When running parallel I/O threads on machines with lots of
CPUs, the incore superblock lock becomes the limiting factor
for buffered write throughput. Make the contended fields in
the incore superblock use per-cpu counters so that there is
no global lock to limit scalability.
Date: Fri Feb 3 17:14:33 AEDT 2006
Workarea: chook.melbourne.sgi.com:/build/dgc/isms/2.6.x-xfs
Inspected by: nathans
The following file(s) were checked into:
longdrop.melbourne.sgi.com:/isms/linux/2.6.x-xfs-melb
Modid: xfs-linux-melb:xfs-kern:25106a
fs/xfs/xfs_vfsops.c - 1.494 - changed
http://oss.sgi.com/cgi-bin/cvsweb.cgi/xfs-linux/xfs_vfsops.c.diff?r1=text&tr1=1.494&r2=text&tr2=1.493&f=h
- Aggregate per-cpu sb counters before reporting current state
to userspace.
fs/xfs/xfs_mount.h - 1.214 - changed
http://oss.sgi.com/cgi-bin/cvsweb.cgi/xfs-linux/xfs_mount.h.diff?r1=text&tr1=1.214&r2=text&tr2=1.213&f=h
- Introduce per-cpu superblock counters.
fs/xfs/xfs_mount.c - 1.370 - changed
http://oss.sgi.com/cgi-bin/cvsweb.cgi/xfs-linux/xfs_mount.c.diff?r1=text&tr1=1.370&r2=text&tr2=1.369&f=h
- Introduce per-cpu superblock counters.
fs/xfs/xfs_fsops.c - 1.112 - changed
http://oss.sgi.com/cgi-bin/cvsweb.cgi/xfs-linux/xfs_fsops.c.diff?r1=text&tr1=1.112&r2=text&tr2=1.111&f=h
- Aggregate per-cpu sb counters before reporting current state
to userspace.
fs/xfs/linux-2.6/xfs_linux.h - 1.139 - changed
http://oss.sgi.com/cgi-bin/cvsweb.cgi/xfs-linux/linux-2.6/xfs_linux.h.diff?r1=text&tr1=1.139&r2=text&tr2=1.138&f=h
- Enable per-cpu sb counters for multi-processor configurations.
fs/xfs/linux-2.4/xfs_linux.h - 1.151 - changed
http://oss.sgi.com/cgi-bin/cvsweb.cgi/xfs-linux/linux-2.4/xfs_linux.h.diff?r1=text&tr1=1.151&r2=text&tr2=1.150&f=h
- Ensure per-cpu sb counters are not enabled.
|