On Tue, Apr 11, 2017 at 04:08:53PM +0200, Kevin Wolf wrote:
> This fixes a regression introduced in commit 9d456654.
>
> aio_co_wake() can only be used to reenter a coroutine that was already
> previously entered, otherwise co->ctx is uninitialised and we access
> garbage. Using it immediately after qemu_coroutine_create() like in
> co_read_response() is wrong and causes segfaults.
>
> Replace the call with aio_co_enter(), which gets an explicit AioContext
> parameter and works even for new coroutines.
>
> Signed-off-by: Kevin Wolf <kw...@redhat.com>
> Tested-by: Kashyap Chamarthy <kcham...@redhat.com>
Thanks.
The test evidence was posted on IRC. But just for archives' sake, I was
here (on Git) when I applied this patch:
$ git describe
v2.9.0-rc3-45-gaa388dd
And that was the specific test:
$ qemu-img convert -t directsync \
./cirros-0.3.5.qcow2 sheepdog:cirros0.3.5vm
After applying this patch, the segfault is gone, and the `qemu-img convert`
happens successfully.
That was my Sheepdog environment:
https://kashyapc.fedorapeople.org/virt/sheepdog-qemu-corosync.txt
> ---
> block/sheepdog.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/block/sheepdog.c b/block/sheepdog.c
> index 1b71fc8..142eb4f 100644
> --- a/block/sheepdog.c
> +++ b/block/sheepdog.c
> @@ -942,7 +942,7 @@ static void co_read_response(void *opaque)
> s->co_recv = qemu_coroutine_create(aio_read_response, opaque);
> }
>
> - aio_co_wake(s->co_recv);
> + aio_co_enter(s->aio_context, s->co_recv);
> }
>
> static void co_write_request(void *opaque)
> --
> 1.8.3.1
>
--
/kashyap
