diff options
Diffstat (limited to 'drivers/accel/qaic')
| -rw-r--r-- | drivers/accel/qaic/Makefile | 5 | ||||
| -rw-r--r-- | drivers/accel/qaic/mhi_controller.c | 38 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic.h | 12 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic_control.c | 2 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic_data.c | 83 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic_debugfs.c | 305 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic_debugfs.h | 20 | ||||
| -rw-r--r-- | drivers/accel/qaic/qaic_drv.c | 177 | ||||
| -rw-r--r-- | drivers/accel/qaic/sahara.c | 822 | ||||
| -rw-r--r-- | drivers/accel/qaic/sahara.h | 10 |
10 files changed, 1371 insertions, 103 deletions
diff --git a/drivers/accel/qaic/Makefile b/drivers/accel/qaic/Makefile index 3f7f6dfde7f2..35e883515629 100644 --- a/drivers/accel/qaic/Makefile +++ b/drivers/accel/qaic/Makefile @@ -10,4 +10,7 @@ qaic-y := \ qaic_control.o \ qaic_data.o \ qaic_drv.o \ - qaic_timesync.o + qaic_timesync.o \ + sahara.o + +qaic-$(CONFIG_DEBUG_FS) += qaic_debugfs.o diff --git a/drivers/accel/qaic/mhi_controller.c b/drivers/accel/qaic/mhi_controller.c index cb77d048ed54..8ab82e78dd94 100644 --- a/drivers/accel/qaic/mhi_controller.c +++ b/drivers/accel/qaic/mhi_controller.c @@ -20,7 +20,7 @@ static unsigned int mhi_timeout_ms = 2000; /* 2 sec default */ module_param(mhi_timeout_ms, uint, 0600); MODULE_PARM_DESC(mhi_timeout_ms, "MHI controller timeout value"); -static struct mhi_channel_config aic100_channels[] = { +static const struct mhi_channel_config aic100_channels[] = { { .name = "QAIC_LOOPBACK", .num = 0, @@ -358,8 +358,8 @@ static struct mhi_channel_config aic100_channels[] = { .wake_capable = false, }, { - .num = 21, .name = "QAIC_TIMESYNC", + .num = 21, .num_elements = 32, .local_elements = 0, .event_ring = 0, @@ -390,8 +390,8 @@ static struct mhi_channel_config aic100_channels[] = { .wake_capable = false, }, { - .num = 23, .name = "QAIC_TIMESYNC_PERIODIC", + .num = 23, .num_elements = 32, .local_elements = 0, .event_ring = 0, @@ -405,6 +405,38 @@ static struct mhi_channel_config aic100_channels[] = { .auto_queue = false, .wake_capable = false, }, + { + .name = "IPCR", + .num = 24, + .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, + }, + { + .name = "IPCR", + .num = 25, + .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, + .offload_channel = false, + .doorbell_mode_switch = false, + .auto_queue = true, + .wake_capable = false, + }, }; static struct mhi_event_config aic100_events[] = { diff --git a/drivers/accel/qaic/qaic.h b/drivers/accel/qaic/qaic.h index 582836f9538f..02561b6cecc6 100644 --- a/drivers/accel/qaic/qaic.h +++ b/drivers/accel/qaic/qaic.h @@ -30,6 +30,7 @@ #define to_qaic_drm_device(dev) container_of(dev, struct qaic_drm_device, drm) #define to_drm(qddev) (&(qddev)->drm) #define to_accel_kdev(qddev) (to_drm(qddev)->accel->kdev) /* Return Linux device of accel node */ +#define to_qaic_device(dev) (to_qaic_drm_device((dev))->qdev) enum __packed dev_states { /* Device is offline or will be very soon */ @@ -152,6 +153,14 @@ struct qaic_device { struct mhi_device *qts_ch; /* Work queue for tasks related to MHI "QAIC_TIMESYNC" channel */ struct workqueue_struct *qts_wq; + /* Head of list of page allocated by MHI bootlog device */ + struct list_head bootlog; + /* MHI bootlog channel device */ + struct mhi_device *bootlog_ch; + /* Work queue for tasks related to MHI bootlog device */ + struct workqueue_struct *bootlog_wq; + /* Synchronizes access of pages in MHI bootlog device */ + struct mutex bootlog_mutex; }; struct qaic_drm_device { @@ -191,8 +200,6 @@ struct qaic_bo { u32 nr_slice; /* Number of slice that have been transferred by DMA engine */ u32 nr_slice_xfer_done; - /* true = BO is queued for execution, true = BO is not queued */ - bool queued; /* * If true then user has attached slicing information to this BO by * calling DRM_IOCTL_QAIC_ATTACH_SLICE_BO ioctl. @@ -281,6 +288,7 @@ int disable_dbc(struct qaic_device *qdev, u32 dbc_id, struct qaic_user *usr); void enable_dbc(struct qaic_device *qdev, u32 dbc_id, struct qaic_user *usr); void wakeup_dbc(struct qaic_device *qdev, u32 dbc_id); void release_dbc(struct qaic_device *qdev, u32 dbc_id); +void qaic_data_get_fifo_info(struct dma_bridge_chan *dbc, u32 *head, u32 *tail); void wake_all_cntl(struct qaic_device *qdev); void qaic_dev_reset_clean_local_state(struct qaic_device *qdev); diff --git a/drivers/accel/qaic/qaic_control.c b/drivers/accel/qaic/qaic_control.c index 9e8a8cbadf6b..d8bdab69f800 100644 --- a/drivers/accel/qaic/qaic_control.c +++ b/drivers/accel/qaic/qaic_control.c @@ -496,7 +496,7 @@ static int encode_addr_size_pairs(struct dma_xfer *xfer, struct wrapper_list *wr nents = sgt->nents; nents_dma = nents; *size = QAIC_MANAGE_EXT_MSG_LENGTH - msg_hdr_len - sizeof(**out_trans); - for_each_sgtable_sg(sgt, sg, i) { + for_each_sgtable_dma_sg(sgt, sg, i) { *size -= sizeof(*asp); /* Save 1K for possible follow-up transactions. */ if (*size < SZ_1K) { diff --git a/drivers/accel/qaic/qaic_data.c b/drivers/accel/qaic/qaic_data.c index 03c9a793da35..43aba57b48f0 100644 --- a/drivers/accel/qaic/qaic_data.c +++ b/drivers/accel/qaic/qaic_data.c @@ -141,6 +141,11 @@ struct dbc_rsp { __le16 status; } __packed; +static inline bool bo_queued(struct qaic_bo *bo) +{ + return !list_empty(&bo->xfer_list); +} + inline int get_dbc_req_elem_size(void) { return sizeof(struct dbc_req); @@ -167,9 +172,10 @@ static void free_slice(struct kref *kref) static int clone_range_of_sgt_for_slice(struct qaic_device *qdev, struct sg_table **sgt_out, struct sg_table *sgt_in, u64 size, u64 offset) { - int total_len, len, nents, offf = 0, offl = 0; struct scatterlist *sg, *sgn, *sgf, *sgl; + unsigned int len, nents, offf, offl; struct sg_table *sgt; + size_t total_len; int ret, j; /* find out number of relevant nents needed for this mem */ @@ -177,9 +183,11 @@ static int clone_range_of_sgt_for_slice(struct qaic_device *qdev, struct sg_tabl sgf = NULL; sgl = NULL; nents = 0; + offf = 0; + offl = 0; size = size ? size : PAGE_SIZE; - for (sg = sgt_in->sgl; sg; sg = sg_next(sg)) { + for_each_sgtable_dma_sg(sgt_in, sg, j) { len = sg_dma_len(sg); if (!len) @@ -216,7 +224,7 @@ static int clone_range_of_sgt_for_slice(struct qaic_device *qdev, struct sg_tabl /* copy relevant sg node and fix page and length */ sgn = sgf; - for_each_sgtable_sg(sgt, sg, j) { + for_each_sgtable_dma_sg(sgt, sg, j) { memcpy(sg, sgn, sizeof(*sg)); if (sgn == sgf) { sg_dma_address(sg) += offf; @@ -296,7 +304,7 @@ static int encode_reqs(struct qaic_device *qdev, struct bo_slice *slice, * fence. */ dev_addr = req->dev_addr; - for_each_sgtable_sg(slice->sgt, sg, i) { + for_each_sgtable_dma_sg(slice->sgt, sg, i) { slice->reqs[i].cmd = cmd; slice->reqs[i].src_addr = cpu_to_le64(slice->dir == DMA_TO_DEVICE ? sg_dma_address(sg) : dev_addr); @@ -549,6 +557,7 @@ static bool invalid_sem(struct qaic_sem *sem) static int qaic_validate_req(struct qaic_device *qdev, struct qaic_attach_slice_entry *slice_ent, u32 count, u64 