From: "Darrick J. Wong" <djwong@kernel.org>
To: Brian Foster <bfoster@redhat.com>
Cc: linux-xfs@vger.kernel.org, linux-fsdevel@vger.kernel.org
Subject: Re: [PATCH v3 2/3] xfs: kick large ioends to completion workqueue
Date: Tue, 25 May 2021 18:20:16 -0700 [thread overview]
Message-ID: <20210526012016.GF202078@locust> (raw)
In-Reply-To: <20210517171722.1266878-3-bfoster@redhat.com>
On Mon, May 17, 2021 at 01:17:21PM -0400, Brian Foster wrote:
> We've had reports of soft lockup warnings in the iomap ioend
> completion path due to very large bios and/or bio chains. This
> occurs because ioend completion touches every page associated with
> the ioend. It generally requires exceedingly large (i.e. multi-GB)
> bios or bio chains to reproduce a soft lockup warning, but even with
> smaller ioends there's really no good reason to incur the cost of
> potential cacheline misses in bio completion context. Divert ioends
> larger than 1MB to the workqueue so completion occurs in non-atomic
> context and can reschedule to avoid soft lockup warnings.
>
> Signed-off-by: Brian Foster <bfoster@redhat.com>
Will give this a spin on the test farm overnight but at least in
principle this seems fine to me:
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
--D
> ---
> fs/xfs/xfs_aops.c | 16 ++++++++++++++--
> 1 file changed, 14 insertions(+), 2 deletions(-)
>
> diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
> index 84cd6cf46b12..05b1bb146f17 100644
> --- a/fs/xfs/xfs_aops.c
> +++ b/fs/xfs/xfs_aops.c
> @@ -30,6 +30,13 @@ XFS_WPC(struct iomap_writepage_ctx *ctx)
> return container_of(ctx, struct xfs_writepage_ctx, ctx);
> }
>
> +/*
> + * Completion touches every page associated with the ioend. Send anything
> + * larger than 1MB (based on 4k pages) or so to the completion workqueue to
> + * avoid this work in bio completion context.
> + */
> +#define XFS_LARGE_IOEND (256ULL << PAGE_SHIFT)
> +
> /*
> * Fast and loose check if this write could update the on-disk inode size.
> */
> @@ -409,9 +416,14 @@ xfs_prepare_ioend(
>
> memalloc_nofs_restore(nofs_flag);
>
> - /* send ioends that might require a transaction to the completion wq */
> + /*
> + * Send ioends that might require a transaction or are large enough that
> + * we don't want to do page processing in bio completion context to the
> + * wq.
> + */
> if (xfs_ioend_is_append(ioend) || ioend->io_type == IOMAP_UNWRITTEN ||
> - (ioend->io_flags & IOMAP_F_SHARED))
> + (ioend->io_flags & IOMAP_F_SHARED) ||
> + ioend->io_size >= XFS_LARGE_IOEND)
> ioend->io_bio->bi_end_io = xfs_end_bio;
> return status;
> }
> --
> 2.26.3
>
next prev parent reply other threads:[~2021-05-26 1:20 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-17 17:17 [PATCH v3 0/3] iomap: avoid soft lockup warnings on large ioends Brian Foster
2021-05-17 17:17 ` [PATCH v3 1/3] iomap: resched ioend completion when in non-atomic context Brian Foster
2021-05-17 17:54 ` Matthew Wilcox
2021-05-18 11:38 ` Brian Foster
2021-05-20 21:58 ` Darrick J. Wong
2021-05-24 11:57 ` Brian Foster
2021-05-24 16:53 ` Darrick J. Wong
2021-05-26 1:19 ` Darrick J. Wong
2021-05-22 7:45 ` Ming Lei
2021-05-24 11:57 ` Brian Foster
2021-05-24 14:11 ` Ming Lei
2021-05-17 17:17 ` [PATCH v3 2/3] xfs: kick large ioends to completion workqueue Brian Foster
2021-05-26 1:20 ` Darrick J. Wong [this message]
2021-05-17 17:17 ` [PATCH RFC v3 3/3] iomap: bound ioend size to 4096 pages Brian Foster
2021-05-19 13:28 ` Christoph Hellwig
2021-05-19 14:52 ` Brian Foster
2021-05-20 23:27 ` Darrick J. Wong
2021-05-24 12:02 ` Brian Foster
2021-05-25 4:20 ` Darrick J. Wong
2021-05-25 4:29 ` Damien Le Moal
2021-05-25 7:13 ` Dave Chinner
2021-05-25 9:07 ` Andreas Gruenbacher
2021-05-26 2:12 ` Matthew Wilcox
2021-05-26 3:32 ` Darrick J. Wong
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210526012016.GF202078@locust \
--to=djwong@kernel.org \
--cc=bfoster@redhat.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-xfs@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).