diff options
Diffstat (limited to 'drivers/accel/qaic/qaic_data.c')
| -rw-r--r-- | drivers/accel/qaic/qaic_data.c | 600 |
1 files changed, 386 insertions, 214 deletions
diff --git a/drivers/accel/qaic/qaic_data.c b/drivers/accel/qaic/qaic_data.c index f4b06792c6f1..60cb4d65d48e 100644 --- a/drivers/accel/qaic/qaic_data.c +++ b/drivers/accel/qaic/qaic_data.c @@ -18,6 +18,7 @@ #include <linux/scatterlist.h> #include <linux/spinlock.h> #include <linux/srcu.h> +#include <linux/string.h> #include <linux/types.h> #include <linux/uaccess.h> #include <linux/wait.h> @@ -51,6 +52,7 @@ }) #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); @@ -140,6 +142,11 @@ 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); @@ -154,20 +161,22 @@ static void free_slice(struct kref *kref) { struct bo_slice *slice = container_of(kref, struct bo_slice, ref_count); + slice->bo->total_slice_nents -= slice->nents; list_del(&slice->slice); drm_gem_object_put(&slice->bo->base); sg_free_table(slice->sgt); kfree(slice->sgt); - kfree(slice->reqs); + kvfree(slice->reqs); kfree(slice); } 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 */ @@ -175,9 +184,11 @@ 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) @@ -214,7 +225,7 @@ 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; @@ -294,7 +305,7 @@ 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); @@ -394,7 +405,7 @@ static int qaic_map_one_slice(struct qaic_device *qdev, struct qaic_bo *bo, goto free_sgt; } - slice->reqs = kcalloc(sgt->nents, sizeof(*slice->reqs), GFP_KERNEL); + slice->reqs = kvcalloc(sgt->nents, sizeof(*slice->reqs), GFP_KERNEL); if (!slice->reqs) { ret = -ENOMEM; goto free_slice; @@ -420,7 +431,7 @@ static int qaic_map_one_slice(struct qaic_device *qdev, struct qaic_bo *bo, return 0; free_req: - kfree(slice->reqs); + kvfree(slice->reqs); free_slice: kfree(slice); free_sgt: @@ -450,7 +461,7 @@ 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; @@ -547,6 +558,7 @@ 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++) { @@ -556,7 +568,8 @@ 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; } @@ -567,6 +580,9 @@ 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)); @@ -579,7 +595,7 @@ static void qaic_gem_print_info(struct drm_printer *p, unsigned int indent, { struct qaic_bo *bo = to_qaic_bo(obj); - drm_printf_indent(p, indent, "user requested size=%llu\n", bo->size); + drm_printf_indent(p, indent, "BO DMA direction %d\n", bo->dir); } static const struct vm_operations_struct drm_vm_ops = { @@ -594,7 +610,7 @@ static int qaic_gem_object_mmap(struct drm_gem_object *obj, struct vm_area_struc struct scatterlist *sg; int ret = 0; - if (obj->import_attach) + if (drm_gem_is_imported(obj)) return -EINVAL; for (sg = bo->sgt->sgl; sg; sg = sg_next(sg)) { @@ -615,7 +631,7 @@ static void qaic_free_object(struct drm_gem_object *obj) { struct qaic_bo *bo = to_qaic_bo(obj); - if (obj->import_attach) { + if (drm_gem_is_imported(obj)) { /* DMABUF/PRIME Path */ drm_prime_gem_destroy(obj, NULL); } else { @@ -623,17 +639,60 @@ static void qaic_free_object(struct drm_gem_object *obj) qaic_free_sgt(bo->sgt); } + mutex_destroy(&bo->lock); drm_gem_object_release(obj); kfree(bo); } +static struct sg_table *qaic_get_sg_table(struct drm_gem_object *obj) +{ + struct qaic_bo *bo = to_qaic_bo(obj); + struct