[PATCH 01/12] drm/vmwgfx: fix guest_memory_dirty bitfield clobbered as size

[PATCH 01/12] drm/vmwgfx: fix guest_memory_dirty bitfield clobbered as size

[Zack Rusin]

Two sites in vmwgfx_resource.c assign boolean literals to
res->guest_memory_size, which is an unsigned long allocation-size
field; the intended target is the adjacent res->guest_memory_dirty
bitfield.  After the assignments the field holds 0 or 1 instead of
the resource's MOB allocation size:

  - vmw_resource_release()       writes 0 (false), and
  - vmw_resource_unbind_list()   writes 1 (true).

Subsequent revalidation paths read guest_memory_size when computing
the dirty page range (vmw_bo_dirty_transfer_to_res()) and the buffer
allocation size (vmw_resource_buf_alloc()), producing zero-length
walks or wrap-around ranges that read or write past the MOB bitmap.
The dirty-tracking intent of the original code (mark the resource as
dirtied since the last sync) is also lost, since guest_memory_dirty
is never updated.

Rename both assignments to guest_memory_dirty.

Fixes: 668b206601c5 ("drm/vmwgfx: Stop using raw ttm_buffer_object's")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_resource.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
index 388011696941..e3a187a2c7a1 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
@@ -136,7 +136,7 @@ static void vmw_resource_release(struct kref *kref)
 			val_buf.num_shared = 0;
 			res->func->unbind(res, false, &val_buf);
 		}
-		res->guest_memory_size = false;
+		res->guest_memory_dirty = false;
 		vmw_resource_mob_detach(res);
 		if (res->dirty)
 			res->func->dirty_free(res);
@@ -773,7 +773,7 @@ void vmw_resource_unbind_list(struct vmw_bo *vbo)
 		if (!WARN_ON_ONCE(!res->func->unbind))
 			(void) res->func->unbind(res, res->res_dirty, &val_buf);
 
-		res->guest_memory_size = true;
+		res->guest_memory_dirty = true;
 		res->res_dirty = false;
 		vmw_resource_mob_detach(res);
 	}
-- 
2.51.0

[PATCH 02/12] drm/vmwgfx: reject DX_BIND_QUERY without a DX context

[Zack Rusin]

vmw_cmd_dx_bind_query() unconditionally dereferences
sw_context->dx_ctx_node->ctx.  Userspace can trigger a NULL pointer
dereference from any render-node fd by submitting an execbuf with
dx_context_handle == SVGA3D_INVALID_ID and a SVGA_3D_CMD_DX_BIND_QUERY
opcode in the command stream: dx_ctx_node is left NULL and the kernel
oopses on the assignment.  The same NULL is then re-read in
vmw_resources_reserve() via vmw_context_get_dx_query_mob().

All sibling DX handlers fail-close on a missing dx_ctx_node using
VMW_GET_CTX_NODE().  Use the same pattern here, returning -EINVAL up
front before any relocation state is published.

Fixes: 9c079b8ce8bf ("drm/vmwgfx: Adapt execbuf to the new validation api")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
index e1f18020170a..b07f052474d0 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
@@ -1272,9 +1272,13 @@ static int vmw_cmd_dx_bind_query(struct vmw_private *dev_priv,
 				 SVGA3dCmdHeader *header)
 {
 	VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdDXBindQuery);
+	struct vmw_ctx_validation_info *ctx_node = VMW_GET_CTX_NODE(sw_context);
 	struct vmw_bo *vmw_bo;
 	int ret;
 
+	if (!ctx_node)
+		return -EINVAL;
+
 	cmd = container_of(header, typeof(*cmd), header);
 
 	/*
@@ -1288,7 +1292,7 @@ static int vmw_cmd_dx_bind_query(struct vmw_private *dev_priv,
 		return ret;
 
 	sw_context->dx_query_mob = vmw_bo;
-	sw_context->dx_query_ctx = sw_context->dx_ctx_node->ctx;
+	sw_context->dx_query_ctx = ctx_node->ctx;
 	return 0;
 }
 
-- 
2.51.0

[PATCH 03/12] drm/vmwgfx: clamp dirty-page range with min, not max

[Zack Rusin]

vmw_bo_dirty_transfer_to_res() and vmw_bo_dirty_clear() compute the
intersection of a resource's page range with the BO's tracked dirty
range, but clamp res_end against dirty->end with max() instead of
min().  When dirty->end exceeds the resource end, the loop walks past
the resource's pages, calls vmw_resource_dirty_update() for ranges
owned by other resources sharing the same backing MOB and clears
their pending dirty bits via bitmap_clear().  The result is silent
loss of writeback for unrelated resources whenever two resources
share a MOB.

