From ed21f6c3fe423b17211fa5a85cd028621a749a2a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Thu, 5 Jan 2023 10:45:36 +0100 Subject: [PATCH 1/9] drm/amdgpu: fix another missing fence reference in the CS code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit drm_sched_job_add_dependency() consumes the references of the gang members. Only triggered by mesh shaders. Signed-off-by: Christian König Fixes: 1728baa7e4e6 ("drm/amdgpu: use scheduler dependencies for CS") Tested-by: Mike Lothian Tested-by: Bert Karwatzki Link: https://patchwork.freedesktop.org/patch/msgid/20230105111703.52695-1-christian.koenig@amd.com Reviewed-by: Alex Deucher Reviewed-by: Luben Tuikov --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 8516c814bc9b..47763ac0d14a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -1254,9 +1254,12 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, continue; fence = &p->jobs[i]->base.s_fence->scheduled; + dma_fence_get(fence); r = drm_sched_job_add_dependency(&leader->base, fence); - if (r) + if (r) { + dma_fence_put(fence); goto error_cleanup; + } } if (p->gang_size > 1) { From 41cc108b2451e0c65e8c9c9f6a3e103b62b60bb3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Thu, 5 Jan 2023 11:27:16 +0100 Subject: [PATCH 2/9] drm/amdgpu: fix missing dma_fence_put in error path MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When the fence can't be added we need to drop the reference. Suggested-by: Bert Karwatzki Signed-off-by: Christian König Link: https://patchwork.freedesktop.org/patch/msgid/20230105111703.52695-2-christian.koenig@amd.com Reviewed-by: Alex Deucher Reviewed-by: Luben Tuikov --- drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c index bac7976975bd..dcd8c066bc1f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c @@ -391,8 +391,10 @@ int amdgpu_sync_push_to_job(struct amdgpu_sync *sync, struct amdgpu_job *job) dma_fence_get(f); r = drm_sched_job_add_dependency(&job->base, f); - if (r) + if (r) { + dma_fence_put(f); return r; + } } return 0; } From 0688773f0710528e1ab302c3d6317e269f2e2e6e Mon Sep 17 00:00:00 2001 From: Patrick Thompson Date: Tue, 20 Dec 2022 15:58:26 -0500 Subject: [PATCH 3/9] drm: Add orientation quirk for Lenovo ideapad D330-10IGL Panel is 800x1280 but mounted on a detachable form factor sideways. Signed-off-by: Patrick Thompson Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20221220205826.178008-1-ptf@google.com --- drivers/gpu/drm/drm_panel_orientation_quirks.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/drivers/gpu/drm/drm_panel_orientation_quirks.c b/drivers/gpu/drm/drm_panel_orientation_quirks.c index 52d8800a8ab8..3659f0465a72 100644 --- a/drivers/gpu/drm/drm_panel_orientation_quirks.c +++ b/drivers/gpu/drm/drm_panel_orientation_quirks.c @@ -304,6 +304,12 @@ static const struct dmi_system_id orientation_data[] = { DMI_EXACT_MATCH(DMI_PRODUCT_VERSION, "Lenovo ideapad D330-10IGM"), }, .driver_data = (void *)&lcd1200x1920_rightside_up, + }, { /* Lenovo Ideapad D330-10IGL (HD) */ + .matches = { + DMI_EXACT_MATCH(DMI_SYS_VENDOR, "LENOVO"), + DMI_EXACT_MATCH(DMI_PRODUCT_VERSION, "Lenovo ideapad D330-10IGL"), + }, + .driver_data = (void *)&lcd800x1280_rightside_up, }, { /* Lenovo Yoga Book X90F / X91F / X91L */ .matches = { /* Non exact match to match all versions */ From 52531258318ed59a2dc5a43df2eaf0eb1d65438e Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Fri, 16 Dec 2022 15:33:55 -0800 Subject: [PATCH 4/9] drm/virtio: Fix GEM handle creation UAF Userspace can guess the handle value and try to race GEM object creation with handle close, resulting in a use-after-free if we dereference the object after dropping the handle's reference. For that reason, dropping the handle's reference must be done *after* we are done dereferencing the object. Signed-off-by: Rob Clark Reviewed-by: Chia-I Wu Fixes: 62fb7a5e1096 ("virtio-gpu: add 3d/virgl support") Cc: stable@vger.kernel.org Signed-off-by: Dmitry Osipenko Link: https://patchwork.freedesktop.org/patch/msgid/20221216233355.542197-2-robdclark@gmail.com --- drivers/gpu/drm/virtio/virtgpu_ioctl.c | 19 +++++++++++++++++-- 1 file changed, 17 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/virtio/virtgpu_ioctl.c b/drivers/gpu/drm/virtio/virtgpu_ioctl.c index 5d05093014ac..9f4a90493aea 100644 --- a/drivers/gpu/drm/virtio/virtgpu_ioctl.c +++ b/drivers/gpu/drm/virtio/virtgpu_ioctl.c @@ -358,10 +358,18 @@ static int virtio_gpu_resource_create_ioctl(struct drm_device *dev, void *data, drm_gem_object_release(obj); return ret; } - drm_gem_object_put(obj); rc->res_handle = qobj->hw_res_handle; /* similiar to a VM address */ rc->bo_handle = handle; + + /* + * The handle owns the reference now. But we must drop our + * remaining reference *after* we no longer need to dereference + * the obj. Otherwise userspace could guess the handle and + * race closing it from another thread. + */ + drm_gem_object_put(obj); + return 0; } @@ -723,11 +731,18 @@ static int virtio_gpu_resource_create_blob_ioctl(struct drm_device *dev, drm_gem_object_release(obj); return ret; } - drm_gem_object_put(obj); rc_blob->res_handle = bo->hw_res_handle; rc_blob->bo_handle = handle; + /* + * The handle owns the reference now. But we must drop our + * remaining reference *after* we no longer need to dereference + * the obj. Otherwise userspace could guess the handle and + * race closing it from another thread. + */ + drm_gem_object_put(obj); + return 0; } From a309c7194e8a2f8bd4539b9449917913f6c2cd50 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 7 Dec 2022 12:29:07 -0500 Subject: [PATCH 5/9] drm/vmwgfx: Remove rcu locks from user resources User resource lookups used rcu to avoid two extra atomics. Unfortunately the rcu paths were buggy and it was easy to make the driver crash by submitting command buffers from two different threads. Because the lookups never show up in performance profiles replace them with a regular spin lock which fixes the races in accesses to those shared resources. Fixes kernel oops'es in IGT's vmwgfx execution_buffer stress test and seen crashes with apps using shared resources. Fixes: e14c02e6b699 ("drm/vmwgfx: Look up objects without taking a reference") Signed-off-by: Zack Rusin Reviewed-by: Martin Krastev Reviewed-by: Maaz