Re: [PATCH] vfs: check dentry is still valid in get_link()

[Christian Brauner]

On Mon, Jan 17, 2022 at 06:10:36PM +0000, Al Viro wrote:
> On Mon, Jan 17, 2022 at 04:28:52PM +0000, Al Viro wrote:
> 
> > IOW, ->free_inode() is RCU-delayed part of ->destroy_inode().  If both
> > are present, ->destroy_inode() will be called synchronously, followed
> > by ->free_inode() from RCU callback, so you can have both - moving just
> > the "finally mark for reuse" part into ->free_inode() would be OK.
> > Any blocking stuff (if any) can be left in ->destroy_inode()...
> 
> BTW, we *do* have a problem with ext4 fast symlinks.  Pathwalk assumes that
> strings it parses are not changing under it.  There are rather delicate
> dances in dcache lookups re possibility of ->d_name contents changing under
> it, but the search key is assumed to be stable.
> 
> What's more, there's a correctness issue even if we do not oops.  Currently
> we do not recheck ->d_seq of symlink dentry when we dismiss the symlink from
> the stack.  After all, we'd just finished traversing what used to be the
> contents of a symlink that used to be in the right place.  It might have been
> unlinked while we'd been traversing it, but that's not a correctness issue.
> 
> But that critically depends upon the contents not getting mangled.  If it
> *can* be screwed by such unlink, we risk successful lookup leading to the

Out of curiosity: whether or not it can get mangled depends on the
filesystem and how it implements fast symlinks or do fast symlinks
currently guarantee that contents are mangled?