Messages in this thread | | | Subject | Re: 2.6.34rc4 NFS writeback regression (bisected): client often fails to delete things it just created | From | Trond Myklebust <> | Date | Sun, 18 Apr 2010 15:59:10 -0400 |
| |
On Sun, 2010-04-18 at 20:27 +0100, Nix wrote: > On 18 Apr 2010, Trond Myklebust verbalised: > > > On Sat, 2010-04-17 at 20:43 +0100, Nix wrote: > >> I suspect that unlink()ing a not otherwise open file for which writeback > >> is still underway is causing the files to be sillyrenamed because > >> writeback is holding them open. If writeback is the only user, they > >> should surely not be held open: nobody cares what their contents are, > >> and a lot of code depends on rm -r of directories containing recently- > >> written-but-still-closed files succeeding. > > > > Did you test with commit b80c3cb628f0ebc241b02e38dd028969fb8026a2 (NFS: > > Ensure that writeback_single_inode() calls write_inode() when syncing)? > > That fixed the above problem on my setup. > > tip-of-tree includes that commit, and it's still happening for me there. > (Just verified again.)
Does a 'sync' clear out the sillyrenamed files?
> (The exported filesystem is coming from a box running 2.6.33 atop ext4, > in case it matters.)
That shouldn't matter.
| |