[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH] block: fix bdrv_ioctl called from coroutine
From: |
Fam Zheng |
Subject: |
Re: [Qemu-devel] [PATCH] block: fix bdrv_ioctl called from coroutine |
Date: |
Thu, 17 Dec 2015 08:59:46 +0800 |
User-agent: |
Mutt/1.5.21 (2010-09-15) |
On Wed, 12/16 19:33, Paolo Bonzini wrote:
> When called from a coroutine, bdrv_ioctl must be asynchronous just like
> e.g. bdrv_flush. The code was incorrectly making it synchronous, fix
> it.
>
> Signed-off-by: Paolo Bonzini <address@hidden>
> ---
> Fam, any reason why you did it this way? I don't see
> any coroutine caller, but it doesn't make much sense. :)
That is a surprising question! From a coroutine, it is bdrv_flush ->
bdrv_flush_co_entry -> bdrv_co_flush, which I think is always synchronous,
especially, noticing the code around calling bs->bdrv_aio_flush:
acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co);
if (acb == NULL) {
ret = -EIO;
} else {
qemu_coroutine_yield();
ret = co.ret;
}
Am I missing something?
Fam
>
> block/io.c | 7 ++++---
> 1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/block/io.c b/block/io.c
> index e00fb5d..841f5b5 100644
> --- a/block/io.c
> +++ b/block/io.c
> @@ -2614,10 +2614,11 @@ int bdrv_ioctl(BlockDriverState *bs, unsigned long
> int req, void *buf)
> bdrv_co_ioctl_entry(&data);
> } else {
> Coroutine *co = qemu_coroutine_create(bdrv_co_ioctl_entry);
> +
> qemu_coroutine_enter(co, &data);
> - }
> - while (data.ret == -EINPROGRESS) {
> - aio_poll(bdrv_get_aio_context(bs), true);
> + while (data.ret == -EINPROGRESS) {
> + aio_poll(bdrv_get_aio_context(bs), true);
> + }
> }
> return data.ret;
> }
> --
> 2.5.0
>
Re: [Qemu-devel] [PATCH] block: fix bdrv_ioctl called from coroutine, Kevin Wolf, 2015/12/18