[Qemu-devel] [PATCH] qxl: fix local renderer crash

[Qemu-devel] [PATCH] qxl: fix local renderer crash

[Gerd Hoffmann]

Make sure we only ask the spice local renderer for display updates in
case we have a valid primary surface.  Without that spice is confused
and throws errors in case a display update request (triggered by
screendump for example) happens in parallel to a mode switch and hits
the race window where the old primary surface is gone and the new isn't
establisted yet.

Fixes: https://bugzilla.redhat.com//show_bug.cgi?id=1567733
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
 hw/display/qxl-render.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/hw/display/qxl-render.c b/hw/display/qxl-render.c
index e7ac4f8789..c62b9a5e75 100644
--- a/hw/display/qxl-render.c
+++ b/hw/display/qxl-render.c
@@ -169,7 +169,8 @@ void qxl_render_update(PCIQXLDevice *qxl)
 
     qemu_mutex_lock(&qxl->ssd.lock);
 
-    if (!runstate_is_running() || !qxl->guest_primary.commands) {
+    if (!runstate_is_running() || !qxl->guest_primary.commands ||
+        qxl->mode == QXL_MODE_UNDEFINED) {
         qxl_render_update_area_unlocked(qxl);
         qemu_mutex_unlock(&qxl->ssd.lock);
         return;
-- 
2.9.3

Re: [Qemu-devel] [PATCH] qxl: fix local renderer crash

[Marc-André Lureau]

Hi

On Fri, Apr 27, 2018 at 1:55 PM, Gerd Hoffmann <kraxel@redhat.com> wrote:
> Make sure we only ask the spice local renderer for display updates in
> case we have a valid primary surface.  Without that spice is confused
> and throws errors in case a display update request (triggered by
> screendump for example) happens in parallel to a mode switch and hits
> the race window where the old primary surface is gone and the new isn't
> establisted yet.
>
> Fixes: https://bugzilla.redhat.com//show_bug.cgi?id=1567733
> Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

Because the update message is handled asynchronously by the spice
worker thread, I wonder if it shouldn't be ignored by spice instead if
the display state is invalid/undefined.

But this patch seems reasonable too:
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>


> ---
>  hw/display/qxl-render.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/hw/display/qxl-render.c b/hw/display/qxl-render.c
> index e7ac4f8789..c62b9a5e75 100644
> --- a/hw/display/qxl-render.c
> +++ b/hw/display/qxl-render.c
> @@ -169,7 +169,8 @@ void qxl_render_update(PCIQXLDevice *qxl)
>
>      qemu_mutex_lock(&qxl->ssd.lock);
>
> -    if (!runstate_is_running() || !qxl->guest_primary.commands) {
> +    if (!runstate_is_running() || !qxl->guest_primary.commands ||
> +        qxl->mode == QXL_MODE_UNDEFINED) {
>          qxl_render_update_area_unlocked(qxl);
>          qemu_mutex_unlock(&qxl->ssd.lock);
>          return;
> --
> 2.9.3
>
>



-- 
Marc-André Lureau