Use min() in both functions so the loop is bounded to the
intersection of the resource and dirty ranges.

Fixes: b7468b15d271 ("drm/vmwgfx: Implement an infrastructure for write-coherent resources")
Fixes: 965544150d1c ("drm/vmwgfx: Refactor cursor handling")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c b/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c
index 45561bc1c9ef..8ab88f388652 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_page_dirty.c
@@ -311,7 +311,7 @@ void vmw_bo_dirty_transfer_to_res(struct vmw_resource *res)
 		return;
 
 	cur = max(res_start, dirty->start);
-	res_end = max(res_end, dirty->end);
+	res_end = min(res_end, dirty->end);
 	while (cur < res_end) {
 		unsigned long num;
 
@@ -347,7 +347,7 @@ void vmw_bo_dirty_clear(struct vmw_bo *vbo)
 		return;
 
 	cur = max(res_start, dirty->start);
-	res_end = max(res_end, dirty->end);
+	res_end = min(res_end, dirty->end);
 	while (cur < res_end) {
 		unsigned long num;
 
-- 
2.51.0

[PATCH 04/12] drm/vmwgfx: take fman->lock around fence list mutation in fifo_down

[Zack Rusin]

vmw_fence_fifo_down() drops fman->lock to wait on a fence and, on
timeout, mutates fman->fence_list via list_del_init() and signals
the fence without re-acquiring the lock.  __vmw_fences_update() walks
and removes entries from the same list under fman->lock from any
other waiter, the fence-IRQ thread, or vmw_fences_update(), so the
unlocked list_del_init() can corrupt the list head.

Re-take fman->lock before manipulating fence->head and use
dma_fence_signal_locked().  Wrap the locked signalling in
dma_fence_begin_signalling() / dma_fence_end_signalling() so the
lockdep annotation that dma_fence_signal() previously provided is
preserved (the same pattern as __vmw_fences_update()).

dma_fence_put() is moved outside the lock to avoid a recursive
acquire from vmw_fence_obj_destroy(), which also takes fman->lock.

Fixes: ae2a104058e2 ("vmwgfx: Implement fence objects")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_fence.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c
index 4ef84ff9b638..384c6736cf6b 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c
@@ -367,13 +367,24 @@ void vmw_fence_fifo_down(struct vmw_fence_manager *fman)
 		ret = vmw_fence_obj_wait(fence, false, false,
 					 VMW_FENCE_WAIT_TIMEOUT);
 
+		spin_lock(&fman->lock);
 		if (unlikely(ret != 0)) {
+			bool cookie = dma_fence_begin_signalling();
+
 			list_del_init(&fence->head);
-			dma_fence_signal(&fence->base);
+			if (fence->waiter_added) {
+				vmw_seqno_waiter_remove(fman->dev_priv);
+				fence->waiter_added = false;
+			}
+			dma_fence_signal_locked(&fence->base);
+			dma_fence_end_signalling(cookie);
 		}
 
 		BUG_ON(!list_empty(&fence->head));
+		spin_unlock(&fman->lock);
+
 		dma_fence_put(&fence->base);
+
 		spin_lock(&fman->lock);
 	}
 	spin_unlock(&fman->lock);
-- 
2.51.0

Re: [PATCH 04/12] drm/vmwgfx: take fman->lock around fence list mutation in fifo_down

[Matthew Brost]

On Tue, May 05, 2026 at 06:22:25PM -0400, Zack Rusin wrote:
> vmw_fence_fifo_down() drops fman->lock to wait on a fence and, on
> timeout, mutates fman->fence_list via list_del_init() and signals
> the fence without re-acquiring the lock.  __vmw_fences_update() walks
> and removes entries from the same list under fman->lock from any
> other waiter, the fence-IRQ thread, or vmw_fences_update(), so the
> unlocked list_del_init() can corrupt the list head.
> 
> Re-take fman->lock before manipulating fence->head and use
> dma_fence_signal_locked().  Wrap the locked signalling in
> dma_fence_begin_signalling() / dma_fence_end_signalling() so the
> lockdep annotation that dma_fence_signal() previously provided is
> preserved (the same pattern as __vmw_fences_update()).
> 
> dma_fence_put() is moved outside the lock to avoid a recursive
> acquire from vmw_fence_obj_destroy(), which also takes fman->lock.
> 

Just looking as someone who is curious about AI - not my driver but this
almost certainly looks like a good fix from quick look at vmwgfx code.

