]> git.proxmox.com Git - mirror_qemu.git/commit
Revert "qed: Implement .bdrv_drain"
authorStefan Hajnoczi <stefanha@redhat.com>
Tue, 8 Mar 2016 10:11:16 +0000 (10:11 +0000)
committerStefan Hajnoczi <stefanha@redhat.com>
Thu, 17 Mar 2016 09:50:14 +0000 (09:50 +0000)
commit1f3ddfcb258e6e0562467a9644daad3f28597da8
treef6cfc2b655d997660b63add838cc64f24124cc65
parent147dfab747dd1720cb6350b6b746ca84e15f9a24
Revert "qed: Implement .bdrv_drain"

This reverts commit df9a681dc9ad41c9cdeb9ecc5d060ba9abd27e01.

Note that commit df9a681dc9ad41c9cdeb9ecc5d060ba9abd27e01 included some
unrelated hunks, possibly due to a merge failure or an overlooked
squash.  This only reverts the qed .bdrv_drain() implementation.

The qed .bdrv_drain() implementation is unsafe and can lead to a double
request completion.

Paolo Bonzini reports:
"The problem is that bdrv_qed_drain calls qed_plug_allocating_write_reqs
unconditionally, but this is not correct if an allocating write is
queued.  In this case, qed_unplug_allocating_write_reqs will restart the
allocating write and possibly cause it to complete.  The aiocb however
is still in use for the L2/L1 table writes, and will then be completed
again as soon as the table writes are stable."

For QEMU 2.6 we can simply revert this commit.  A full solution for the
qed need check timer may be added if the bdrv_drain() implementation is
extended.

Reported-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1457431876-8475-1-git-send-email-stefanha@redhat.com
block/qed.c