[PATCH] drm/vboxvideo: Vmap/vunmap cursor BO in prepare_fb and cleanup_fb
Hans de Goede
hdegoede at redhat.com
Wed Feb 3 11:50:46 UTC 2021
Hi,
On 2/3/21 12:14 PM, Thomas Zimmermann wrote:
> Hi
>
> Am 03.02.21 um 11:44 schrieb Daniel Vetter:
>> On Wed, Feb 03, 2021 at 11:34:21AM +0100, Thomas Zimmermann wrote:
>>> Hi
>>>
>>> Am 03.02.21 um 11:29 schrieb Daniel Vetter:
>>>> On Wed, Jan 27, 2021 at 03:05:03PM +0100, Thomas Zimmermann wrote:
>>>>> Functions in the atomic commit tail are not allowed to acquire the
>>>>> dmabuf's reservation lock. So we cannot legally call the GEM object's
>>>>> vmap functionality in atomic_update.
>>>>>
>>>>> Instead vmap the cursor BO in prepare_fb and vunmap it in cleanup_fb.
>>>>> The cursor plane state stores the mapping's address. The pinning of the
>>>>> BO is implicitly done by vmap.
>>>>>
>>>>> As an extra benefit, there's no source of runtime errors left in
>>>>> atomic_update.
>>>>>
>>>>> Signed-off-by: Thomas Zimmermann <tzimmermann at suse.de>
>>>>
>>>> Did you test this with my dma_fence_signalling annotations patches?
>>>
>>> Not with vbox. I did test similar code in my recent ast patchset. But I
>>> think there's still a bug here, as there's no custom plane-reset function.
>>
>> Do right, KASAN should complain when you load the driver because the first
>> state is a bit too small. But probably still within the kmalloc'ed block
>> by sheer luck. Worth confirming that KASAN can catch this.
>
> I have KASAN enabled and I might just have missed the error message. I later saw the error with another driver after I already posted the vbox and ast patches.
>
> If you have the time, a look at the first half of the ast patchset might still be worth it. It removes the cursor-code abstraction and shouldn't be affected by the issue.
I must say I'm not entirely following this thread.
If I understand things correctly, there is some memory corruption
introduced by this patch, so there will be a v2 fixing this ?
The reason why I'm asking is because I always try to test vboxvideo patches
inside a vbox vm, but if a v2 is coming, there is not much use in me testing
this version, correct ?
Regards,
Hans
>>>>> ---
>>>>> drivers/gpu/drm/vboxvideo/vbox_mode.c | 102 +++++++++++++++++++++-----
>>>>> 1 file changed, 82 insertions(+), 20 deletions(-)
>>>>>
>>>>> diff --git a/drivers/gpu/drm/vboxvideo/vbox_mode.c b/drivers/gpu/drm/vboxvideo/vbox_mode.c
>>>>> index dbc0dd53c69e..b5625a7d6cef 100644
>>>>> --- a/drivers/gpu/drm/vboxvideo/vbox_mode.c
>>>>> +++ b/drivers/gpu/drm/vboxvideo/vbox_mode.c
>>>>> @@ -324,6 +324,19 @@ static void vbox_primary_atomic_disable(struct drm_plane *plane,
>>>>> old_state->src_y >> 16);
>>>>> }
>>>>> +struct vbox_cursor_plane_state {
>>>>> + struct drm_plane_state base;
>>>>> +
>>>>> + /* Transitional state - do not export or duplicate */
>>>>> +
>>>>> + struct dma_buf_map map;
>>>>> +};
>>>>> +