total_size) { + u64 total; int i; for (i = 0; i < count; i++) { @@ -558,7 +567,8 @@ static int qaic_validate_req(struct qaic_device *qdev, struct qaic_attach_slice_ invalid_sem(&slice_ent[i].sem2) || invalid_sem(&slice_ent[i].sem3)) return -EINVAL; - if (slice_ent[i].offset + slice_ent[i].size > total_size) + if (check_add_overflow(slice_ent[i].offset, slice_ent[i].size, &total) || + total > total_size) return -EINVAL; } @@ -569,6 +579,9 @@ static void qaic_free_sgt(struct sg_table *sgt) { struct scatterlist *sg; + if (!sgt) + return; + for (sg = sgt->sgl; sg; sg = sg_next(sg)) if (sg_page(sg)) __free_pages(sg_page(sg), get_order(sg->length)); @@ -648,6 +661,7 @@ static void qaic_init_bo(struct qaic_bo *bo, bool reinit) } complete_all(&bo->xfer_done); INIT_LIST_HEAD(&bo->slices); + INIT_LIST_HEAD(&bo->xfer_list); } static struct qaic_bo *qaic_alloc_init_bo(void) @@ -709,9 +723,13 @@ int qaic_create_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *fi if (ret) goto free_bo; + ret = drm_gem_create_mmap_offset(obj); + if (ret) + goto free_bo; + ret = drm_gem_handle_create(file_priv, obj, &args->handle); if (ret) - goto free_sgt; + goto free_bo; bo->handle = args->handle; drm_gem_object_put(obj); @@ -720,10 +738,8 @@ int qaic_create_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *fi return 0; -free_sgt: - qaic_free_sgt(bo->sgt); free_bo: - kfree(bo); + drm_gem_object_put(obj); unlock_dev_srcu: srcu_read_unlock(&qdev->dev_lock, qdev_rcu_id); unlock_usr_srcu: @@ -738,7 +754,7 @@ int qaic_mmap_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *file struct drm_gem_object *obj; struct qaic_device *qdev; struct qaic_user *usr; - int ret; + int ret = 0; usr = file_priv->driver_priv; usr_rcu_id = srcu_read_lock(&usr->qddev_lock); @@ -760,9 +776,7 @@ int qaic_mmap_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *file goto unlock_dev_srcu; } - ret = drm_gem_create_mmap_offset(obj); - if (ret == 0) - args->offset = drm_vma_node_offset_addr(&obj->vma_node); + args->offset = drm_vma_node_offset_addr(&obj->vma_node); drm_gem_object_put(obj); @@ -828,9 +842,6 @@ static int qaic_prepare_import_bo(struct qaic_bo *bo, struct qaic_attach_slice_h struct sg_table *sgt; int ret; - if (obj->import_attach->dmabuf->size < hdr->size) - return -EINVAL; - sgt = dma_buf_map_attachment(obj->import_attach, hdr->dir); if (IS_ERR(sgt)) { ret = PTR_ERR(sgt); @@ -847,9 +858,6 @@ static int qaic_prepare_export_bo(struct qaic_device *qdev, struct qaic_bo *bo, { int ret; - if (bo->base.size < hdr->size) - return -EINVAL; - ret = dma_map_sgtable(&qdev->pdev->dev, bo->sgt, hdr->dir, 0); if (ret) return -EFAULT; @@ -950,9 +958,6 @@ int qaic_attach_slice_bo_ioctl(struct drm_device *dev, void *data, struct drm_fi if (arg_size / args->hdr.count != sizeof(*slice_ent)) return -EINVAL; - if (args->hdr.size == 0) - return -EINVAL; - if (!(args->hdr.dir == DMA_TO_DEVICE || args->hdr.dir == DMA_FROM_DEVICE)) return -EINVAL; @@ -992,16 +997,16 @@ int qaic_attach_slice_bo_ioctl(struct drm_device *dev, void *data, struct drm_fi goto free_slice_ent; } - ret = qaic_validate_req(qdev, slice_ent, args->hdr.count, args->hdr.size); - if (ret) - goto free_slice_ent; - obj = drm_gem_object_lookup(file_priv, args->hdr.handle); if (!obj) { ret = -ENOENT; goto free_slice_ent; } + ret = qaic_validate_req(qdev, slice_ent, args->hdr.count, obj->size); + if (ret) + goto put_bo; + bo = to_qaic_bo(obj); ret = mutex_lock_interruptible(&bo->lock); if (ret) @@ -1173,7 +1178,6 @@ static int send_bo_list_to_device(struct qaic_device *qdev, struct drm_file *fil struct bo_slice *slice; unsigned long flags; struct qaic_bo *bo; - bool queued; int i, j; int ret; @@ -1205,9 +1209,7 @@ static int send_bo_list_to_device(struct qaic_device *qdev, struct drm_file *fil } spin_lock_irqsave(&dbc->xfer_lock, flags); - queued = bo->queued; - bo->queued = true; - if (queued) { + if (bo_queued(bo)) { spin_unlock_irqrestore(&dbc->xfer_lock, flags); ret = -EINVAL; goto unlock_bo; @@ -1230,7 +1232,6 @@ static int send_bo_list_to_device(struct qaic_device *qdev, struct drm_file *fil else ret = copy_exec_reqs(qdev, slice, dbc->id, head, tail); if (ret) { - bo->queued = false; spin_unlock_irqrestore(&dbc->xfer_lock, flags); goto unlock_bo; } @@ -1253,8 +1254,7 @@ failed_to_send_bo: spin_lock_irqsave(&dbc->xfer_lock, flags); bo = list_last_entry(&dbc->xfer_list, struct qaic_bo, xfer_list); obj = &bo->base; - bo->queued = false; - list_del(&bo->xfer_list); + list_del_init(&bo->xfer_list); spin_unlock_irqrestore(&dbc->xfer_lock, flags); dma_sync_sgtable_for_cpu(&qdev->pdev->dev, bo->sgt, bo->dir); drm_gem_object_put(obj); @@ -1615,8 +1615,7 @@ read_fifo: */ dma_sync_sgtable_for_cpu(&qdev->pdev->dev, bo->sgt, bo->dir); bo->nr_slice_xfer_done = 0; - bo->queued = false; - list_del(&bo->xfer_list); + list_del_init(&bo->xfer_list); bo->perf_stats.req_processed_ts = ktime_get_ns(); complete_all(&bo->xfer_done); drm_gem_object_put(&bo->base); @@ -1875,7 +1874,7 @@ int qaic_detach_slice_bo_ioctl(struct drm_device *dev, void *data, struct drm_fi /* Check if BO is committed to H/W for DMA */ spin_lock_irqsave(&dbc->xfer_lock, flags); - if (bo->queued) { + if (bo_queued(bo)) { spin_unlock_irqrestore(&dbc->xfer_lock, flags); ret = -EBUSY; goto unlock_ch_srcu; @@ -1905,8 +1904,7 @@ static void empty_xfer_list(struct qaic_device *qdev, struct dma_bridge_chan *db spin_lock_irqsave(&dbc->xfer_lock, flags); while (!list_empty(&dbc->xfer_list)) { bo = list_first_entry(&dbc->xfer_list, typeof(*bo), xfer_list); - bo->queued = false; - list_del(&bo->xfer_list); + list_del_init(&bo->xfer_list); spin_unlock_irqrestore(&dbc->xfer_lock, flags); bo->nr_slice_xfer_done = 0; bo->req_id = 0; @@ -1988,3 +1986,12 @@ void release_dbc(struct qaic_device *qdev, u32 dbc_id) dbc->in_use = false; wake_up(&dbc->dbc_release); } + +void qaic_data_get_fifo_info(struct dma_bridge_chan *dbc, u32 *head, u32 *tail) +{ + if (!dbc || !head || !tail) + return; + + *head = readl(dbc->dbc_base + REQHP_OFF); + *tail = readl(dbc->dbc_base + REQTP_OFF); +} diff --git a/drivers/accel/qaic/qaic_debugfs.c b/drivers/accel/qaic/qaic_debugfs.c new file mode 100644 index 000000000000..ba0cf2f94732 --- /dev/null +++ b/drivers/accel/qaic/qaic_debugfs.c @@ -0,0 +1,305 @@ +// SPDX-License-Identifier: GPL-2.0-only + +/* Copyright (c) 2020, The Linux Foundation. All rights reserved. */ +/* Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved. */ + +#include <linux/debugfs.h> +#include <linux/device.h> +#include <linux/fs.h> +#include <linux/list.h> +#include <linux/mhi.h> +#include <linux/mutex.h> +#include <linux/overflow.h> +#include <linux/pci.h> +#include <linux/seq_file.h> +#include <linux/sprintf.h> +#include <linux/string.h> +#include <linux/types.h> +#include <linux/workqueue.h> + +#include "qaic.h" +#include "qaic_debugfs.h" + +#define BOOTLOG_POOL_SIZE 16 +#define BOOTLOG_MSG_SIZE 512 +#define QAIC_DBC_DIR_NAME 9 + +struct bootlog_msg { + /* Buffer for bootlog messages */ + char str[BOOTLOG_MSG_SIZE]; + /* Root struct of device, used to access device resources */ + struct qaic_device *qdev; + /* Work struct to schedule work coming on QAIC_LOGGING channel */ + struct work_struct work; +}; + +struct bootlog_page { + /* Node in list of bootlog pages maintained by root device struct */ + struct list_head node; + /* Total size of the buffer that holds the bootlogs. It is PAGE_SIZE */ + unsigned int size; + /* Offset for the next bootlog */ + unsigned int offset; +}; + +static int bootlog_show(struct seq_file *s, void *unused) +{ + struct bootlog_page *page; + struct qaic_device *qdev; + void *page_end; + void *log; + + qdev = s->private; + mutex_lock(&qdev->bootlog_mutex); + list_for_each_entry(page, &qdev->bootlog, node) { + log = page + 1; + page_end = (void *)page + page->offset; + while (log < page_end) { + seq_printf(s, "%s", (char *)log); + log += strlen(log) + 1; + } + } + mutex_unlock(&qdev->bootlog_mutex); + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(bootlog); + +static int fifo_size_show(struct seq_file *s, void *unused) +{ + struct dma_bridge_chan *dbc = s->private; + + seq_printf(s, "%u\n", dbc->nelem); + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(fifo_size); + +static int queued_show(struct seq_file *s, void *unused) +{ + struct dma_bridge_chan *dbc = s->private; + u32 tail = 0, head = 0; + + qaic_data_get_fifo_info(dbc, &head, &tail); + + if (head == U32_MAX || tail == U32_MAX) + seq_printf(s, "%u\n", 0); + else if (head > tail) + seq_printf(s, "%u\n", dbc->nelem - head + tail); + else + seq_printf(s, "%u\n", tail - head); + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(queued); + +void qaic_debugfs_init(struct qaic_drm_device *qddev) +{ + struct qaic_device *qdev = qddev->qdev; + struct dentry *debugfs_root; + struct dentry *debugfs_dir; + char name[QAIC_DBC_DIR_NAME]; + u32 i; + + debugfs_root = to_drm(qddev)->debugfs_root; + + debugfs_create_file("bootlog", 0400, debugfs_root, qdev, &bootlog_fops); + /* + * 256 dbcs per device is likely the max we will ever see and lets static checking see a + * reasonable range. + */ + for (i = 0; i < qdev->num_dbc && i < 256; ++i) { + snprintf(name, QAIC_DBC_DIR_NAME, "dbc%03u", i); + debugfs_dir = debugfs_create_dir(name, debugfs_root); + debugfs_create_file("fifo_size", 0400, debugfs_dir, &qdev->dbc[i], &fifo_size_fops); + debugfs_create_file("queued", 0400, debugfs_dir, &qdev->dbc[i], &queued_fops); + } +} + +static struct bootlog_page *alloc_bootlog_page(struct qaic_device *qdev) +{ + struct bootlog_page *page; + + page = (struct bootlog_page *)devm_get_free_pages(&qdev->pdev->dev, GFP_KERNEL, 0); + if (!page) + return page; + + page->size = PAGE_SIZE; + page->offset = sizeof(*page); + list_add_tail(&page->node, &qdev->bootlog); + + return page; +} + +static int reset_bootlog(struct qaic_device *qdev) +{ + struct bootlog_page *page; + struct bootlog_page *i; + + mutex_lock(&qdev->bootlog_mutex); + list_for_each_entry_safe(page, i, &qdev->bootlog, node) { + list_del(&page->node); + devm_free_pages(&qdev->pdev->dev, (unsigned long)page); + } + + page = alloc_bootlog_page(qdev); + mutex_unlock(&qdev->bootlog_mutex); + if (!page) + return -ENOMEM; + + return 0; +} + +static void *bootlog_get_space(struct qaic_device *qdev, unsigned int size) +{ + struct bootlog_page *page; + + page = list_last_entry(&qdev->bootlog, struct bootlog_page, node); + + if (size_add(size, sizeof(*page)) > page->size) + return NULL; + + if (page->offset + size > page->size) { + page = alloc_bootlog_page(qdev); + if (!page) + return NULL; + } + + return (void *)page + page->offset; +} + +static void bootlog_commit(struct qaic_device *qdev, unsigned int size) +{ + struct bootlog_page *page; + + page = list_last_entry(&qdev->bootlog, struct bootlog_page, node); + + page->offset += size; +} + +static void bootlog_log(struct work_struct *work) +{ + struct bootlog_msg *msg = container_of(work, struct bootlog_msg, work); + unsigned int len = strlen(msg->str) + 1; + struct qaic_device *qdev = msg->qdev; + void *log; + + mutex_lock(&qdev->bootlog_mutex); + log = bootlog_get_space(qdev, len); + if (log) { + memcpy(log, msg, len); + bootlog_commit(qdev, len); + } + mutex_unlock(&qdev->bootlog_mutex); + + if (mhi_queue_buf(qdev->bootlog_ch, DMA_FROM_DEVICE, msg, BOOTLOG_MSG_SIZE, MHI_EOT)) + devm_kfree(&qdev->pdev->dev, msg); +} + +static int qaic_bootlog_mhi_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 bootlog_msg *msg; + int i, ret; + + qdev->bootlog_wq = alloc_ordered_workqueue("qaic_bootlog", 0); + if (!qdev->bootlog_wq) { + ret = -ENOMEM; + goto out; + } + + ret = reset_bootlog(qdev); + if (ret) + goto destroy_workqueue; + + ret = mhi_prepare_for_transfer(mhi_dev); + if (ret) + goto destroy_workqueue; + + for (i = 0; i < BOOTLOG_POOL_SIZE; i++) { + msg = devm_kzalloc(&qdev->pdev->dev, sizeof(*msg), GFP_KERNEL); + if (!msg) { + ret = -ENOMEM; + goto mhi_unprepare; + } + + msg->qdev = qdev; + INIT_WORK(&msg->work, bootlog_log); + + ret = mhi_queue_buf(mhi_dev, DMA_FROM_DEVICE, msg, BOOTLOG_MSG_SIZE, MHI_EOT); + if (ret) + goto mhi_unprepare; + } + + dev_set_drvdata(&mhi_dev->dev, qdev); + qdev->bootlog_ch = mhi_dev; + return 0; + +mhi_unprepare: + mhi_unprepare_from_transfer(mhi_dev); +destroy_workqueue: + flush_workqueue(qdev->bootlog_wq); + destroy_workqueue(qdev->bootlog_wq); +out: + return ret; +} + +static void qaic_bootlog_mhi_remove(struct mhi_device *mhi_dev) +{ + struct qaic_device *qdev; + + qdev = dev_get_drvdata(&mhi_dev->dev); + + mhi_unprepare_from_transfer(qdev->bootlog_ch); + flush_workqueue(qdev->bootlog_wq); + destroy_workqueue(qdev->bootlog_wq); + qdev->bootlog_ch = NULL; +} + +static void qaic_bootlog_mhi_ul_xfer_cb(struct mhi_device *mhi_dev, struct mhi_result *mhi_result) +{ +} + +static void qaic_bootlog_mhi_dl_xfer_cb(struct mhi_device *mhi_dev, struct mhi_result *mhi_result) +{ + struct qaic_device *qdev = dev_get_drvdata(&mhi_dev->dev); + struct bootlog_msg *msg = mhi_result->buf_addr; + + if (mhi_result->transaction_status) { + devm_kfree(&qdev->pdev->dev, msg); + return; + } + + /* Force a null at the end of the transferred string */ + msg->str[mhi_result->bytes_xferd - 1] = 0; + + queue_work(qdev->bootlog_wq, &msg->work); +} + +static const struct mhi_device_id qaic_bootlog_mhi_match_table[] = { + { .chan = "QAIC_LOGGING", }, + {}, +}; + +static struct mhi_driver qaic_bootlog_mhi_driver = { + .id_table = qaic_bootlog_mhi_match_table, + .remove = qaic_bootlog_mhi_remove, + .probe = qaic_bootlog_mhi_probe, + .ul_xfer_cb = qaic_bootlog_mhi_ul_xfer_cb, + .dl_xfer_cb = qaic_bootlog_mhi_dl_xfer_cb, + .driver = { + .name = "qaic_bootlog", + }, +}; + +int qaic_bootlog_register(void) +{ + return mhi_driver_register(&qaic_bootlog_mhi_driver); +} + +void qaic_bootlog_unregister(void) +{ + mhi_driver_unregister(&qaic_bootlog_mhi_driver); +} diff --git a/drivers/accel/qaic/qaic_debugfs.h b/drivers/accel/qaic/qaic_debugfs.h new file mode 100644 index 000000000000..05e74f84cf9f --- /dev/null +++ b/drivers/accel/qaic/qaic_debugfs.h @@ -0,0 +1,20 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ + +/* Copyright (c) 2020, The Linux Foundation. All