Dmitry reported a handful of vhost-user-gpu issues (#4846-4849): crash when a command waits for a reply and a fence is signaled (#4848), guest losing display when the device is restarted (#4849), segfault in vga_update_display() after a guest device reset (#4846), or main loop spinning on bad descriptor (#4847). The first patches address those.
The rest are problems found while working through the code: console is not resized on dmabuf scanout, the 2d scanout path frees a surface it does not own, and the exported dmabuf is imported with the wrong stride in some setup. Signed-off-by: Marc-André Lureau <[email protected]> --- Changes in v3: - fix crash on finalize - replace Dmitry email address - Link to v2: https://lore.kernel.org/qemu-devel/[email protected] Changes in v2: - add "hw/display/vhost-user-gpu: reset scanouts on the main loop" patch, similar to virtio-gpu - add Tested-by tags - Link to v1: https://lore.kernel.org/qemu-devel/[email protected] --- Marc-André Lureau (8): contrib/vhost-user-gpu: don't fence commands waiting for a reply contrib/vhost-user-gpu: reset the frontend socket on stop and start hw/display/vhost-user-gpu: reset the console scanout on device reset hw/display/vhost-user-gpu: reset scanouts on the main loop hw/display/vhost-user-gpu: don't set an fd handler without a socket hw/display/vhost-user-gpu: resize the console on dmabuf scanout hw/display/vhost-user-gpu: fix scanout bookkeeping on the 2D path contrib/vhost-user-gpu: use the stride of the exported dmabuf contrib/vhost-user-gpu/vhost-user-gpu.c | 169 ++++++++++++++++++++++++++------ contrib/vhost-user-gpu/virgl.c | 70 +++++++++++-- contrib/vhost-user-gpu/vugpu.h | 7 +- hw/display/vhost-user-gpu.c | 106 ++++++++++++++++++-- include/hw/virtio/virtio-gpu.h | 4 + 5 files changed, 304 insertions(+), 52 deletions(-) --- base-commit: 4bcd55a1f4b926ae86872087f7f3bc9f9e1d065e change-id: 20261112-issue-4848-d49d6bd8e7b4 Best regards, -- Marc-André Lureau <[email protected]>
