Re: Backport of "xfs: open code ioend needs workqueue helper" to 5.10?

From: Amir Goldstein <amir73il@gmail.com>
To: Christian Theune <ct@flyingcircus.io>
Cc: linux-xfs@vger.kernel.org, Christoph Hellwig <hch@lst.de>,
	Brian Foster <bfoster@redhat.com>,
	Chandan Babu R <chandan.babu@oracle.com>
Subject: Re: Backport of "xfs: open code ioend needs workqueue helper" to 5.10?
Date: Thu, 13 Apr 2023 10:44:43 +0300	[thread overview]
Message-ID: <CAOQ4uxjj2UqA0h4Y31NbmpHksMhVrXfXjLG4Tnz3zq_UR-3gSA@mail.gmail.com> (raw)
In-Reply-To: <6AB6497D-18E5-41C4-B688-4DED6703534F@flyingcircus.io>

On Wed, Apr 12, 2023 at 6:58 PM Christian Theune <ct@flyingcircus.io> wrote:
>
> Hi,
>
> ugh. Sorry, looks like I jumped the gun. Mea culpa.
>
> We experienced a hang like this:
>
> Apr 05 11:51:27 kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> Apr 05 11:51:27 kernel: task:xfs-conv/vdc1   state:D stack:    0 pid:  606 ppid:     2 flags:0x00004080
> Apr 05 11:51:27 kernel: Workqueue: xfs-conv/vdc1 xfs_end_io [xfs]
> Apr 05 11:51:27 kernel: Call Trace:
> Apr 05 11:51:27 kernel:  __schedule+0x274/0x870
> Apr 05 11:51:27 kernel:  schedule+0x46/0xb0
> Apr 05 11:51:27 kernel:  xlog_grant_head_wait+0xc5/0x1d0 [xfs]
> Apr 05 11:51:27 kernel:  xlog_grant_head_check+0xde/0x100 [xfs]
> Apr 05 11:51:27 kernel:  xfs_log_reserve+0xbe/0x1b0 [xfs]
> Apr 05 11:51:27 kernel:  xfs_trans_reserve+0x143/0x180 [xfs]
> Apr 05 11:51:27 kernel:  xfs_trans_alloc+0xee/0x1a0 [xfs]
> Apr 05 11:51:27 kernel:  xfs_iomap_write_unwritten+0x120/0x2e0 [xfs]
> Apr 05 11:51:27 kernel:  ? record_times+0x15/0x90
> Apr 05 11:51:27 kernel:  xfs_end_ioend+0xd8/0x140 [xfs]
> Apr 05 11:51:27 kernel:  xfs_end_io+0xb8/0xf0 [xfs]
> Apr 05 11:51:27 kernel:  process_one_work+0x1b6/0x350
> Apr 05 11:51:27 kernel:  rescuer_thread+0x1d1/0x3a0
> Apr 05 11:51:27 kernel:  ? worker_thread+0x3e0/0x3e0
> Apr 05 11:51:27 kernel:  kthread+0x11b/0x140
> Apr 05 11:51:27 kernel:  ? kthread_associate_blkcg+0xb0/0xb0
> Apr 05 11:51:27 kernel:  ret_from_fork+0x22/0x30
>
> Which seems to be similar to this:
> https://bugs.launchpad.net/bugs/1996269
>
> I followed their patchset here:
> https://review.opendev.org/c/starlingx/kernel/+/864257
>
> And I was under the impression that I picked the right one to ask for backporting, but it seems that was incorrect. I went through the list again and I think the following patches are the ones missing from 5.10:
>
> 8182ec00803085354761bbadf0287cad7eac0e2f - https://review.opendev.org/c/starlingx/kernel/+/864257/5/kernel-std/centos/patches/0035-xfs-drop-submit-side-trans-alloc-for-append-ioends.patch
> edbf1eb9032b84631031d9b43570e262f3461c24 - https://review.opendev.org/c/starlingx/kernel/+/864257/5/kernel-std/centos/patches/0036-xfs-open-code-ioend-needs-workqueue-helper.patch
> 170e31793806ce5e5a9647b6340954536244518e - https://review.opendev.org/c/starlingx/kernel/+/864257/5/kernel-std/centos/patches/0037-xfs-drop-unused-ioend-private-merge-and-setfilesize-.patch
> 2fd609b6c90a88630a50fb317473b210759b3873 - https://review.opendev.org/c/starlingx/kernel/+/864257/5/kernel-std/centos/patches/0038-xfs-drop-unnecessary-setfilesize-helper.patch
>

The only commit that fixes the bug is:
7cd3099f4925 xfs: drop submit side trans alloc for append ioends

The rest are just code cleanups.

That fix was missed in my original backports from v5.13 because of a tool error,
so thank you for pointing it out.

I have added it to my test branch and will follow up with posting to
stable later on.

Chandan,

Please make sure you include this fix when you get to considering
fixes from v5.13 to 5.4.y.

I will wait with posting this fix to 5.10.y until I get the v5.13
backports wish list from you.

Thanks,
Amir.