rights reserved. */ +/* Copyright (c) 2024 Qualcomm Innovation Center, Inc. All rights reserved. */ + +#ifndef __QAIC_DEBUGFS_H__ +#define __QAIC_DEBUGFS_H__ + +#include <drm/drm_file.h> + +#ifdef CONFIG_DEBUG_FS +int qaic_bootlog_register(void); +void qaic_bootlog_unregister(void); +void qaic_debugfs_init(struct qaic_drm_device *qddev); +#else +static inline int qaic_bootlog_register(void) { return 0; } +static inline void qaic_bootlog_unregister(void) {} +static inline void qaic_debugfs_init(struct qaic_drm_device *qddev) {} +#endif /* CONFIG_DEBUG_FS */ +#endif /* __QAIC_DEBUGFS_H__ */ diff --git a/drivers/accel/qaic/qaic_drv.c b/drivers/accel/qaic/qaic_drv.c index 2a313eb69b12..81819b9ef8d4 100644 --- a/drivers/accel/qaic/qaic_drv.c +++ b/drivers/accel/qaic/qaic_drv.c @@ -28,10 +28,13 @@ #include "mhi_controller.h" #include "qaic.h" +#include "qaic_debugfs.h" #include "qaic_timesync.h" +#include "sahara.h" -MODULE_IMPORT_NS(DMA_BUF); +MODULE_IMPORT_NS("DMA_BUF"); +#define PCI_DEV_AIC080 0xa080 #define PCI_DEV_AIC100 0xa100 #define QAIC_NAME "qaic" #define QAIC_DESC "Qualcomm Cloud AI Accelerators" @@ -44,6 +47,53 @@ MODULE_PARM_DESC(datapath_polling, "Operate the datapath in polling mode"); static bool link_up; static DEFINE_IDA(qaic_usrs); +static void qaicm_wq_release(struct drm_device *dev, void *res) +{ + struct workqueue_struct *wq = res; + + destroy_workqueue(wq); +} + +static struct workqueue_struct *qaicm_wq_init(struct drm_device *dev, const char *name) +{ + struct workqueue_struct *wq; + int ret; + + wq = alloc_workqueue("%s", WQ_UNBOUND, 0, name); + if (!wq) + return ERR_PTR(-ENOMEM); + ret = drmm_add_action_or_reset(dev, qaicm_wq_release, wq); + if (ret) + return ERR_PTR(ret); + + return wq; +} + +static void qaicm_srcu_release(struct drm_device *dev, void *res) +{ + struct srcu_struct *lock = res; + + cleanup_srcu_struct(lock); +} + +static int qaicm_srcu_init(struct drm_device *dev, struct srcu_struct *lock) +{ + int ret; + + ret = init_srcu_struct(lock); + if (ret) + return ret; + + return drmm_add_action_or_reset(dev, qaicm_srcu_release, lock); +} + +static void qaicm_pci_release(struct drm_device *dev, void *res) +{ + struct qaic_device *qdev = to_qaic_device(dev); + + pci_set_drvdata(qdev->pdev, NULL); +} + static void free_usr(struct kref *kref) { struct qaic_user *usr = container_of(kref, struct qaic_user, ref_count); @@ -158,7 +208,6 @@ static const struct drm_driver qaic_accel_driver = { .name = QAIC_NAME, .desc = QAIC_DESC, - .date = "20190618", .fops = &qaic_accel_fops, .open = qaic_open, @@ -182,8 +231,12 @@ static int qaic_create_drm_device(struct qaic_device *qdev, s32 partition_id) qddev->partition_id = partition_id; ret = drm_dev_register(drm, 0); - if (ret) + if (ret) { pci_dbg(qdev->pdev, "drm_dev_register failed %d\n", ret); + return ret; + } + + qaic_debugfs_init(qddev); return ret; } @@ -299,74 +352,77 @@ void qaic_dev_reset_clean_local_state(struct qaic_device *qdev) release_dbc(qdev, i); } -static void cleanup_qdev(struct qaic_device *qdev) -{ - int i; - - for (i = 0; i < qdev->num_dbc; ++i) - cleanup_srcu_struct(&qdev->dbc[i].ch_lock); - 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) { + struct device *dev = &pdev->dev; struct qaic_drm_device *qddev; struct qaic_device *qdev; - int i; + struct drm_device *drm; + int i, ret; - qdev = devm_kzalloc(&pdev->dev, sizeof(*qdev), GFP_KERNEL); + qdev = devm_kzalloc(dev, sizeof(*qdev), GFP_KERNEL); if (!qdev) return NULL; qdev->dev_state = QAIC_OFFLINE; - if (id->device == PCI_DEV_AIC100) { + if (id->device == PCI_DEV_AIC080 || id->device == PCI_DEV_AIC100) { qdev->num_dbc = 16; - qdev->dbc = devm_kcalloc(&pdev->dev, qdev->num_dbc, sizeof(*qdev->dbc), GFP_KERNEL); + qdev->dbc = devm_kcalloc(dev, qdev->num_dbc, sizeof(*qdev->dbc), GFP_KERNEL); if (!qdev->dbc) return NULL; } - qdev->cntl_wq = alloc_workqueue("qaic_cntl", WQ_UNBOUND, 0); - if (!qdev->cntl_wq) + qddev = devm_drm_dev_alloc(&pdev->dev, &qaic_accel_driver, struct qaic_drm_device, drm); + if (IS_ERR(qddev)) return NULL; - qdev->qts_wq = alloc_workqueue("qaic_ts", WQ_UNBOUND, 0); - if (!qdev->qts_wq) { - destroy_workqueue(qdev->cntl_wq); + drm = to_drm(qddev); + pci_set_drvdata(pdev, qdev); + + ret = drmm_mutex_init(drm, &qddev->users_mutex); + if (ret) + return NULL; + ret = drmm_add_action_or_reset(drm, qaicm_pci_release, NULL); + if (ret) + return NULL; + ret = drmm_mutex_init(drm, &qdev->cntl_mutex); + if (ret) + return NULL; + ret = drmm_mutex_init(drm, &qdev->bootlog_mutex); + if (ret) return NULL; - } - pci_set_drvdata(pdev, qdev); + qdev->cntl_wq = qaicm_wq_init(drm, "qaic_cntl"); + if (IS_ERR(qdev->cntl_wq)) + return NULL; + qdev->qts_wq = qaicm_wq_init(drm, "qaic_ts"); + if (IS_ERR(qdev->qts_wq)) + return NULL; + + ret = qaicm_srcu_init(drm, &qdev->dev_lock); + if (ret) + return NULL; + + qdev->qddev = qddev; qdev->pdev = pdev; + qddev->qdev = qdev; - mutex_init(&qdev->cntl_mutex); INIT_LIST_HEAD(&qdev->cntl_xfer_list); - init_srcu_struct(&qdev->dev_lock); + INIT_LIST_HEAD(&qdev->bootlog); + INIT_LIST_HEAD(&qddev->users); for (i = 0; i < qdev->num_dbc; ++i) { spin_lock_init(&qdev->dbc[i].xfer_lock); qdev->dbc[i].qdev = qdev; qdev->dbc[i].id = i; INIT_LIST_HEAD(&qdev->dbc[i].xfer_list); - init_srcu_struct(&qdev->dbc[i].ch_lock); + ret = qaicm_srcu_init(drm, &qdev->dbc[i].ch_lock); + if (ret) + return NULL; init_waitqueue_head(&qdev->dbc[i].dbc_release); INIT_LIST_HEAD(&qdev->dbc[i].bo_lists); } - qddev = devm_drm_dev_alloc(&pdev->dev, &qaic_accel_driver, struct qaic_drm_device, drm); - if (IS_ERR(qddev)) { - cleanup_qdev(qdev); - return NULL; - } - - drmm_mutex_init(to_drm(qddev), &qddev->users_mutex); - INIT_LIST_HEAD(&qddev->users); - qddev->qdev = qdev; - qdev->qddev = qddev; - return qdev; } @@ -391,9 +447,7 @@ static int init_pci(struct qaic_device *qdev, struct pci_dev *pdev) ret = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64)); if (ret) return ret; - ret = dma_set_max_seg_size(&pdev->dev, UINT_MAX); - if (ret) - return ret; + dma_set_max_seg_size(&pdev->dev, UINT_MAX); qdev->bar_0 = devm_ioremap_resource(&pdev->dev, &pdev->resource[0]); if (IS_ERR(qdev->bar_0)) @@ -472,35 +526,28 @@ static int qaic_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id) ret = init_pci(qdev, pdev); if (ret) - goto cleanup_qdev; + return ret; for (i = 0; i < qdev->num_dbc; ++i) qdev->dbc[i].dbc_base = qdev->bar_2 + QAIC_DBC_OFF(i); mhi_irq = init_msi(qdev, pdev); - if (mhi_irq < 0) { - ret = mhi_irq; - goto cleanup_qdev; - } + if (mhi_irq < 0) + return mhi_irq; ret = qaic_create_drm_device(qdev, QAIC_NO_PARTITION); if (ret) - goto cleanup_qdev; + return ret; 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_drm_dev; + qaic_destroy_drm_device(qdev, QAIC_NO_PARTITION); + return ret; } return 0; - -cleanup_drm_dev: - qaic_destroy_drm_device(qdev, QAIC_NO_PARTITION); -cleanup_qdev: - cleanup_qdev(qdev); - return ret; } static void qaic_pci_remove(struct pci_dev *pdev) @@ -511,9 +558,8 @@ static void qaic_pci_remove(struct pci_dev *pdev) return; 