f2fs: fix f2fs_bug_on when uninstalling filesystem call f2fs_evict_in…

…ode.

creating a large files during checkpoint disable until it runs out of
space and then delete it, then remount to enable checkpoint again, and
then unmount the filesystem triggers the f2fs_bug_on as below:

------------[ cut here ]------------
kernel BUG at fs/f2fs/inode.c:896!
CPU: 2 UID: 0 PID: 1286 Comm: umount Not tainted 6.11.0-rc7-dirty #360
Oops: invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
RIP: 0010:f2fs_evict_inode+0x58c/0x610
Call Trace:
 __die_body+0x15/0x60
 die+0x33/0x50
 do_trap+0x10a/0x120
 f2fs_evict_inode+0x58c/0x610
 do_error_trap+0x60/0x80
 f2fs_evict_inode+0x58c/0x610
 exc_invalid_op+0x53/0x60
 f2fs_evict_inode+0x58c/0x610
 asm_exc_invalid_op+0x16/0x20
 f2fs_evict_inode+0x58c/0x610
 evict+0x101/0x260
 dispose_list+0x30/0x50
 evict_inodes+0x140/0x190
 generic_shutdown_super+0x2f/0x150
 kill_block_super+0x11/0x40
 kill_f2fs_super+0x7d/0x140
 deactivate_locked_super+0x2a/0x70
 cleanup_mnt+0xb3/0x140
 task_work_run+0x61/0x90

The root cause is: creating large files during disable checkpoint
period results in not enough free segments, so when writing back root
inode will failed in f2fs_enable_checkpoint. When umount the file
system after enabling checkpoint, the root inode is dirty in
f2fs_evict_inode function, which triggers BUG_ON. The steps to
reproduce are as follows:

dd if=/dev/zero of=f2fs.img bs=1M count=55
mount f2fs.img f2fs_dir -o checkpoint=disable:10%
dd if=/dev/zero of=big bs=1M count=50
sync
rm big
mount -o remount,checkpoint=enable f2fs_dir
umount f2fs_dir

Let's redirty inode when there is not free segments during checkpoint
is disable.

Signed-off-by: Qi Han <[email protected]>
Reviewed-by: Chao Yu <[email protected]>
Signed-off-by: Jaegeuk Kim <[email protected]>

fs/f2fs/inode.c

@@ -775,8 +775,10 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
 		!is_inode_flag_set(inode, FI_DIRTY_INODE))
 		return 0;
 
-	if (!f2fs_is_checkpoint_ready(sbi))
+	if (!f2fs_is_checkpoint_ready(sbi)) {
+		f2fs_mark_inode_dirty_sync(inode, true);
 		return -ENOSPC;
+	}
 
 	/*
 	 * We need to balance fs here to prevent from producing dirty node pages