scatterlist *sg, *sg_in; + struct sg_table *sgt, *sgt_in; + int i; + + sgt_in = bo->sgt; + + sgt = kmalloc(sizeof(*sgt), GFP_KERNEL); + if (!sgt) + return ERR_PTR(-ENOMEM); + + if (sg_alloc_table(sgt, sgt_in->orig_nents, GFP_KERNEL)) { + kfree(sgt); + return ERR_PTR(-ENOMEM); + } + + sg = sgt->sgl; + for_each_sgtable_sg(sgt_in, sg_in, i) { + memcpy(sg, sg_in, sizeof(*sg)); + sg = sg_next(sg); + } + + return sgt; +} + static const struct drm_gem_object_funcs qaic_gem_funcs = { .free = qaic_free_object, + .get_sg_table = qaic_get_sg_table, .print_info = qaic_gem_print_info, .mmap = qaic_gem_object_mmap, .vm_ops = &drm_vm_ops, }; +static void qaic_init_bo(struct qaic_bo *bo, bool reinit) +{ + if (reinit) { + bo->sliced = false; + reinit_completion(&bo->xfer_done); + } else { + mutex_init(&bo->lock); + init_completion(&bo->xfer_done); + } + 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) { struct qaic_bo *bo; @@ -642,9 +701,7 @@ static struct qaic_bo *qaic_alloc_init_bo(void) if (!bo) return ERR_PTR(-ENOMEM); - INIT_LIST_HEAD(&bo->slices); - init_completion(&bo->xfer_done); - complete_all(&bo->xfer_done); + qaic_init_bo(bo, false); return bo; } @@ -676,7 +733,7 @@ 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; } @@ -695,23 +752,22 @@ int qaic_create_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *fi if (ret) goto free_bo; - bo->size = args->size; + 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); srcu_read_unlock(&qdev->dev_lock, qdev_rcu_id); srcu_read_unlock(&usr->qddev_lock, usr_rcu_id); 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: @@ -726,7 +782,7 @@ 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); @@ -737,7 +793,7 @@ 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; } @@ -748,9 +804,7 @@ 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); @@ -766,7 +820,6 @@ struct drm_gem_object *qaic_gem_prime_import(struct drm_device *dev, struct dma_ struct dma_buf_attachment *attach; struct drm_gem_object *obj; struct qaic_bo *bo; - size_t size; int ret; bo = qaic_alloc_init_bo(); @@ -784,13 +837,12 @@ struct drm_gem_object *qaic_gem_prime_import(struct drm_device *dev, struct dma_ goto attach_fail; } - size = PAGE_ALIGN(attach->dmabuf->size); - if (size == 0) { + if (!attach->dmabuf->size) { ret = -EINVAL; goto size_align_fail; } - drm_gem_private_object_init(dev, obj, size); + drm_gem_private_object_init(dev, obj, attach->dmabuf->size); /* * skipping dma_buf_map_attachment() as we do not know the direction * just yet. Once the direction is known in the subsequent IOCTL to @@ -818,9 +870,6 @@ 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); @@ -828,7 +877,6 @@ static int qaic_prepare_import_bo(struct qaic_bo *bo, struct qaic_attach_slice_h } bo->sgt = sgt; - bo->size = hdr->size; return 0; } @@ -838,9 +886,6 @@ static int qaic_prepare_export_bo(struct qaic_device *qdev, struct qaic_bo *bo, { int ret; - if (bo->size != hdr->size) - return -EINVAL; - ret = dma_map_sgtable(&qdev->pdev->dev, bo->sgt, hdr->dir, 0); if (ret) return -EFAULT; @@ -853,13 +898,13 @@ static int qaic_prepare_bo(struct qaic_device *qdev, struct qaic_bo *bo, { int ret; - if (bo->base.import_attach) + if (drm_gem_is_imported(&bo->base)) ret = qaic_prepare_import_bo(bo, hdr); else ret = qaic_prepare_export_bo(qdev, bo, hdr); - - if (ret == 0) - bo->dir = hdr->dir; + bo->dir = hdr->dir; + bo->dbc = &qdev->dbc[hdr->dbc_id]; + bo->nr_slice = hdr->count; return ret; } @@ -868,7 +913,6 @@ static void qaic_unprepare_import_bo(struct qaic_bo *bo) { dma_buf_unmap_attachment(bo->base.import_attach, bo->sgt, bo->dir); bo->sgt = NULL; - bo->size = 0; } static void qaic_unprepare_export_bo(struct qaic_device *qdev, struct qaic_bo *bo) @@ -878,12 +922,14 @@ static void qaic_unprepare_export_bo(struct qaic_device *qdev, struct qaic_bo *b static void qaic_unprepare_bo(struct qaic_device *qdev, struct qaic_bo *bo) { - if (bo->base.import_attach) + if (drm_gem_is_imported(&bo->base)) qaic_unprepare_import_bo(bo); else qaic_unprepare_export_bo(qdev, bo); bo->dir = 0; + bo->dbc = NULL; + bo->nr_slice = 0; } static void qaic_free_slices_bo(struct qaic_bo *bo) @@ -892,6 +938,9 @@ static void qaic_free_slices_bo(struct qaic_bo *bo) list_for_each_entry_safe(slice, temp, &bo->slices, slice) kref_put(&slice->ref_count, free_slice); + if (WARN_ON_ONCE(bo->total_slice_nents != 0)) + bo->total_slice_nents = 0; + bo->nr_slice = 0; } static int qaic_attach_slicing_bo(struct qaic_device *qdev, struct qaic_bo *bo, @@ -908,15 +957,11 @@ static int qaic_attach_slicing_bo(struct qaic_device *qdev, struct qaic_bo *bo, } } - if (bo->total_slice_nents > qdev->dbc[hdr->dbc_id].nelem) { + if (bo->total_slice_nents > bo->dbc->nelem) { qaic_free_slices_bo(bo); return -ENOSPC; } - bo->sliced = true; - bo->nr_slice = hdr->count; - list_add_tail(&bo->bo_list, &qdev->dbc[hdr->dbc_id].bo_lists); - return 0; } @@ -937,11 +982,9 @@ int qaic_attach_slice_bo_ioctl(struct drm_device *dev, void *data, struct drm_fi if (args->hdr.count == 0) return -EINVAL; - arg_size = args->hdr.count * sizeof(*slice_ent); - if (arg_size / args->hdr.count != sizeof(*slice_ent)) - return -EINVAL; - - if (args->hdr.size == 0) + if (check_mul_overflow((unsigned long)args->hdr.count, + (unsigned long)sizeof(*slice_ent), + &arg_size)) return -EINVAL; if (!(args->hdr.dir == DMA_TO_DEVICE || args->hdr.dir == DMA_FROM_DEVICE)) @@ -959,7 +1002,7 @@ 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; } @@ -971,33 +1014,30 @@ int qaic_attach_slice_bo_ioctl(struct drm_device *dev, void *data, struct drm_fi user_data = u64_to_user_ptr(args->data); - slice_ent = kzalloc(arg_size, GFP_KERNEL); - if (!slice_ent) { - ret = -EINVAL; + slice_ent = memdup_user(user_data, arg_size); + if (IS_ERR(slice_ent)) { + ret = PTR_ERR(slice_ent); goto unlock_dev_srcu; } - ret = copy_from_user(slice_ent, user_data, arg_size); - if (ret) { - ret = -EFAULT; - 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) + goto put_bo; if (bo->sliced) { ret = -EINVAL; - goto put_bo; + goto unlock_bo; } dbc = &qdev->dbc[args->hdr.dbc_id]; @@ -1007,6 +1047,11 @@ int qaic_attach_slice_bo_ioctl(struct drm_device *dev, void *data, struct drm_fi goto unlock_ch_srcu; } + if (dbc->id == qdev->ssr_dbc) { + ret = -EPIPE; + goto unlock_ch_srcu; + } + ret = qaic_prepare_bo(qdev, bo, &args->hdr); if (ret) goto unlock_ch_srcu; @@ -1018,9 +1063,10 @@ int qaic_attach_slice_bo_ioctl(struct drm_device *dev, void *data, struct drm_fi if (args->hdr.dir == DMA_TO_DEVICE) dma_sync_sgtable_for_cpu(&qdev->pdev->dev, bo->sgt, args->hdr.dir); - bo->dbc = dbc; + bo->sliced = true; + list_add_tail(&bo->bo_list, &bo->dbc->bo_lists); srcu_read_unlock(&dbc->ch_lock, rcu_id); - drm_gem_object_put(obj); + mutex_unlock(&bo->lock); kfree(slice_ent); srcu_read_unlock(&qdev->dev_lock, qdev_rcu_id); srcu_read_unlock(&usr->qddev_lock, usr_rcu_id); @@ -1031,6 +1077,8 @@ unprepare_bo: qaic_unprepare_bo(qdev, bo); unlock_ch_srcu: srcu_read_unlock(&dbc->ch_lock, rcu_id); +unlock_bo: + mutex_unlock(&bo->lock); put_bo: drm_gem_object_put(obj); free_slice_ent: @@ -1042,6 +1090,16 @@ 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) { @@ -1050,27 +1108,20 @@ 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; @@ -1078,46 +1129,31 @@ 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; @@ -1126,22 +1162,21 @@ 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)); /* @@ -1152,6 +1187,9 @@ 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; @@ -1168,7 +1206,6 @@ 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; @@ -1185,65 +1222,59 @@ static int send_bo_list_to_device(struct qaic_device *qdev, struct drm_file *fil } bo = to_qaic_bo(obj); + ret = mutex_lock_interruptible(&bo->lock); + if (ret) + goto failed_to_send_bo; if (!bo->sliced) { ret = -EINVAL; - goto failed_to_send_bo; + goto unlock_bo; } - if (is_partial && pexec[i].resize > bo->size) { + if (is_partial && pexec[i].resize > bo->base.size) { ret = -EINVAL; - goto failed_to_send_bo; + goto unlock_bo; } 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 failed_to_send_bo; + goto unlock_bo; } 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) { - bo->queued = false; spin_unlock_irqrestore(&dbc->xfer_lock, flags); - goto failed_to_send_bo; + goto unlock_bo; } } reinit_completion(&bo->xfer_done); list_add_tail(&bo->xfer_list, &dbc->xfer_list); spin_unlock_irqrestore(&dbc->xfer_lock, flags); dma_sync_sgtable_for_device(&qdev->pdev->dev, bo->sgt, bo->dir); + mutex_unlock(&bo->lock); } return 0; +unlock_bo: + mutex_unlock(&bo->lock); failed_to_send_bo: if (likely(obj)) drm_gem_object_put(obj); @@ -1251,8 +1282,7 @@ 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); @@ -1299,8 +1329,6 @@ static int __qaic_execute_bo_ioctl(struct drm_device *dev, void *data, struct dr int usr_rcu_id, qdev_rcu_id; struct qaic_device *qdev; struct qaic_user *usr; - u8 __user *user_data; - unsigned long n; u64 received_ts; u32 queue_level; u64 submit_ts; @@ -1313,20 +1341,12 @@ static int __qaic_execute_bo_ioctl(struct drm_device *dev, void *data, struct dr received_ts = ktime_get_ns(); size = is_partial ? sizeof(struct qaic_partial_execute_entry) : sizeof(*exec); - n = (unsigned long)size * args->hdr.count; - if (args->hdr.count == 0 || n / args->hdr.count != size) + if (args->hdr.count == 0) return -EINVAL; - user_data = u64_to_user_ptr(args->data); - - exec = kcalloc(args->hdr.count, size, GFP_KERNEL); - if (!exec) - return -ENOMEM; - - if (copy_from_user(exec, user_data, n)) { - ret = -EFAULT; - goto free_exec; - } + exec = memdup_array_user(u64_to_user_ptr(args->data), args->hdr.count, size); + if (IS_ERR(exec)) + return PTR_ERR(exec); usr = file_priv->driver_priv; usr_rcu_id = srcu_read_lock(&usr->qddev_lock); @@ -1337,7 +1357,7 @@ 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; } @@ -1355,13 +1375,22 @@ static int __qaic_execute_bo_ioctl(struct drm_device *dev, void *data, struct dr goto release_ch_rcu; } + if (dbc->id == qdev->ssr_dbc) { + ret = -EPIPE; + goto release_ch_rcu; + } + + ret = mutex_lock_interruptible(&dbc->req_lock); + if (ret) + goto release_ch_rcu; + head = readl(dbc->dbc_base + REQHP_OFF); tail = readl(dbc->dbc_base + REQTP_OFF); if (head == U32_MAX || tail == U32_MAX) { /* PCI link error */ ret = -ENODEV; - goto release_ch_rcu; + goto unlock_req_lock; } queue_level = head <= tail ? tail - head : dbc->nelem - (head - tail); @@ -1369,11 +1398,12 @@ static int __qaic_execute_bo_ioctl(struct drm_device *dev, void *data, struct dr ret = send_bo_list_to_device(qdev, file_priv, exec, args->hdr.count, is_partial, dbc, head, &tail); if (ret) - goto release_ch_rcu; + goto unlock_req_lock; /* Finalize commit to hardware */ submit_ts = ktime_get_ns(); writel(tail, dbc->dbc_base + REQTP_OFF); + mutex_unlock(&dbc->req_lock); update_profiling_data(file_priv, exec, args->hdr.count, is_partial, received_ts, submit_ts, queue_level); @@ -1381,13 +1411,15 @@ static int __qaic_execute_bo_ioctl(struct drm_device *dev, void *data, struct dr if (datapath_polling) schedule_work(&dbc->poll_work); +unlock_req_lock: + if (ret) + mutex_unlock(&dbc->req_lock); release_ch_rcu: srcu_read_unlock(&dbc->ch_lock, rcu_id); unlock_dev_srcu: srcu_read_unlock(&qdev->dev_lock, qdev_rcu_id); unlock_usr_srcu: srcu_read_unlock(&usr->qddev_lock, usr_rcu_id); -free_exec: kfree(exec); return ret; } @@ -1444,6 +1476,16 @@ 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; @@ -1466,12 +1508,13 @@ 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; } -void irq_polling_work(struct work_struct *work) +void qaic_irq_polling_work(struct work_struct *work) { struct dma_bridge_chan *dbc = container_of(work, struct dma_bridge_chan, poll_work); unsigned long flags; @@ -1482,7 +1525,7 @@ 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; } @@ -1537,12 +1580,12 @@ 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) { @@ -1602,8 +1645,7 @@ 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); @@ -1623,14 +1665,14 @@ 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; } @@ -1639,9 +1681,9 @@ 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; @@ -1672,7 +1714,7 @@ 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; } @@ -1690,6 +1732,11 @@ int qaic_wait_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *file goto unlock_ch_srcu; } + if (dbc->id == qdev->ssr_dbc) { + ret = -EPIPE; + goto unlock_ch_srcu; + } + obj = drm_gem_object_lookup(file_priv, args->handle); if (!obj) { ret = -ENOENT; @@ -1710,6 +1757,9 @@ int qaic_wait_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *file if (!dbc->usr) ret = -EPERM; + if (dbc->id == qdev->ssr_dbc) + ret = -EPIPE; + put_obj: drm_gem_object_put(obj); unlock_ch_srcu: @@ -1730,7 +1780,8 @@ int qaic_perf_stats_bo_ioctl(struct drm_device *dev, void *data, struct drm_file struct qaic_device *qdev; struct qaic_user *usr; struct qaic_bo *bo; - int ret, i; + int ret = 0; + int i; usr = file_priv->driver_priv; usr_rcu_id = srcu_read_lock(&usr->qddev_lock); @@ -1741,7 +1792,7 @@ 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; } @@ -1751,18 +1802,12 @@ int qaic_perf_stats_bo_ioctl(struct drm_device *dev, void *data, struct drm_file goto unlock_dev_srcu; } - ent = kcalloc(args->hdr.count, sizeof(*ent), GFP_KERNEL); - if (!ent) { - ret = -EINVAL; + ent = memdup_array_user(u64_to_user_ptr(args->data), args->hdr.count, sizeof(*ent)); + if (IS_ERR(ent)) { + ret = PTR_ERR(ent); goto unlock_dev_srcu; } - ret = copy_from_user(ent, u64_to_user_ptr(args->data), args->hdr.count * sizeof(*ent)); - if (ret) { - ret = -EFAULT; - goto free_ent; - } - for (i = 0; i < args->hdr.count; i++) { obj = drm_gem_object_lookup(file_priv, ent[i].handle); if (!obj) { @@ -1770,6 +1815,16 @@ int qaic_perf_stats_bo_ioctl(struct drm_device *dev, void *data, struct drm_file goto free_ent; } bo = to_qaic_bo(obj); + if (!bo->sliced) { + drm_gem_object_put(obj); + ret = -EINVAL; + goto free_ent; + } + if (bo->dbc->id != args->hdr.dbc_id) { + drm_gem_object_put(obj); + ret = -EINVAL; + goto free_ent; + } /* * perf stats ioctl is called before wait ioctl is complete then * the latency information is invalid. @@ -1799,6 +1854,91 @@ unlock_usr_srcu: return ret; } +static void detach_slice_bo(struct qaic_device *qdev, struct qaic_bo *bo) +{ + qaic_free_slices_bo(bo); + qaic_unprepare_bo(qdev, bo); + qaic_init_bo(bo, true); + list_del(&bo->bo_list); + drm_gem_object_put(&bo->base); +} + +int qaic_detach_slice_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) +{ + struct qaic_detach_slice *args = data; + int rcu_id, usr_rcu_id, qdev_rcu_id; + struct dma_bridge_chan *dbc; + struct drm_gem_object *obj; + struct qaic_device *qdev; + struct qaic_user *usr; + unsigned long flags; + struct qaic_bo *bo; + int ret; + + if (args->pad != 0) + return -EINVAL; + + usr = file_priv->driver_priv; + usr_rcu_id = srcu_read_lock(&usr->qddev_lock); + if (!usr->qddev) { + ret = -ENODEV; + goto unlock_usr_srcu; + } + + qdev = usr->qddev->qdev; + qdev_rcu_id = srcu_read_lock(&qdev->dev_lock); + if (qdev->dev_state != QAIC_ONLINE) { + ret = -ENODEV; + goto unlock_dev_srcu; + } + + obj = drm_gem_object_lookup(file_priv, args->handle); + if (!obj) { + ret = -ENOENT; + goto unlock_dev_srcu; + } + + bo = to_qaic_bo(obj); + ret = mutex_lock_interruptible(&bo->lock); + if (ret) + goto put_bo; + + if (!bo->sliced) { + ret = -EINVAL; + goto unlock_bo; + } + + dbc = bo->dbc; + rcu_id = srcu_read_lock(&dbc->ch_lock); + if (dbc->usr != usr) { + ret = -EINVAL; + goto unlock_ch_srcu; + } + + /* Check if BO is committed to H/W for DMA */ + spin_lock_irqsave(&dbc->xfer_lock, flags); + if (bo_queued(bo)) { + spin_unlock_irqrestore(&dbc->xfer_lock, flags); + ret = -EBUSY; + goto unlock_ch_srcu; + } + spin_unlock_irqrestore(&dbc->xfer_lock, flags); + + detach_slice_bo(qdev, bo); + +unlock_ch_srcu: + srcu_read_unlock(&dbc->ch_lock, rcu_id); +unlock_bo: + mutex_unlock(&bo->lock); +put_bo: + drm_gem_object_put(obj); +unlock_dev_srcu: + srcu_read_unlock(&qdev->dev_lock, qdev_rcu_id); +unlock_usr_srcu: + srcu_read_unlock(&usr->qddev_lock, usr_rcu_id); + return ret; +} + static void empty_xfer_list(struct qaic_device *qdev, struct dma_bridge_chan *dbc) { unsigned long flags; @@ -1807,9 +1947,14 @@ 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; + bo->perf_stats.req_received_ts = 0; + bo->perf_stats.req_submit_ts = 0; + bo->perf_stats.req_processed_ts = 0; + bo->perf_stats.queue_level_before = 0; dma_sync_sgtable_for_cpu(&qdev->pdev->dev, bo->sgt, bo->dir); complete_all(&bo->xfer_done); drm_gem_object_put(&bo->base); @@ -1818,6 +1963,17 @@ static void