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); + qaic_destroy_drm_device(qdev, QAIC_NO_PARTITION); } static void qaic_pci_shutdown(struct pci_dev *pdev) @@ -561,6 +607,7 @@ static struct mhi_driver qaic_mhi_driver = { }; static const struct pci_device_id qaic_ids[] = { + { PCI_DEVICE(PCI_VENDOR_ID_QCOM, PCI_DEV_AIC080), }, { PCI_DEVICE(PCI_VENDOR_ID_QCOM, PCI_DEV_AIC100), }, { } }; @@ -597,12 +644,24 @@ static int __init qaic_init(void) goto free_pci; } + ret = sahara_register(); + if (ret) { + pr_debug("qaic: sahara_register failed %d\n", ret); + goto free_mhi; + } + ret = qaic_timesync_init(); if (ret) pr_debug("qaic: qaic_timesync_init failed %d\n", ret); + ret = qaic_bootlog_register(); + if (ret) + pr_debug("qaic: qaic_bootlog_register failed %d\n", ret); + return 0; +free_mhi: + mhi_driver_unregister(&qaic_mhi_driver); free_pci: pci_unregister_driver(&qaic_pci_driver); return ret; @@ -626,7 +685,9 @@ static void __exit qaic_exit(void) * reinitializing the link_up state after the cleanup is done. */ link_up = true; + qaic_bootlog_unregister(); qaic_timesync_deinit(); + sahara_unregister(); mhi_driver_unregister(&qaic_mhi_driver); pci_unregister_driver(&qaic_pci_driver); } diff --git a/drivers/accel/qaic/sahara.c b/drivers/accel/qaic/sahara.c new file mode 100644 index 000000000000..21d58aed0deb --- /dev/null +++ b/drivers/accel/qaic/sahara.c @@ -0,0 +1,822 @@ +// SPDX-License-Identifier: GPL-2.0-only + +/* Copyright (c) 2024 Qualcomm Innovation Center, Inc. All rights reserved. */ + +#include <linux/devcoredump.h> +#include <linux/firmware.h> +#include <linux/limits.h> +#include <linux/mhi.h> +#include <linux/minmax.h> +#include <linux/mod_devicetable.h> +#include <linux/overflow.h> +#include <linux/types.h> +#include <linux/vmalloc.h> +#include <linux/workqueue.h> + +#include "sahara.h" + +#define SAHARA_HELLO_CMD 0x1 /* Min protocol version 1.0 */ +#define SAHARA_HELLO_RESP_CMD 0x2 /* Min protocol version 1.0 */ +#define SAHARA_READ_DATA_CMD 0x3 /* Min protocol version 1.0 */ +#define SAHARA_END_OF_IMAGE_CMD 0x4 /* Min protocol version 1.0 */ +#define SAHARA_DONE_CMD 0x5 /* Min protocol version 1.0 */ +#define SAHARA_DONE_RESP_CMD 0x6 /* Min protocol version 1.0 */ +#define SAHARA_RESET_CMD 0x7 /* Min protocol version 1.0 */ +#define SAHARA_RESET_RESP_CMD 0x8 /* Min protocol version 1.0 */ +#define SAHARA_MEM_DEBUG_CMD 0x9 /* Min protocol version 2.0 */ +#define SAHARA_MEM_READ_CMD 0xa /* Min protocol version 2.0 */ +#define SAHARA_CMD_READY_CMD 0xb /* Min protocol version 2.1 */ +#define SAHARA_SWITCH_MODE_CMD 0xc /* Min protocol version 2.1 */ +#define SAHARA_EXECUTE_CMD 0xd /* Min protocol version 2.1 */ +#define SAHARA_EXECUTE_RESP_CMD 0xe /* Min protocol version 2.1 */ +#define SAHARA_EXECUTE_DATA_CMD 0xf /* Min protocol version 2.1 */ +#define SAHARA_MEM_DEBUG64_CMD 0x10 /* Min protocol version 2.5 */ +#define SAHARA_MEM_READ64_CMD 0x11 /* Min protocol version 2.5 */ +#define SAHARA_READ_DATA64_CMD 0x12 /* Min protocol version 2.8 */ +#define SAHARA_RESET_STATE_CMD 0x13 /* Min protocol version 2.9 */ +#define SAHARA_WRITE_DATA_CMD 0x14 /* Min protocol version 3.0 */ + +#define SAHARA_PACKET_MAX_SIZE 0xffffU /* MHI_MAX_MTU */ +#define SAHARA_TRANSFER_MAX_SIZE 0x80000 +#define SAHARA_READ_MAX_SIZE 0xfff0U /* Avoid unaligned requests */ +#define SAHARA_NUM_TX_BUF DIV_ROUND_UP(SAHARA_TRANSFER_MAX_SIZE,\ + SAHARA_PACKET_MAX_SIZE) +#define SAHARA_IMAGE_ID_NONE U32_MAX + +#define SAHARA_VERSION 2 +#define SAHARA_SUCCESS 0 +#define SAHARA_TABLE_ENTRY_STR_LEN 20 + +#define SAHARA_MODE_IMAGE_TX_PENDING 0x0 +#define SAHARA_MODE_IMAGE_TX_COMPLETE 0x1 +#define SAHARA_MODE_MEMORY_DEBUG 0x2 +#define SAHARA_MODE_COMMAND 0x3 + +#define SAHARA_HELLO_LENGTH 0x30 +#define SAHARA_READ_DATA_LENGTH 0x14 +#define SAHARA_END_OF_IMAGE_LENGTH 0x10 +#define SAHARA_DONE_LENGTH 0x8 +#define SAHARA_RESET_LENGTH 0x8 +#define SAHARA_MEM_DEBUG64_LENGTH 0x18 +#define SAHARA_MEM_READ64_LENGTH 0x18 + +struct sahara_packet { + __le32 cmd; + __le32 length; + + union { + struct { + __le32 version; + __le32 version_compat; + __le32 max_length; + __le32 mode; + } hello; + struct { + __le32 version; + __le32 version_compat; + __le32 status; + __le32 mode; + } hello_resp; + struct { + __le32 image; + __le32 offset; + __le32 length; + } read_data; + struct { + __le32 image; + __le32 status; + } end_of_image; + struct { + __le64 table_address; + __le64 table_length; + } memory_debug64; + struct { + __le64 memory_address; + __le64 memory_length; + } memory_read64; + }; +}; + +struct sahara_debug_table_entry64 { + __le64 type; + __le64 address; + __le64 length; + char description[SAHARA_TABLE_ENTRY_STR_LEN]; + char filename[SAHARA_TABLE_ENTRY_STR_LEN]; +}; + +struct sahara_dump_table_entry { + u64 type; + u64 address; + u64 length; + char description[SAHARA_TABLE_ENTRY_STR_LEN]; + char filename[SAHARA_TABLE_ENTRY_STR_LEN]; +}; + +#define SAHARA_DUMP_V1_MAGIC 0x1234567890abcdef +#define SAHARA_DUMP_V1_VER 1 +struct sahara_memory_dump_meta_v1 { + u64 magic; + u64 version; + u64 dump_size; + u64 table_size; +}; + +/* + * Layout of crashdump provided to user via devcoredump + * +------------------------------------------+ + * | Crashdump Meta structure | + * | type: struct sahara_memory_dump_meta_v1 | + * +------------------------------------------+ + * | Crashdump Table | + * | type: array of struct | + * | sahara_dump_table_entry | + * | | + * | | + * +------------------------------------------+ + * | Crashdump | + * | | + * | | + * | | + * | | + * | | + * +------------------------------------------+ + * + * First is the metadata header. Userspace can use the magic number to verify + * the content type, and then check the version for the rest of the format. + * New versions should keep the magic number location/value, and version + * location, but increment the version value. + * + * For v1, the metadata lists the size of the entire dump (header + table + + * dump) and the size of the table. Then the dump image table, which describes + * the contents of the dump. Finally all the images are listed in order, with + * no deadspace in between. Userspace can use the sizes listed in the image + * table to reconstruct the individual images. + */ + +struct sahara_context { + struct sahara_packet *tx[SAHARA_NUM_TX_BUF]; + struct sahara_packet *rx; + struct work_struct fw_work; + struct work_struct dump_work; + struct mhi_device *mhi_dev; + const char **image_table; + u32 table_size; + u32 active_image_id; + const struct firmware *firmware; + u64 dump_table_address; + u64 dump_table_length; + size_t rx_size; + size_t rx_size_requested; + void *mem_dump; + size_t mem_dump_sz; + struct sahara_dump_table_entry *dump_image; + u64 dump_image_offset; + void *mem_dump_freespace; + u64 dump_images_left; + bool is_mem_dump_mode; +}; + +static const char *aic100_image_table[] = { + [1] = "qcom/aic100/fw1.bin", + [2] = "qcom/aic100/fw2.bin", + [4] = "qcom/aic100/fw4.bin", + [5] = "qcom/aic100/fw5.bin", + [6] = "qcom/aic100/fw6.bin", + [8] = "qcom/aic100/fw8.bin", + [9] = "qcom/aic100/fw9.bin", + [10] = "qcom/aic100/fw10.bin", +}; + +static