0; j < slice->nents; j++) slice->reqs[j].req_id = cpu_to_le16(bo->req_id); - /* - * If it is a partial execute ioctl call then check if - * resize has cut this slice short then do a partial copy - * else do complete copy - */ - if (is_partial && pexec[i].resize && - pexec[i].resize < slice->offset + slice->size) + if (is_partial && (!pexec[i].resize || pexec[i].resize <= slice->offset)) + /* Configure the slice for no DMA transfer */ + ret = copy_partial_exec_reqs(qdev, slice, 0, dbc, head, tail); + else if (is_partial && pexec[i].resize < slice->offset + slice->size) + /* Configure the slice to be partially DMA transferred */ ret = copy_partial_exec_reqs(qdev, slice, - pexec[i].resize - slice->offset, - dbc->id, head, tail); + pexec[i].resize - slice->offset, dbc, + head, tail); else ret = copy_exec_reqs(qdev, slice, dbc->id, head, tail); if (ret) { @@ -1357,7 +1339,7 @@ static int __qaic_execute_bo_ioctl(struct drm_device *dev, void *data, struct dr qdev = usr->qddev->qdev; qdev_rcu_id = srcu_read_lock(&qdev->dev_lock); - if (qdev->in_reset) { + if (qdev->dev_state != QAIC_ONLINE) { ret = -ENODEV; goto unlock_dev_srcu; } @@ -1464,6 +1446,16 @@ irqreturn_t dbc_irq_handler(int irq, void *data) rcu_id = srcu_read_lock(&dbc->ch_lock); + if (datapath_polling) { + srcu_read_unlock(&dbc->ch_lock, rcu_id); + /* + * Normally datapath_polling will not have irqs enabled, but + * when running with only one MSI the interrupt is shared with + * MHI so it cannot be disabled. Return ASAP instead. + */ + return IRQ_HANDLED; + } + if (!dbc->usr) { srcu_read_unlock(&dbc->ch_lock, rcu_id); return IRQ_HANDLED; @@ -1486,7 +1478,8 @@ irqreturn_t dbc_irq_handler(int irq, void *data) return IRQ_NONE; } - disable_irq_nosync(irq); + if (!dbc->qdev->single_msi) + disable_irq_nosync(irq); srcu_read_unlock(&dbc->ch_lock, rcu_id); return IRQ_WAKE_THREAD; } @@ -1502,7 +1495,7 @@ void irq_polling_work(struct work_struct *work) rcu_id = srcu_read_lock(&dbc->ch_lock); while (1) { - if (dbc->qdev->in_reset) { + if (dbc->qdev->dev_state != QAIC_ONLINE) { srcu_read_unlock(&dbc->ch_lock, rcu_id); return; } @@ -1557,12 +1550,12 @@ irqreturn_t dbc_irq_threaded_fn(int irq, void *data) u32 tail; rcu_id = srcu_read_lock(&dbc->ch_lock); + qdev = dbc->qdev; head = readl(dbc->dbc_base + RSPHP_OFF); if (head == U32_MAX) /* PCI link error */ goto error_out; - qdev = dbc->qdev; read_fifo: if (!event_count) { @@ -1643,14 +1636,14 @@ read_fifo: goto read_fifo; normal_out: - if (likely(!datapath_polling)) + if (!qdev->single_msi && likely(!datapath_polling)) enable_irq(irq); - else + else if (unlikely(datapath_polling)) schedule_work(&dbc->poll_work); /* checking the fifo and enabling irqs is a race, missed event check */ tail = readl(dbc->dbc_base + RSPTP_OFF); if (tail != U32_MAX && head != tail) { - if (likely(!datapath_polling)) + if (!qdev->single_msi && likely(!datapath_polling)) disable_irq_nosync(irq); goto read_fifo; } @@ -1659,9 +1652,9 @@ normal_out: error_out: srcu_read_unlock(&dbc->ch_lock, rcu_id); - if (likely(!datapath_polling)) + if (!qdev->single_msi && likely(!datapath_polling)) enable_irq(irq); - else + else if (unlikely(datapath_polling)) schedule_work(&dbc->poll_work); return IRQ_HANDLED; @@ -1692,7 +1685,7 @@ int qaic_wait_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *file qdev = usr->qddev->qdev; qdev_rcu_id = srcu_read_lock(&qdev->dev_lock); - if (qdev->in_reset) { + if (qdev->dev_state != QAIC_ONLINE) { ret = -ENODEV; goto unlock_dev_srcu; } @@ -1761,7 +1754,7 @@ int qaic_perf_stats_bo_ioctl(struct drm_device *dev, void *data, struct drm_file qdev = usr->qddev->qdev; qdev_rcu_id = srcu_read_lock(&qdev->dev_lock); - if (qdev->in_reset) { + if (qdev->dev_state != QAIC_ONLINE) { ret = -ENODEV; goto unlock_dev_srcu; } @@ -1852,7 +1845,7 @@ int qaic_detach_slice_bo_ioctl(struct drm_device *dev, void *data, struct drm_fi qdev = usr->qddev->qdev; qdev_rcu_id = srcu_read_lock(&qdev->dev_lock); - if (qdev->in_reset) { + if (qdev->dev_state != QAIC_ONLINE) { ret = -ENODEV; goto unlock_dev_srcu; } diff --git a/drivers/accel/qaic/qaic_drv.c b/drivers/accel/qaic/qaic_drv.c index 6f58095767..2a313eb69b 100644 --- a/drivers/accel/qaic/qaic_drv.c +++ b/drivers/accel/qaic/qaic_drv.c @@ -8,6 +8,7 @@ #include <linux/idr.h> #include <linux/interrupt.h> #include <linux/list.h> +#include <linux/kobject.h> #include <linux/kref.h> #include <linux/mhi.h> #include <linux/module.h> @@ -27,6 +28,7 @@ #include "mhi_controller.h" #include "qaic.h" +#include "qaic_timesync.h" MODULE_IMPORT_NS(DMA_BUF); @@ -42,9 +44,6 @@ MODULE_PARM_DESC(datapath_polling, "Operate the datapath in polling mode"); static bool link_up; static DEFINE_IDA(qaic_usrs); -static int qaic_create_drm_device(struct qaic_device *qdev, s32 partition_id); -static void qaic_destroy_drm_device(struct qaic_device *qdev, s32 partition_id); - static void free_usr(struct kref *kref) { struct qaic_user *usr = container_of(kref, struct qaic_user, ref_count); @@ -63,7 +62,7 @@ static int qaic_open(struct drm_device *dev, struct drm_file *file) int ret; rcu_id = srcu_read_lock(&qdev->dev_lock); - if (qdev->in_reset) { + if (qdev->dev_state != QAIC_ONLINE) { ret = -ENODEV; goto dev_unlock; } @@ -120,7 +119,7 @@ static void qaic_postclose(struct drm_device *dev, struct drm_file *file) if (qddev) { qdev = qddev->qdev; qdev_rcu_id = srcu_read_lock(&qdev->dev_lock); - if (!qdev->in_reset) { + if (qdev->dev_state == QAIC_ONLINE) { qaic_release_usr(qdev, usr); for (i = 0; i < qdev->num_dbc; ++i) if (qdev->dbc[i].usr && qdev->dbc[i].usr->handle == usr->handle) @@ -182,13 +181,6 @@ static int qaic_create_drm_device(struct qaic_device *qdev, s32 partition_id) qddev->partition_id = partition_id; - /* - * drm_dev_unregister() sets the driver data to NULL and - * drm_dev_register() does not update the driver data. During a SOC - * reset drm dev is unregistered and registered again leaving the - * driver data to NULL. - */ - dev_set_drvdata(to_accel_kdev(qddev), drm->accel); ret = drm_dev_register(drm, 0); if (ret) pci_dbg(qdev->pdev, "drm_dev_register failed %d\n", ret); @@ -202,7 +194,6 @@ static void qaic_destroy_drm_device(struct qaic_device *qdev, s32 partition_id) struct drm_device *drm = to_drm(qddev); struct qaic_user *usr; - drm_dev_get(drm); drm_dev_unregister(drm); qddev->partition_id = 0; /* @@ -231,7 +222,6 @@ static void qaic_destroy_drm_device(struct qaic_device *qdev, s32 partition_id) mutex_lock(&qddev->users_mutex); } mutex_unlock(&qddev->users_mutex); - drm_dev_put(drm); } static int qaic_mhi_probe(struct mhi_device *mhi_dev, const struct mhi_device_id *id) @@ -253,8 +243,6 @@ static int qaic_mhi_probe(struct mhi_device *mhi_dev, const struct mhi_device_id qdev = pci_get_drvdata(to_pci_dev(mhi_dev->mhi_cntrl->cntrl_dev)); - qdev->in_reset = false; - dev_set_drvdata(&mhi_dev->dev, qdev); qdev->cntl_ch = mhi_dev; @@ -264,6 +252,7 @@ static int qaic_mhi_probe(struct mhi_device *mhi_dev, const struct mhi_device_id return ret; } + qdev->dev_state = QAIC_BOOT; ret = get_cntl_version(qdev, NULL, &major, &minor); if (ret || major != CNTL_MAJOR || minor > CNTL_MINOR) { pci_err(qdev->pdev, "%s: Control protocol version (%d.%d) not supported. Supported version is (%d.%d). Ret: %d\n", @@ -271,8 +260,8 @@ static int qaic_mhi_probe(struct mhi_device *mhi_dev, const struct mhi_device_id ret = -EINVAL; goto close_control; } - - ret = qaic_create_drm_device(qdev, QAIC_NO_PARTITION); + qdev->dev_state = QAIC_ONLINE; + kobject_uevent(&(to_accel_kdev(qdev->qddev))->kobj, KOBJ_ONLINE); return ret; @@ -290,7 +279,8 @@ static void qaic_notify_reset(struct qaic_device *qdev) { int i; - qdev->in_reset = true; + kobject_uevent(&(to_accel_kdev(qdev->qddev))->kobj, KOBJ_OFFLINE); + qdev->dev_state = QAIC_OFFLINE; /* wake up any waiters to avoid waiting for timeouts at sync */ wake_all_cntl(qdev); for (i = 0; i < qdev->num_dbc; ++i) @@ -298,21 +288,15 @@ static void qaic_notify_reset(struct qaic_device *qdev) synchronize_srcu(&qdev->dev_lock); } -void qaic_dev_reset_clean_local_state(struct qaic_device *qdev, bool exit_reset) +void qaic_dev_reset_clean_local_state(struct qaic_device *qdev) { int i; qaic_notify_reset(qdev); - /* remove drmdevs to prevent new users from coming in */ - qaic_destroy_drm_device(qdev, QAIC_NO_PARTITION); - /* start tearing things down */ for (i = 0; i < qdev->num_dbc; ++i) release_dbc(qdev, i); - - if (exit_reset) - qdev->in_reset = false; } static void cleanup_qdev(struct qaic_device *qdev) @@ -324,6 +308,7 @@ static void cleanup_qdev(struct qaic_device *qdev) cleanup_srcu_struct(&qdev->dev_lock); pci_set_drvdata(qdev->pdev, NULL); destroy_workqueue(qdev->cntl_wq); + destroy_workqueue(qdev->qts_wq); } static struct qaic_device *create_qdev(struct pci_dev *pdev, const struct pci_device_id *id) @@ -336,6 +321,7 @@ static struct qaic_device *create_qdev(struct pci_dev *pdev, const struct pci_de if (!qdev) return NULL; + qdev->dev_state = QAIC_OFFLINE; if (id->device == PCI_DEV_AIC100) { qdev->num_dbc = 16; qdev->dbc = devm_kcalloc(&pdev->dev, qdev->num_dbc, sizeof(*qdev->dbc), GFP_KERNEL); @@ -347,6 +333,12 @@ static struct qaic_device *create_qdev(struct pci_dev *pdev, const struct pci_de if (!qdev->cntl_wq) return NULL; + qdev->qts_wq = alloc_workqueue("qaic_ts", WQ_UNBOUND, 0); + if (!qdev->qts_wq) { + destroy_workqueue(qdev->cntl_wq); + return NULL; + } + pci_set_drvdata(pdev, qdev); qdev->pdev = pdev; @@ -424,14 +416,24 @@ static int init_msi(struct qaic_device *qdev, struct pci_dev *pdev) int i; /* Managed release since we use pcim_enable_device */ - ret = pci_alloc_irq_vectors(pdev, 1, 32, PCI_IRQ_MSI); - if (ret < 0) - return ret; + ret = pci_alloc_irq_vectors(pdev, 32, 32, PCI_IRQ_MSI); + if (ret == -ENOSPC) { + ret = pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_MSI); + if (ret < 0) + return ret; - if (ret < 32) { - pci_err(pdev, "%s: Requested 32 MSIs. Obtained %d MSIs which is less than the 32 required.\n", - __func__, ret); - return -ENODEV; + /* + * Operate in one MSI mode. All interrupts will be directed to + * MSI0; every interrupt will wake up all the interrupt handlers + * (MHI and DBC[0-15]). Since the interrupt is now shared, it is + * not disabled during DBC threaded handler, but only one thread + * will be allowed to run per DBC, so while it can be + * interrupted, it shouldn't race with itself. + */ + qdev->single_msi = true; + pci_info(pdev, "Allocating 32 MSIs failed, operating in 1 MSI mode. Performance may be impacted.\n"); + } else if (ret < 0) { + return ret; } mhi_irq = pci_irq_vector(pdev, 0); @@ -439,15 +441,17 @@ static int init_msi(struct qaic_device *qdev, struct pci_dev *pdev) return mhi_irq; for (i = 0; i < qdev->num_dbc; ++i) { - ret = devm_request_threaded_irq(&pdev->dev, pci_irq_vector(pdev, i + 1), + ret = devm_request_threaded_irq(&pdev->dev, + pci_irq_vector(pdev, qdev->single_msi ? 0 : i + 1), dbc_irq_handler, dbc_irq_threaded_fn, IRQF_SHARED, "qaic_dbc", &qdev->dbc[i]); if (ret) return ret; if (datapath_polling) { - qdev->dbc[i].irq = pci_irq_vector(pdev, i + 1); - disable_irq_nosync(qdev->dbc[i].irq); + qdev->dbc[i].irq = pci_irq_vector(pdev, qdev->single_msi ? 0 : i + 1); + if (!qdev->single_msi) + disable_irq_nosync(qdev->dbc[i].irq); INIT_WORK(&qdev->dbc[i].poll_work, irq_polling_work); } } @@ -479,14 +483,21 @@ static int qaic_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id) goto cleanup_qdev; } - qdev->mhi_cntrl = qaic_mhi_register_controller(pdev, qdev->bar_0, mhi_irq); + ret = qaic_create_drm_device(qdev, QAIC_NO_PARTITION); + if (ret) + goto cleanup_qdev; + + qdev->mhi_cntrl = qaic_mhi_register_controller(pdev, qdev->bar_0, mhi_irq, + qdev->single_msi); if (IS_ERR(qdev->mhi_cntrl)) { ret = PTR_ERR(qdev->mhi_cntrl); - goto cleanup_qdev; + goto cleanup_drm_dev; } return 0; +cleanup_drm_dev: + qaic_destroy_drm_device(qdev, QAIC_NO_PARTITION); cleanup_qdev: cleanup_qdev(qdev); return ret; @@ -499,7 +510,8 @@ static void qaic_pci_remove(struct pci_dev *pdev) if (!qdev) return; - qaic_dev_reset_clean_local_state(qdev, false); + qaic_dev_reset_clean_local_state(qdev); + qaic_destroy_drm_device(qdev, QAIC_NO_PARTITION); qaic_mhi_free_controller(qdev->mhi_cntrl, link_up); cleanup_qdev(qdev); } @@ -522,14 +534,13 @@ static void qaic_pci_reset_prepare(struct pci_dev *pdev) qaic_notify_reset(qdev); qaic_mhi_start_reset(qdev->mhi_cntrl); - qaic_dev_reset_clean_local_state(qdev, false); + qaic_dev_reset_clean_local_state(qdev); } static void qaic_pci_reset_done(struct pci_dev *pdev) { struct qaic_device *qdev = pci_get_drvdata(pdev); - qdev->in_reset = false; qaic_mhi_reset_done(qdev->mhi_cntrl); } @@ -586,6 +597,10 @@ static int __init qaic_init(void) goto free_pci; } + ret = qaic_timesync_init(); + if (ret) + pr_debug("qaic: qaic_timesync_init failed %d\n", ret); + return 0; free_pci: @@ -611,6 +626,7 @@ static void __exit qaic_exit(void) * reinitializing the link_up state after the cleanup is done. */ link_up = true; + qaic_timesync_deinit(); mhi_driver_unregister(&qaic_mhi_driver); pci_unregister_driver(&qaic_pci_driver); } diff --git a/drivers/accel/qaic/qaic_timesync.c b/drivers/accel/qaic/qaic_timesync.c new file mode 100644 index 0000000000..301f4462d5 --- /dev/null +++ b/drivers/accel/qaic/qaic_timesync.c @@ -0,0 +1,395 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright (c) 2023 Qualcomm Innovation Center, Inc. All rights reserved. */ + +#include <linux/io.h> +#include <linux/kernel.h> +#include <linux/math64.h> +#include <linux/mhi.h> +#include <linux/mod_devicetable.h> +#include <linux/module.h> +#include <linux/time64.h> +#include <linux/timer.h> + +#include "qaic.h" +#include "qaic_timesync.h" + +#define QTIMER_REG_OFFSET 0xa28 +#define QAIC_TIMESYNC_SIGNATURE 0x55aa +#define QAIC_CONV_QTIMER_TO_US(qtimer) (mul_u64_u32_div(qtimer, 10, 192)) + +static unsigned int timesync_delay_ms = 1000; /* 1 sec default */ +module_param(timesync_delay_ms, uint, 0600); +MODULE_PARM_DESC(timesync_delay_ms, "Delay in ms between two consecutive timesync operations"); + +enum qts_msg_type { + QAIC_TS_CMD_TO_HOST, + QAIC_TS_SYNC_REQ, + QAIC_TS_ACK_TO_HOST, + QAIC_TS_MSG_TYPE_MAX +}; + +/** + * struct qts_hdr - Timesync message header structure. + * @signature: Unique signature to identify the timesync message. + * @reserved_1: Reserved for future use. + * @reserved_2: Reserved for future use. + * @msg_type: sub-type of the timesync message. + * @reserved_3: Reserved for future use. + */ +struct qts_hdr { + __le16 signature; + __le16 reserved_1; + u8 reserved_2; + u8 msg_type; + __le16 reserved_3; +} __packed; + +/** + * struct qts_timeval - Structure to carry time information. + * @tv_sec: Seconds part of the time. + * @tv_usec: uS (microseconds) part of the time. + */ +struct qts_timeval { + __le64 tv_sec; + __le64 tv_usec; +} __packed; + +/** + * struct qts_host_time_sync_msg_data - Structure to denote the timesync message. + * @header: Header of the timesync message. + * @data: Time information. + */ +struct qts_host_time_sync_msg_data { + struct qts_hdr header; + struct qts_timeval data; +} __packed; + +/** + * struct mqts_dev - MHI QAIC Timesync Control device. + * @qdev: Pointer to the root device struct driven by QAIC driver. + * @mhi_dev: Pointer to associated MHI device. + * @timer: Timer handle used for timesync. + * @qtimer_addr: Device QTimer register pointer. + * @buff_in_use: atomic variable to track if the sync_msg