empty_xfer_list(struct qaic_device *qdev, struct dma_bridge_chan *db spin_unlock_irqrestore(&dbc->xfer_lock, flags); } +static void sync_empty_xfer_list(struct qaic_device *qdev, struct dma_bridge_chan *dbc) +{ + empty_xfer_list(qdev, dbc); + synchronize_srcu(&dbc->ch_lock); + /* + * Threads holding channel lock, may add more elements in the xfer_list. + * Flush out these elements from xfer_list. + */ + empty_xfer_list(qdev, dbc); +} + int disable_dbc(struct qaic_device *qdev, u32 dbc_id, struct qaic_user *usr) { if (!qdev->dbc[dbc_id].usr || qdev->dbc[dbc_id].usr->handle != usr->handle) @@ -1832,7 +1988,7 @@ int disable_dbc(struct qaic_device *qdev, u32 dbc_id, struct qaic_user *usr) * enable_dbc - Enable the DBC. DBCs are disabled by removing the context of * user. Add user context back to DBC to enable it. This function trusts the * DBC ID passed and expects the DBC to be disabled. - * @qdev: Qranium device handle + * @qdev: qaic device handle * @dbc_id: ID of the DBC * @usr: User context */ @@ -1846,18 +2002,11 @@ void wakeup_dbc(struct qaic_device *qdev, u32 dbc_id) struct dma_bridge_chan *dbc = &qdev->dbc[dbc_id]; dbc->usr = NULL; - empty_xfer_list(qdev, dbc); - synchronize_srcu(&dbc->ch_lock); - /* - * Threads holding channel lock, may add more elements in the xfer_list. - * Flush out these elements from xfer_list. - */ - empty_xfer_list(qdev, dbc); + sync_empty_xfer_list(qdev, dbc); } void release_dbc(struct qaic_device *qdev, u32 dbc_id) { - struct bo_slice *slice, *slice_temp; struct qaic_bo *bo, *bo_temp; struct dma_bridge_chan *dbc; @@ -1875,26 +2024,49 @@ void release_dbc(struct qaic_device *qdev, u32 dbc_id) dbc->usr = NULL; list_for_each_entry_safe(bo, bo_temp, &dbc->bo_lists, bo_list) { - list_for_each_entry_safe(slice, slice_temp, &bo->slices, slice) - kref_put(&slice->ref_count, free_slice); - bo->sliced = false; - INIT_LIST_HEAD(&bo->slices); - bo->total_slice_nents = 0; - bo->dir = 0; - bo->dbc = NULL; - bo->nr_slice = 0; - bo->nr_slice_xfer_done = 0; - bo->queued = false; - bo->req_id = 0; - init_completion(&bo->xfer_done); - complete_all(&bo->xfer_done); - list_del(&bo->bo_list); - bo->perf_stats.req_received_ts = 0; - bo->perf_stats.req_submit_ts = 0; - bo->perf_stats.req_processed_ts = 0; - bo->perf_stats.queue_level_before = 0; + drm_gem_object_get(&bo->base); + mutex_lock(&bo->lock); + detach_slice_bo(qdev, bo); + mutex_unlock(&bo->lock); + drm_gem_object_put(&bo->base); } 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); +} + +/* + * qaic_dbc_enter_ssr - Prepare to enter in sub system reset(SSR) for given DBC ID. + * @qdev: qaic device handle + * @dbc_id: ID of the DBC which will enter SSR + * + * The device will automatically deactivate the workload as not + * all errors can be silently recovered. The user will be + * notified and will need to decide the required recovery + * action to take. + */ +void qaic_dbc_enter_ssr(struct qaic_device *qdev, u32 dbc_id) +{ + qdev->ssr_dbc = dbc_id; + release_dbc(qdev, dbc_id); +} + +/* + * qaic_dbc_exit_ssr - Prepare to exit from sub system reset(SSR) for given DBC ID. + * @qdev: qaic device handle + * + * The DBC returns to an operational state and begins accepting work after exiting SSR. + */ +void qaic_dbc_exit_ssr(struct qaic_device *qdev) +{ + qdev->ssr_dbc = QAIC_SSR_DBC_SENTINEL; +} |