>> This makes it all the more important that the "fast symlink" checking be
>> divorced from i_blocks, and instead depend on i_size. Otherwise, we will
>> have yet another bunch of problems as large xattrs on a fast symlink will
>> incorrectly result in problems with the symlink.
>>
> Totally agree. The changes to i_blocks accounting for ea_inode feature
> likely broke ext4_inode_is_fast_symlink() in kernel. I will try to fix
> that first.
FYI, I sent out a patch to fix e2fsprogs part too: ("[PATCH]
e2fsprogs: add ext2fs_is_fast_symlink() function")
