Skip to contentSkip to conversation

virtio-gpu: do not replace surface when scanout is disabled

PATCH In review Dongwon Kim 2 participants 3 messages Jun 27, 2023 – Jul 6, 2023

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 virtio-gpu: do not replace surface when scanout is disabled

Surface is replaced with a place holder whenever the surface res is unreferenced by the guest message. With this logic, there is very frequent switching between guest display and the place holder image, which is looking like a flickering display if the guest driver is designed to unref the current scanout resource before sending out a new scanout resource. So it is better to leave the current scanout image until there is a new one flushed by the guest.

Cc Gerd Hoffmann <[email protected]>
Cc Marc-André Lureau <[email protected]>
Cc Vivek Kasireddy <[email protected]>
---
 hw/display/virtio-gpu.c | 1 -
 1 file changed, 1 deletion(-)
hw/display/virtio-gpu.c +0 -1
index 66cddd94d9..9d3e922c8f 100644
@@ -387,7 +387,6 @@ static void virtio_gpu_disable_scanout(VirtIOGPU *g, int scanout_id)
         res->scanout_bitmask &= ~(1 << scanout_id);
     }
 
-    dpy_gfx_replace_surface(scanout->con, NULL);
     scanout->resource_id = 0;
     scanout->ds = NULL;
     scanout->width = 0;
Signature · 1 line
2.34.1
#1 lore raw

Hi

Quoted 8 lines · Dongwon Kim

Surface is replaced with a place holder whenever the surface res is unreferenced by the guest message. With this logic, there is very frequent switching between guest display and the place holder image, which is looking like a flickering display if the guest driver is designed to unref the current scanout resource before sending out a new scanout resource. So it is better to leave the current scanout image until there is a new one flushed by the guest.

Quoted 5 lines
Cc Gerd Hoffmann <[email protected]>
Cc Marc-André Lureau <[email protected]>
Cc Vivek Kasireddy <[email protected]>

Why is the driver not setting a different scanout before destroying the resource?

I think it's wrong to not replace the surface, as the associated scanout resource may be destroyed or explicitly disabled for various purposes, and we don't want to display garbage either.

---

Quoted 21 lines
 hw/display/virtio-gpu.c | 1 -
 1 file changed, 1 deletion(-)
hw/display/virtio-gpu.c +0 -1
index 66cddd94d9..9d3e922c8f 100644
@@ -387,7 +387,6 @@ static void virtio_gpu_disable_scanout(VirtIOGPU *g,
 int scanout_id)
         res->scanout_bitmask &= ~(1 << scanout_id);
     }
 
-    dpy_gfx_replace_surface(scanout->con, NULL);
     scanout->resource_id = 0;
     scanout->ds = NULL;
     scanout->width = 0;

--
2.34.1

Signature · 1 line
Marc-André Lureau
#2 lore raw
Quoted 18 lines · Marc-André Lureau

Hi

On Wed, Jun 28, 2023 at 12:32 AM Dongwon Kim <[email protected]> wrote:

Surface is replaced with a place holder whenever the surface res is unreferenced by the guest message. With this logic, there is very frequent switching between guest display and the place holder image, which is looking like a flickering display if the guest driver is designed to unref the current scanout resource before sending out a new scanout resource. So it is better to leave the current scanout image until there is a new one flushed by the guest.

Cc Gerd Hoffmann <[email protected]>
Cc Marc-André Lureau <[email protected]>
Cc Vivek Kasireddy <[email protected]>
Quoted 3 lines

Why is the driver not setting a different scanout before destroying the resource?

I think it's wrong to not replace the surface, as the associated scanout resource may be destroyed or explicitly disabled for various purposes, and we don't want to display garbage either.

Yeah..I got your point. This is to address very specific to our use-case with windows guest that runs virtio-gpu like driver that does unref before the next framebuffer is set. And I agree that this sequence doesn't look right. Let me check if we could change the sequence in the guest driver.

Quoted 26 lines
    ---
     hw/display/virtio-gpu.c | 1 -
     1 file changed, 1 deletion(-)

    diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c
    index 66cddd94d9..9d3e922c8f 100644
    --- a/hw/display/virtio-gpu.c
    +++ b/hw/display/virtio-gpu.c
    @@ -387,7 +387,6 @@ static void
    virtio_gpu_disable_scanout(VirtIOGPU *g, int scanout_id)
             res->scanout_bitmask &= ~(1 << scanout_id);
         }

    -    dpy_gfx_replace_surface(scanout->con, NULL);
         scanout->resource_id = 0;
         scanout->ds = NULL;
         scanout->width = 0;
    -- 
    2.34.1
Signature · 1 line
Marc-André Lureau