buffer is in use. + * @dev: Device pointer to qdev->pdev->dev stored for easy access. + * @sync_msg: Buffer used to send timesync message over MHI. + */ +struct mqts_dev { + struct qaic_device *qdev; + struct mhi_device *mhi_dev; + struct timer_list timer; + void __iomem *qtimer_addr; + atomic_t buff_in_use; + struct device *dev; + struct qts_host_time_sync_msg_data *sync_msg; +}; + +struct qts_resp_msg { + struct qts_hdr hdr; +} __packed; + +struct qts_resp { + struct qts_resp_msg data; + struct work_struct work; + struct qaic_device *qdev; +}; + +#ifdef readq +static u64 read_qtimer(const volatile void __iomem *addr) +{ + return readq(addr); +} +#else +static u64 read_qtimer(const volatile void __iomem *addr) +{ + u64 low, high; + + low = readl(addr); + high = readl(addr + sizeof(u32)); + return low | (high << 32); +} +#endif + +static void qaic_timesync_ul_xfer_cb(struct mhi_device *mhi_dev, struct mhi_result *mhi_result) +{ + struct mqts_dev *mqtsdev = dev_get_drvdata(&mhi_dev->dev); + + dev_dbg(mqtsdev->dev, "%s status: %d xfer_len: %zu\n", __func__, + mhi_result->transaction_status, mhi_result->bytes_xferd); + + atomic_set(&mqtsdev->buff_in_use, 0); +} + +static void qaic_timesync_dl_xfer_cb(struct mhi_device *mhi_dev, struct mhi_result *mhi_result) +{ + struct mqts_dev *mqtsdev = dev_get_drvdata(&mhi_dev->dev); + + dev_err(mqtsdev->dev, "%s no data expected on dl channel\n", __func__); +} + +static void qaic_timesync_timer(struct timer_list *t) +{ + struct mqts_dev *mqtsdev = from_timer(mqtsdev, t, timer); + struct qts_host_time_sync_msg_data *sync_msg; + u64 device_qtimer_us; + u64 device_qtimer; + u64 host_time_us; + u64 offset_us; + u64 host_sec; + int ret; + + if (atomic_read(&mqtsdev->buff_in_use)) { + dev_dbg(mqtsdev->dev, "%s buffer not free, schedule next cycle\n", __func__); + goto mod_timer; + } + atomic_set(&mqtsdev->buff_in_use, 1); + + sync_msg = mqtsdev->sync_msg; + sync_msg->header.signature = cpu_to_le16(QAIC_TIMESYNC_SIGNATURE); + sync_msg->header.msg_type = QAIC_TS_SYNC_REQ; + /* Read host UTC time and convert to uS*/ + host_time_us = div_u64(ktime_get_real_ns(), NSEC_PER_USEC); + device_qtimer = read_qtimer(mqtsdev->qtimer_addr); + device_qtimer_us = QAIC_CONV_QTIMER_TO_US(device_qtimer); + /* Offset between host UTC and device time */ + offset_us = host_time_us - device_qtimer_us; + + host_sec = div_u64(offset_us, USEC_PER_SEC); + sync_msg->data.tv_usec = cpu_to_le64(offset_us - host_sec * USEC_PER_SEC); + sync_msg->data.tv_sec = cpu_to_le64(host_sec); + ret = mhi_queue_buf(mqtsdev->mhi_dev, DMA_TO_DEVICE, sync_msg, sizeof(*sync_msg), MHI_EOT); + if (ret && (ret != -EAGAIN)) { + dev_err(mqtsdev->dev, "%s unable to queue to mhi:%d\n", __func__, ret); + return; + } else if (ret == -EAGAIN) { + atomic_set(&mqtsdev->buff_in_use, 0); + } + +mod_timer: + ret = mod_timer(t, jiffies + msecs_to_jiffies(timesync_delay_ms)); + if (ret) + dev_err(mqtsdev->dev, "%s mod_timer error:%d\n", __func__, ret); +} + +static int qaic_timesync_probe(struct mhi_device *mhi_dev, const struct mhi_device_id *id) +{ + struct