The original code has only one spin_lock 'qgroup_lock' to protect quota
configurations on memory. If we want to add a BTRFS_QGROUP_INFO_KEY,
it will be added to Btree firstly and then update quota configurations
on memory, however,a race condition may happen between these operations.
For example:
->add_qqroup_info_item()
->add_qgroup_rb()
For the above case, del_qgroup_info_item() may happen before add_qgroup_rb().
What's worse, when we want to add a qgroup relations:
->add_qgroup_relation_item()
->add_qgroup_relations()
We don't have any checks whether 'src' and 'dst' exists before
add_qgroup_relation_item(), a race condition can also happen for the above case.
To avoid race conditions and have all the necessay checks, we introduce
a mutex 'quota_lock', and we make all the user change operations protected by
the mutex_lock.
The benefit of mutex_lock is more, with a mutex lock, we can remove
some spin_lock usages thus easing the burden of spin_lock 'qgroup_lock'.
V1->V2:
use quota configurations on memory to speed up ioctl checks
Wang Shilong (6):
Btrfs: introduce a mutex lock for btrfs quota operations
Btrfs: remove some unnecessary spin_lock usages
Btrfs: fix missing check before updating qgroup limit
Btrfs: fix missing check before creating/destroying a qgroup
Btrfs: fix missing check before assigning/removing qgroup relation
Btrfs: fix missing check before btrfs_qgroup_inherit()
fs/btrfs/ctree.h | 10 +++
fs/btrfs/disk-io.c | 1 +
fs/btrfs/ioctl.c | 43 ++++++++++----
fs/btrfs/qgroup.c | 167 +++++++++++++++++++++++++++++++++-------------------
4 files changed, 150 insertions(+), 71 deletions(-)
--
1.7.7.6
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
