f2fs: give a chance to mount again when encountering errors

This patch gives another chance to try mount process when we encounter an error.
This makes an effect on the roll-forward recovery failures as well.

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

diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 7a5477915d993f4d48b3dbd30e0e9717bd6a97da..e161e13958e28bd86d6447dce9056f721814933f 100644 (file)
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -902,8 +902,10 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
        struct buffer_head *raw_super_buf;
        struct inode *root;
        long err = -EINVAL;
+       bool retry = true;
        int i;
 
+try_onemore:
        /* allocate memory for f2fs-specific super block info */
        sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
        if (!sbi)
@@ -1083,9 +1085,11 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
        /* recover fsynced data */
        if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
                err = recover_fsync_data(sbi);
-               if (err)
+               if (err) {
                        f2fs_msg(sb, KERN_ERR,
                                "Cannot recover all fsync data errno=%ld", err);
+                       goto free_kobj;
+               }
        }
 
        /*
@@ -1126,6 +1130,13 @@ free_sb_buf:
        brelse(raw_super_buf);
 free_sbi:
        kfree(sbi);
+
+       /* give only one another chance */
+       if (retry) {
+               retry = !retry;
+               shrink_dcache_sb(sb);
+               goto try_onemore;
+       }
        return err;
 }