qaic_device *qdev = pci_get_drvdata(to_pci_dev(mhi_dev->mhi_cntrl->cntrl_dev)); + struct mqts_dev *mqtsdev; + struct timer_list *timer; + int ret; + + mqtsdev = kzalloc(sizeof(*mqtsdev), GFP_KERNEL); + if (!mqtsdev) { + ret = -ENOMEM; + goto out; + } + + timer = &mqtsdev->timer; + mqtsdev->mhi_dev = mhi_dev; + mqtsdev->qdev = qdev; + mqtsdev->dev = &qdev->pdev->dev; + + mqtsdev->sync_msg = kzalloc(sizeof(*mqtsdev->sync_msg), GFP_KERNEL); + if (!mqtsdev->sync_msg) { + ret = -ENOMEM; + goto free_mqts_dev; + } + atomic_set(&mqtsdev->buff_in_use, 0); + + ret = mhi_prepare_for_transfer(mhi_dev); + if (ret) + goto free_sync_msg; + + /* Qtimer register pointer */ + mqtsdev->qtimer_addr = qdev->bar_0 + QTIMER_REG_OFFSET; + timer_setup(timer, qaic_timesync_timer, 0); + timer->expires = jiffies + msecs_to_jiffies(timesync_delay_ms); + add_timer(timer); + dev_set_drvdata(&mhi_dev->dev, mqtsdev); + + return 0; + +free_sync_msg: + kfree(mqtsdev->sync_msg); +free_mqts_dev: + kfree(mqtsdev); +out: + return ret; +}; + +static void qaic_timesync_remove(struct mhi_device *mhi_dev) +{ + struct mqts_dev *mqtsdev = dev_get_drvdata(&mhi_dev->dev); + + del_timer_sync(&mqtsdev->timer); + mhi_unprepare_from_transfer(mqtsdev->mhi_dev); + kfree(mqtsdev->sync_msg); + kfree(mqtsdev); +} + +static const struct mhi_device_id qaic_timesync_match_table[] = { + { .chan = "QAIC_TIMESYNC_PERIODIC"}, + {}, +}; + +MODULE_DEVICE_TABLE(mhi, qaic_timesync_match_table); + +static struct mhi_driver qaic_timesync_driver = { + .id_table = qaic_timesync_match_table, + .remove = qaic_timesync_remove, + .probe = qaic_timesync_probe, + .ul_xfer_cb = qaic_timesync_ul_xfer_cb, + .dl_xfer_cb = qaic_timesync_dl_xfer_cb, + .driver = { + .name = "qaic_timesync_periodic", + }, +}; + +static void qaic_boot_timesync_worker(struct work_struct *work) +{ + struct qts_resp *resp = container_of(work, struct qts_resp, work); + struct qts_host_time_sync_msg_data *req; + struct qts_resp_msg data = resp->data; + struct qaic_device *qdev = resp->qdev; + struct mhi_device *mhi_dev; + struct timespec64 ts; + int ret; + + mhi_dev = qdev->qts_ch; + /* Queue the response message beforehand to avoid race conditions */ + ret = mhi_queue_buf(mhi_dev, DMA_FROM_DEVICE, &resp->data, sizeof(resp->data), MHI_EOT); + if (ret) { + kfree(resp); + dev_warn(&mhi_dev->dev, "Failed to re-queue response buffer %d\n", ret); + return; + } + + switch (data.hdr.msg_type) { + case QAIC_TS_CMD_TO_HOST: + req = kzalloc(sizeof(*req), GFP_KERNEL); + if (!req) + break; + + req->header = data.hdr; + req->header.msg_type = QAIC_TS_SYNC_REQ; + ktime_get_real_ts64(&ts); + req->data.tv_sec = cpu_to_le64(ts.tv_sec); + req->data.tv_usec = cpu_to_le64(div_u64(ts.tv_nsec, NSEC_PER_USEC)); + + ret = mhi_queue_buf(mhi_dev, DMA_TO_DEVICE, req, sizeof(*req), MHI_EOT); + if (ret) { + kfree(req); + dev_dbg(&mhi_dev->dev, "Failed to send request message. Error %d\n", ret); + } + break; + case QAIC_TS_ACK_TO_HOST: + dev_dbg(&mhi_dev->dev, "ACK received from device\n"); + break; + default: + dev_err(&mhi_dev->dev, "Invalid message type %u.