xfs: atomic drop extent entries when inactiving attr

Offering: HULK
hulk inclusion
category: bugfix
bugzilla: https://gitee.com/openeuler/kernel/issues/I76JSK



--------------------------------

When inactiving an unlinked inode and it's attrs, if xlog is shutdown
either during or just after the process of recurse deleting attribute
nodes/leafs in xfs_attr3_root_inactive(), the log will records some
buffer cancel items, but doesn't contain the corresponding extent
entries and inode updates, this is incomplete and inconsistent. Because
of the inactiving process is not completed and the unlinked inode is
still in the agi_unlinked table, it will continue to be inactived after
replaying the log on the next mount, the attr node/leaf blocks' created
record before the cancel items could not be replayed but the inode
does. So we could get corrupted data when reading the canceled blocks.

 XFS (pmem0): Metadata corruption detected at
 xfs_da3_node_read_verify+0x53/0x220, xfs_da3_node block 0x78
 XFS (pmem0): Unmount and run xfs_repair
 XFS (pmem0): First 128 bytes of corrupted metadata buffer:
 00000000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
 00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
 00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
 00000030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
 00000040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
 00000050: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
 00000060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
 00000070: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
 XFS (pmem0): metadata I/O error in "xfs_da_read_buf+0x104/0x190" at daddr 0x78 len 8 error 117

In order to fix the issue, we need to remove the extent entries, update
inode and attr btree atomically when staling attr node/leaf blocks. And
note that we may also need to log and update the parent attr node
entry when removing child or leaf attr block. Fortunately, it doesn't
have to be so complicated, we could leave the removed entres as
holes and skip them if we need to do re-inactiving, the whole node tree
will be removed completely in the end.

Cc: stable@vger.kernel.org
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Signed-off-by: Long Li <leo.lilong@huawei.com>