Mombasawala Link: https://patchwork.freedesktop.org/patch/msgid/20221207172907.959037-1-zack@kde.org --- drivers/gpu/drm/vmwgfx/ttm_object.c | 41 +---- drivers/gpu/drm/vmwgfx/ttm_object.h | 14 -- drivers/gpu/drm/vmwgfx/vmwgfx_bo.c | 38 ----- drivers/gpu/drm/vmwgfx/vmwgfx_drv.h | 18 +-- drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c | 182 +++++++++++------------ drivers/gpu/drm/vmwgfx/vmwgfx_resource.c | 33 ---- 6 files changed, 90 insertions(+), 236 deletions(-) diff --git a/drivers/gpu/drm/vmwgfx/ttm_object.c b/drivers/gpu/drm/vmwgfx/ttm_object.c index 932b125ebf3d..ddf8373c1d77 100644 --- a/drivers/gpu/drm/vmwgfx/ttm_object.c +++ b/drivers/gpu/drm/vmwgfx/ttm_object.c @@ -254,40 +254,6 @@ void ttm_base_object_unref(struct ttm_base_object **p_base) kref_put(&base->refcount, ttm_release_base); } -/** - * ttm_base_object_noref_lookup - look up a base object without reference - * @tfile: The struct ttm_object_file the object is registered with. - * @key: The object handle. - * - * This function looks up a ttm base object and returns a pointer to it - * without refcounting the pointer. The returned pointer is only valid - * until ttm_base_object_noref_release() is called, and the object - * pointed to by the returned pointer may be doomed. Any persistent usage - * of the object requires a refcount to be taken using kref_get_unless_zero(). - * Iff this function returns successfully it needs to be paired with - * ttm_base_object_noref_release() and no sleeping- or scheduling functions - * may be called inbetween these function callse. - * - * Return: A pointer to the object if successful or NULL otherwise. - */ -struct ttm_base_object * -ttm_base_object_noref_lookup(struct ttm_object_file *tfile, uint64_t key) -{ - struct vmwgfx_hash_item *hash; - int ret; - - rcu_read_lock(); - ret = ttm_tfile_find_ref_rcu(tfile, key, &hash); - if (ret) { - rcu_read_unlock(); - return NULL; - } - - __release(RCU); - return hlist_entry(hash, struct ttm_ref_object, hash)->obj; -} -EXPORT_SYMBOL(ttm_base_object_noref_lookup); - struct ttm_base_object *ttm_base_object_lookup(struct ttm_object_file *tfile, uint64_t key) { @@ -295,15 +261,16 @@ struct ttm_base_object *ttm_base_object_lookup(struct ttm_object_file *tfile, struct vmwgfx_hash_item *hash; int ret; - rcu_read_lock(); - ret = ttm_tfile_find_ref_rcu(tfile, key, &hash); + spin_lock(&tfile->lock); + ret = ttm_tfile_find_ref(tfile, key, &hash); if (likely(ret == 0)) { base = hlist_entry(hash, struct ttm_ref_object, hash)->obj; if (!kref_get_unless_zero(&base->refcount)) base = NULL; } - rcu_read_unlock(); + spin_unlock(&tfile->lock); + return base; } diff --git a/drivers/gpu/drm/vmwgfx/ttm_object.h b/drivers/gpu/drm/vmwgfx/ttm_object.h index f0ebbe340ad6..8098a3846bae 100644 --- a/drivers/gpu/drm/vmwgfx/ttm_object.h +++ b/drivers/gpu/drm/vmwgfx/ttm_object.h @@ -307,18 +307,4 @@ extern int ttm_prime_handle_to_fd(struct ttm_object_file *tfile, #define ttm_prime_object_kfree(__obj, __prime) \ kfree_rcu(__obj, __prime.base.rhead) -struct ttm_base_object * -ttm_base_object_noref_lookup(struct ttm_object_file *tfile, uint64_t key); - -/** - * ttm_base_object_noref_release - release a base object pointer looked up - * without reference - * - * Releases a base object pointer looked up with ttm_base_object_noref_lookup(). - */ -static inline void ttm_base_object_noref_release(void) -{ - __acquire(RCU); - rcu_read_unlock(); -} #endif diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c index 321c551784a1..aa1cd5126a32 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c @@ -715,44 +715,6 @@ int vmw_user_bo_lookup(struct drm_file *filp, return 0; } -/** - * vmw_user_bo_noref_lookup - Look up a vmw user buffer object without reference - * @filp: The TTM object file the handle is registered with. - * @handle: The user buffer object handle. - * - * This function looks up a struct vmw_bo and returns a pointer to the - * struct vmw_buffer_object it derives from without refcounting the pointer. - * The returned pointer is only valid until vmw_user_bo_noref_release() is - * called, and the object pointed to by the returned pointer may be doomed. - * Any persistent usage of the object requires a refcount to be taken using - * ttm_bo_reference_unless_doomed(). Iff this function returns successfully it - * needs to be paired with vmw_user_bo_noref_release() and no sleeping- - * or scheduling functions may be called in between these function calls. - * - * Return: A struct vmw_buffer_object pointer if successful or negative - * error pointer on failure. - */ -struct vmw_buffer_object * -vmw_user_bo_noref_lookup(struct drm_file *filp, u32 handle) -{ - struct vmw_buffer_object *vmw_bo; - struct ttm_buffer_object *bo; - struct drm_gem_object *gobj = drm_gem_object_lookup(filp, handle); - - if (!gobj) { - DRM_ERROR("Invalid buffer object handle 0x%08lx.\n", - (unsigned long)handle); - return ERR_PTR(-ESRCH); - } - vmw_bo = gem_to_vmw_bo(gobj); - bo = ttm_bo_get_unless_zero(&vmw_bo->base); - vmw_bo = vmw_buffer_object(bo); - drm_gem_object_put(gobj); - - return vmw_bo; -} - - /** * vmw_bo_fence_single - Utility function to fence a single TTM buffer * object without unreserving it. diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h index b062b020b378..5acbf5849b27 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h @@ -830,12 +830,7 @@ extern int vmw_user_resource_lookup_handle( uint32_t handle, const struct vmw_user_resource_conv *converter, struct vmw_resource **p_res); -extern struct vmw_resource * -vmw_user_resource_noref_lookup_handle(struct vmw_private *dev_priv, - struct ttm_object_file *tfile, - uint32_t handle, - const struct vmw_user_resource_conv * - converter); + extern int vmw_stream_claim_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv); extern int vmw_stream_unref_ioctl(struct drm_device *dev, void *data, @@ -874,15 +869,6 @@ static inline bool vmw_resource_mob_attached(const struct vmw_resource *res) return !RB_EMPTY_NODE(&res->mob_node); } -/** - * vmw_user_resource_noref_release - release a user resource pointer looked up - * without reference - */ -static inline void vmw_user_resource_noref_release(void) -{ - ttm_base_object_noref_release(); -} - /** * Buffer object helper functions - vmwgfx_bo.c */ @@ -934,8 +920,6 @@ extern void vmw_bo_unmap(struct vmw_buffer_object *vbo); extern void vmw_bo_move_notify(struct ttm_buffer_object *bo, struct ttm_resource *mem); extern void vmw_bo_swap_notify(struct ttm_buffer_object *bo); -extern struct vmw_buffer_object * -vmw_user_bo_noref_lookup(struct drm_file *filp, u32 handle); /** * vmw_bo_adjust_prio - Adjust the buffer object eviction priority diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c index a5379f6fb5ab..a44d53e33cdb 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c @@ -290,20 +290,26 @@ static void vmw_execbuf_rcache_update(struct vmw_res_cache_entry *rcache, rcache->valid_handle = 0; } +enum vmw_val_add_flags { + vmw_val_add_flag_none = 0, + vmw_val_add_flag_noctx = 1 << 0, +}; + /** - * vmw_execbuf_res_noref_val_add - Add a resource described by an unreferenced - * rcu-protected pointer to the validation list. + * vmw_execbuf_res_val_add - Add a resource to the validation list. * * @sw_context: Pointer to the software context. * @res: Unreferenced rcu-protected pointer to the resource. * @dirty: Whether to change dirty status. + * @flags: specifies whether to use the context or not * * Returns: 0 on success. Negative error code on failure. Typical error codes * are %-EINVAL on inconsistency and %-ESRCH if the resource was doomed. */ -static int vmw_execbuf_res_noref_val_add(struct vmw_sw_context *sw_context, - struct vmw_resource *res, - u32 dirty) +static int vmw_execbuf_res_val_add(struct vmw_sw_context *sw_context, + struct vmw_resource *res, + u32 dirty, + u32 flags) { struct vmw_private *dev_priv = res->dev_priv; int ret; @@ -318,24 +324,30 @@ static int vmw_execbuf_res_noref_val_add(struct vmw_sw_context *sw_context, if (dirty) vmw_validation_res_set_dirty(sw_context->ctx, rcache->private, dirty); - vmw_user_resource_noref_release(); return 0; } - priv_size = vmw_execbuf_res_size(dev_priv, res_type); - ret = vmw_validation_add_resource(sw_context->ctx, res, priv_size, - dirty, (void **)&ctx_info, - &first_usage); - vmw_user_resource_noref_release(); - if (ret) - return ret; - - if (priv_size && first_usage) { - ret = vmw_cmd_ctx_first_setup(dev_priv, sw_context, res, - ctx_info); - if (ret) { - VMW_DEBUG_USER("Failed first usage context setup.\n"); + if ((flags & vmw_val_add_flag_noctx) != 0) { + ret = vmw_validation_add_resource(sw_context->ctx, res, 0, dirty, + (void **)&ctx_info, NULL); + if (ret) return ret; + + } else { + priv_size = vmw_execbuf_res_size(dev_priv, res_type); + ret = vmw_validation_add_resource(sw_context->ctx, res, priv_size, + dirty, (void **)&ctx_info, + &first_usage); + if (ret) + return ret; + + if (priv_size && first_usage) { + ret = vmw_cmd_ctx_first_setup(dev_priv, sw_context, res, + ctx_info); + if (ret) { + VMW_DEBUG_USER("Failed first usage context setup.\n"); + return ret; + } } } @@ -343,43 +355,6 @@ static int vmw_execbuf_res_noref_val_add(struct vmw_sw_context *sw_context, return 0; } -/** - * vmw_execbuf_res_noctx_val_add - Add a non-context resource to the resource - * validation list if it's not already on it - * - * @sw_context: Pointer to the software context. - * @res: Pointer to the resource. - * @dirty: Whether to change dirty status. - * - * Returns: Zero on success. Negative error code on failure. - */ -static int vmw_execbuf_res_noctx_val_add(struct vmw_sw_context *sw_context, - struct vmw_resource *res, - u32 dirty) -{ - struct vmw_res_cache_entry *rcache; - enum vmw_res_type res_type = vmw_res_type(res); - void *ptr; - int ret; - - rcache = &sw_context->res_cache[res_type]; - if (likely(rcache->valid && rcache->res == res)) { - if (dirty) - vmw_validation_res_set_dirty(sw_context->ctx, - rcache->private, dirty); - return 0; - } - - ret = vmw_validation_add_resource(sw_context->ctx, res, 0, dirty, - &ptr, NULL); - if (ret) - return ret; - - vmw_execbuf_rcache_update(rcache, res, ptr); - - return 0; -} - /** * vmw_view_res_val_add - Add a view and the surface it's pointing to to the * validation list @@ -398,13 +373,13 @@ static int vmw_view_res_val_add(struct vmw_sw_context *sw_context, * First add the resource the view is pointing to, otherwise it may be * swapped out when the view is validated. */ - ret = vmw_execbuf_res_noctx_val_add(sw_context, vmw_view_srf(view), - vmw_view_dirtying(view)); + ret = vmw_execbuf_res_val_add(sw_context, vmw_view_srf(view), + vmw_view_dirtying(view), vmw_val_add_flag_noctx); if (ret) return ret; - return vmw_execbuf_res_noctx_val_add(sw_context, view, - VMW_RES_DIRTY_NONE); + return vmw_execbuf_res_val_add(sw_context, view, VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); } /** @@ -475,8 +450,9 @@ static int vmw_resource_context_res_add(struct vmw_private *dev_priv, if (IS_ERR(res)) continue; - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_SET); + ret = vmw_execbuf_res_val_add(sw_context, res, + VMW_RES_DIRTY_SET, + vmw_val_add_flag_noctx); if (unlikely(ret != 0)) return ret; } @@ -490,9 +466,9 @@ static int vmw_resource_context_res_add(struct vmw_private *dev_priv, if (vmw_res_type(entry->res) == vmw_res_view) ret = vmw_view_res_val_add(sw_context, entry->res); else - ret = vmw_execbuf_res_noctx_val_add - (sw_context, entry->res, - vmw_binding_dirtying(entry->bt)); + ret = vmw_execbuf_res_val_add(sw_context, entry->res, + vmw_binding_dirtying(entry->bt), + vmw_val_add_flag_noctx); if (unlikely(ret != 0)) break; } @@ -658,7 +634,8 @@ vmw_cmd_res_check(struct vmw_private *dev_priv, { struct vmw_res_cache_entry *rcache = &sw_context->res_cache[res_type]; struct vmw_resource *res; - int ret; + int ret = 0; + bool needs_unref = false; if (p_res) *p_res = NULL; @@ -683,17 +660,18 @@ vmw_cmd_res_check(struct vmw_private *dev_priv, if (ret) return ret; - res = vmw_user_resource_noref_lookup_handle - (dev_priv, sw_context->fp->tfile, *id_loc, converter); - if (IS_ERR(res)) { + ret = vmw_user_resource_lookup_handle + (dev_priv, sw_context->fp->tfile, *id_loc, converter, &res); + if (ret != 0) { VMW_DEBUG_USER("Could not find/use resource 0x%08x.