> Fixes: ae2a104058e2 ("vmwgfx: Implement fence objects")
> Cc: stable@vger.kernel.org
> Assisted-by: Claude:claude-opus-4.7
> Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
> ---
>  drivers/gpu/drm/vmwgfx/vmwgfx_fence.c | 13 ++++++++++++-
>  1 file changed, 12 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c
> index 4ef84ff9b638..384c6736cf6b 100644
> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c
> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c
> @@ -367,13 +367,24 @@ void vmw_fence_fifo_down(struct vmw_fence_manager *fman)
>  		ret = vmw_fence_obj_wait(fence, false, false,
>  					 VMW_FENCE_WAIT_TIMEOUT);
>  
> +		spin_lock(&fman->lock);
>  		if (unlikely(ret != 0)) {
> +			bool cookie = dma_fence_begin_signalling();
> +
>  			list_del_init(&fence->head);
> -			dma_fence_signal(&fence->base);
> +			if (fence->waiter_added) {
> +				vmw_seqno_waiter_remove(fman->dev_priv);
> +				fence->waiter_added = false;
> +			}
> +			dma_fence_signal_locked(&fence->base);
> +			dma_fence_end_signalling(cookie);
>  		}
>  
>  		BUG_ON(!list_empty(&fence->head));
> +		spin_unlock(&fman->lock);
> +

You likely can drop spin_unlock/spin_lock dance around the put here as a
put is just ref count move or vmw_fence_obj_destroy on final which seems
to resolve to kfree_rcu in any case. ofc, if you want to be parnoid it
is perfectly fine to drop the reacquire the lock.

Matt

>  		dma_fence_put(&fence->base);
> +
>  		spin_lock(&fman->lock);
>  	}
>  	spin_unlock(&fman->lock);
> -- 
> 2.51.0
> 

[PATCH 05/12] drm/vmwgfx: drop dma_buf reference on foreign-fd prime import

[Zack Rusin]

ttm_prime_fd_to_handle() returns -ENOSYS when the imported fd's
dma_buf->ops do not match the ttm_object_device's ops, but does so
without releasing the reference acquired by dma_buf_get().  Any
unprivileged renderD client passing a non-vmwgfx prime fd through the
DRM_VMW_GB_SURFACE_REF{,_EXT} path leaks one dma_buf reference per
call and indefinitely pins the foreign exporter's GEM resources.

Funnel the error path through the existing dma_buf_put() so the
reference is always dropped.

Fixes: 65981f7681ab ("drm/ttm: Add a minimal prime implementation for ttm base objects")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/ttm_object.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/vmwgfx/ttm_object.c b/drivers/gpu/drm/vmwgfx/ttm_object.c
index 2421b0dd057c..f9042bafdc93 100644
--- a/drivers/gpu/drm/vmwgfx/ttm_object.c
+++ b/drivers/gpu/drm/vmwgfx/ttm_object.c
@@ -547,14 +547,17 @@ int ttm_prime_fd_to_handle(struct ttm_object_file *tfile,
 	if (IS_ERR(dma_buf))
 		return PTR_ERR(dma_buf);
 
-	if (dma_buf->ops != &tdev->ops)
-		return -ENOSYS;
+	if (dma_buf->ops != &tdev->ops) {
+		ret = -ENOSYS;
+		goto out;
+	}
 
 	prime = (struct ttm_prime_object *) dma_buf->priv;
 	base = &prime->base;
 	*handle = base->handle;
 	ret = ttm_ref_object_add(tfile, base, NULL, false);
 
+out:
 	dma_buf_put(dma_buf);
 
 	return ret;
-- 
2.51.0

[PATCH 06/12] drm/vmwgfx: validate DRAW_PRIMITIVES header size before division

[Zack Rusin]

vmw_cmd_draw() computes

	maxnum = (header->size - sizeof(cmd->body)) / sizeof(*decl);

where header->size is u32 and is taken straight from the user-supplied
command stream.  When header->size is less than sizeof(cmd->body) the
unsigned subtraction wraps to nearly 4 GiB, producing a huge maxnum.
Any user-controlled cmd->body.numVertexDecls then passes the bound and
the loop dereferences decl[i] far past the end of the kernel command
bounce buffer, producing an out-of-bounds read of kernel memory.