\n", data.hdr.msg_type); + } +} + +static int qaic_boot_timesync_queue_resp(struct mhi_device *mhi_dev, struct qaic_device *qdev) +{ + struct qts_resp *resp; + int ret; + + resp = kzalloc(sizeof(*resp), GFP_KERNEL); + if (!resp) + return -ENOMEM; + + resp->qdev = qdev; + INIT_WORK(&resp->work, qaic_boot_timesync_worker); + + ret = mhi_queue_buf(mhi_dev, DMA_FROM_DEVICE, &resp->data, sizeof(resp->data), MHI_EOT); + if (ret) { + kfree(resp); + dev_warn(&mhi_dev->dev, "Failed to queue response buffer %d\n", ret); + return ret; + } + + return 0; +} + +static void qaic_boot_timesync_remove(struct mhi_device *mhi_dev) +{ + struct qaic_device *qdev; + + qdev = dev_get_drvdata(&mhi_dev->dev); + mhi_unprepare_from_transfer(qdev->qts_ch); + qdev->qts_ch = NULL; +} + +static int qaic_boot_timesync_probe(struct mhi_device *mhi_dev, const struct mhi_device_id *id) +{ + struct qaic_device *qdev = pci_get_drvdata(to_pci_dev(mhi_dev->mhi_cntrl->cntrl_dev)); + int ret; + + ret = mhi_prepare_for_transfer(mhi_dev); + if (ret) + return ret; + + qdev->qts_ch = mhi_dev; + dev_set_drvdata(&mhi_dev->dev, qdev); + + ret = qaic_boot_timesync_queue_resp(mhi_dev, qdev); + if (ret) { + dev_set_drvdata(&mhi_dev->dev, NULL); + qdev->qts_ch = NULL; + mhi_unprepare_from_transfer(mhi_dev); + } + + return ret; +} + +static void qaic_boot_timesync_ul_xfer_cb(struct mhi_device *mhi_dev, struct mhi_result *mhi_result) +{ + kfree(mhi_result->buf_addr); +} + +static void qaic_boot_timesync_dl_xfer_cb(struct mhi_device *mhi_dev, struct mhi_result *mhi_result) +{ + struct qts_resp *resp = container_of(mhi_result->buf_addr, struct qts_resp, data); + + if (mhi_result->transaction_status || mhi_result->bytes_xferd != sizeof(resp->data)) { + kfree(resp); + return; + } + + queue_work(resp->qdev->qts_wq, &resp->work); +} + +static const struct mhi_device_id qaic_boot_timesync_match_table[] = { + { .chan = "QAIC_TIMESYNC"}, + {}, +}; + +static struct mhi_driver qaic_boot_timesync_driver = { + .id_table = qaic_boot_timesync_match_table, + .remove = qaic_boot_timesync_remove, + .probe = qaic_boot_timesync_probe, + .ul_xfer_cb = qaic_boot_timesync_ul_xfer_cb, + .dl_xfer_cb = qaic_boot_timesync_dl_xfer_cb, + .driver = { + .name = "qaic_timesync", + }, +}; + +int qaic_timesync_init(void) +{ + int ret; + + ret = mhi_driver_register(&qaic_timesync_driver); + if (ret) + return ret; + ret = mhi_driver_register(&qaic_boot_timesync_driver); + + return ret; +} + +void qaic_timesync_deinit(void) +{ + mhi_driver_unregister(&qaic_boot_timesync_driver); + mhi_driver_unregister(&qaic_timesync_driver); +} diff --git a/drivers/accel/qaic/qaic_timesync.h b/drivers/accel/qaic/qaic_timesync.h new file mode 100644 index 0000000000..851b7acd43 --- /dev/null +++ b/drivers/accel/qaic/qaic_timesync.h @@ -0,0 +1,11 @@ +/* SPDX-License-Identifier: GPL-2.0-only + * + * Copyright (c) 2023 Qualcomm Innovation Center, Inc. All rights reserved. + */ + +#ifndef __QAIC_TIMESYNC_H__ +#define __QAIC_TIMESYNC_H__ + +int qaic_timesync_init(void); +void qaic_timesync_deinit(void); +#endif /* __QAIC_TIMESYNC_H__ */ |