\n", (unsigned int) *id_loc); - return PTR_ERR(res); - } - - ret = vmw_execbuf_res_noref_val_add(sw_context, res, dirty); - if (unlikely(ret != 0)) return ret; + } + needs_unref = true; + + ret = vmw_execbuf_res_val_add(sw_context, res, dirty, vmw_val_add_flag_none); + if (unlikely(ret != 0)) + goto res_check_done; if (rcache->valid && rcache->res == res) { rcache->valid_handle = true; @@ -708,7 +686,11 @@ vmw_cmd_res_check(struct vmw_private *dev_priv, if (p_res) *p_res = res; - return 0; +res_check_done: + if (needs_unref) + vmw_resource_unreference(&res); + + return ret; } /** @@ -1171,9 +1153,9 @@ static int vmw_translate_mob_ptr(struct vmw_private *dev_priv, int ret; vmw_validation_preload_bo(sw_context->ctx); - vmw_bo = vmw_user_bo_noref_lookup(sw_context->filp, handle); - if (IS_ERR(vmw_bo)) { - VMW_DEBUG_USER("Could not find or use MOB buffer.\n"); + ret = vmw_user_bo_lookup(sw_context->filp, handle, &vmw_bo); + if (ret != 0) { + drm_dbg(&dev_priv->drm, "Could not find or use MOB buffer.\n"); return PTR_ERR(vmw_bo); } ret = vmw_validation_add_bo(sw_context->ctx, vmw_bo, true, false); @@ -1225,9 +1207,9 @@ static int vmw_translate_guest_ptr(struct vmw_private *dev_priv, int ret; vmw_validation_preload_bo(sw_context->ctx); - vmw_bo = vmw_user_bo_noref_lookup(sw_context->filp, handle); - if (IS_ERR(vmw_bo)) { - VMW_DEBUG_USER("Could not find or use GMR region.\n"); + ret = vmw_user_bo_lookup(sw_context->filp, handle, &vmw_bo); + if (ret != 0) { + drm_dbg(&dev_priv->drm, "Could not find or use GMR region.\n"); return PTR_ERR(vmw_bo); } ret = vmw_validation_add_bo(sw_context->ctx, vmw_bo, false, false); @@ -2025,8 +2007,9 @@ static int vmw_cmd_set_shader(struct vmw_private *dev_priv, res = vmw_shader_lookup(vmw_context_res_man(ctx), cmd->body.shid, cmd->body.type); if (!IS_ERR(res)) { - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_NONE); + ret = vmw_execbuf_res_val_add(sw_context, res, + VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); if (unlikely(ret != 0)) return ret; @@ -2273,8 +2256,9 @@ static int vmw_cmd_dx_set_shader(struct vmw_private *dev_priv, return PTR_ERR(res); } - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_NONE); + ret = vmw_execbuf_res_val_add(sw_context, res, + VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); if (ret) return ret; } @@ -2777,8 +2761,8 @@ static int vmw_cmd_dx_bind_shader(struct vmw_private *dev_priv, return PTR_ERR(res); } - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_NONE); + ret = vmw_execbuf_res_val_add(sw_context, res, VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); if (ret) { VMW_DEBUG_USER("Error creating resource validation node.\n"); return ret; @@ -3098,8 +3082,8 @@ static int vmw_cmd_dx_bind_streamoutput(struct vmw_private *dev_priv, vmw_dx_streamoutput_set_size(res, cmd->body.sizeInBytes); - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_NONE); + ret = vmw_execbuf_res_val_add(sw_context, res, VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); if (ret) { DRM_ERROR("Error creating resource validation node.\n"); return ret; @@ -3148,8 +3132,8 @@ static int vmw_cmd_dx_set_streamoutput(struct vmw_private *dev_priv, return 0; } - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_NONE); + ret = vmw_execbuf_res_val_add(sw_context, res, VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); if (ret) { DRM_ERROR("Error creating resource validation node.\n"); return ret; @@ -4066,22 +4050,26 @@ static int vmw_execbuf_tie_context(struct vmw_private *dev_priv, if (ret) return ret; - res = vmw_user_resource_noref_lookup_handle + ret = vmw_user_resource_lookup_handle (dev_priv, sw_context->fp->tfile, handle, - user_context_converter); - if (IS_ERR(res)) { + user_context_converter, &res); + if (ret != 0) { VMW_DEBUG_USER("Could not find or user DX context 0x%08x.\n", (unsigned int) handle); - return PTR_ERR(res); + return ret; } - ret = vmw_execbuf_res_noref_val_add(sw_context, res, VMW_RES_DIRTY_SET); - if (unlikely(ret != 0)) + ret = vmw_execbuf_res_val_add(sw_context, res, VMW_RES_DIRTY_SET, + vmw_val_add_flag_none); + if (unlikely(ret != 0)) { + vmw_resource_unreference(&res); return ret; + } sw_context->dx_ctx_node = vmw_execbuf_info_from_res(sw_context, res); sw_context->man = vmw_context_res_man(res); + vmw_resource_unreference(&res); return 0; } diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c index f66caa540e14..c7d645e5ec7b 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c @@ -281,39 +281,6 @@ out_bad_resource: return ret; } -/** - * vmw_user_resource_noref_lookup_handle - lookup a struct resource from a - * TTM user-space handle and perform basic type checks - * - * @dev_priv: Pointer to a device private struct - * @tfile: Pointer to a struct ttm_object_file identifying the caller - * @handle: The TTM user-space handle - * @converter: Pointer to an object describing the resource type - * - * If the handle can't be found or is associated with an incorrect resource - * type, -EINVAL will be returned. - */ -struct vmw_resource * -vmw_user_resource_noref_lookup_handle(struct vmw_private *dev_priv, - struct ttm_object_file *tfile, - uint32_t handle, - const struct vmw_user_resource_conv - *converter) -{ - struct ttm_base_object *base; - - base = ttm_base_object_noref_lookup(tfile, handle); - if (!base) - return ERR_PTR(-ESRCH); - - if (unlikely(ttm_base_object_type(base) != converter->object_type)) { - ttm_base_object_noref_release(); - return ERR_PTR(-EINVAL); - } - - return converter->base_obj_to_res(base); -} - /* * Helper function that looks either a surface or bo. * From 3bd68b32c911a3a610ad782bb04d1a7bfc440638 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Mon, 9 Jan 2023 09:06:03 +0100 Subject: [PATCH 6/9] drm/amdgpu: fix pipeline sync v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This fixes a potential memory leak of dma_fence objects in the CS code as well as glitches in firefox because of missing pipeline sync. v2: use the scheduler instead of the fence context Signed-off-by: Christian König Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/2323 Tested-by: Michal Kubecek mkubecek@suse.cz Tested-by: Vlastimil Babka Acked-by: Alex Deucher Link: https://patchwork.freedesktop.org/patch/msgid/20230109130120.73389-1-christian.koenig@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 46 +++++++++++++++++--------- 1 file changed, 30 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 47763ac0d14a..7b5ce00f0602 