Reject undersized headers up front.

Fixes: 7a73ba7469cb ("drm/vmwgfx: Use TTM handles instead of SIDs as user-space surface handles.")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
index b07f052474d0..2410d53a75aa 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
@@ -1571,11 +1571,17 @@ static int vmw_cmd_draw(struct vmw_private *dev_priv,
 	uint32_t maxnum;
 	int ret;
 
+	cmd = container_of(header, typeof(*cmd), header);
+
+	if (unlikely(header->size < sizeof(cmd->body))) {
+		VMW_DEBUG_USER("Illegal DRAW_PRIMITIVES header size.\n");
+		return -EINVAL;
+	}
+
 	ret = vmw_cmd_cid_check(dev_priv, sw_context, header);
 	if (unlikely(ret != 0))
 		return ret;
 
-	cmd = container_of(header, typeof(*cmd), header);
 	maxnum = (header->size - sizeof(cmd->body)) / sizeof(*decl);
 
 	if (unlikely(cmd->body.numVertexDecls > maxnum)) {
-- 
2.51.0

[PATCH 07/12] drm/vmwgfx: bound DMA command body size against suffix pointer

[Zack Rusin]

vmw_cmd_dma() locates the DMA suffix at

	(unsigned long) &cmd->body + header->size - sizeof(*suffix)

without checking that header->size is large enough to contain both
cmd->body and the suffix.  An undersized header makes the suffix
pointer underflow back into the previous command in the bounce
buffer.  The verifier later writes suffix->maximumOffset, clobbering
verified fields of an already-relocated earlier command -- a TOCTOU
on the device-visible command stream that lets one command rewrite
another's GMR id, surface id, or other authenticated fields.

Reject the command if the body is too small for the suffix to fit.

Fixes: 4e4ddd477743 ("drm/vmwgfx: Fix queries if no dma buffer thrashing is occuring.")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
index 2410d53a75aa..a9136a6523cb 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
@@ -1510,6 +1510,12 @@ static int vmw_cmd_dma(struct vmw_private *dev_priv,
 	bool dirty;
 
 	cmd = container_of(header, typeof(*cmd), header);
+
+	if (unlikely(header->size < sizeof(cmd->body) + sizeof(*suffix))) {
+		VMW_DEBUG_USER("Illegal SVGA_3D_CMD_SURFACE_DMA size.\n");
+		return -EINVAL;
+	}
+
 	suffix = (SVGA3dCmdSurfaceDMASuffix *)((unsigned long) &cmd->body +
 					       header->size - sizeof(*suffix));
 
-- 
2.51.0

[PATCH 08/12] drm/vmwgfx: avoid destroy_workqueue(NULL) on vkms init failure

[Zack Rusin]

Two paths through vmw_vkms_init() can leave vmw->crc_workq NULL while
still leaving the rest of the driver in a state that calls
vmw_vkms_cleanup() at module unload:

  1. vmw_host_get_guestinfo(GUESTINFO_VBLANK, ...) failing or
     returning an oversized buffer -- the common case on hosts
     without a VBLANK guestinfo entry -- early-returned before the
     workqueue allocation.
  2. alloc_ordered_workqueue() returning NULL on memory pressure.

vmw_vkms_cleanup() then calls destroy_workqueue(NULL), which
dereferences wq->name and panics.

Fix the first case by removing the early return: vmw->vkms_enabled
is already false on the rpci-failure path so no work will ever be
queued, and allocating the workqueue unconditionally keeps the
control flow simple.  Fix the second case by guarding the cleanup
with a NULL check, since alloc_ordered_workqueue() can still fail
under low memory.

