[PATCH 0/6] virtio-blk: iothread-vq-mapping cleanups

Stefan Hajnoczi posted 6 patches 10 months, 1 week ago
Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/20240119135748.270944-1-stefanha@redhat.com
Maintainers: Stefan Hajnoczi <stefanha@redhat.com>, Kevin Wolf <kwolf@redhat.com>, Hanna Reitz <hreitz@redhat.com>, "Michael S. Tsirkin" <mst@redhat.com>, Paolo Bonzini <pbonzini@redhat.com>, "Marc-André Lureau" <marcandre.lureau@redhat.com>, "Daniel P. Berrangé" <berrange@redhat.com>, Thomas Huth <thuth@redhat.com>, "Philippe Mathieu-Daudé" <philmd@linaro.org>
There is a newer version of this series
meson.build                     |   1 -
hw/block/dataplane/trace.h      |   1 -
hw/block/dataplane/virtio-blk.h |  34 ---
include/hw/virtio/virtio-blk.h  |  16 +-
hw/block/dataplane/virtio-blk.c | 404 -------------------------------
hw/block/virtio-blk.c           | 412 +++++++++++++++++++++++++++++---
hw/block/dataplane/meson.build  |   1 -
hw/block/dataplane/trace-events |   5 -
8 files changed, 391 insertions(+), 483 deletions(-)
delete mode 100644 hw/block/dataplane/trace.h
delete mode 100644 hw/block/dataplane/virtio-blk.h
delete mode 100644 hw/block/dataplane/virtio-blk.c
delete mode 100644 hw/block/dataplane/trace-events
[PATCH 0/6] virtio-blk: iothread-vq-mapping cleanups
Posted by Stefan Hajnoczi 10 months, 1 week ago
Kevin Wolf identified some issues with the virtio-blk iothread-vq-mapping patch
series that was merged at the end of 2023:
1. s->rq is restarted from one AioContext and races with the other
   iothread-vq-mapping AioContexts.
2. Failure to set the AioContext is no longer fatal since the IO_CODE APIs can
   be called from any thread. We can relax the error handling behavior.
3. Starting dataplane must self-trigger the ioeventfd even in a drained
   section. Failure to do so could lead to an unresponsive virtio-blk device.

This patch series addresses these issues. The first few patches merge the
hw/block/dataplane/virtio-blk.c code into hw/block/virtio-blk.c so that s->rq
can easily be restarted in the correct AioContexts.

Stefan Hajnoczi (6):
  virtio-blk: move dataplane code into virtio-blk.c
  virtio-blk: rename dataplane create/destroy functions
  virtio-blk: rename dataplane to ioeventfd
  virtio-blk: restart s->rq reqs in vq AioContexts
  virtio-blk: tolerate failure to set BlockBackend AioContext
  virtio-blk: always set ioeventfd during startup

 meson.build                     |   1 -
 hw/block/dataplane/trace.h      |   1 -
 hw/block/dataplane/virtio-blk.h |  34 ---
 include/hw/virtio/virtio-blk.h  |  16 +-
 hw/block/dataplane/virtio-blk.c | 404 -------------------------------
 hw/block/virtio-blk.c           | 412 +++++++++++++++++++++++++++++---
 hw/block/dataplane/meson.build  |   1 -
 hw/block/dataplane/trace-events |   5 -
 8 files changed, 391 insertions(+), 483 deletions(-)
 delete mode 100644 hw/block/dataplane/trace.h
 delete mode 100644 hw/block/dataplane/virtio-blk.h
 delete mode 100644 hw/block/dataplane/virtio-blk.c
 delete mode 100644 hw/block/dataplane/trace-events

-- 
2.43.0
Re: [PATCH 0/6] virtio-blk: iothread-vq-mapping cleanups
Posted by Kevin Wolf 10 months, 1 week ago
Am 19.01.2024 um 14:57 hat Stefan Hajnoczi geschrieben:
> Kevin Wolf identified some issues with the virtio-blk iothread-vq-mapping patch
> series that was merged at the end of 2023:
> 1. s->rq is restarted from one AioContext and races with the other
>    iothread-vq-mapping AioContexts.
> 2. Failure to set the AioContext is no longer fatal since the IO_CODE APIs can
>    be called from any thread. We can relax the error handling behavior.
> 3. Starting dataplane must self-trigger the ioeventfd even in a drained
>    section. Failure to do so could lead to an unresponsive virtio-blk device.
> 
> This patch series addresses these issues. The first few patches merge the
> hw/block/dataplane/virtio-blk.c code into hw/block/virtio-blk.c so that s->rq
> can easily be restarted in the correct AioContexts.

Thanks, applied to the block branch.

Kevin