>>>>> +static struct vbox_cursor_plane_state *to_vbox_cursor_plane_state(struct drm_plane_state *state)
>>>>> +{
>>>>> + return container_of(state, struct vbox_cursor_plane_state, base);
>>>>> +}
>>>>> +
>>>>> static int vbox_cursor_atomic_check(struct drm_plane *plane,
>>>>> struct drm_plane_state *new_state)
>>>>> {
>>>>> @@ -381,14 +394,13 @@ static void vbox_cursor_atomic_update(struct drm_plane *plane,
>>>>> container_of(plane->dev, struct vbox_private, ddev);
>>>>> struct vbox_crtc *vbox_crtc = to_vbox_crtc(plane->state->crtc);
>>>>> struct drm_framebuffer *fb = plane->state->fb;
>>>>> - struct drm_gem_vram_object *gbo = drm_gem_vram_of_gem(fb->obj[0]);
>>>>> u32 width = plane->state->crtc_w;
>>>>> u32 height = plane->state->crtc_h;
>>>>> + struct vbox_cursor_plane_state *vbox_state = to_vbox_cursor_plane_state(plane->state);
>>>>> + struct dma_buf_map map = vbox_state->map;
>>>>> + u8 *src = map.vaddr; /* TODO: Use mapping abstraction properly */
>>>>> size_t data_size, mask_size;
>>>>> u32 flags;
>>>>> - struct dma_buf_map map;
>>>>> - int ret;
>>>>> - u8 *src;
>>>>> /*
>>>>> * VirtualBox uses the host windowing system to draw the cursor so
>>>>> @@ -401,17 +413,6 @@ static void vbox_cursor_atomic_update(struct drm_plane *plane,
>>>>> vbox_crtc->cursor_enabled = true;
>>>>> - ret = drm_gem_vram_vmap(gbo, &map);
>>>>> - if (ret) {
>>>>> - /*
>>>>> - * BUG: we should have pinned the BO in prepare_fb().
>>>>> - */
>>>>> - mutex_unlock(&vbox->hw_mutex);
>>>>> - DRM_WARN("Could not map cursor bo, skipping update\n");
>>>>> - return;
>>>>> - }
>>>>> - src = map.vaddr; /* TODO: Use mapping abstraction properly */
>>>>> -
>>>>> /*
>>>>> * The mask must be calculated based on the alpha
>>>>> * channel, one bit per ARGB word, and must be 32-bit
>>>>> @@ -421,7 +422,6 @@ static void vbox_cursor_atomic_update(struct drm_plane *plane,
>>>>> data_size = width * height * 4 + mask_size;
>>>>> copy_cursor_image(src, vbox->cursor_data, width, height, mask_size);
>>>>> - drm_gem_vram_vunmap(gbo, &map);
>>>>> flags = VBOX_MOUSE_POINTER_VISIBLE | VBOX_MOUSE_POINTER_SHAPE |
>>>>> VBOX_MOUSE_POINTER_ALPHA;
>>>>> @@ -458,6 +458,43 @@ static void vbox_cursor_atomic_disable(struct drm_plane *plane,
>>>>> mutex_unlock(&vbox->hw_mutex);
>>>>> }
>>>>> +static int vbox_cursor_prepare_fb(struct drm_plane *plane, struct drm_plane_state *new_state)
>>>>> +{
>>>>> + struct vbox_cursor_plane_state *new_vbox_state = to_vbox_cursor_plane_state(new_state);
>>>>> + struct drm_framebuffer *fb = new_state->fb;
>>>>> + struct drm_gem_vram_object *gbo;
>>>>> + struct dma_buf_map map;
>>>>> + int ret;
>>>>> +
>>>>> + if (!fb)
>>>>> + return 0;
>>>>> +
>>>>> + gbo = drm_gem_vram_of_gem(fb->obj[0]);
>>>>> +
>>>>> + ret = drm_gem_vram_vmap(gbo, &map);
>>>>> + if (ret)
>>>>> + return ret;
>>>>> +
>>>>> + new_vbox_state->map = map;
>>>>> +
>>>>> + return 0;
>>>>> +}
>>>>> +
>>>>> +static void vbox_cursor_cleanup_fb(struct drm_plane *plane, struct drm_plane_state *old_state)