Fixes: 7b0062036c3b ("drm/vmwgfx: Implement virtual crc generation")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_vkms.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_vkms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_vkms.c
index 5abd7f5ad2db..0d499917682d 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_vkms.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_vkms.c
@@ -214,14 +214,14 @@ vmw_vkms_init(struct vmw_private *vmw)
 	vmw->vkms_enabled = false;
 
 	ret = vmw_host_get_guestinfo(GUESTINFO_VBLANK, buffer, &buf_len);
-	if (ret || buf_len > max_buf_len)
-		return;
-	buffer[buf_len] = '\0';
+	if (!ret && buf_len <= max_buf_len) {
+		buffer[buf_len] = '\0';
 
-	ret = kstrtobool(buffer, &vmw->vkms_enabled);
-	if (!ret && vmw->vkms_enabled) {
-		ret = drm_vblank_init(&vmw->drm, VMWGFX_NUM_DISPLAY_UNITS);
-		vmw->vkms_enabled = (ret == 0);
+		ret = kstrtobool(buffer, &vmw->vkms_enabled);
+		if (!ret && vmw->vkms_enabled) {
+			ret = drm_vblank_init(&vmw->drm, VMWGFX_NUM_DISPLAY_UNITS);
+			vmw->vkms_enabled = (ret == 0);
+		}
 	}
 
 	vmw->crc_workq = alloc_ordered_workqueue("vmwgfx_crc_generator", 0);
@@ -236,7 +236,8 @@ vmw_vkms_init(struct vmw_private *vmw)
 void
 vmw_vkms_cleanup(struct vmw_private *vmw)
 {
-	destroy_workqueue(vmw->crc_workq);
+	if (vmw->crc_workq)
+		destroy_workqueue(vmw->crc_workq);
 }
 
 bool
-- 
2.51.0

[PATCH 09/12] drm/vmwgfx: enforce cursor size limits for MOB cursors

[Zack Rusin]

vmw_cursor_plane_atomic_check() bounds cursor width and height only
on the legacy update path; the SVGA_CAP2_CURSOR_MOB path -- the
default on modern hosts -- accepts any size.  When the requested size
exceeds SVGA_REG_CURSOR_MAX_DIMENSION or SVGA_REG_MOB_MAX_SIZE,
vmw_cursor_mob_get() returns -EINVAL and leaves vps->cursor.mob NULL.
Its return value is then discarded in vmw_cursor_plane_prepare_fb(),
so the subsequent vmw_cursor_update_mob() calls
vmw_bo_map_and_cache(NULL) and oopses inside
vmw_bo_map_and_cache_size() on the tbo.base.size load.

Reachable from any DRM master via DRM_IOCTL_MODE_CURSOR2 with a
sufficiently large width or height (e.g. cursor_max_dim + 1).

Reject oversized cursors in atomic_check for both MOB-backed cursor
update types.  The MOB byte-size limit only applies to the
SVGA_CAP2_CURSOR_MOB path (vmw_cursor_mob_size() returns 0 for
GB_ONLY); compute the required MOB size in 64-bit to avoid overflow
when very large dimensions are requested.

In prepare_fb only call vmw_cursor_mob_get()/_map() for
VMW_CURSOR_UPDATE_MOB -- the GB_ONLY path uses bo->map.virtual
directly and would otherwise be silently downgraded to NONE on hosts
without SVGA_CAP2_CURSOR_MOB (where vmw_cursor_mob_get() always
returns -EINVAL).  Degrade the update to NONE if vmw_cursor_mob_get()
or vmw_cursor_mob_map() fails so the update path does not run with a
NULL backing MOB.

Fixes: 965544150d1c ("drm/vmwgfx: Refactor cursor handling")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_cursor_plane.c | 49 ++++++++++++++++++--
 1 file changed, 44 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_cursor_plane.c b/drivers/gpu/drm/vmwgfx/vmwgfx_cursor_plane.c
index c46f17ba7236..c53bb9376b36 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_cursor_plane.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_cursor_plane.c
@@ -432,6 +432,7 @@ vmw_cursor_mob_map(struct vmw_plane_state *vps)
 	u32 size = vmw_cursor_mob_size(vps->cursor.update_type,
 				       vps->base.crtc_w, vps->base.crtc_h);
 	struct vmw_bo *vbo = vps->cursor.mob;
+	void *map;
 
 	if (!vbo)
 		return -EINVAL;
@@ -446,11 +447,15 @@ vmw_cursor_mob_map(struct vmw_plane_state *vps)
 	if (unlikely(ret != 0))
 		return -ENOMEM;
 
