mbox series

[0/2] Revert "monitor: use aio_co_reschedule_self()"

Message ID 20240506190622.56095-1-stefanha@redhat.com
Headers show
Series Revert "monitor: use aio_co_reschedule_self()" | expand

Message

Stefan Hajnoczi May 6, 2024, 7:06 p.m. UTC
This series fixes RHEL-34618 "qemu crash on Assertion `luringcb->co->ctx ==
s->aio_context' failed when do block_resize on hotplug disk with aio=io_uring":
https://issues.redhat.com/browse/RHEL-34618

Kevin identified commit 1f25c172f837 ("monitor: use aio_co_reschedule_self()")
as the root cause. There is a subtlety regarding how
qemu_get_current_aio_context() returns qemu_aio_context even though we may be
running in iohandler_ctx.

Revert commit 1f25c172f837, it was just intended as a code cleanup.

Stefan Hajnoczi (2):
  Revert "monitor: use aio_co_reschedule_self()"
  aio: warn about iohandler_ctx special casing

 include/block/aio.h | 6 ++++++
 qapi/qmp-dispatch.c | 7 +++++--
 2 files changed, 11 insertions(+), 2 deletions(-)

Comments

Kevin Wolf May 28, 2024, 3:54 p.m. UTC | #1
Am 06.05.2024 um 21:06 hat Stefan Hajnoczi geschrieben:
> This series fixes RHEL-34618 "qemu crash on Assertion `luringcb->co->ctx ==
> s->aio_context' failed when do block_resize on hotplug disk with aio=io_uring":
> https://issues.redhat.com/browse/RHEL-34618
> 
> Kevin identified commit 1f25c172f837 ("monitor: use aio_co_reschedule_self()")
> as the root cause. There is a subtlety regarding how
> qemu_get_current_aio_context() returns qemu_aio_context even though we may be
> running in iohandler_ctx.
> 
> Revert commit 1f25c172f837, it was just intended as a code cleanup.
> 
> Stefan Hajnoczi (2):
>   Revert "monitor: use aio_co_reschedule_self()"
>   aio: warn about iohandler_ctx special casing

Thanks, applied to the block branch.

Kevin