Re: [BUG] cow_file_range failed and kernel panic on zoned device

On Tue, Jul 21, 2026 at 07:46:53PM +0800, HAN Yuwei wrote:
> Hi all,
> I have received a report from a user stating kernel panic when writing to
> HM-SMR HDD in 1TiB. The user used "mkfs.btrfs -f /dev/sdX" to create a new
> volume. And he tested xfs on the same device without any issue.
> 
> Reported-By: LUO Xiang <[email protected]>
> HDD Model:Seagate ST18000NM009J
> uname -a:Linux pro69 7.1.3-arch2-2 #1 SMP PREEMPT_DYNAMIC Thu, 16 Jul 2026
> 17:41:28 +0000 x86_64 GNU/Linux
> btrfs-progs: local/btrfs-progs 7.1-1
> dmesg:
> [redacted]

[snip]

> Jul 19 21:28:37 pro69 kernel: BTRFS: error (device sde state A) in
> cleanup_transaction:2067: errno=-11 unknown
> Jul 19 21:28:37 pro69 kernel: BTRFS info (device sde state EA): forced
> readonly
> Jul 19 21:29:08 pro69 kernel: BTRFS error (device sde state EA):
> cow_file_range failed, root=5 inode=268 start=68295892992 len=688128
> cur_offset=68295892992 cur_alloc_size=0: -30
> Jul 19 21:29:08 pro69 kernel: writepage_delalloc: 383 callbacks suppressed
> Jul 19 21:29:08 pro69 kernel: BTRFS error (device sde state EA): failed to
> run delalloc range, root=5 ino=268 folio=68295892992 submit_bitmap=0
> start=68295892992 len=688128: -30
> 
> the full dmesg can be retreived if interested.

I /think/ I know the source of the problem. Errno 11 is EAGAIN and from
writepage_delalloc() it means we've hit a "hole" or out-of-order
metadata. This is exactly the kind of bug I'm trying to fix for 2 weeks
now. Not sure why it surfaces now but I see it failing in my fstests
setups as well.

There's some hotfix attempts to it and I *hope* I have something
upstream acceptible as well soon.

This is a BTRFS specific sequential metadata thing XFS will not see
because XFS is not putting metadata on sequential zones. 

I hope I can give you something to try by the end of the week.

Byte,
	Johannes