-	vmw_bo_map_and_cache(vbo);
+	map = vmw_bo_map_and_cache(vbo);
+	if (!map) {
+		vmw_bo_unmap(vbo);
+		ret = -ENOMEM;
+	}
 
 	ttm_bo_unreserve(&vbo->tbo);
 
-	return 0;
+	return ret;
 }
 
 /**
@@ -663,9 +668,15 @@ int vmw_cursor_plane_prepare_fb(struct drm_plane *plane,
 			    !vmw_cursor_buffer_changed(vps, old_vps)) {
 				vps->cursor.update_type =
 					VMW_CURSOR_UPDATE_NONE;
-			} else {
-				vmw_cursor_mob_get(vcp, vps);
-				vmw_cursor_mob_map(vps);
+			} else if (vps->cursor.update_type ==
+				   VMW_CURSOR_UPDATE_MOB &&
+				   (vmw_cursor_mob_get(vcp, vps) ||
+				    vmw_cursor_mob_map(vps))) {
+				/*
+				 * Reset the cursor to avoid crashes later.
+				 */
+				vps->cursor.update_type =
+					VMW_CURSOR_UPDATE_NONE;
 			}
 		}
 	}
@@ -732,6 +743,34 @@ int vmw_cursor_plane_atomic_check(struct drm_plane *plane,
 				 "surface not suitable for cursor\n");
 			return -EINVAL;
 		}
+	} else if (update_type == VMW_CURSOR_UPDATE_GB_ONLY ||
+		   update_type == VMW_CURSOR_UPDATE_MOB) {
+		u32 cursor_max_dim =
+			vmw_read(vmw, SVGA_REG_CURSOR_MAX_DIMENSION);
+
+		if (new_state->crtc_w > cursor_max_dim ||
+		    new_state->crtc_h > cursor_max_dim) {
+			drm_warn(&vmw->drm,
+				 "Cursor dimensions (%d, %d) exceed device max %u\n",
+				 new_state->crtc_w, new_state->crtc_h,
+				 cursor_max_dim);
+			return -EINVAL;
+		}
+
+		if (update_type == VMW_CURSOR_UPDATE_MOB) {
+			u32 mob_max_size =
+				vmw_read(vmw, SVGA_REG_MOB_MAX_SIZE);
+			u64 mob_size = (u64)new_state->crtc_w *
+				       new_state->crtc_h * sizeof(u32) +
+				       sizeof(SVGAGBCursorHeader);
+
+			if (mob_size > mob_max_size) {
+				drm_warn(&vmw->drm,
+					 "Cursor MOB size %llu exceeds device max %u\n",
+					 mob_size, mob_max_size);
+				return -EINVAL;
+			}
+		}
 	}
 
 	return 0;
-- 
2.51.0

[PATCH 10/12] drm/vmwgfx: skip hash_del_rcu when validation context has no hash table

[Zack Rusin]

vmw_validation_add_resource() conditionally calls hash_add_rcu() only
when ctx->sw_context is non-NULL, but the doomed-resource error path
calls hash_del_rcu() unconditionally.

The KMS validation contexts created with DECLARE_VAL_CONTEXT(_, NULL,
0) in vmwgfx_kms.c, vmwgfx_scrn.c, and vmwgfx_stdu.c never add the
node to a hash chain, so the resulting hlist_del_rcu() writes through
node->hash.head.pprev which is freshly allocated and uninitialized,
corrupting whatever happens to lie at that address.

Mirror the conditional from the add side in the cleanup path so the
node is only unlinked from the hash table when it was actually added.

Fixes: dfe1323ab3c8 ("drm/vmwgfx: Fix Use-after-free in validation")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_validation.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c b/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c
index 35dc94c3db39..45fde7ec514f 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c
@@ -309,7 +309,8 @@ int vmw_validation_add_resource(struct vmw_validation_context *ctx,
 	}
 	node->res = vmw_resource_reference_unless_doomed(res);
 	if (!node->res) {
-		hash_del_rcu(&node->hash.head);
+		if (ctx->sw_context)
+			hash_del_rcu(&node->hash.head);
 		return -ESRCH;
 	}
 
-- 
2.51.0

[PATCH 11/12] drm/vmwgfx: use check_add_overflow for shader size+offset bound

[Zack Rusin]

vmw_shader_define() validates the user-supplied shader window against
its backing buffer with

	(u64)buffer->tbo.base.size < (u64)size + (u64)offset

drm_vmw_shader_create_arg::offset is __u64 in the uapi; when it is
near U64_MAX the unsigned addition wraps and the resulting tiny value
passes the check.  The unbounded offset is then stored in
res->guest_memory_offset and forwarded to host SVGA shader-create
commands.

