Btrfs fixes
From: Filipe Manana <[email protected]>
Hi Chris,
please consider the following fixes for the 4.4 merge window (they were all previously sent to the mailing list already).
One fixes a sleep inside atomic context issue, introduced by one patch in the integration-4.4 branch. Another two fix races regarding waiting for qgroup rescan worker to finish and a race between the qgroup rescan worker and unmounting the filesystem (leading to crashes). The remaining patch fixes an issue with partial direct IO writes, which has been introduced in the 4.0 kernel, and results either in an assertion failure (BUG_ON) when CONFIG_BTRFS_ASSERT=y or arithmetic underflow of an inode's outstanding extents counter (used for proper space reservation) when assertions are disabled.
Two test cases for fstests were sent recently to cover the issues regarding the races and the direct IO partial write regression.
Thanks.
The following changes since commit 2959a32a858a2c44bbbce83d19c158d54cc5998a:
Btrfs: fix hole punching when using the no-holes feature (2015-11-03 07:44:20 -0800)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/fdmanana/linux.git integration-4.4
for you to fetch changes up to 3b2ba7b31d56c3d8f57cd5d32b8fb5101ab446e4:
Btrfs: fix sleeping inside atomic context in qgroup rescan worker (2015-11-05 11:02:22 +0000)
----------------------------------------------------------------
Filipe Manana (3):
Btrfs: fix extent accounting for partial direct IO writes
Btrfs: fix race waiting for qgroup rescan worker
Btrfs: fix sleeping inside atomic context in qgroup rescan worker
Justin Maggard (1):
btrfs: qgroup: exit the rescan worker during umount
fs/btrfs/disk-io.c | 3 +++ fs/btrfs/inode.c | 52 +++++++++++++++++++++++++++++++++++++--------------- fs/btrfs/qgroup.c | 13 +++++++------ 3 files changed, 47 insertions(+), 21 deletions(-)
Signature · 1 line
2.1.3
Quoted 17 lines · On Thu, Nov 05, 2015 at 11:20:37AM +0000, [email protected] wrote:
From: Filipe Manana <[email protected]>
Hi Chris,
please consider the following fixes for the 4.4 merge window (they were all previously sent to the mailing list already).
One fixes a sleep inside atomic context issue, introduced by one patch in the integration-4.4 branch. Another two fix races regarding waiting for qgroup rescan worker to finish and a race between the qgroup rescan worker and unmounting the filesystem (leading to crashes). The remaining patch fixes an issue with partial direct IO writes, which has been introduced in the 4.0 kernel, and results either in an assertion failure (
BUG_ON) whenCONFIG_BTRFS_ASSERT=yor arithmetic underflow of an inode's outstanding extents counter (used for proper space reservation) when assertions are disabled.
Quoted 2 lines
Two test cases for fstests were sent recently to cover the issues regarding the races and the direct IO partial write regression.
Great, thanks Filipe.
I'll send these for my second pull (probably next Wed).
-chris