The following reproducer causes a panic.
Reproducer:
-------------------------------------------------------------------------------
#!/bin/sh
echo "a" > /tmp/a # file under an ext4 filesystem
chattr +j /tmp/a
echo "a" >> /tmp/a
chattr -j /tmp/a
-------------------------------------------------------------------------------
The timing when the panic occurs is a time of executing "chattr -j".
Because "chattr -j" calls ioctl(EXT4_IOC_SETFLAGS) which executes
ext4_change_inode_journal_flag() to change the attribute of "/tmp/a".
The detailed explanation is as follows:
ext4_change_inode_journal_flag() needs to write all the unprocessed data of
this inode (/tmp/a) by jbd2_journal_flush() first. However, if an ext4
filesystem mounts with "delalloc" mode, jbd2_journal_flush() cannot completely
flush all the unprocessed delayed allocation's buffers of this inode before
calling jbd2_log_do_checkpoint().
Therefore, a panic can occur by "BUG_ON(buffer_delay(bh))" of submit_bh() via
__flush_batch() called from jbd2_log_do_checkpoint() because some unprocessed
delayed allocation's buffers may remain though jbd2_journal_flush() flushed all
transactions.
The following is an example backtrace():
-------------------------------------------------------------------------------
crash> bt
PID: 4591 TASK: ffff88007980a0c0 CPU: 2 COMMAND: "chattr"
#0 [ffff88003795f900] machine_kexec at ffffffff810356e5
#1 [ffff88003795f980] crash_kexec at ffffffff810a5c64
#2 [ffff88003795fa50] oops_end at ffffffff8149bf7b
#3 [ffff88003795fa80] die at ffffffff810163cb
#4 [ffff88003795fab0] do_trap at ffffffff8149b960
#5 [ffff88003795fb00] do_invalid_op at ffffffff810143c5
#6 [ffff88003795fba0] invalid_op at ffffffff814a4b2b
[exception RIP: submit_bh+288]
RIP: ffffffff8117e280 RSP: ffff88003795fc58 RFLAGS: 00010202
RAX: 000000000004c225 RBX: ffff880071890f00 RCX: 0000000000000002
RDX: ffff88003795ffd8 RSI: ffff880071890f00 RDI: 0000000000000211
RBP: ffff88003795fc78 R8: 0000000091827364 R9: 0000000000000000
R10: 0000000000000002 R11: 0000000000000000 R12: 0000000000000211
R13: 0000000000000211 R14: ffff88003795fca8 R15: 0000000000000004
ORIG_RAX: ffffffffffffffff CS: 0010 SS: 0018
#7 [ffff88003795fc50] __sync_dirty_buffer at ffffffff811817dd
#8 [ffff88003795fc80] write_dirty_buffer at ffffffff8117ede7
#9 [ffff88003795fca0] __flush_batch at ffffffffa01af8f8 [jbd2]
#10 [ffff88003795fd00] jbd2_log_do_checkpoint at ffffffffa01afb72 [jbd2]
#11 [ffff88003795fd60] jbd2_journal_flush at ffffffffa01b3e2f [jbd2]
#12 [ffff88003795fda0] ext4_change_inode_journal_flag at ffffffffa01d0ad3 [ext4]
#13 [ffff88003795fdd0] ext4_ioctl at ffffffffa01d66b6 [ext4]
#14 [ffff88003795fea0] vfs_ioctl at ffffffff81162b1d
#15 [ffff88003795feb0] do_vfs_ioctl at ffffffff81163145
#16 [ffff88003795ff30] sys_ioctl at ffffffff81163644
#17 [ffff88003795ff80] system_call_fastpath at ffffffff814a2b42
RIP: 00000034930d95f7 RSP: 00007fff656ceae8 RFLAGS: 00010213
RAX: 0000000000000010 RBX: ffffffff814a2b42 RCX: 00000034930d95f7
RDX: 00007fff656ceb8c RSI: 0000000040086602 RDI: 0000000000000003
RBP: 0000000000080000 R8: 0000000000000000 R9: 0000000000000000
R10: 0000000000000001 R11: 0000000000000246 R12: 0000000000004000
R13: 0000000000000001 R14: 0000000000000000 R15: 0000000000000003
ORIG_RAX: 0000000000000010 CS: 0033 SS: 002b
-------------------------------------------------------------------------------
Therefore, we must flush all unprocessed delayed allocation's buffers and turn
"delalloc" mode off before jbd2_journal_flush() calls. Then, we must turn
"delalloc" mode on after jbd2_journal_flush() finishes.
Signed-off-by: Toshiyuki Okajima <toshi.okajima@xxxxxxxxxxxxxx>
---
fs/ext4/ext4.h | 3 +++
fs/ext4/inode.c | 18 ++++++++++++++++++
2 files changed, 21 insertions(+), 0 deletions(-)
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 5b0e26a..3eeb68b 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1249,6 +1249,9 @@ struct ext4_sb_info {
/* record the last minlen when FITRIM is called. */
atomic_t s_last_trim_minblks;
+
+ /* to sync the delayed allocation's buffers */
+ atomic_t s_need_da_sync;
};
static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 848f436..8d317a3 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -2368,6 +2368,14 @@ static int ext4_nonda_switch(struct super_block *sb)
*/
return 1;
}
+
+ /*
+ * Second case to switch to non dealloc mode, if we want to write all the ext4
+ * data out.
+ */
+ if (atomic_read(&sbi->s_need_da_sync) > 0)
+ return 1;
+
/*
* Even if we don't switch but are nearing capacity,
* start pushing delalloc when 1/2 of free blocks are dirty.
@@ -4665,6 +4673,8 @@ int ext4_change_inode_journal_flag(struct inode *inode, int val)
journal_t *journal;
handle_t *handle;
int err;
+ struct super_block *sb = inode->i_sb;
+ struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
/*
* We have to be very careful here: changing a data block's
@@ -4682,6 +4692,13 @@ int ext4_change_inode_journal_flag(struct inode *inode, int val)
if (is_journal_aborted(journal))
return -EROFS;
+ /*
+ * To write all delayed allocation's buffers out before calling
+ * jbd2_journal_flush(). Because it cannot flush the buffers.
+ */
+ atomic_inc(&sbi->s_need_da_sync);
+ sync_filesystem(sb);
+
jbd2_journal_lock_updates(journal);
jbd2_journal_flush(journal);
@@ -4700,6 +4717,7 @@ int ext4_change_inode_journal_flag(struct inode *inode, int val)
ext4_set_aops(inode);
jbd2_journal_unlock_updates(journal);
+ atomic_dec(&sbi->s_need_da_sync);
/* Finally we can mark the inode as dirty. */
--
1.5.5.6
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
