Messages in this thread | | | Date | Mon, 23 Jul 2012 10:03:04 +0100 | From | Al Viro <> | Subject | Re: [git pull] vfs.git; pile 1 |
| |
On Mon, Jul 23, 2012 at 11:20:25AM +0300, Boaz Harrosh wrote: > On 07/22/2012 11:20 PM, Al Viro wrote: > > > I think the least painful solution is this: I've created a new branch > > (for-linus-2) in there, growing off the parent of merge in nfs.git. > > I've put the fixup to kern_path_locked() there as a separate commit > > + stuff that went in for-linus after that point. > > > > > Does this means that between "kern_path_locked() patch" and it's fixup > the tree is not bisectable? or is it just an enhancement/simplification? > > Not having any opinion, just wants to know
It is bisectable, all right; lookup_one_len() is OK there (note that places converted to kern_path_locked() used to use it before the conversion. It's just that lookup_one_len() is an overkill - we have already checked exec permissions on parent and we'd already calculated the full qstr for last component - both length and hash. All that is left to do is actual __lookup_hash()...
| |