Use check_add_overflow() to detect the wrap and compare the resulting
endpoint against the buffer size.

Fixes: 668b206601c5 ("drm/vmwgfx: Stop using raw ttm_buffer_object's")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_shader.c | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c b/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
index eca4e3e97eb4..39811cf19db1 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
@@ -25,6 +25,8 @@
  *
  **************************************************************************/
 
+#include <linux/overflow.h>
+
 #include <drm/ttm/ttm_placement.h>
 
 #include "vmwgfx_binding.h"
@@ -685,7 +687,7 @@ int vmw_shader_destroy_ioctl(struct drm_device *dev, void *data,
 static int vmw_user_shader_alloc(struct vmw_private *dev_priv,
 				 struct vmw_bo *buffer,
 				 size_t shader_size,
-				 size_t offset,
+				 u64 offset,
 				 SVGA3dShaderType shader_type,
 				 uint8_t num_input_sig,
 				 uint8_t num_output_sig,
@@ -739,7 +741,7 @@ static int vmw_user_shader_alloc(struct vmw_private *dev_priv,
 static struct vmw_resource *vmw_shader_alloc(struct vmw_private *dev_priv,
 					     struct vmw_bo *buffer,
 					     size_t shader_size,
-					     size_t offset,
+					     u64 offset,
 					     SVGA3dShaderType shader_type)
 {
 	struct vmw_shader *shader;
@@ -768,7 +770,7 @@ static struct vmw_resource *vmw_shader_alloc(struct vmw_private *dev_priv,
 
 static int vmw_shader_define(struct drm_device *dev, struct drm_file *file_priv,
 			     enum drm_vmw_shader_type shader_type_drm,
-			     u32 buffer_handle, size_t size, size_t offset,
+			     u32 buffer_handle, size_t size, u64 offset,
 			     uint8_t num_input_sig, uint8_t num_output_sig,
 			     uint32_t *shader_handle)
 {
@@ -779,13 +781,16 @@ static int vmw_shader_define(struct drm_device *dev, struct drm_file *file_priv,
 	int ret;
 
 	if (buffer_handle != SVGA3D_INVALID_ID) {
+		u64 end;
+
 		ret = vmw_user_bo_lookup(file_priv, buffer_handle, &buffer);
 		if (unlikely(ret != 0)) {
 			VMW_DEBUG_USER("Couldn't find buffer for shader creation.\n");
 			return ret;
 		}
 
-		if ((u64)buffer->tbo.base.size < (u64)size + (u64)offset) {
+		if (check_add_overflow((u64)size, (u64)offset, &end) ||
+		    end > buffer->tbo.base.size) {
 			VMW_DEBUG_USER("Illegal buffer- or shader size.\n");
 			ret = -EINVAL;
 			goto out_bad_arg;
-- 
2.51.0

[PATCH 12/12] drm/vmwgfx: validate external BO copy bounds for both stride paths

[Zack Rusin]

vmw_external_bo_copy() trusts caller-supplied offsets, strides, and
heights and operates on imported dma-buf vmaps:

  - The equal-stride memcpy() bound was clamped after subtracting the
    offsets from dst_size and src_size; an offset larger than the BO
    size wraps the unsigned subtraction to a huge value and the
    resulting memcpy() runs off the end of the vmap.  dst_stride *
    height is also a u32 multiplication that can overflow.
  - The non-equal-stride row-by-row path had no bound at all.  The
    loop touches bytes through offset + (height - 1) * stride +
    width_in_bytes, with only a WARN_ON(dst_stride < width_in_bytes),
    and could likewise step past the end of either mapping.

The offsets and strides are derived from STDU/SOU plane state, so a
configured CRTC submitting a crafted atomic commit on an imported
framebuffer can reach this path.