int sahara_find_image(struct sahara_context *context, u32 image_id) +{ + int ret; + + if (image_id == context->active_image_id) + return 0; + + if (context->active_image_id != SAHARA_IMAGE_ID_NONE) { + dev_err(&context->mhi_dev->dev, "image id %d is not valid as %d is active\n", + image_id, context->active_image_id); + return -EINVAL; + } + + if (image_id >= context->table_size || !context->image_table[image_id]) { + dev_err(&context->mhi_dev->dev, "request for unknown image: %d\n", image_id); + return -EINVAL; + } + + /* + * This image might be optional. The device may continue without it. + * Only the device knows. Suppress error messages that could suggest an + * a problem when we were actually able to continue. + */ + ret = firmware_request_nowarn(&context->firmware, + context->image_table[image_id], + &context->mhi_dev->dev); + if (ret) { + dev_dbg(&context->mhi_dev->dev, "request for image id %d / file %s failed %d\n", + image_id, context->image_table[image_id], ret); + return ret; + } + + context->active_image_id = image_id; + + return 0; +} + +static void sahara_release_image(struct sahara_context *context) +{ + if (context->active_image_id != SAHARA_IMAGE_ID_NONE) + release_firmware(context->firmware); + context->active_image_id = SAHARA_IMAGE_ID_NONE; +} + +static void sahara_send_reset(struct sahara_context *context) +{ + int ret; + + context->is_mem_dump_mode = false; + + context->tx[0]->cmd = cpu_to_le32(SAHARA_RESET_CMD); + context->tx[0]->length = cpu_to_le32(SAHARA_RESET_LENGTH); + + ret = mhi_queue_buf(context->mhi_dev, DMA_TO_DEVICE, context->tx[0], + SAHARA_RESET_LENGTH, MHI_EOT); + if (ret) + dev_err(&context->mhi_dev->dev, "Unable to send reset response %d\n", ret); +} + +static void sahara_hello(struct sahara_context *context) +{ + int ret; + + dev_dbg(&context->mhi_dev->dev, + "HELLO cmd received. length:%d version:%d version_compat:%d max_length:%d mode:%d\n", + le32_to_cpu(context->rx->length), + le32_to_cpu(context->rx->hello.version), + le32_to_cpu(context->rx->hello.version_compat), + le32_to_cpu(context->rx->hello.max_length), + le32_to_cpu(context->rx->hello.mode)); + + if (le32_to_cpu(context->rx->length) != SAHARA_HELLO_LENGTH) { + dev_err(&context->mhi_dev->dev, "Malformed hello packet - length %d\n", + le32_to_cpu(context->rx->length)); + return; + } + if (le32_to_cpu(context->rx->hello.version) != SAHARA_VERSION) { + dev_err(&context->mhi_dev->dev, "Unsupported hello packet - version %d\n", + le32_to_cpu(context->rx->hello.version)); + return; + } + + if (le32_to_cpu(context->rx->hello.mode) != SAHARA_MODE_IMAGE_TX_PENDING && + le32_to_cpu(context->rx->hello.mode) != SAHARA_MODE_IMAGE_TX_COMPLETE && + le32_to_cpu(context->rx->hello.mode) != SAHARA_MODE_MEMORY_DEBUG) { + dev_err(&context->mhi_dev->dev, "Unsupported hello packet - mode %d\n", + le32_to_cpu(context->rx->hello.mode)); + return; + } + + context->tx[0]->cmd = cpu_to_le32(SAHARA_HELLO_RESP_CMD); + context->tx[0]->length = cpu_to_le32(SAHARA_HELLO_LENGTH); + context->tx[0]->hello_resp.version = cpu_to_le32(SAHARA_VERSION); + context->tx[0]->hello_resp.version_compat = cpu_to_le32(SAHARA_VERSION); + context->tx[0]->hello_resp.status = cpu_to_le32(SAHARA_SUCCESS); + context->tx[0]->hello_resp.mode = context->rx->hello_resp.mode; + + ret = mhi_queue_buf(context->mhi_dev, DMA_TO_DEVICE, context->tx[0], + SAHARA_HELLO_LENGTH, MHI_EOT); + if (ret) + dev_err(&context->mhi_dev->dev, "Unable to send hello response %d\n", ret); +} + +static void sahara_read_data(struct sahara_context *context) +{ + u32 image_id, data_offset, data_len, pkt_data_len; + int ret; + int i; + + dev_dbg(&context->mhi_dev->dev, + "READ_DATA cmd received. length:%d image:%d offset:%d data_length:%d\n", + le32_to_cpu(context->rx->length), + le32_to_cpu(context->rx->read_data.image), + le32_to_cpu(context->rx->read_data.offset), + le32_to_cpu(context->rx->read_data.length)); + + if (le32_to_cpu(context->rx->length) != SAHARA_READ_DATA_LENGTH) { + dev_err(&context->mhi_dev->dev, "Malformed read_data packet - length %d\n", + le32_to_cpu(context->rx->length)); + return; + } + + image_id = le32_to_cpu(context->rx->read_data.image); + data_offset = le32_to_cpu(context->rx->read_data.offset); + data_len = le32_to_cpu(context->rx->read_data.length); + + ret = sahara_find_image(context, image_id); + if (ret) { + sahara_send_reset(context); + return; + } + + /* + * Image is released when the device is done with it via + * SAHARA_END_OF_IMAGE_CMD. sahara_send_reset() will either cause the + * device to retry the operation with a modification, or decide to be + * done with the image and trigger SAHARA_END_OF_IMAGE_CMD. + * release_image() is called from SAHARA_END_OF_IMAGE_CMD. processing + * and is not needed here on error. + */ + + if (data_len > SAHARA_TRANSFER_MAX_SIZE) { + dev_err(&context->mhi_dev->dev, "Malformed read_data packet - data len %d exceeds max xfer size %d\n", + data_len, SAHARA_TRANSFER_MAX_SIZE); + sahara_send_reset(context); + return; + } + + if (data_offset >= context->firmware->size) { + dev_err(&context->mhi_dev->dev, "Malformed read_data packet - data offset %d exceeds file size %zu\n", + data_offset, context->firmware->size); + sahara_send_reset(context); + return; + } + + if (size_add(data_offset, data_len) > context->firmware->size) { + dev_err(&context->mhi_dev->dev, "Malformed read_data packet - data offset %d and length %d exceeds file size %zu\n", + data_offset, data_len, context->firmware->size); + sahara_send_reset(context); + return; + } + + for (i = 0; i < SAHARA_NUM_TX_BUF && data_len; ++i) { + pkt_data_len = min(data_len, SAHARA_PACKET_MAX_SIZE); + + memcpy(context->tx[i], &context->firmware->data[data_offset], pkt_data_len); + + data_offset += pkt_data_len; + data_len -= pkt_data_len; + + ret = mhi_queue_buf(context->mhi_dev, DMA_TO_DEVICE, + context->tx[i], pkt_data_len, + !data_len ? MHI_EOT : MHI_CHAIN); + if (ret) { + dev_err(&context->mhi_dev->dev, "Unable to send read_data response %d\n", + ret); + return; + } + } +} + +static void sahara_end_of_image(struct sahara_context *context) +{ + int ret; + + dev_dbg(&context->mhi_dev->dev, + "END_OF_IMAGE cmd received. length:%d image:%d status:%d\n", + le32_to_cpu(context->rx->length), + le32_to_cpu(context->rx->end_of_image.image), + le32_to_cpu(context->rx->end_of_image.status)); + + if (le32_to_cpu(context->rx->length) != SAHARA_END_OF_IMAGE_LENGTH) { + dev_err(&context->mhi_dev->dev, "Malformed end_of_image packet - length %d\n", + le32_to_cpu(context->rx->length)); + return; + } + + if (context->active_image_id != SAHARA_IMAGE_ID_NONE && + le32_to_cpu(context->rx->end_of_image.image) != context->active_image_id) { + dev_err(&context->mhi_dev->dev, "Malformed end_of_image packet - image %d is not the active image\n", + le32_to_cpu(context->rx->end_of_image.image)); + return; + } + + sahara_release_image(context); + + if (le32_to_cpu(context->rx->end_of_image.status)) + return; + + context->tx[0]->cmd = cpu_to_le32(SAHARA_DONE_CMD); + context->tx[0]->length = cpu_to_le32(SAHARA_DONE_LENGTH); + + ret = mhi_queue_buf(context->mhi_dev, DMA_TO_DEVICE, context->tx[0], + SAHARA_DONE_LENGTH, MHI_EOT); + if (ret) + dev_dbg(&context->mhi_dev->dev, "Unable to send done response %d\n", ret); +} + +static void sahara_memory_debug64(struct sahara_context *context) +{ + int ret; + + dev_dbg(&context->mhi_dev->dev, + "MEMORY DEBUG64 cmd received. length:%d table_address:%#llx table_length:%#llx\n", + le32_to_cpu(context->rx->length), + le64_to_cpu(context->rx->memory_debug64.table_address), + le64_to_cpu(context->rx->memory_debug64.table_length)); + + if (le32_to_cpu(context->rx->length) != SAHARA_MEM_DEBUG64_LENGTH) { + dev_err(&context->mhi_dev->dev, "Malformed memory debug64 packet - length %d\n", + le32_to_cpu(context->rx->length)); + return; + } + + context->dump_table_address = le64_to_cpu(context->rx->memory_debug64.table_address); + context->dump_table_length = le64_to_cpu(context->rx->memory_debug64.table_length); + + if (context->dump_table_length % sizeof(struct sahara_debug_table_entry64) != 0 || + !context->dump_table_length) { + dev_err(&context->mhi_dev->dev, "Malformed memory debug64 packet - table length %lld\n", + context->dump_table_length); + return; + } + + /* + * From this point, the protocol flips. We make memory_read requests to + * the device, and the device responds with the raw data. If the device + * has an error, it will send an End of Image command. First we need to + * request the memory dump table so that we know where all the pieces + * of the dump are that we can consume. + */ + + context->is_mem_dump_mode = true; + + /* + * Assume that the table is smaller than our MTU so that we can read it + * in one shot. The spec does not put an upper limit on the table, but + * no known device will exceed this. + */ + if (context->dump_table_length > SAHARA_PACKET_MAX_SIZE) { + dev_err(&context->mhi_dev->dev, "Memory dump table length %lld exceeds supported size. Discarding dump\n", + context->dump_table_length); + sahara_send_reset(context); + return; + } + + context->tx[0]->cmd = cpu_to_le32(SAHARA_MEM_READ64_CMD); + context->tx[0]->length = cpu_to_le32(SAHARA_MEM_READ64_LENGTH); + context->tx[0]->memory_read64.memory_address = cpu_to_le64(context->dump_table_address); + context->tx[0]->memory_read64.memory_length = cpu_to_le64(context->dump_table_length); + + context->rx_size_requested = context->dump_table_length; + + ret = mhi_queue_buf(context->mhi_dev, DMA_TO_DEVICE, context->tx[0], + SAHARA_MEM_READ64_LENGTH, MHI_EOT); + if (ret) + dev_err(&context->mhi_dev->dev, "Unable to send read for dump table %d\n", ret); +} + +static void sahara_processing(struct work_struct *work) +{ + struct sahara_context *context = container_of(work, struct sahara_context, fw_work); + int ret; + + switch (le32_to_cpu(context->rx->cmd)) { + case SAHARA_HELLO_CMD: + sahara_hello(context); + break; + case SAHARA_READ_DATA_CMD: + sahara_read_data(context); + break; + case SAHARA_END_OF_IMAGE_CMD: + sahara_end_of_image(context); + break; + case SAHARA_DONE_RESP_CMD: + /* Intentional do nothing as we don't need to exit an app */ + break; + case SAHARA_RESET_RESP_CMD: + /* Intentional do nothing as we don't need to exit an app */ + break; + case SAHARA_MEM_DEBUG64_CMD: + sahara_memory_debug64(context); + break; + default: + dev_err(&context->mhi_dev->dev, "Unknown command %d\n", + le32_to_cpu(context->rx->cmd)); + break; + } + + ret = mhi_queue_buf(context->mhi_dev, DMA_FROM_DEVICE, context->rx, + SAHARA_PACKET_MAX_SIZE, MHI_EOT); + if (ret) + dev_err(&context->mhi_dev->dev, "Unable to requeue rx buf %d\n", ret); +} + +static void sahara_parse_dump_table(struct sahara_context *context) +{ + struct sahara_dump_table_entry *image_out_table; + struct sahara_debug_table_entry64 *dev_table; + struct sahara_memory_dump_meta_v1 *dump_meta; + u64 table_nents; + u64 dump_length; + int ret; + u64 i; + + table_nents = context->dump_table_length / sizeof(*dev_table); + context->dump_images_left = table_nents; + dump_length = 0; + + dev_table = (struct sahara_debug_table_entry64 *)(context->rx); + for (i = 0; i < table_nents; ++i) { + /* Do not trust the device, ensure the strings are terminated */ + dev_table[i].description[SAHARA_TABLE_ENTRY_STR_LEN - 1] = 0; + dev_table[i].filename[SAHARA_TABLE_ENTRY_STR_LEN - 1] = 0; + + dump_length = size_add(dump_length, le64_to_cpu(dev_table[i].length)); + if (dump_length == SIZE_MAX) { + /* Discard the dump */ + sahara_send_reset(context); + return; + } + + dev_dbg(&context->mhi_dev->dev, + "Memory dump table entry %lld type: %lld address: %#llx length: %#llx description: \"%s\" filename \"%s\"\n", + i, + le64_to_cpu(dev_table[i].type), + le64_to_cpu(dev_table[i].address), + le64_to_cpu(dev_table[i].length), + dev_table[i].description, + dev_table[i].filename); + } + + dump_length = size_add(dump_length, sizeof(*dump_meta)); + if (dump_length == SIZE_MAX) { + /* Discard the dump */ + sahara_send_reset(context); + return; + } + dump_length = size_add(dump_length, size_mul(sizeof(*image_out_table), table_nents)); + if (dump_length == SIZE_MAX) { + /* Discard the dump */ + sahara_send_reset(context); + return; + } + + context->mem_dump_sz = dump_length; + context->mem_dump = vzalloc(dump_length); + if (!context->mem_dump) { + /* Discard the dump */ + sahara_send_reset(context); + return; + } + + /* Populate the dump metadata and table for userspace */ + dump_meta = context->mem_dump; + dump_meta->magic = SAHARA_DUMP_V1_MAGIC; + dump_meta->version = SAHARA_DUMP_V1_VER; + dump_meta->dump_size = dump_length; + dump_meta->table_size = context->dump_table_length; + + image_out_table = context->mem_dump + sizeof(*dump_meta); + for (i = 0; i < table_nents; ++i) { + image_out_table[i].type = le64_to_cpu(dev_table[i].type); + image_out_table[i].address = le64_to_cpu(dev_table[i].address); + image_out_table[i].length = le64_to_cpu(dev_table[i].length); + strscpy(image_out_table[i].description, dev_table[i].description, + SAHARA_TABLE_ENTRY_STR_LEN); + strscpy(image_out_table[i].filename, + dev_table[i].filename, + SAHARA_TABLE_ENTRY_STR_LEN); + } + + context->mem_dump_freespace = &image_out_table[i]; + + /* Done parsing the table, switch to image dump mode */ + context->dump_table_length = 0; + + /* Request the first chunk of the first image */ + context->dump_image = &image_out_table[0]; + dump_length = min(context->dump_image->length, SAHARA_READ_MAX_SIZE); + /* Avoid requesting EOI sized data so that we can identify errors */ + if (dump_length == SAHARA_END_OF_IMAGE_LENGTH) + dump_length = SAHARA_END_OF_IMAGE_LENGTH / 2; + + context->dump_image_offset = dump_length; + + context->tx[0]->cmd = cpu_to_le32(SAHARA_MEM_READ64_CMD); + context->tx[0]->length = cpu_to_le32(SAHARA_MEM_READ64_LENGTH); + context->tx[0]->memory_read64.memory_address = cpu_to_le64(context->dump_image->address); + context->tx[0]->memory_read64.memory_length = cpu_to_le64(dump_length); + + context->rx_size_requested = dump_length; + + ret = mhi_queue_buf(context->mhi_dev, DMA_TO_DEVICE, context->tx[0], + SAHARA_MEM_READ64_LENGTH, MHI_EOT); + if (ret) + dev_err(&context->mhi_dev->dev, "Unable to send read for dump content %d\n", ret); +} + +static void sahara_parse_dump_image(struct sahara_context *context) +{ + u64 dump_length; + int ret; + + memcpy(context->mem_dump_freespace, context->rx, context->rx_size); + context->mem_dump_freespace += context->rx_size; + + if (context->dump_image_offset >= context->dump_image->length) { + /* Need to move to next image */ + context->dump_image++; + context->dump_images_left--; + context->dump_image_offset = 0; + + if (!context->dump_images_left) { + /* Dump done */ + dev_coredumpv(context->mhi_dev->mhi_cntrl->cntrl_dev, + context->mem_dump, + context->mem_dump_sz, + GFP_KERNEL); + context->mem_dump = NULL; + sahara_send_reset(context); + return; + } + } + + /* Get next image chunk */ + dump_length = context->dump_image->length - context->dump_image_offset; + dump_length = min(dump_length, SAHARA_READ_MAX_SIZE); + /* Avoid requesting EOI sized data so that we can identify errors */ + if (dump_length == SAHARA_END_OF_IMAGE_LENGTH) + dump_length = SAHARA_END_OF_IMAGE_LENGTH / 2; + + context->tx[0]->cmd = cpu_to_le32(SAHARA_MEM_READ64_CMD); + context->tx[0]->length = cpu_to_le32(SAHARA_MEM_READ64_LENGTH); + context->tx[0]->memory_read64.memory_address = + cpu_to_le64(context->dump_image->address + context->dump_image_offset); + context->tx[0]->memory_read64.memory_length = cpu_to_le64(dump_length); + + context->dump_image_offset += dump_length; + context->rx_size_requested = dump_length; + + ret = mhi_queue_buf(context->mhi_dev, DMA_TO_DEVICE, context->tx[0], + SAHARA_MEM_READ64_LENGTH, MHI_EOT); + if (ret) + dev_err(&context->mhi_dev->dev, + "Unable to send read for dump content %d\n", ret); +} + +static void sahara_dump_processing(struct work_struct *work) +{ + struct sahara_context *context = container_of(work, struct sahara_context, dump_work); + int ret; + + /* + * We should get the expected raw data, but if the device has an error + * it is supposed to send EOI with an error code. + */ + if (context->rx_size != context->rx_size_requested && + context->rx_size != SAHARA_END_OF_IMAGE_LENGTH) { + dev_err(&context->mhi_dev->dev, + "Unexpected response to read_data. Expected size: %#zx got: %#zx\n", + context->rx_size_requested, + context->rx_size); + goto error; + } + + if (context->rx_size == SAHARA_END_OF_IMAGE_LENGTH && + le32_to_cpu(context->rx->cmd) == SAHARA_END_OF_IMAGE_CMD) { + dev_err(&context->mhi_dev->dev, + "Unexpected EOI response to read_data. Status: %d\n", + le32_to_cpu(context->rx->end_of_image.status)); + goto error; + } + + if (context->rx_size == SAHARA_END_OF_IMAGE_LENGTH && + le32_to_cpu(context->rx->cmd) != SAHARA_END_OF_IMAGE_CMD) { + dev_err(&context->mhi_dev->dev, + "Invalid EOI response to read_data. CMD: %d\n", + le32_to_cpu(context->rx->cmd)); + goto error; + } + + /* + * Need to know if we received the dump table, or part of a dump image. + * Since we get raw data, we cannot tell from the data itself. Instead, + * we use the stored dump_table_length, which we zero after we read and + * process the entire table. + */ + if (context->dump_table_length) + sahara_parse_dump_table(context); + else + sahara_parse_dump_image(context); + + ret = mhi_queue_buf(context->mhi_dev, DMA_FROM_DEVICE, context->rx, + SAHARA_PACKET_MAX_SIZE, MHI_EOT); + if (ret) + dev_err(&context->mhi_dev->dev, "Unable to requeue rx buf %d\n", ret); + + return; + +error: + vfree(context->mem_dump); + context->mem_dump = NULL; + sahara_send_reset(context); +} + +static int sahara_mhi_probe(struct mhi_device *mhi_dev, const struct mhi_device_id *id) +{ + struct sahara_context *context; + int ret; + int i; + + context = devm_kzalloc(&mhi_dev->dev, sizeof(*context), GFP_KERNEL); + if (!context) + return -ENOMEM; + + context->rx = devm_kzalloc(&mhi_dev->dev, SAHARA_PACKET_MAX_SIZE, GFP_KERNEL); + if (!context->rx) + return -ENOMEM; + + /* + * AIC100 defines SAHARA_TRANSFER_MAX_SIZE as the largest value it + * will request for READ_DATA. This is larger than + * SAHARA_PACKET_MAX_SIZE, and we need 9x SAHARA_PACKET_MAX_SIZE to + * cover SAHARA_TRANSFER_MAX_SIZE. When the remote side issues a + * READ_DATA, it requires a transfer of the exact size requested. We + * can use MHI_CHAIN to link multiple buffers into a single transfer + * but the remote side will not consume the buffers until it sees an + * EOT, thus we need to allocate enough buffers to put in the tx fifo + * to cover an entire READ_DATA request of the max size. + */ + for (i = 0; i < SAHARA_NUM_TX_BUF; ++i) { + context->tx[i] = devm_kzalloc(&mhi_dev->dev, SAHARA_PACKET_MAX_SIZE, GFP_KERNEL); + if (!context->tx[i]) + return -ENOMEM; + } + + context->mhi_dev = mhi_dev; + INIT_WORK(&context->fw_work, sahara_processing); + INIT_WORK(&context->dump_work, sahara_dump_processing); + context->image_table = aic100_image_table; + context->table_size = ARRAY_SIZE(aic100_image_table); + context->active_image_id = SAHARA_IMAGE_ID_NONE; + dev_set_drvdata(&mhi_dev->dev, context); + + ret = mhi_prepare_for_transfer(mhi_dev); + if (ret) + return ret; + + ret = mhi_queue_buf(mhi_dev, DMA_FROM_DEVICE, context->rx, SAHARA_PACKET_MAX_SIZE, MHI_EOT); + if (ret) { + mhi_unprepare_from_transfer(mhi_dev); + return ret; + } + + return 0; +} + +static void sahara_mhi_remove(struct mhi_device *mhi_dev) +{ + struct sahara_context *context = dev_get_drvdata(&mhi_dev->dev); + + cancel_work_sync(&context->fw_work); + cancel_work_sync(&context->dump_work); + vfree(context->mem_dump); + sahara_release_image(context); + mhi_unprepare_from_transfer(mhi_dev); +} + +static void sahara_mhi_ul_xfer_cb(struct mhi_device *mhi_dev, struct mhi_result *mhi_result) +{ +} + +static void sahara_mhi_dl_xfer_cb(struct mhi_device *mhi_dev, struct mhi_result *mhi_result) +{ + struct sahara_context *context = dev_get_drvdata(&mhi_dev->dev); + + if (!mhi_result->transaction_status) { + context->rx_size = mhi_result->bytes_xferd; + if (context->is_mem_dump_mode) + schedule_work(&context->dump_work); + else + schedule_work(&context->fw_work); + } + +} + +static const struct mhi_device_id sahara_mhi_match_table[] = { + { .chan = "QAIC_SAHARA", }, + {}, +}; + +static struct mhi_driver sahara_mhi_driver = { + .id_table = sahara_mhi_match_table, + .remove = sahara_mhi_remove, + .probe = sahara_mhi_probe, + .ul_xfer_cb = sahara_mhi_ul_xfer_cb, + .dl_xfer_cb = sahara_mhi_dl_xfer_cb, + .driver = { + .name = "sahara", + }, +}; + +int sahara_register(void) +{ + return mhi_driver_register(&sahara_mhi_driver); +} + +void sahara_unregister(void) +{ + mhi_driver_unregister(&sahara_mhi_driver); +} diff --git a/drivers/accel/qaic/sahara.h b/drivers/accel/qaic/sahara.h new file mode 100644 index 000000000000..640208acc0d1 --- /dev/null +++ b/drivers/accel/qaic/sahara.h @@ -0,0 +1,10 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ + +/* Copyright (c) 2024 Qualcomm Innovation Center, Inc. All rights reserved. */ + +#ifndef __SAHARA_H__ +#define __SAHARA_H__ + +int sahara_register(void); +void sahara_unregister(void); +#endif /* __SAHARA_H__ */ |