From: robbieko <robbieko@synology.com>
To: linux-btrfs@vger.kernel.org
Cc: Robbie Ko <robbieko@synology.com>
Subject: [PATCH v2 0/3] btrfs: fix unexpected -ENOMEM with percpu_counter_init when create snapshot
Date: Wed, 14 Dec 2022 10:11:22 +0800 [thread overview]
Message-ID: <20221214021125.28289-1-robbieko@synology.com> (raw)
From: Robbie Ko <robbieko@synology.com>
[Issue]
When creating subvolume/snapshot, the transaction may be abort due to -ENOMEM
WARNING: CPU: 1 PID: 411 at fs/btrfs/transaction.c:1937 create_pending_snapshot+0xe30/0xe70 [btrfs]()
CPU: 1 PID: 411 Comm: btrfs Tainted: P O 4.4.180+ #42661
Call Trace:
create_pending_snapshot+0xe30/0xe70 [btrfs]
create_pending_snapshots+0x89/0xb0 [btrfs]
btrfs_commit_transaction+0x469/0xc60 [btrfs]
btrfs_mksubvol+0x5bd/0x690 [btrfs]
btrfs_mksnapshot+0x102/0x170 [btrfs]
btrfs_ioctl_snap_create_transid+0x1ad/0x1c0 [btrfs]
btrfs_ioctl_snap_create_v2+0x102/0x160 [btrfs]
btrfs_ioctl+0x2111/0x3130 [btrfs]
do_vfs_ioctl+0x7ea/0xa80
SyS_ioctl+0xa1/0xb0
entry_SYSCALL_64_fastpath+0x1e/0x8e
---[ end trace 910c8f86780ca385 ]---
BTRFS: error (device dm-2) in create_pending_snapshot:1937: errno=-12 Out of memory
[Cause]
During creating a subvolume/snapshot, it is necessary to allocate memory for Initializing fs root.
Therefore, it can only use GFP_NOFS to allocate memory to avoid deadlock issues.
However, atomic allocation is required when processing percpu_counter_init
without GFP_KERNEL due to the unique structure of percpu_counter.
In this situation, allocating memory for initializing fs root may cause
unexpected -ENOMEM when free memory is low and causes btrfs transaction to abort.
[Fix]
We allocate memory at the beginning of creating a subvolume/snapshot.
This way, we can ensure the memory is enough when initializing fs root.
Even -ENOMEM happens at the beginning of creating a subvolume/snapshot,
the transaction won’t abort since it hasn’t started yet.
Robbie Ko (3):
btrfs: refactor anon_dev with new_fs_root_args for create
subvolume/snapshot
btrfs: change snapshot_lock to dynamic pointer
btrfs: add snapshot_lock to new_fs_root_args
fs/btrfs/ctree.h | 2 +-
fs/btrfs/disk-io.c | 107 ++++++++++++++++++++++++++++++-----------
fs/btrfs/disk-io.h | 12 ++++-
fs/btrfs/file.c | 8 +--
fs/btrfs/inode.c | 12 ++---
fs/btrfs/ioctl.c | 38 +++++++--------
fs/btrfs/transaction.c | 2 +-
fs/btrfs/transaction.h | 5 +-
8 files changed, 123 insertions(+), 63 deletions(-)
--
2.17.1
next reply other threads:[~2022-12-14 2:11 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-12-14 2:11 robbieko [this message]
2022-12-14 2:11 ` [PATCH v2 1/3] btrfs: refactor anon_dev with new_fs_root_args for create subvolume/snapshot robbieko
2022-12-14 2:11 ` [PATCH v2 2/3] btrfs: change snapshot_lock to dynamic pointer robbieko
2022-12-14 2:11 ` [PATCH v2 3/3] btrfs: add snapshot_lock to new_fs_root_args robbieko
2022-12-14 16:59 ` [PATCH v2 0/3] btrfs: fix unexpected -ENOMEM with percpu_counter_init when create snapshot Josef Bacik
2022-12-14 18:07 ` David Sterba
2022-12-19 6:54 ` robbieko
2023-01-10 15:08 ` David Sterba
2023-03-09 18:47 ` David Sterba
2023-03-10 1:07 ` robbieko
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=20221214021125.28289-1-robbieko@synology.com \
--to=robbieko@synology.com \
--cc=linux-btrfs@vger.kernel.org \
/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;
as well as URLs for NNTP newsgroup(s).