Re: [Devil-Linux-discuss] [Devil-Linux 1.4 0000066]: virtio-access produce IO-Errors

[Udo L. <udo...@al...>]

Hi Serge,
sorry for the delayed answer. I'm also not a kernel hacker, but you are
right, the part looks different (but alos different like the patch).
For reproducing it's enough to run devil-linux with an virtio-disk in a
kvm-session - all work's but the error fill /var/log/messages. The easies
way is using proxmox for this (http://pve.proxmox.com/wiki/Downloads).
But you need a 64bit computer with hardwarevirtualisation and a blank disk
(proxmox-pve erase the whole disk). With proxmox-pve you can create and
manage the kvm (and openvz) guest from a web-gui.

I have tried the svn-version but i failed at building... (the posting
before of me).

And thanks a lot for the work at devil-linux - it's a very nice distribution.

Best regards

Udo

> Udo,
>
> you are welcome.
> I've perused the patch and the bug description, then compared it with the
> recent
> kernel patch 2.6.32.10. The patch seems to be integrated.
>
> Please take a look:
>
> Patch:
> -----------------------------------------------------------------------
> #  end_request: I/O error, dev vda, sector 0
> #  end_request: I/O error, dev vda, sector 0
> #  commit 52b1fd5a27c625c78373e024bf570af3c9d44a79
> #  Author: Mikulas Patocka <mpa...@re...>
> #    dm: send empty barriers to targets in dm_flush
> #  https://bugzilla.redhat.com/514901
> # block/blk-core.c |    3 +--
> # 1 files changed, 1 insertions(+), 2 deletions(-)
> --- a/block/blk-core.c
> +++ a/block/blk-core.c
> @@ -1163,8 +1163,7 @@ static int __make_request(struct request_queue *q,
> struct
> bio *bio)
>         const int unplug = bio_unplug(bio);
>         int rw_flags;
>
> -       if (bio_barrier(bio) && bio_has_data(bio) &&
> -           (q->next_ordered == QUEUE_ORDERED_NONE)) {
> +       if (bio_barrier(bio) && (q->next_ordered == QUEUE_ORDERED_NONE)) {
>                 bio_endio(bio, -EOPNOTSUPP);
>                 return 0;
>         }
>
>
> Kernel code:
> -----------------------------------------------------------------------
>         int rw_flags;
>
>         if (bio_rw_flagged(bio, BIO_RW_BARRIER) &&
>             (q->next_ordered == QUEUE_ORDERED_NONE)) {
>                 bio_endio(bio, -EOPNOTSUPP);
>                 return 0;
>         }
> -----------------------------------------------------------------------
> I'm not a kernel hacker though and can be wrong...
>
>  Could you please help me with the issue reproducing?
>
> Serge
>
>