linux-stable/net
Dipankar Sarma 529bf6be5c [PATCH] fix file counting
I have benchmarked this on an x86_64 NUMA system and see no significant
performance difference on kernbench.  Tested on both x86_64 and powerpc.

The way we do file struct accounting is not very suitable for batched
freeing.  For scalability reasons, file accounting was
constructor/destructor based.  This meant that nr_files was decremented
only when the object was removed from the slab cache.  This is susceptible
to slab fragmentation.  With RCU based file structure, consequent batched
freeing and a test program like Serge's, we just speed this up and end up
with a very fragmented slab -

llm22:~ # cat /proc/sys/fs/file-nr
587730  0       758844

At the same time, I see only a 2000+ objects in filp cache.  The following
patch I fixes this problem.

This patch changes the file counting by removing the filp_count_lock.
Instead we use a separate percpu counter, nr_files, for now and all
accesses to it are through get_nr_files() api.  In the sysctl handler for
nr_files, we populate files_stat.nr_files before returning to user.

Counting files as an when they are created and destroyed (as opposed to
inside slab) allows us to correctly count open files with RCU.

Signed-off-by: Dipankar Sarma <dipankar@in.ibm.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-08 14:14:01 -08:00
..
802 [P8023]: Fix tainting of kernel. 2006-02-13 15:38:42 -08:00
8021q [PATCH] capable/capability.h (net/) 2006-01-11 18:42:14 -08:00
appletalk [PATCH] capable/capability.h (net/) 2006-01-11 18:42:14 -08:00
atm [ATM]: keep atmsvc failure messages quiet 2006-03-04 21:06:25 -08:00
ax25 [PATCH] capable/capability.h (net/) 2006-01-11 18:42:14 -08:00
bluetooth [Bluetooth] Fix NULL pointer dereferences of the HCI socket 2006-02-13 11:40:03 +01:00
bridge [BRIDGE]: generate kobject remove event 2006-03-04 21:06:23 -08:00
core [REQSK]: Don't reset rskq_defer_accept in reqsk_queue_alloc 2006-02-27 13:30:43 -08:00
dccp [DCCP] ccid3: Divide by zero fix 2006-03-04 21:06:29 -08:00
decnet [PATCH] capable/capability.h (net/) 2006-01-11 18:42:14 -08:00
econet [ECONET]: Use macro for spinlock_t definition. 2006-01-04 13:56:08 -08:00
ethernet [NET] ethernet: Fix first packet goes out with MAC 00:00:00:00:00:00 2006-02-23 16:18:01 -08:00
ieee80211 [PATCH] ieee80211_rx.c: is_beacon 2006-02-27 20:12:02 -05:00
ipv4 [NETFILTER] ip_queue: Fix wrong skb->len == nlmsg_len assumption 2006-03-07 14:56:12 -08:00
ipv6 [NETFILTER] ip_queue: Fix wrong skb->len == nlmsg_len assumption 2006-03-07 14:56:12 -08:00
ipx [PATCH] capable/capability.h (net/) 2006-01-11 18:42:14 -08:00
irda [IRDA]: out of range array access 2006-02-09 16:59:48 -08:00
key [XFRM]: Eliminate refcounting confusion by creating __xfrm_state_put(). 2006-02-23 16:10:53 -08:00
lapb [NET]: Kill skb->list 2005-08-29 15:31:14 -07:00
llc [NET]: Add a dev_ioctl() fallback to sock_ioctl() 2006-01-03 14:18:33 -08:00
netfilter [NETFILTER]: nf_queue: fix end-of-list check 2006-02-27 13:03:55 -08:00
netlink [NETLINK] genetlink: Fix bugs spotted by Andrew Morton. 2006-02-13 15:51:24 -08:00
netrom [PATCH] capable/capability.h (net/) 2006-01-11 18:42:14 -08:00
packet [NET]: Fix some whitespace issues in af_packet.c 2006-01-23 16:28:02 -08:00
rose [PATCH] capable/capability.h (net/) 2006-01-11 18:42:14 -08:00
rxrpc [NET]: "signed long" -> "long" 2006-01-17 13:03:54 -08:00
sched [PKT_SCHED]: Handle SCTP/DCCP in sfq_hash 2006-01-17 13:01:06 -08:00
sctp [SCTP]: Fix 'fast retransmit' to send a TSN only once. 2006-02-02 16:57:31 -08:00
sunrpc SUNRPC: Move upcall out of auth->au_ops->crcreate() 2006-02-01 12:52:25 -05:00
tipc [TIPC] Avoid polluting the global namespace 2006-01-18 00:45:16 +01:00
unix [PATCH] fix file counting 2006-03-08 14:14:01 -08:00
wanrouter [PATCH] capable/capability.h (net/) 2006-01-11 18:42:14 -08:00
x25 [PATCH] capable/capability.h (net/) 2006-01-11 18:42:14 -08:00
xfrm [IPSEC]: Kill post_input hook and do NAT-T in esp_input directly 2006-02-27 13:00:40 -08:00
compat.c [PATCH] Fix 32bit sendmsg() flaw 2005-09-08 08:14:11 -07:00
Kconfig [NET]: Add CONFIG_NETDEBUG to suppress bad packet messages. 2006-02-02 20:40:09 -08:00
Makefile [TIPC] Initial merge 2006-01-12 14:06:31 -08:00
nonet.c [NET]: Fix sock_init() return value. 2006-01-03 13:11:17 -08:00
socket.c [PATCH] percpu data: only iterate over possible CPUs 2006-02-05 11:06:51 -08:00
sysctl_net.c [NET]: Fix "sysctl_net.c:36: error: 'core_table' undeclared here" 2005-10-03 14:16:34 -07:00
TUNABLE Linux-2.6.12-rc2 2005-04-16 15:20:36 -07:00