Re: small writeback fixes v2

[Josef Bacik <josef@toxicpanda.com>]

On Wed, Aug 02, 2023 at 05:16:43PM +0200, Christoph Hellwig wrote:
> On Wed, Aug 02, 2023 at 08:49:56AM -0400, Josef Bacik wrote:
> > I ran this through the CI
> 
> Thanks a lot!
> 
> > [ 3461.147888] assertion failed: block_group->io_ctl.inode == NULL, in
> > fs/btrfs/block-group.c:4256
> 
> Hmm, this looks so unrelated that it leaves me puzzled.  How confident
> are you that this is a new issue based on the overall test setup?
> 
> > I also got an EBUSY trying to umount $SCRATCH_MNT with generic/475 with
> 
> > on an ARM machine with 64kib pagesize.  Though I'm pretty sure you're not to
> > blame for that last failure.  Thanks,
> 
> Yes, I've seen EBUSY in 475 quite regulary even without the changes,
> I think I also mentioned it in reply to the other 475-related discussion
> we had.  I tried to debug it for a while but didn't manage to get far.
> 

[  594.911885] ------------[ cut here ]------------
[  594.911892] WARNING: CPU: 1 PID: 1833 at fs/namespace.c:179
mnt_add_count+0xa4/0xb8
[  594.911897] Modules linked in: rfkill vfat fat virtio_net net_failover
virtio_balloon failover loop zram virtiofs fuse virtio_console virtio_blk
qemu_fw_cfg
[  594.911908] CPU: 1 PID: 1833 Comm: kworker/1:7 Tainted: G        W
6.5.0-rc3+ #31
[  594.911910] Hardware name: QEMU KVM Virtual Machine, BIOS
edk2-20230301gitf80f052277c8-26.fc38 03/01/2023
[  594.911911] Workqueue: events delayed_fput
[  594.911916] pstate: 21400005 (nzCv daif +PAN -UAO -TCO +DIT -SSBS BTYPE=--)
[  594.911918] pc : mnt_add_count+0xa4/0xb8
[  594.911920] lr : mnt_add_count+0x98/0xb8
[  594.911922] sp : ffff800093ea3c60
[  594.911923] x29: ffff800093ea3c60 x28: 0000000000000000 x27: 0000000000000000
[  594.911926] x26: ffff6b947f39d505 x25: ffffbdbf2bd9b1c8 x24: ffff6b9459997d40
[  594.911929] x23: ffff6b9456365728 x22: cd9fbdbf29b30b60 x21: ffff800093ea3ce8
[  594.911931] x20: ffff6b94596af5c0 x19: 00000000ffffffff x18: 0000000000003f75
[  594.911934] x17: 0000000000000000 x16: 0000000000000000 x15: ffff6b94497dc000
[  594.911936] x14: 0000000000000000 x13: 0000000000000030 x12: 0000000000000000
[  594.911939] x11: ffffbdbf32e69204 x10: ffffbdbf2b6dae50 x9 : ffffbdbf2f49f000
[  594.911942] x8 : ffff800093ea3c08 x7 : 0000000000000000 x6 : 00000000000fffff
[  594.911944] x5 : 00000000002b60c6 x4 : ffff6b94405716c0 x3 : ffff6b94405716c0
[  594.911947] x2 : ffffadd553c90000 x1 : ffff6b9441625900 x0 : 0000000060001020
[  594.911949] Call trace:
[  594.911950]  mnt_add_count+0xa4/0xb8
[  594.911952]  mntput_no_expire+0x88/0x4e0
[  594.911955]  mntput+0x28/0x48
[  594.911957]  __fput+0x134/0x2a0
[  594.911960]  delayed_fput+0x4c/0x68
[  594.911964]  process_one_work+0x290/0x5e0
[  594.911967]  worker_thread+0x7c/0x428
[  594.911970]  kthread+0x124/0x130
[  594.911972]  ret_from_fork+0x10/0x20
[  594.911974] irq event stamp: 27982
[  594.911975] hardirqs last  enabled at (27981): [<ffffbdbf2982162c>]
__call_rcu_common.constprop.0+0x16c/0x680
[  594.911980] hardirqs last disabled at (27982): [<ffffbdbf2a7642e4>]
el1_dbg+0x24/0x90
[  594.911984] softirqs last  enabled at (24348): [<ffffbdbf2a4d8650>]
neigh_managed_work+0xd8/0xf8
[  594.911989] softirqs last disabled at (24344): [<ffffbdbf2a4d85a8>]
neigh_managed_work+0x30/0xf8
[  594.911993] ---[ end trace 0000000000000000 ]---

So it's the delayed fput thing that we already know about.  I talked to Jens and
he's on his way to a fix but isn't there yet.

I'll drop a patch in our fstests to just umount in a loop, this test uncovers
lots of things so I don't want to exclude it from the CI runs.  I'll go look at
the other failures now.  Thanks,

Josef