linux/fs/proc
Nick Piggin fa0d7e3de6 fs: icache RCU free inodes
RCU free the struct inode. This will allow:

- Subsequent store-free path walking patch. The inode must be consulted for
  permissions when walking, so an RCU inode reference is a must.
- sb_inode_list_lock to be moved inside i_lock because sb list walkers who want
  to take i_lock no longer need to take sb_inode_list_lock to walk the list in
  the first place. This will simplify and optimize locking.
- Could remove some nested trylock loops in dcache code
- Could potentially simplify things a bit in VM land. Do not need to take the
  page lock to follow page->mapping.

The downsides of this is the performance cost of using RCU. In a simple
creat/unlink microbenchmark, performance drops by about 10% due to inability to
reuse cache-hot slab objects. As iterations increase and RCU freeing starts
kicking over, this increases to about 20%.

In cases where inode lifetimes are longer (ie. many inodes may be allocated
during the average life span of a single inode), a lot of this cache reuse is
not applicable, so the regression caused by this patch is smaller.

The cache-hot regression could largely be avoided by using SLAB_DESTROY_BY_RCU,
however this adds some complexity to list walking and store-free path walking,
so I prefer to implement this at a later date, if it is shown to be a win in
real situations. I haven't found a regression in any non-micro benchmark so I
doubt it will be a problem.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>
2011-01-07 17:50:26 +11:00
..
array.c CRED: Fix get_task_cred() and task_state() to not resurrect dead credentials 2010-07-29 15:16:17 -07:00
base.c fs: change d_delete semantics 2011-01-07 17:50:18 +11:00
cmdline.c proc: switch /proc/cmdline to seq_file 2008-10-23 14:29:04 +04:00
cpuinfo.c proc: move /proc/cpuinfo code to fs/proc/cpuinfo.c 2008-10-23 15:05:11 +04:00
devices.c proc: move /proc/devices code to fs/proc/devices.c 2008-10-23 15:02:18 +04:00
generic.c fs: change d_delete semantics 2011-01-07 17:50:18 +11:00
inode.c fs: icache RCU free inodes 2011-01-07 17:50:26 +11:00
internal.h proc: rename de_get() to pde_get() and inline it 2009-12-16 07:19:57 -08:00
interrupts.c proc: move /proc/interrupts boilerplate code to fs/proc/interrupts.c 2008-10-23 15:15:46 +04:00
Kconfig vmcore: it is not experimental any more 2010-10-26 16:52:05 -07:00
kcore.c kcore: add _text to KCORE_TEXT 2010-05-27 09:12:47 -07:00
kmsg.c procfs: Use generic_file_llseek in /proc/kmsg 2010-04-09 16:35:41 +02:00
loadavg.c sched, timers: cleanup avenrun users 2009-05-15 15:32:45 +02:00
Makefile procfs: simplify conditional processing of fs/proc.o. 2010-08-11 08:59:20 -07:00
meminfo.c hwpoison: fix/proc/meminfo alignment 2009-10-29 07:39:25 -07:00
mmu.c fs/proc/mmu.c: headers butchery 2007-10-17 08:42:48 -07:00
nommu.c include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h 2010-03-30 22:02:32 +09:00
page.c proc: export uncached bit properly in /proc/kpageflags 2010-09-09 18:57:23 -07:00
proc_devtree.c of: Drop properties with "/" in their name 2010-06-13 18:12:24 -06:00
proc_net.c include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h 2010-03-30 22:02:32 +09:00
proc_sysctl.c fs: change d_compare for rcu-walk 2011-01-07 17:50:19 +11:00
proc_tty.c Revert "tty: Add a new file /proc/tty/consoles" 2010-10-23 08:14:12 -07:00
root.c switch procfs to ->mount() 2010-10-29 04:17:01 -04:00
softirqs.c procfs: fix /proc/softirqs formatting 2010-10-27 18:03:13 -07:00
stat.c /proc/stat: fix scalability of irq sum of all cpu 2010-10-27 18:03:13 -07:00
task_mmu.c pagemap: set pagemap walk limit to PMD boundary 2010-11-25 06:50:46 +09:00
task_nommu.c nommu: add '[stack]' label to /proc/pid/maps output 2010-06-29 15:29:30 -07:00
uptime.c [PATCH] Fix idle time field in /proc/uptime 2009-09-24 10:16:24 +02:00
version.c proc: switch /proc/version to seq_file 2008-10-23 14:19:58 +04:00
vmcore.c /proc/vmcore: fix seeking 2010-09-22 17:22:38 -07:00