пятница

[Bug 2174473] [NEW] Btrfs zoned: ENOSPC/EAGAIN transaction abort and forced readonly on WD host-managed SMR (Ubuntu 26.04)

Public bug reported: **Impact** Ubuntu Server 26.04 LTS experiences recurring Btrfs zoned filesystem failures on a 27 TB Western Digital host-managed SMR drive. Write operations fail despite substantial available space, and the filesystem can abort a transaction and switch to read-only. This prevents reliable use of the filesystem and presents a risk of incomplete writes or data loss. **Hardware and filesystem** - Drive: WDC WSH722870ALE604, 27 TB, host-managed SMR - Filesystem: Btrfs zoned, single-device - Zone size: 256 MiB - Filesystem UUID: 42bc4deb-1694-419f-96ea-1843a0a4f8d2 - Distribution: Ubuntu Server 26.04 LTS - Storage controller: Broadcom / LSI SAS9305-16i (SAS3224), using the mpt3sas kernel driver **Affected kernel versions** - 7.0.0-34-generic: transaction abort with EAGAIN (-11), followed by forced read-only - 7.0.0-42-generic: repeated ENOSPC (-28) errors during writes of large files, followed by filesystem failure - 7.0.0-43-generic: installed but not tested for this issue **Observed behavior** On November 7, 2026, kernel 7.0.0-34-generic reported: BTRFS error (device sdg): error while writing out transaction: -11 BTRFS warning (device sdg): Skipping commit of aborted transaction. BTRFS: Transaction aborted (error -11) BTRFS: error (device sdg state A) in cleanup_transaction:2060: errno=-11 unknown BTRFS info (device sdg state EA): forced readonly Previously, on kernel 7.0.0-42-generic, writing large media files repeatedly produced: BTRFS error (device sdg): cow_file_range failed ...: -28 BTRFS error (device sdg): failed to run delalloc range ...: -28 At the time of the November 7 failure, Btrfs reported: Device size: 24.56 TiB Device unallocated: 10.65 TiB Device zone unusable: 251.31 GiB Free (estimated): 10.86 TiB All Btrfs device error counters were zero, including read, write, flush, corruption and generation errors. Dynamic and periodic block-group reclaim were disabled. The corresponding reclaim counters were zero. **Expected behavior** Btrfs should successfully allocate space and complete transactions when sufficient usable space is available, without unexpectedly aborting transactions or forcing the filesystem into read-only mode. **Actual behavior** Write failures and transaction aborts occur intermittently during filesystem activity. The issue has been observed on multiple Ubuntu kernel versions. After the transaction abort, the filesystem becomes read-only and cannot accept further writes. **Related upstream development** I have reported the issue directly to the Linux Btrfs developers: https://lore-kernel.gnuweeb.org/linux- btrfs/e3261fe5-4b98-47e6-ae66-d4af05caa621@gmail.com/ Related Btrfs zoned patch discussion: https://lore-kernel.gnuweeb.org/linux- btrfs/20261110080932.1205332-1-johannes.thumshirn@wdc.com/ A broader Btrfs zoned fixes series by Dongjiang Zhu has also shown promising testing results on similar host-managed SMR hardware. The precise fix or combination of fixes required for this Ubuntu system has not yet been confirmed. **Request for Ubuntu Kernel Team** Please investigate this issue and evaluate backporting the relevant upstream Btrfs zoned fixes to the Ubuntu 26.04 LTS (Resolute) kernel through the Stable Release Update (SRU) process once the fixes have been accepted and validated upstream. I can provide complete kernel traces, filesystem usage statistics and additional read-only diagnostics. The affected filesystem contains important data without a complete backup, so testing experimental kernels or potentially destructive recovery procedures is not currently possible. ** Affects: linux (Ubuntu) Importance: Undecided Status: New ** Tags: btrfs eagain enospc smr zoned ** Attachment added: "Full Btrfs kernel error log from November 7, 2026 (Ubuntu 7.0.0-34): EAGAIN (-11) transaction abort, stack trace and forced read-only." https://bugs.launchpad.net/bugs/2174473/+attachment/6018019/+files/btrfs-error.log -- You received this bug notification because you are subscribed to linux in Ubuntu. Matching subscriptions: Bgg, Bmail, Nb https://bugs.launchpad.net/bugs/2174473 Title: Btrfs zoned: ENOSPC/EAGAIN transaction abort and forced readonly on WD host-managed SMR (Ubuntu 26.04) Status in linux package in Ubuntu: New Bug description: **Impact** Ubuntu Server 26.04 LTS experiences recurring Btrfs zoned filesystem failures on a 27 TB Western Digital host-managed SMR drive. Write operations fail despite substantial available space, and the filesystem can abort a transaction and switch to read-only. This prevents reliable use of the filesystem and presents a risk of incomplete writes or data loss. **Hardware and filesystem** - Drive: WDC WSH722870ALE604, 27 TB, host-managed SMR - Filesystem: Btrfs zoned, single-device - Zone size: 256 MiB - Filesystem UUID: 42bc4deb-1694-419f-96ea-1843a0a4f8d2 - Distribution: Ubuntu Server 26.04 LTS - Storage controller: Broadcom / LSI SAS9305-16i (SAS3224), using the mpt3sas kernel driver **Affected kernel versions** - 7.0.0-34-generic: transaction abort with EAGAIN (-11), followed by forced read-only - 7.0.0-42-generic: repeated ENOSPC (-28) errors during writes of large files, followed by filesystem failure - 7.0.0-43-generic: installed but not tested for this issue **Observed behavior** On November 7, 2026, kernel 7.0.0-34-generic reported: BTRFS error (device sdg): error while writing out transaction: -11 BTRFS warning (device sdg): Skipping commit of aborted transaction. BTRFS: Transaction aborted (error -11) BTRFS: error (device sdg state A) in cleanup_transaction:2060: errno=-11 unknown BTRFS info (device sdg state EA): forced readonly Previously, on kernel 7.0.0-42-generic, writing large media files repeatedly produced: BTRFS error (device sdg): cow_file_range failed ...: -28 BTRFS error (device sdg): failed to run delalloc range ...: -28 At the time of the November 7 failure, Btrfs reported: Device size: 24.56 TiB Device unallocated: 10.65 TiB Device zone unusable: 251.31 GiB Free (estimated): 10.86 TiB All Btrfs device error counters were zero, including read, write, flush, corruption and generation errors. Dynamic and periodic block-group reclaim were disabled. The corresponding reclaim counters were zero. **Expected behavior** Btrfs should successfully allocate space and complete transactions when sufficient usable space is available, without unexpectedly aborting transactions or forcing the filesystem into read-only mode. **Actual behavior** Write failures and transaction aborts occur intermittently during filesystem activity. The issue has been observed on multiple Ubuntu kernel versions. After the transaction abort, the filesystem becomes read-only and cannot accept further writes. **Related upstream development** I have reported the issue directly to the Linux Btrfs developers: https://lore-kernel.gnuweeb.org/linux- btrfs/e3261fe5-4b98-47e6-ae66-d4af05caa621@gmail.com/ Related Btrfs zoned patch discussion: https://lore-kernel.gnuweeb.org/linux- btrfs/20261110080932.1205332-1-johannes.thumshirn@wdc.com/ A broader Btrfs zoned fixes series by Dongjiang Zhu has also shown promising testing results on similar host-managed SMR hardware. The precise fix or combination of fixes required for this Ubuntu system has not yet been confirmed. **Request for Ubuntu Kernel Team** Please investigate this issue and evaluate backporting the relevant upstream Btrfs zoned fixes to the Ubuntu 26.04 LTS (Resolute) kernel through the Stable Release Update (SRU) process once the fixes have been accepted and validated upstream. I can provide complete kernel traces, filesystem usage statistics and additional read-only diagnostics. The affected filesystem contains important data without a complete backup, so testing experimental kernels or potentially destructive recovery procedures is not currently possible. To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2174473/+subscriptions

Комментариев нет:

Отправить комментарий