linux-kernel - Re: [PATCH 3/5] direct-io: honor dio->boundary a little more strictly

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20100512001347.GB27011@dhcp231-156.rdu.redhat.com>
Date:	Tue, 11 May 2010 20:13:48 -0400
From:	Josef Bacik <josef@...hat.com>
To:	Josef Bacik <josef@...hat.com>
Cc:	linux-btrfs@...r.kernel.org, linux-kernel@...r.kernel.org,
	linux-fsdevel@...r.kernel.org, linux-mm@...ck.org,
	hch@...radead.org, akpm@...ux-foundation.org
Subject: Re: [PATCH 3/5] direct-io: honor dio->boundary a little more
	strictly

On Fri, May 07, 2010 at 01:41:04PM -0400, Josef Bacik wrote:
> Because BTRFS needs to be able to lookup checksums when we submit the bio's, we
> need to be able to look up the logical offset in the inode we're submitting the
> bio for.  The way we do this is in our get_blocks function is return the map_bh
> with b_blocknr of the logical offset in the file, and then in the submit path
> turn that into an actual block number on the device.  This causes problems with
> the DIO stuff since it will try and merge requests that look like they are
> contiguous, even though they are not actually contiguous on disk.  So BTRFS sets
> buffer_boundary on the map_bh.  Unfortunately if there is not a bio already
> setup in the DIO stuff, dio->boundary gets cleared and then the next time a
> request is made they will get merged.  So instead of clearing dio->boundary in
> dio_new_bio, save the boundary value before doing anything, that way if
> dio->boundary gets cleared, we still submit the IO.  Thanks,
> 
> Signed-off-by: Josef Bacik <josef@...hat.com>
> ---
>  fs/direct-io.c |    5 ++++-
>  1 files changed, 4 insertions(+), 1 deletions(-)
> 
> diff --git a/fs/direct-io.c b/fs/direct-io.c
> index 2dbf2e9..98f6f42 100644
> --- a/fs/direct-io.c
> +++ b/fs/direct-io.c
> @@ -615,6 +615,7 @@ static int dio_bio_add_page(struct dio *dio)
>   */
>  static int dio_send_cur_page(struct dio *dio)
>  {
> +	int boundary = dio->boundary;
>  	int ret = 0;
>  
>  	if (dio->bio) {
> @@ -627,7 +628,7 @@ static int dio_send_cur_page(struct dio *dio)
>  		 * Submit now if the underlying fs is about to perform a
>  		 * metadata read
>  		 */
> -		if (dio->boundary)
> +		if (boundary)
>  			dio_bio_submit(dio);
>  	}
>  
> @@ -644,6 +645,8 @@ static int dio_send_cur_page(struct dio *dio)
>  			ret = dio_bio_add_page(dio);
>  			BUG_ON(ret != 0);
>  		}
> +	} else if (boundary) {
> +		dio_bio_submit(dio);
>  	}
>  out:
>  	return ret;

Self-NACK on this one.  Seems to have an unwanted side-effect of forcing every
page to be submitted individually.  I'm going to fix this a different way.
Thanks,

Josef
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/