Comment # 26 on bug 1111991 from Wenruo Qu

OK, it turns out that, it's a real corruption here.

In slot 75:

        item 75 key (10550 DIR_ITEM 108382495) itemoff 13225 itemsize 72
                location key (74333 INODE_ITEM 0) type FILE
                transid 58 data_len 0 name_len 6
                name: newg
                location key (8869339 INODE_ITEM 0) type FILE
                transid 346840 data_len 0 name_len 6
                name: newgrp

The first DIR_ITEM, name_len 6 with name: "newg\0\0".
Its hash won't match with 108382495.

And .16 kernel detects such problem, then refuse to read it.
(early kernel doesn't has such error detection, so they just let them pass)

Considering the transid, it's pretty old, so could be caused by early kernel
bugs.

In that case, we need to ensure there is only one corruption for the fs.

Please run "btrfs check --readonly" on that fs, and attach the result as a
file.

After that, we can determine how to solve the problem.

Thanks,
Qu