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -61,6 +61,8 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, amdgpu_ctx_put(p->ctx); return -ECANCELED; } + + amdgpu_sync_create(&p->sync); return 0; } @@ -452,18 +454,6 @@ static int amdgpu_syncobj_lookup_and_add(struct amdgpu_cs_parser *p, } r = amdgpu_sync_fence(&p->sync, fence); - if (r) - goto error; - - /* - * When we have an explicit dependency it might be necessary to insert a - * pipeline sync to make sure that all caches etc are flushed and the - * next job actually sees the results from the previous one. - */ - if (fence->context == p->gang_leader->base.entity->fence_context) - r = amdgpu_sync_fence(&p->gang_leader->explicit_sync, fence); - -error: dma_fence_put(fence); return r; } @@ -1188,10 +1178,19 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p) { struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + struct drm_gpu_scheduler *sched; struct amdgpu_bo_list_entry *e; + struct dma_fence *fence; unsigned int i; int r; + r = amdgpu_ctx_wait_prev_fence(p->ctx, p->entities[p->gang_leader_idx]); + if (r) { + if (r != -ERESTARTSYS) + DRM_ERROR("amdgpu_ctx_wait_prev_fence failed.\n"); + return r; + } + list_for_each_entry(e, &p->validated, tv.head) { struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); struct dma_resv *resv = bo->tbo.base.resv; @@ -1211,10 +1210,24 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p) return r; } - r = amdgpu_ctx_wait_prev_fence(p->ctx, p->entities[p->gang_leader_idx]); - if (r && r != -ERESTARTSYS) - DRM_ERROR("amdgpu_ctx_wait_prev_fence failed.\n"); - return r; + sched = p->gang_leader->base.entity->rq->sched; + while ((fence = amdgpu_sync_get_fence(&p->sync))) { + struct drm_sched_fence *s_fence = to_drm_sched_fence(fence); + + /* + * When we have an dependency it might be necessary to insert a + * pipeline sync to make sure that all caches etc are flushed and the + * next job actually sees the results from the previous one + * before we start executing on the same scheduler ring. + */ + if (!s_fence || s_fence->sched != sched) + continue; + + r = amdgpu_sync_fence(&p->gang_leader->explicit_sync, fence); + if (r) + return r; + } + return 0; } static void amdgpu_cs_post_dependencies(struct amdgpu_cs_parser *p) @@ -1347,6 +1360,7 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser) { unsigned i; + amdgpu_sync_free(&parser->sync); for (i = 0; i < parser->num_post_deps; i++) { drm_syncobj_put(parser->post_deps[i].syncobj); kfree(parser->post_deps[i].chain); From 4e2ec2500bfc5cf429ddcfe78b49cb76b36bc46d Mon Sep 17 00:00:00 2001 From: Thomas Zimmermann Date: Tue, 10 Jan 2023 13:35:26 +0100 Subject: [PATCH 7/9] drm/nouveau: Remove file nouveau_fbcon.c Commit 4a16dd9d18a0 ("drm/nouveau/kms: switch to drm fbdev helpers") converted nouveau to generic fbdev emulation. The driver's internal implementation later got accidentally restored during a merge commit. Remove the file from the driver. No functional changes. v2: * point Fixes tag to merge commit (Alex) Signed-off-by: Thomas Zimmermann Reviewed-by: Alex Deucher Fixes: 4e291f2f5853 ("Merge tag 'drm-misc-next-2022-11-10-1' of git://anongit.freedesktop.org/drm/drm-misc into drm-next") Cc: Ben Skeggs Cc: Karol Herbst Cc: Lyude Paul Cc: Thomas Zimmermann Cc: Javier Martinez Canillas Cc: Sam Ravnborg Cc: Jani Nikula Cc: Dave Airlie Cc: dri-devel@lists.freedesktop.org Cc: nouveau@lists.freedesktop.org Link: https://patchwork.freedesktop.org/patch/msgid/20230110123526.28770-1-tzimmermann@suse.de --- drivers/gpu/drm/nouveau/nouveau_fbcon.c | 613 ------------------------ 1 file changed, 613 deletions(-) delete mode 100644 drivers/gpu/drm/nouveau/nouveau_fbcon.c diff --git a/drivers/gpu/drm/nouveau/nouveau_fbcon.c b/drivers/gpu/drm/nouveau/nouveau_fbcon.c deleted file mode 100644 index e87de7906f78..000000000000 --- a/drivers/gpu/drm/nouveau/nouveau_fbcon.c +++ /dev/null @@ -1,613 +0,0 @@ -/* - * Copyright © 2007 David Airlie - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - * - * Authors: - * David Airlie - */ - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include -#include -#include -#include - -#include "nouveau_drv.h" -#include "nouveau_gem.h" -#include "nouveau_bo.h" -#include "nouveau_fbcon.h" -#include "nouveau_chan.h" -#include "nouveau_vmm.h" - -#include "nouveau_crtc.h" - -MODULE_PARM_DESC(nofbaccel, "Disable fbcon acceleration"); -int nouveau_nofbaccel = 0; -module_param_named(nofbaccel, nouveau_nofbaccel, int, 0400); - -MODULE_PARM_DESC(fbcon_bpp, "fbcon bits-per-pixel (default: auto)"); -static int nouveau_fbcon_bpp; -module_param_named(fbcon_bpp, nouveau_fbcon_bpp, int, 0400); - -static void -nouveau_fbcon_fillrect(struct fb_info *info, const struct fb_fillrect *rect) -{ - struct nouveau_fbdev *fbcon = info->par; - struct nouveau_drm *drm = nouveau_drm(fbcon->helper.dev); - struct nvif_device *device = &drm->client.device; - int ret; - - if (info->state != FBINFO_STATE_RUNNING) - return; - - ret = -ENODEV; - if (!in_interrupt() && !(info->flags & FBINFO_HWACCEL_DISABLED) && - mutex_trylock(&drm->client.mutex)) { - if (device->info.family < NV_DEVICE_INFO_V0_TESLA) - ret = nv04_fbcon_fillrect(info, rect); - else - if (device->info.family < NV_DEVICE_INFO_V0_FERMI) - ret = nv50_fbcon_fillrect(info, rect); - else - ret = nvc0_fbcon_fillrect(info, rect); - mutex_unlock(&drm->client.mutex); - } - - if (ret == 0) - return; - - if (ret != -ENODEV) - nouveau_fbcon_gpu_lockup(info); - drm_fb_helper_cfb_fillrect(info, rect); -} - -static void -nouveau_fbcon_copyarea(struct fb_info *info, const struct fb_copyarea *image) -{ - struct nouveau_fbdev *fbcon = info->par; - struct nouveau_drm *drm = nouveau_drm(fbcon->helper.dev); - struct nvif_device *device = &drm->client.device; - int ret; - - if (info->state != FBINFO_STATE_RUNNING) - return; - - ret = -ENODEV; - if (!in_interrupt() && !