]> pilppa.com Git - linux-2.6-omap-h63xx.git/commit
[PATCH] knfsd: make nfsd readahead params cache SMP-friendly
authorGreg Banks <gnb@melbourne.sgi.com>
Wed, 4 Oct 2006 09:15:49 +0000 (02:15 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Wed, 4 Oct 2006 14:55:16 +0000 (07:55 -0700)
commitfce1456a19f5c08b688c29f00ef90fdfa074c79b
tree789afb4efe2764cabbd65f7d7069acb538711e8b
parent596bbe53eb3abfe7326b2f5e8afd614265c319c8
[PATCH] knfsd: make nfsd readahead params cache SMP-friendly

Make the nfsd read-ahead params cache more SMP-friendly by changing the single
global list and lock into a fixed 16-bucket hashtable with per-bucket locks.
This reduces spinlock contention in nfsd_read() on read-heavy workloads on
multiprocessor servers.

Testing was on a 4 CPU 4 NIC Altix using 4 IRIX clients each doing 1K
streaming reads at full line rate.  The server had 128 nfsd threads, which
sizes the RA cache at 256 entries, of which only a handful were used.  Flat
profiling shows nfsd_read(), including the inlined nfsd_get_raparms(), taking
10.4% of each CPU.  This patch drops the contribution from nfsd() to 1.71% for
each CPU.

Signed-off-by: Greg Banks <gnb@melbourne.sgi.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
fs/nfsd/vfs.c