On Mon, Dec 23, 2013 at 01:35:51PM +0200, Alexander Motin wrote: > On 23.12.2013 13:14, Konstantin Belousov wrote: > > On Mon, Dec 23, 2013 at 08:43:17AM +0000, Alexander Motin wrote: > >> Author: mav > >> Date: Mon Dec 23 08:43:16 2013 > >> New Revision: 259765 > >> URL: http://svnweb.freebsd.org/changeset/base/259765 > >> > >> Log: > >> Fix RPC server threads file handle affinity to work better with ZFS. > >> > >> Instead of taking 8 specific bytes of file handle to identify file > >> during > >> RPC thread affitinity handling, use trivial hash of the full file > >> handle. > >> ZFS's struct zfid_short does not have padding field after the length > >> field, > >> as result, originally picked 8 bytes are loosing lower 16 bits of > >> object ID, > >> causing many false matches and unneeded requests affinity to same > >> thread. > >> This fix substantially improves NFS server latency and scalability in > >> SPEC > >> NFS benchmark by more flexible use of multiple NFS threads. > >> > >> Sponsored by: iXsystems, Inc. > > > > Did you audited all other filesystems to ensure that struct fid. > > fid_data0 is filled (zeroed) consistently ? > > Yes, I did a tree search and in all found cases the structure was > pre-erased. I.e. you checked all in-tree VOP_VPTOFH implementations, right ?
> > > Also, this constitues subtle VFS KBI change. > > In what way it touched VFS KBI? I indeed modified RPC FHA KPI to make it > more safe, but I don't see how it may touch anything else then two of > our NFS servers. See above, it is now required to ensure that fid_data0 is zeroed, for filesystems which only fill fid_data.
pgpB04mivGTKC.pgp
Description: PGP signature
