f2fs: avoid data race when deciding checkpoin in f2fs_sync_file

When fs utilization is almost full, f2fs_sync_file should do checkpoint if
there is not enough space for roll-forward later. (i.e. space_for_roll_forward)
So, currently we have no lock for sbi->alloc_valid_block_count, resulting in
race condition.

In rare case, we can get -ENOSPC when doing roll-forward which triggers

	if (is_valid_blkaddr(sbi, dest, META_POR)) {
		if (src == NULL_ADDR) {
			err = reserve_new_block(&dn);
			f2fs_bug_on(sbi, err);
			...
		}
		...
	}
in do_recover_data.

So, this patch avoids that situation in advance.

Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 7a57279b2c54425b8109f3ab3de68a0c1a09aadb..30981094dff84ee1d8f57016d0d8ea8423bfad3b 100644 (file)
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -1147,24 +1147,33 @@ static inline void f2fs_i_blocks_write(struct inode *, blkcnt_t, bool);
 static inline bool inc_valid_block_count(struct f2fs_sb_info *sbi,
                                 struct inode *inode, blkcnt_t *count)
 {
+       blkcnt_t diff;
+
 #ifdef CONFIG_F2FS_FAULT_INJECTION
        if (time_to_inject(FAULT_BLOCK))
                return false;
 #endif
+       /*
+        * let's increase this in prior to actual block count change in order
+        * for f2fs_sync_file to avoid data races when deciding checkpoint.
+        */
+       percpu_counter_add(&sbi->alloc_valid_block_count, (*count));
+
        spin_lock(&sbi->stat_lock);
        sbi->total_valid_block_count += (block_t)(*count);
        if (unlikely(sbi->total_valid_block_count > sbi->user_block_count)) {
-               *count -= sbi->total_valid_block_count - sbi->user_block_count;
+               diff = sbi->total_valid_block_count - sbi->user_block_count;
+               *count -= diff;
                sbi->total_valid_block_count = sbi->user_block_count;
                if (!*count) {
                        spin_unlock(&sbi->stat_lock);
+                       percpu_counter_sub(&sbi->alloc_valid_block_count, diff);
                        return false;
                }
        }
        spin_unlock(&sbi->stat_lock);
 
        f2fs_i_blocks_write(inode, *count, true);
-       percpu_counter_add(&sbi->alloc_valid_block_count, (*count));
        return true;
 }