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: NDipankar Sarma <dipankar@in.ibm.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

This patch converts the inode semaphore to a mutex. I have tested it on
XFS and compiled as much as one can consider on an ia64. Anyway your
luck with it might be different.
Modified-by: NIngo Molnar <mingo@elte.hu>

(finished the conversion)
Signed-off-by: NJes Sorensen <jes@sgi.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Currently all network protocols need to call dev_ioctl as the default
fallback in their ioctl implementations.  This patch adds a fallback
to dev_ioctl to sock_ioctl if the protocol returned -ENOIOCTLCMD.
This way all the procotol ioctl handlers can be simplified and we don't
need to export dev_ioctl.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

From: Benjamin LaHaise <bcrl@kvack.org>

In af_unix, a rwlock is used to protect internal state.  At least on my 
P4 with HT it is faster to use a spinlock due to the simpler memory 
barrier used to unlock.  This patch raises bw_unix to ~690K/s.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

I noticed that some of 'struct proto_ops' used in the kernel may share
a cache line used by locks or other heavily modified data. (default
linker alignement is 32 bytes, and L1_CACHE_LINE is 64 or 128 at
least)

This patch makes sure a 'struct proto_ops' can be declared as const,
so that all cpus can share all parts of it without false sharing.

This is not mandatory : a driver can still use a read/write structure
if it needs to (and eventually a __read_mostly)

I made a global stubstitute to change all existing occurences to make
them const.

This should reduce the possibility of false sharing on SMP, and
speedup some socket system calls.
Signed-off-by: NEric Dumazet <dada1@cosmosbay.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This lock is actually taken mostly as a writer,
so using a rwlock actually just makes performance
worse especially on chips like the Intel P4.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

AF_UNIX stream socket performance on P4 CPUs tends to suffer due to a
lot of pipeline flushes from atomic operations.  The patch below
removes the sock_hold() and sock_put() in unix_stream_sendmsg().  This
should be safe as the socket still holds a reference to its peer which
is only released after the file descriptor's final user invokes
unix_release_sock().  The only consideration is that we must add a
memory barrier before setting the peer initially.
Signed-off-by: NBenjamin LaHaise <benjamin.c.lahaise@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Most permission() calls have a struct nameidata * available.  This helper
takes that as an argument and thus makes sure we pass it down for lookup
intents and prepares for per-mount read-only support where we need a struct
vfsmount for checking whether a file is writeable.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Of this type, mostly:

CHECK   net/ipv6/netfilter.c
net/ipv6/netfilter.c:96:12: warning: symbol 'ipv6_netfilter_init' was not declared. Should it be static?
net/ipv6/netfilter.c:101:6: warning: symbol 'ipv6_netfilter_fini' was not declared. Should it be static?
Signed-off-by: NArnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Lots of places just needs the states, not even linux/tcp.h, where this
enum was, needs it.

This speeds up development of the refactorings as less sources are
rebuilt when things get moved from net/tcp.h.
Signed-off-by: NArnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This is part of the grand scheme to eliminate the qlen
member of skb_queue_head, and subsequently remove the
'list' member of sk_buff.

Most users of skb_queue_len() want to know if the queue is
empty or not, and that's trivially done with skb_queue_empty()
which doesn't use the skb_queue_head->qlen member and instead
uses the queue list emptyness as the test.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

currently it opencodes it, but that's in the way of chaning the
lookup_hash interface.

I'd prefer to disallow modular af_unix over exporting lookup_create,
but I'll leave that to you.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

	A lot of places in there are including major.h for no reason
whatsoever.  Removed.  And yes, it still builds.

	The history of that stuff is often amusing.  E.g. for net/core/sock.c
the story looks so, as far as I've been able to reconstruct it: we used to
need major.h in net/socket.c circa 1.1.early.  In 1.1.13 that need had
disappeared, along with register_chrdev(SOCKET_MAJOR, "socket", &net_fops)
in sock_init().  Include had not.  When 1.2 -> 1.3 reorg of net/* had moved
a lot of stuff from net/socket.c to net/core/sock.c, this crap had followed...
Signed-off-by: NAl Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

A lot of places in there are including major.h for no reason whatsoever.
Removed.  And yes, it still builds. 

The history of that stuff is often amusing.  E.g.  for net/core/sock.c
the story looks so, as far as I've been able to reconstruct it: we used
to need major.h in net/socket.c circa 1.1.early.  In 1.1.13 that need
had disappeared, along with register_chrdev(SOCKET_MAJOR, "socket",
&net_fops) in sock_init().  Include had not.  When 1.2 -> 1.3 reorg of
net/* had moved a lot of stuff from net/socket.c to net/core/sock.c,
this crap had followed... 
Signed-off-by: NAl Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.

Let it rip!