(info->flags & FBINFO_HWACCEL_DISABLED) && - mutex_trylock(&drm->client.mutex)) { - if (device->info.family < NV_DEVICE_INFO_V0_TESLA) - ret = nv04_fbcon_copyarea(info, image); - else - if (device->info.family < NV_DEVICE_INFO_V0_FERMI) - ret = nv50_fbcon_copyarea(info, image); - else - ret = nvc0_fbcon_copyarea(info, image); - mutex_unlock(&drm->client.mutex); - } - - if (ret == 0) - return; - - if (ret != -ENODEV) - nouveau_fbcon_gpu_lockup(info); - drm_fb_helper_cfb_copyarea(info, image); -} - -static void -nouveau_fbcon_imageblit(struct fb_info *info, const struct fb_image *image) -{ - struct nouveau_fbdev *fbcon = info->par; - struct nouveau_drm *drm = nouveau_drm(fbcon->helper.dev); - struct nvif_device *device = &drm->client.device; - int ret; - - if (info->state != FBINFO_STATE_RUNNING) - return; - - ret = -ENODEV; - if (!in_interrupt() && !(info->flags & FBINFO_HWACCEL_DISABLED) && - mutex_trylock(&drm->client.mutex)) { - if (device->info.family < NV_DEVICE_INFO_V0_TESLA) - ret = nv04_fbcon_imageblit(info, image); - else - if (device->info.family < NV_DEVICE_INFO_V0_FERMI) - ret = nv50_fbcon_imageblit(info, image); - else - ret = nvc0_fbcon_imageblit(info, image); - mutex_unlock(&drm->client.mutex); - } - - if (ret == 0) - return; - - if (ret != -ENODEV) - nouveau_fbcon_gpu_lockup(info); - drm_fb_helper_cfb_imageblit(info, image); -} - -static int -nouveau_fbcon_sync(struct fb_info *info) -{ - struct nouveau_fbdev *fbcon = info->par; - struct nouveau_drm *drm = nouveau_drm(fbcon->helper.dev); - struct nouveau_channel *chan = drm->channel; - int ret; - - if (!chan || !chan->accel_done || in_interrupt() || - info->state != FBINFO_STATE_RUNNING || - info->flags & FBINFO_HWACCEL_DISABLED) - return 0; - - if (!mutex_trylock(&drm->client.mutex)) - return 0; - - ret = nouveau_channel_idle(chan); - mutex_unlock(&drm->client.mutex); - if (ret) { - nouveau_fbcon_gpu_lockup(info); - return 0; - } - - chan->accel_done = false; - return 0; -} - -static int -nouveau_fbcon_open(struct fb_info *info, int user) -{ - struct nouveau_fbdev *fbcon = info->par; - struct nouveau_drm *drm = nouveau_drm(fbcon->helper.dev); - int ret = pm_runtime_get_sync(drm->dev->dev); - if (ret < 0 && ret != -EACCES) { - pm_runtime_put(drm->dev->dev); - return ret; - } - return 0; -} - -static int -nouveau_fbcon_release(struct fb_info *info, int user) -{ - struct nouveau_fbdev *fbcon = info->par; - struct nouveau_drm *drm = nouveau_drm(fbcon->helper.dev); - pm_runtime_put(drm->dev->dev); - return 0; -} - -static const struct fb_ops nouveau_fbcon_ops = { - .owner = THIS_MODULE, - DRM_FB_HELPER_DEFAULT_OPS, - .fb_open = nouveau_fbcon_open, - .fb_release = nouveau_fbcon_release, - .fb_fillrect = nouveau_fbcon_fillrect, - .fb_copyarea = nouveau_fbcon_copyarea, - .fb_imageblit = nouveau_fbcon_imageblit, - .fb_sync = nouveau_fbcon_sync, -}; - -static const struct fb_ops nouveau_fbcon_sw_ops = { - .owner = THIS_MODULE, - DRM_FB_HELPER_DEFAULT_OPS, - .fb_open = nouveau_fbcon_open, - .fb_release = nouveau_fbcon_release, - .fb_fillrect = drm_fb_helper_cfb_fillrect, - .fb_copyarea = drm_fb_helper_cfb_copyarea, - .fb_imageblit = drm_fb_helper_cfb_imageblit, -}; - -void -nouveau_fbcon_accel_save_disable(struct drm_device *dev) -{ - struct nouveau_drm *drm = nouveau_drm(dev); - if (drm->fbcon && drm->fbcon->helper.info) { - drm->fbcon->saved_flags = drm->fbcon->helper.info->flags; - drm->fbcon->helper.info->flags |= FBINFO_HWACCEL_DISABLED; - } -} - -void -nouveau_fbcon_accel_restore(struct drm_device *dev) -{ - struct nouveau_drm *drm = nouveau_drm(dev); - if (drm->fbcon && drm->fbcon->helper.info) - drm->fbcon->helper.info->flags = drm->fbcon->saved_flags; -} - -static void -nouveau_fbcon_accel_fini(struct drm_device *dev) -{ - struct nouveau_drm *drm = nouveau_drm(dev); - struct nouveau_fbdev *fbcon = drm->fbcon; - if (fbcon && drm->channel) { - console_lock(); - if (fbcon->helper.info) - fbcon->helper.info->flags |= FBINFO_HWACCEL_DISABLED; - console_unlock(); - nouveau_channel_idle(drm->channel); - nvif_object_dtor(&fbcon->twod); - nvif_object_dtor(&fbcon->blit); - nvif_object_dtor(&fbcon->gdi); - nvif_object_dtor(&fbcon->patt); - nvif_object_dtor(&fbcon->rop); - nvif_object_dtor(&fbcon->clip); - nvif_object_dtor(&fbcon->surf2d); - } -} - -static void -nouveau_fbcon_accel_init(struct drm_device *dev) -{ - struct nouveau_drm *drm = nouveau_drm(dev); - struct nouveau_fbdev *fbcon = drm->fbcon; - struct fb_info *info = fbcon->helper.info; - int ret; - - if (drm->client.device.info.family < NV_DEVICE_INFO_V0_TESLA) - ret = nv04_fbcon_accel_init(info); - else - if (drm->client.device.info.family < NV_DEVICE_INFO_V0_FERMI) - ret = nv50_fbcon_accel_init(info); - else - ret = nvc0_fbcon_accel_init(info); - - if (ret == 0) - info->fbops = &nouveau_fbcon_ops; -} - -static void -nouveau_fbcon_zfill(struct drm_device *dev, struct nouveau_fbdev *fbcon) -{ - struct fb_info *info = fbcon->helper.info; - struct fb_fillrect rect; - - /* Clear the entire fbcon. The drm will program every connector - * with it's preferred mode. If the sizes differ, one display will - * quite likely have garbage around the console. - */ - rect.dx = rect.dy = 0; - rect.width = info->var.xres_virtual; - rect.height = info->var.yres_virtual; - rect.color = 0; - rect.rop = ROP_COPY; - info->fbops->fb_fillrect(info, &rect); -} - -static int -nouveau_fbcon_create(struct drm_fb_helper *helper, - struct drm_fb_helper_surface_size *sizes) -{ - struct nouveau_fbdev *fbcon = - container_of(helper, struct nouveau_fbdev, helper); - struct drm_device *dev = fbcon->helper.dev; - struct nouveau_drm *drm = nouveau_drm(dev); - struct nvif_device *device = &drm->client.device; - struct fb_info *info; - struct drm_framebuffer *fb; - struct nouveau_channel *chan; - struct nouveau_bo *nvbo; - struct drm_mode_fb_cmd2 mode_cmd = {}; - int ret; - - mode_cmd.width = sizes->surface_width; - mode_cmd.height = sizes->surface_height; - - mode_cmd.pitches[0] = mode_cmd.width * (sizes->surface_bpp >> 3); - mode_cmd.pitches[0] = roundup(mode_cmd.pitches[0], 256); - - mode_cmd.pixel_format = drm_mode_legacy_fb_format(sizes->surface_bpp, - sizes->surface_depth); - - ret = nouveau_gem_new(&drm->client, mode_cmd.pitches[0] * - mode_cmd.height, 0, NOUVEAU_GEM_DOMAIN_VRAM, - 0, 0x0000, &nvbo); - if (ret) { - NV_ERROR(drm, "failed to allocate framebuffer\n"); - goto out; - } - - ret = nouveau_framebuffer_new(dev, &mode_cmd, &nvbo->bo.base, &fb); - if (ret) - goto out_unref; - - ret = nouveau_bo_pin(nvbo, NOUVEAU_GEM_DOMAIN_VRAM, false); - if (ret) { - NV_ERROR(drm, "failed to pin fb: %d\n", ret); - goto out_unref; - } - - ret = nouveau_bo_map(nvbo); - if (ret) { - NV_ERROR(drm, "failed to map fb: %d\n", ret); - goto out_unpin; - } - - chan = nouveau_nofbaccel ? NULL : drm->channel; - if (chan && device->info.family >= NV_DEVICE_INFO_V0_TESLA) { - ret = nouveau_vma_new(nvbo, chan->vmm, &fbcon->vma); - if (ret) { - NV_ERROR(drm, "failed to map fb into chan: %d\n", ret); - chan = NULL; - } - } - - info = drm_fb_helper_alloc_info(helper); - if (IS_ERR(info)) { - ret = PTR_ERR(info); - goto out_unlock; - } - - /* setup helper */ - fbcon->helper.fb = fb; - - if (!chan) - info->flags = FBINFO_HWACCEL_DISABLED; - else - info->flags = FBINFO_HWACCEL_COPYAREA | - FBINFO_HWACCEL_FILLRECT | - FBINFO_HWACCEL_IMAGEBLIT; - info->fbops = &nouveau_fbcon_sw_ops; - info->fix.smem_start = nvbo->bo.resource->bus.offset; - info->fix.smem_len = nvbo->bo.base.size; - - info->screen_base = nvbo_kmap_obj_iovirtual(nvbo); - info->screen_size = nvbo->bo.base.size; - - drm_fb_helper_fill_info(info, &fbcon->helper, sizes); - - /* Use default scratch pixmap (info->pixmap.flags = FB_PIXMAP_SYSTEM) */ - - if (chan) - nouveau_fbcon_accel_init(dev); - nouveau_fbcon_zfill(dev, fbcon); - - /* To allow resizeing without swapping buffers */ - NV_INFO(drm, "allocated %dx%d fb: 0x%llx, bo %p\n", - fb->width, fb->height, nvbo->offset, nvbo); - - if (dev_is_pci(dev->dev)) - vga_switcheroo_client_fb_set(to_pci_dev(dev->dev), info); - - return 0; - -out_unlock: - if (chan) - nouveau_vma_del(&fbcon->vma); - nouveau_bo_unmap(nvbo); -out_unpin: - nouveau_bo_unpin(nvbo); -out_unref: - nouveau_bo_ref(NULL, &nvbo); -out: - return ret; -} - -static int -nouveau_fbcon_destroy(struct drm_device *dev, struct nouveau_fbdev *fbcon) -{ - struct drm_framebuffer *fb = fbcon->helper.fb; - struct nouveau_bo *nvbo; - - drm_fb_helper_unregister_info(&fbcon->helper); - drm_fb_helper_fini(&fbcon->helper); - - if (fb && fb->obj[0]) { - nvbo = nouveau_gem_object(fb->obj[0]); - nouveau_vma_del(&fbcon->vma); - nouveau_bo_unmap(nvbo); - nouveau_bo_unpin(nvbo); - drm_framebuffer_put(fb); - } - - return 0; -} - -void nouveau_fbcon_gpu_lockup(struct fb_info *info) -{ - struct nouveau_fbdev *fbcon = info->par; - struct nouveau_drm *drm = nouveau_drm(fbcon->helper.dev); - - NV_ERROR(drm, "GPU lockup - switching to software fbcon\n"); - info->flags |= FBINFO_HWACCEL_DISABLED; -} - -static const struct drm_fb_helper_funcs nouveau_fbcon_helper_funcs = { - .fb_probe = nouveau_fbcon_create, -}; - -static void -nouveau_fbcon_set_suspend_work(struct work_struct *work) -{ - struct nouveau_drm *drm = container_of(work, typeof(*drm), fbcon_work); - int state = READ_ONCE(drm->fbcon_new_state); - - if (state == FBINFO_STATE_RUNNING) - pm_runtime_get_sync(drm->dev->dev); - - console_lock(); - if (state == FBINFO_STATE_RUNNING) - nouveau_fbcon_accel_restore(drm->dev); - drm_fb_helper_set_suspend(&drm->fbcon->helper, state); - if (state != FBINFO_STATE_RUNNING) - nouveau_fbcon_accel_save_disable(drm->dev); - console_unlock(); - - if (state == FBINFO_STATE_RUNNING) { - nouveau_fbcon_hotplug_resume(drm->fbcon); - pm_runtime_mark_last_busy(drm->dev->dev); - pm_runtime_put_autosuspend(drm->dev->dev); - } -} - -void -nouveau_fbcon_set_suspend(struct drm_device *dev, int state) -{ - struct nouveau_drm *drm = nouveau_drm(dev); - - if (!drm->fbcon) - return; - - drm->fbcon_new_state = state; - /* Since runtime resume can happen as a result of a sysfs operation, - * it's possible we already have the console locked. So handle fbcon - * init/deinit from a seperate work thread - */ - schedule_work(&drm->fbcon_work); -} - -void -nouveau_fbcon_output_poll_changed(struct drm_device *dev) -{ - struct nouveau_drm *drm = nouveau_drm(dev); - struct nouveau_fbdev *fbcon = drm->fbcon; - int ret; - - if (!fbcon) - return; - - mutex_lock(&fbcon->hotplug_lock); - - ret = pm_runtime_get(dev->dev); - if (ret == 1 || ret == -EACCES) { - drm_fb_helper_hotplug_event(&fbcon->helper); - - pm_runtime_mark_last_busy(dev->dev); - pm_runtime_put_autosuspend(dev->dev); - } else if (ret == 0) { - /* If the GPU was already in the process of suspending before - * this event happened, then we can't block here as we'll - * deadlock the runtime pmops since they wait for us to - * finish. So, just defer this event for when we runtime - * resume again. It will be handled by fbcon_work. - */ - NV_DEBUG(drm, "fbcon HPD event deferred until runtime resume\n"); - fbcon->hotplug_waiting = true; - pm_runtime_put_noidle(drm->dev->dev); - } else { - DRM_WARN("fbcon HPD event lost due to RPM failure: %d\n", - ret); - } - - mutex_unlock(&fbcon->hotplug_lock); -} - -void -nouveau_fbcon_hotplug_resume(struct nouveau_fbdev *fbcon) -{ - struct nouveau_drm *drm; - - if (!fbcon) - return; - drm = nouveau_drm(fbcon->helper.dev); - - mutex_lock(&fbcon->hotplug_lock); - if (fbcon->hotplug_waiting) { - fbcon->hotplug_waiting = false; - - NV_DEBUG(drm, "Handling deferred fbcon HPD events\n"); - drm_fb_helper_hotplug_event(&fbcon->helper); - } - mutex_unlock(&fbcon->hotplug_lock); -} - -int -nouveau_fbcon_init(struct drm_device *dev) -{ - struct nouveau_drm *drm = nouveau_drm(dev); - struct nouveau_fbdev *fbcon; - int preferred_bpp = nouveau_fbcon_bpp; - int ret; - - if (!dev->mode_config.num_crtc || - (to_pci_dev(dev->dev)->class >> 8) != PCI_CLASS_DISPLAY_VGA) - return 0; - - fbcon = kzalloc(sizeof(struct nouveau_fbdev), GFP_KERNEL); - if (!fbcon) - return -ENOMEM; - - drm->fbcon = fbcon; - INIT_WORK(&drm->fbcon_work, nouveau_fbcon_set_suspend_work); - mutex_init(&fbcon->hotplug_lock); - - drm_fb_helper_prepare(dev, &fbcon->helper, &nouveau_fbcon_helper_funcs); - - ret = drm_fb_helper_init(dev, &fbcon->helper); - if (ret) - goto free; - - if (preferred_bpp != 8 && preferred_bpp != 16 && preferred_bpp != 32) { - if (drm->client.device.info.ram_size <= 32 * 1024 * 1024) - preferred_bpp = 8; - else - if (drm->client.device.info.ram_size <= 64 * 1024 * 1024) - preferred_bpp = 16; - else - preferred_bpp = 32; - } - - /* disable all the possible outputs/crtcs before entering KMS mode */ - if (!drm_drv_uses_atomic_modeset(dev)) - drm_helper_disable_unused_functions(dev); - - ret = drm_fb_helper_initial_config(&fbcon->helper, preferred_bpp); - if (ret) - goto fini; - - if (fbcon->helper.info) - fbcon->helper.info->pixmap.buf_align = 4; - return 0; - -fini: - drm_fb_helper_fini(&fbcon->helper); -free: - kfree(fbcon); - drm->fbcon = NULL; - return ret; -} - -void -nouveau_fbcon_fini(struct drm_device *dev) -{ - struct nouveau_drm *drm = nouveau_drm(dev); - - if (!drm->fbcon) - return; - - drm_kms_helper_poll_fini(dev); - nouveau_fbcon_accel_fini(dev); - nouveau_fbcon_destroy(dev, drm->fbcon); - kfree(drm->fbcon); - drm->fbcon = NULL; -} From 040b35c19bf2bdbb8ba5f8742b4e199ace3cbdc3 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 11 Jan 2023 12:50:15 -0500 Subject: [PATCH 8/9] drm/ttm: Fix a regression causing kernel oops'es MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The branch is explicitly taken if ttm == NULL which means that to avoid a null pointer reference the ttm object can not be used inside. Switch back to dst_mem to avoid kernel oops'es. This fixes kernel oops'es with any buffer objects which don't have ttm_tt, e.g. with vram based screen objects on vmwgfx. Signed-off-by: Zack Rusin Fixes: e3c92eb4a84f ("drm/ttm: rework on ttm_resource to use size_t type") Cc: Somalapuram Amaranath Cc: Christian König Link: https://patchwork.freedesktop.org/patch/msgid/20230111175015.1134923-1-zack@kde.org Signed-off-by: Christian König --- drivers/gpu/drm/ttm/ttm_bo_util.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c index ba3aa0a0fc43..da5493f789df 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_util.c +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c @@ -173,7 +173,7 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo, clear = src_iter->ops->maps_tt && (!ttm || !ttm_tt_is_populated(ttm)); if (!(clear && ttm && !(ttm->page_flags & TTM_TT_FLAG_ZERO_ALLOC))) - ttm_move_memcpy(clear, ttm->num_pages, dst_iter, src_iter); + ttm_move_memcpy(clear, PFN_UP(dst_mem->size), dst_iter, src_iter); if (!src_iter->ops->maps_tt) ttm_kmap_iter_linear_io_fini(&_src_iter.io, bdev, src_mem); From 5640e81607152d7f2d2558227c0f6cb78b8f39cf Mon Sep 17 00:00:00 2001 From: Arunpravin Paneer Selvam Date: Thu, 12 Jan 2023 04:00:27 -0800 Subject: [PATCH 9/9] drm: Optimize drm buddy top-down allocation method MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We are observing performance drop in many usecases which include games, 3D benchmark applications,etc.. To solve this problem, We are strictly not allowing top down flag enabled allocations to steal the memory space from cpu visible region. The idea is, we are sorting each order list entries in ascending order and compare the last entry of each order list in the freelist and return the max block. This patch improves the 3D benchmark scores and solves fragmentation issues. All drm buddy selftests are verfied. drm_buddy: pass:6 fail:0 skip:0 total:6 Signed-off-by: Arunpravin Paneer Selvam Acked-by: Christian König Acked-by: Alex Deucher Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20230112120027.3072-1-Arunpravin.PaneerSelvam@amd.com Signed-off-by: Christian König CC: Cc: stable@vger.kernel.org # 5.18+ --- drivers/gpu/drm/drm_buddy.c | 81 ++++++++++++++++++++++++------------- 1 file changed, 54 insertions(+), 27 deletions(-) diff --git a/drivers/gpu/drm/drm_buddy.c b/drivers/gpu/drm/drm_buddy.c index 11bb59399471..3d1f50f481cf 100644 --- a/drivers/gpu/drm/drm_buddy.c +++ b/drivers/gpu/drm/drm_buddy.c @@ -38,6 +38,25 @@ static void drm_block_free(struct drm_buddy *mm, kmem_cache_free(slab_blocks, block); } +static void list_insert_sorted(struct drm_buddy *mm, + struct drm_buddy_block *block) +{ + struct drm_buddy_block *node; + struct list_head *head; + + head = &mm->free_list[drm_buddy_block_order(block)]; + if (list_empty(head)) { + list_add(&block->link, head); + return; + } + + list_for_each_entry(node, head, link) + if (drm_buddy_block_offset(block) < drm_buddy_block_offset(node)) + break; + + __list_add(&block->link, node->link.prev, &node->link); +} + static void mark_allocated(struct drm_buddy_block *block) { block->header &= ~DRM_BUDDY_HEADER_STATE; @@ -52,8 +71,7 @@ static void mark_free(struct drm_buddy *mm, block->header &= ~DRM_BUDDY_HEADER_STATE; block->header |= DRM_BUDDY_FREE; - list_add(&block->link, - &mm->free_list[drm_buddy_block_order(block)]); + list_insert_sorted(mm, block); } static void mark_split(struct drm_buddy_block *block) @@ -387,20 +405,26 @@ err_undo: } static struct drm_buddy_block * -get_maxblock(struct list_head *head) +get_maxblock(struct drm_buddy *mm, unsigned int order) { struct drm_buddy_block *max_block = NULL, *node; + unsigned int i; - max_block = list_first_entry_or_null(head, - struct drm_buddy_block, - link); - if (!max_block) - return NULL; + for (i = order; i <= mm->max_order; ++i) { + if (!list_empty(&mm->free_list[i])) { + node = list_last_entry(&mm->free_list[i], + struct drm_buddy_block, + link); + if (!max_block) { + max_block = node; + continue; + } - list_for_each_entry(node, head, link) { - if (drm_buddy_block_offset(node) > - drm_buddy_block_offset(max_block)) - max_block = node; + if (drm_buddy_block_offset(node) > + drm_buddy_block_offset(max_block)) { + max_block = node; + } + } } return max_block; @@ -412,20 +436,23 @@ alloc_from_freelist(struct drm_buddy *mm, unsigned long flags) { struct drm_buddy_block *block = NULL; - unsigned int i; + unsigned int tmp; int err; - for (i = order; i <= mm->max_order; ++i) { - if (flags & DRM_BUDDY_TOPDOWN_ALLOCATION) { - block = get_maxblock(&mm->free_list[i]); - if (block) - break; - } else { - block = list_first_entry_or_null(&mm->free_list[i], - struct drm_buddy_block, - link); - if (block) - break; + if (flags & DRM_BUDDY_TOPDOWN_ALLOCATION) { + block = get_maxblock(mm, order); + if (block) + /* Store the obtained block order */ + tmp = drm_buddy_block_order(block); + } else { + for (tmp = order; tmp <= mm->max_order; ++tmp) { + if (!list_empty(&mm->free_list[tmp])) { + block = list_last_entry(&mm->free_list[tmp], + struct drm_buddy_block, + link); + if (block) + break; + } } } @@ -434,18 +461,18 @@ alloc_from_freelist(struct drm_buddy *mm, BUG_ON(!drm_buddy_block_is_free(block)); - while (i != order) { + while (tmp != order) { err = split_block(mm, block); if (unlikely(err)) goto err_undo; block = block->right; - i--; + tmp--; } return block; err_undo: - if (i != order) + if (tmp != order) __drm_buddy_free(mm, block); return ERR_PTR(err); }