Validate the exact row-copy endpoint against each BO's size up front
using check_mul_overflow() and check_add_overflow().  Use the bulk
memcpy() path only when width_in_bytes covers the whole stride;
otherwise copy one row at a time so partial-row updates near the bottom
of a framebuffer remain valid.  Also reject zero strides and stride <
width_in_bytes, both of which the row-by-row path cannot represent
safely.

Fixes: 50f119925091 ("drm/vmwgfx: Fix prime with external buffers")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4.7
Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_blit.c | 39 ++++++++++++++++++++++------
 1 file changed, 31 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_blit.c b/drivers/gpu/drm/vmwgfx/vmwgfx_blit.c
index 135b75a3e013..56f965ec99dc 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_blit.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_blit.c
@@ -30,6 +30,7 @@
 
 #include "vmwgfx_bo.h"
 #include <linux/highmem.h>
+#include <linux/overflow.h>
 
 /*
  * Template that implements find_first_diff() for a generic
@@ -463,19 +464,42 @@ static int vmw_external_bo_copy(struct vmw_bo *dst, u32 dst_offset,
 		container_of(dst->tbo.bdev, struct vmw_private, bdev);
 	size_t dst_size = dst->tbo.resource->size;
 	size_t src_size = src->tbo.resource->size;
+	size_t dst_end, src_end;
 	struct iosys_map dst_map = {0};
 	struct iosys_map src_map = {0};
+	bool dst_mapped = false;
+	bool src_mapped = false;
 	int ret, i;
 	int x_in_bytes;
 	u8 *vsrc;
 	u8 *vdst;
 
+	if (!height || !width_in_bytes)
+		return 0;
+
+	if (!dst_stride || !src_stride)
+		return -EINVAL;
+	if (dst_stride < width_in_bytes || src_stride < width_in_bytes)
+		return -EINVAL;
+	if (check_mul_overflow((size_t)dst_stride, (size_t)height - 1, &dst_end) ||
+	    check_add_overflow(dst_end, (size_t)width_in_bytes, &dst_end) ||
+	    check_add_overflow((size_t)dst_offset, dst_end, &dst_end) ||
+	    dst_end > dst_size ||
+	    check_mul_overflow((size_t)src_stride, (size_t)height - 1, &src_end) ||
+	    check_add_overflow(src_end, (size_t)width_in_bytes, &src_end) ||
+	    check_add_overflow((size_t)src_offset, src_end, &src_end) ||
+	    src_end > src_size) {
+		drm_dbg_driver(&vmw->drm, "Out-of-bounds external BO copy\n");
+		return -EINVAL;
+	}
+
 	vsrc = map_external(src, &src_map);
 	if (!vsrc) {
 		drm_dbg_driver(&vmw->drm, "Wasn't able to map src\n");
 		ret = -ENOMEM;
 		goto out;
 	}
+	src_mapped = true;
 
 	vdst = map_external(dst, &dst_map);
 	if (!vdst) {
@@ -483,16 +507,13 @@ static int vmw_external_bo_copy(struct vmw_bo *dst, u32 dst_offset,
 		ret = -ENOMEM;
 		goto out;
 	}
+	dst_mapped = true;
 
 	vsrc += src_offset;
 	vdst += dst_offset;
-	if (src_stride == dst_stride) {
-		dst_size -= dst_offset;
-		src_size -= src_offset;
-		memcpy(vdst, vsrc,
-		       min(dst_stride * height, min(dst_size, src_size)));
+	if (src_stride == dst_stride && width_in_bytes == dst_stride) {
+		memcpy(vdst, vsrc, dst_stride * (size_t)height);
 	} else {
-		WARN_ON(dst_stride < width_in_bytes);
 		for (i = 0; i < height; ++i) {
 			memcpy(vdst, vsrc, width_in_bytes);
 			vsrc += src_stride;
@@ -508,8 +529,10 @@ static int vmw_external_bo_copy(struct vmw_bo *dst, u32 dst_offset,
 
 	ret = 0;
 out:
-	unmap_external(src, &src_map);
-	unmap_external(dst, &dst_map);
+	if (src_mapped)
+		unmap_external(src, &src_map);
+	if (dst_mapped)
+		unmap_external(dst, &dst_map);
 
 	return ret;
 }
-- 
2.51.0