From: Ryusuke Konishi <konishi.ryusuke@gmail.com>
To: Zheng Zheng <fengaojigui88@gmail.com>
Cc: slava@dubeyko.com, linux-nilfs@vger.kernel.org,
linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
syzkaller-bugs@googlegroups.com
Subject: Re: [BUG] INFO: task hung in nilfs_transaction_begin (fs/nilfs2/segment.c:221, v7.4-rc3)
Date: Fri, 20 Nov 2026 20:24:39 +0900 [thread overview]
Message-ID: <CAKFNMokyeDzig1aJMjTN0nid4DGdnFPKC9G-xk3L30yMikgpRQ@mail.gmail.com> (raw)
In-Reply-To: <CAO-4RZ2pjEOpTBzqca5UxZHtaFS2Dj9zmNj1BtjvKm1yFE7_Jg@mail.gmail.com>
On Fri, Nov 20, 2026 at 7:59 PM Zheng Zheng wrote:
>
> Hi Ryusuke, Viacheslav,
>
> We found a reproducible NILFS2 hang on v7.4-rc3 with syzkaller. The
> supplied recipe reproduced on independent fresh VMs.
>
> Note: This hang may share the same underlying root cause as the
> previously reported nilfs_segctor_thread hang (issue #182):
> https://github.com/nilfs-dev/nilfs2/issues/182
>
> Repository and reproducer:
>
> https://github.com/Zheng-zheng-codes/nilfs2-transaction-begin-hang
> commit f1065fd59515bd55bbb5141b45e979bc0b3c9993
>
> Hung-task report
> ----------------
>
> INFO: task syz.7.24:13081 blocked for more than 142 seconds.
> task:syz.7.24 state:D stack:23176 pid:13081 tgid:13075 ppid:10344
> Call Trace:
> rwsem_down_read_slowpath
> __down_read
> down_read
> nilfs_transaction_begin+0x323/0xa00 fs/nilfs2/segment.c:221
> nilfs_write_begin+0x78/0x1b0 fs/nilfs2/inode.c:228
> generic_perform_write
> vfs_write
> ksys_write
>
> A second thread of the same process (PID 13109) is blocked on the
> f_pos_lock held by PID 13081.
>
> An NMI backtrace of the semaphore holder (PID 13190) shows that it is
> executing the NILFS2 segment-constructor path through
> nilfs_clean_segments(), including btree/bmap propagation:
>
> nilfs_segctor_apply_buffers fs/nilfs2/segment.c:1010
> nilfs_collect_file_data fs/nilfs2/segment.c:589
> nilfs_bmap_propagate fs/nilfs2/bmap.c:329
> nilfs_btree_propagate fs/nilfs2/btree.c:2128
>
> The lock dump identifies the semaphore holder:
>
> locks held by syz.7.24/13190: 3:
> #0: (sb_writers#17) at: nilfs_ioctl_clean_segments
> fs/nilfs2/ioctl.c:869
> #1: (&nilfs->ns_segctor_sem) WRITE at: nilfs_transaction_lock
> fs/nilfs2/segment.c:357
> #2: (&bmap->b_sem) at: nilfs_bmap_assign
> fs/nilfs2/bmap.c:373
>
> So a NILFS_IOCTL_CLEAN_SEGMENTS task is inside the segment-constructor
> path, holding ns_segctor_sem for writing while doing bmap assignment,
> and normal writers cannot acquire the semaphore for reading. In this
> report the segctord thread itself is not among the blocked tasks.
>
> Relation to the earlier nilfs_segctor_thread hang
> -------------------------------------------------
>
> We previously reported a related NILFS2 hang:
>
> INFO: task hung in nilfs_segctor_thread (fs/nilfs2/segment.c:357, v7.4-rc3)
> https://github.com/nilfs-dev/nilfs2/issues/182
> https://github.com/Zheng-zheng-codes/nilfs2-segctor-thread-hang
>
> Its hung-task report showed the constructor thread blocked acquiring
> the same semaphore for writing:
>
> INFO: task segctord:35823 blocked for more than 183 seconds.