>>>>> +{
>>>>> + struct vbox_cursor_plane_state *old_vbox_state = to_vbox_cursor_plane_state(old_state);
>>>>> + struct drm_framebuffer *fb = old_state->fb;
>>>>> + struct dma_buf_map map = old_vbox_state->map;
>>>>> + struct drm_gem_vram_object *gbo;
>>>>> +
>>>>> + if (!fb)
>>>>> + return;
>>>>> +
>>>>> + gbo = drm_gem_vram_of_gem(fb->obj[0]);
>>>>> +
>>>>> + drm_gem_vram_vunmap(gbo, &map);
>>>>> +}
>>>>> +
>>>>> static const u32 vbox_cursor_plane_formats[] = {
>>>>> DRM_FORMAT_ARGB8888,
>>>>> };
>>>>> @@ -466,17 +503,42 @@ static const struct drm_plane_helper_funcs vbox_cursor_helper_funcs = {
>>>>> .atomic_check = vbox_cursor_atomic_check,
>>>>> .atomic_update = vbox_cursor_atomic_update,
>>>>> .atomic_disable = vbox_cursor_atomic_disable,
>>>>> - .prepare_fb = drm_gem_vram_plane_helper_prepare_fb,
>>>>> - .cleanup_fb = drm_gem_vram_plane_helper_cleanup_fb,
>>>>> + .prepare_fb = vbox_cursor_prepare_fb,
>>>>> + .cleanup_fb = vbox_cursor_cleanup_fb,
>>>>> };
>>>>> +static struct drm_plane_state *vbox_cursor_atomic_duplicate_state(struct drm_plane *plane)
>>>>> +{
>>>>> + struct vbox_cursor_plane_state *new_vbox_state;
>>>>> + struct drm_device *dev = plane->dev;
>>>>> +
>>>>> + if (drm_WARN_ON(dev, !plane->state))
>>>>> + return NULL;
>>>>> +
>>>>> + new_vbox_state = kzalloc(sizeof(*new_vbox_state), GFP_KERNEL);
>>>>> + if (!new_vbox_state)
>>>>> + return NULL;
>>>>> + __drm_atomic_helper_plane_duplicate_state(plane, &new_vbox_state->base);
>>>>> +
>>>>> + return &new_vbox_state->base;
>>>>> +}
>>>>> +
>>>>> +static void vbox_cursor_atomic_destroy_state(struct drm_plane *plane,
>>>>> + struct drm_plane_state *state)
>>>>> +{
>>>>> + struct vbox_cursor_plane_state *vbox_state = to_vbox_cursor_plane_state(state);
>>>>> +
>>>>> + __drm_atomic_helper_plane_destroy_state(&vbox_state->base);
>>>>> + kfree(vbox_state);
>>>>> +}
>>>>> +
>>>>> static const struct drm_plane_funcs vbox_cursor_plane_funcs = {
>>>>> .update_plane = drm_atomic_helper_update_plane,
>>>>> .disable_plane = drm_atomic_helper_disable_plane,
>>>>> .destroy = drm_primary_helper_destroy,
>>>>> .reset = drm_atomic_helper_plane_reset,
>>>>> - .atomic_duplicate_state = drm_atomic_helper_plane_duplicate_state,
>>>>> - .atomic_destroy_state = drm_atomic_helper_plane_destroy_state,
>>>>> + .atomic_duplicate_state = vbox_cursor_atomic_duplicate_state,
>>>>> + .atomic_destroy_state = vbox_cursor_atomic_destroy_state,
>>>>> };
>>>>> static const u32 vbox_primary_plane_formats[] = {
>>>>>
>>>>> base-commit: 3836b7fdfad40e2bac5dc882332f42bed6942cf4
>>>>> prerequisite-patch-id: c2b2f08f0eccc9f5df0c0da49fa1d36267deb11d
>>>>> --
>>>>> 2.30.0
>>>>>
>>>>
>>>
>>> --
>>> Thomas Zimmermann
>>> Graphics Driver Developer
>>> SUSE Software Solutions Germany GmbH
>>> Maxfeldstr. 5, 90409 Nürnberg, Germany
>>> (HRB 36809, AG Nürnberg)
>>> Geschäftsführer: Felix Imendörffer
>>>
>>
>>
>>
>>
>
More information about the dri-devel
mailing list