Re: [PATCH v3 1/2] ocfs2: snapshot dentry names in file-operation traces

[Cen Zhang]

Hi Joseph,

> Seems ocfs2_setattr() also risks the same race, and it is missed in last
> review comments.

Thanks for catching this. I've added a name snapshot around
trace_ocfs2_setattr(), passing both the name and length from it.

The inode lock does not exclude a same-directory rename of a source
directory, so this trace also needs the snapshot.

> This looks wried, the inode is dereferenced above in OCFS2_I(inode)->ip_blkno,
> but checked below.
> file_inode() returns filp->f_inode, so it seems inode is never NULL here.

Agreed. I've removed the redundant NULL-inode check from
ocfs2_file_read_iter().

Both changes are included in v4.

Best regards,
Cen Zhang