> > Hmm, this may hold dcache_lock for some time. Isn't it enough to use
> > rcu_readlock() and read_seqbegin(&rename_lock) ?
>
> Maybe. I just blindly copied the d_path() code without thinking much.
> Is this worth optimizing? I mean, getting RCU wrong is easy, and it
> makes my head hurt to think about it. And dcache_lock shouldn't be
> used in any performance critical code paths anyway.
But I'm an idiot. This is a much simpler and faster implementation
that uses is_subdir(), which essentially does what you suggested:
/*
* Return true if 'path' is reachable from 'root'
*/
static bool is_path_reachable(const struct path *path, const struct path *root)
{
struct dentry *dentry = path->dentry;
struct vfsmount *mnt = path->mnt;
bool res = false;
spin_lock(&vfsmount_lock);
while (mnt != root->mnt && mnt->mnt_parent != mnt) {
dentry = mnt->mnt_mountpoint;
mnt = mnt->mnt_parent;
}
if (mnt == root->mnt && is_subdir(dentry, root->dentry))
res = true;
spin_unlock(&vfsmount_lock);
return res;
}
Miklos
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