> task:segctord state:D
> Call Trace:
> rwsem_down_write_slowpath
> __down_write
> down_write
> nilfs_transaction_lock+0x25d/0x410 fs/nilfs2/segment.c:357
> nilfs_segctor_thread_construct fs/nilfs2/segment.c:2601 [inline]
> nilfs_segctor_thread+0x600/0xa60 fs/nilfs2/segment.c:2716
> kthread
>
> The two reports involve different blocked tasks:
>
> - In the earlier report, the segctord kernel thread waits
> for ns_segctor_sem in write mode in nilfs_transaction_lock().
>
> - In this report, a normal file writer waits for the same
> semaphore in read mode in nilfs_transaction_begin().
>
> In both reports, the lock dump identifies a task executing
> NILFS_IOCTL_CLEAN_SEGMENTS as the semaphore holder.
>
> The new external verification logs also repeatedly report:
>
> NILFS (loop0): error -28 cleaning segments
>
> This resembles the repeated -ENOSPC errors observed in the
> earlier report.
>
> We suspect that both hangs may be manifestations of the same
> GC error-handling problem, but the root cause has not yet
> been established.
>
> We would appreciate your analysis of this additional
> reproduction alongside issue #182.
>
> Reproduction
> ------------
>
> git clone https://github.com/Zheng-zheng-codes/nilfs2-transaction-begin-hang
> cd nilfs2-transaction-begin-hang
>
> # boot v7.4-rc3 with the supplied kernel config
> # using a Debian trixie x86_64 guest
>
> # in the guest, as root:
> syz-execprog -procs=1 -threaded -repeat=1 \
> -executor ./syz-executor \
> reproducer/repro.kernsight.prog
>
> The recipe requires sequential execution with -procs=1.
> Using -procs=8 did not reproduce the hang under our tests.
>
> The recipe is a state-reconstruction sequence (72 replay programs plus
> one model-selected priming program, followed by the crashing testcase).
> The crash depends on the kernel state built by the recipe; the testcase
> alone is not sufficient.
>
> Verification:
>
> - reproduction run: two attempts verified (confirmed=true,
> status=verified), 2026-11-20T07:17:00Z and 2026-11-20T08:05:31Z
> - independent fresh-VM verification inside that run: 2/2 target hits
> (2026-11-20T07:46:46Z and 07:50:15Z)
> - stability on fresh VMs: 5/5 reproductions, 187-194 s to the
> hung-task dump in every run (2026-11-20, external-verify/)
> - evidence: reproducer/repro.outcomes.jsonl, external-verify/,
> logs/external-manager-session.log
> - C reproducer: not generated; the syzkaller recipe is the primary
> reproducer
>
> Kernel/guest details:
>
> kernel: v7.4-rc3, bzImage sha256
> 0c16409f8aec508ac44a63e631c2bd352536401f3fd0856df1d839fb953a1882
> guest: Debian trixie x86_64, 6 GiB RAM, 4 vCPU, QEMU/KVM
> boot: oops=panic panic_on_warn=1 hung_task_panic=0
> config: CONFIG_KASAN=y, CONFIG_PROVE_LOCKING=y, CONFIG_NILFS2_FS=y
>
> Thanks,
> Zheng
Hi Zheng Zheng,
I have filed it separately at:
https://github.com/nilfs-dev/nilfs2/issues/183
Since these GC ioctl reports involve destructive corner cases that do
not occur under normal userland usage, we are addressing them
step-by-step at a pace that does not disrupt other tasks.
As these reports seem to stem from the same underlying error-handling
issues that I am about to investigate, it would be helpful if you
could hold off on further reports until the initial fixes land. That
will help avoid duplicates and allow us to process the reports more
effectively.
Thanks,
Ryusuke Konishi
next prev parent reply other threads:[~2026-11-20 11:25 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-20 10:59 Zheng Zheng
2026-11-20 11:24 ` Ryusuke Konishi [this message]
2026-11-20 11:40 ` Zheng Zheng
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=CAKFNMokyeDzig1aJMjTN0nid4DGdnFPKC9G-xk3L30yMikgpRQ@mail.gmail.com \
--to=konishi.ryusuke@gmail.com \
--cc=fengaojigui88@gmail.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-nilfs@vger.kernel.org \
--cc=slava@dubeyko.com \
--cc=syzkaller-bugs@googlegroups.com \
/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
all inboxes | Powered by JetHome®