[Qemu-devel] [Bug 1784900] Re: QEMU (frontend) crashes upon warm reboot with virtio-gpu device and vga=775 on Linux cmdline

[Stefan Berger]

This patch here fixes the issue, but is likely introducing inefficiency.
There are two if statements above the patch that should set full_update
= 1 due to 'some change', but none of them triggers it. So I think the
surface is wrong and needs to be recreated.

diff --git a/hw/display/vga.c b/hw/display/vga.c
index ed476e4e80..71b5684994 100644
--- a/hw/display/vga.c
+++ b/hw/display/vga.c
@@ -1571,6 +1571,7 @@ static void vga_draw_graphic(VGACommonState *s, int 
full_update)
          * must be updated with the new base address */
         full_update = 1;
     }
+    full_update = 1;

     if (full_update) {
         if (share_surface) {


A better solution may be this one here:

diff --git a/hw/display/vga.c b/hw/display/vga.c
index ed476e4e80..4f365b6d43 100644
--- a/hw/display/vga.c
+++ b/hw/display/vga.c
@@ -1566,7 +1566,7 @@ static void vga_draw_graphic(VGACommonState *s, int 
full_update)
         full_update = 1;
     }
     if (surface_data(surface) != s->vram_ptr + (s->start_addr * 4)
-        && is_buffer_shared(surface)) {
+        /*&& is_buffer_shared(surface)*/) {
         /* base address changed (page flip) -> shared display surfaces
          * must be updated with the new base address */
         full_update = 1;

-- 
You received this bug notification because you are a member of qemu-
devel-ml, which is subscribed to QEMU.
https://bugs.launchpad.net/bugs/1784900

Title:
  QEMU (frontend) crashes upon warm reboot with virtio-gpu device and
  vga=775 on Linux cmdline

Status in QEMU:
  New

Bug description:
  With vga=775 on the Linux command line a first boot of the VM running
  Linux works fine. After a warm reboot it crashes during Linux boot.
  The VM was used remotely via virt-manager and VNC.

  Bisecting the code lead to the following patch that introduced the
  bug:

  commit 1fccd7c5a9a722a9cbf1bc91693f4618034f01ac (HEAD, refs/bisect/bad)
  Author: Gerd Hoffmann <address@hidden>
  Date:   Mon Jul 2 18:24:43 2018 +0200

      virtio-gpu: disable scanout when backing resource is destroyed

      Signed-off-by: Gerd Hoffmann <address@hidden>
      Reviewed-by: Marc-André Lureau <address@hidden>
      Message-id: address@hidden

  diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c
  index 336dc59007..08cd567218 100644
  --- a/hw/display/virtio-gpu.c
  +++ b/hw/display/virtio-gpu.c
  @@ -430,6 +430,16 @@ static void virtio_gpu_disable_scanout(VirtIOGPU *g, int 
scanout_id)
   static void virtio_gpu_resource_destroy(VirtIOGPU *g,
                                           struct virtio_gpu_simple_resource 
*res)
   {
  +    int i;
  +
  +    if (res->scanout_bitmask) {
  +        for (i = 0; i < g->conf.max_outputs; i++) {
  +            if (res->scanout_bitmask & (1 << i)) {
  +                virtio_gpu_disable_scanout(g, i);
  +            }
  +        }
  +    }
  +
       pixman_image_unref(res->image);
       virtio_gpu_cleanup_mapping(res);
       QTAILQ_REMOVE(&g->reslist, res, next);

  
  Reported backtraces can be found here:  
https://paste.fedoraproject.org/paste/OUDEfCk1IY7xiy0I0PDlkw

To manage notifications about this bug go to:
https://bugs.launchpad.net/qemu/+bug/1784900/+subscriptions