Comparing with kernel 2.6.24, tbench result has regression with
2.6.25-rc1.
1) On 2 quad-core processor stoakley: 4%.
2) On 4 quad-core processor tigerton: more than 30%.
bisect located below patch.
b4ce92775c is first bad commit
commit b4ce92775c
Author: Herbert Xu <herbert@gondor.apana.org.au>
Date: Tue Nov 13 21:33:32 2007 -0800
[IPV6]: Move nfheader_len into rt6_info
The dst member nfheader_len is only used by IPv6. It's also currently
creating a rather ugly alignment hole in struct dst. Therefore this patch
moves it from there into struct rt6_info.
Above patch changes the cache line alignment, especially member
__refcnt. I did a testing by adding 2 unsigned long pading before
lastuse, so the 3 members, lastuse/__refcnt/__use, are moved to next
cache line. The performance is recovered.
I created a patch to rearrange the members in struct dst_entry.
With Eric and Valdis Kletnieks's suggestion, I made finer arrangement.
1) Move tclassid under ops in case CONFIG_NET_CLS_ROUTE=y. So
sizeof(dst_entry)=200 no matter if CONFIG_NET_CLS_ROUTE=y/n. I
tested many patches on my 16-core tigerton by moving tclassid to
different place. It looks like tclassid could also have impact on
performance. If moving tclassid before metrics, or just don't move
tclassid, the performance isn't good. So I move it behind metrics.
2) Add comments before __refcnt.
On 16-core tigerton:
If CONFIG_NET_CLS_ROUTE=y, the result with below patch is about 18%
better than the one without the patch;
If CONFIG_NET_CLS_ROUTE=n, the result with below patch is about 30%
better than the one without the patch.
With 32bit 2.6.25-rc1 on 8-core stoakley, the new patch doesn't
introduce regression.
Thank Eric, Valdis, and David!
Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
Acked-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There's a horrible slab abuse in net/netfilter/nf_conntrack_extend.c
that can be replaced with a call to ksize().
Cc: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current /proc/net is done with so called "shadows", but current
implementation is broken and has little chances to get fixed.
The problem is that dentries subtree of /proc/net directory has
fancy revalidation rules to make processes living in different
net namespaces see different entries in /proc/net subtree, but
currently, tasks see in the /proc/net subdir the contents of any
other namespace, depending on who opened the file first.
The proposed fix is to turn /proc/net into a symlink, which points
to /proc/self/net, which in turn shows what previously was in
/proc/net - the network-related info, from the net namespace the
appropriate task lives in.
# ls -l /proc/net
lrwxrwxrwx 1 root root 8 Mar 5 15:17 /proc/net -> self/net
In other words - this behaves like /proc/mounts, but unlike
"mounts", "net" is not a file, but a directory.
Changes from v2:
* Fixed discrepancy of /proc/net nlink count and selinux labeling
screwup pointed out by Stephen.
To get the correct nlink count the ->getattr callback for /proc/net
is overridden to read one from the net->proc_net entry.
To make selinux still work the net->proc_net entry is initialized
properly, i.e. with the "net" name and the proc_net parent.
Selinux fixes are
Acked-by: Stephen Smalley <sds@tycho.nsa.gov>
Changes from v1:
* Fixed a task_struct leak in get_proc_task_net, pointed out by Paul.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
hci_sock_cleanup() always returns 0 and its return value isn't used
anywhere in the code.
Compile-tested with 'make allyesconfig && make net/bluetooth/bluetooth.ko'
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
If all of the entropy is in the local and foreign addresses,
but xor'ing together would cancel out that entropy, the
current hash performs poorly.
Suggested by Cosmin Ratiu:
Basically, the situation is as follows: There is a client
machine and a server machine. Both create 15000 virtual
interfaces, open up a socket for each pair of interfaces and
do SIP traffic. By profiling I noticed that there is a lot of
time spent walking the established hash chains with this
particular setup.
The addresses were distributed like this: client interfaces
were 198.18.0.1/16 with increments of 1 and server interfaces
were 198.18.128.1/16 with increments of 1. As I said, there
were 15000 interfaces. Source and destination ports were 5060
for each connection. So in this case, ports don't matter for
hashing purposes, and the bits from the address pairs used
cancel each other, meaning there are no differences in the
whole lot of pairs, so they all end up in the same hash chain.
Signed-off-by: David S. Miller <davem@davemloft.net>
The new SCTP socket api (draft 16) updates the AUTH API structures.
We never exported these since we knew they would change.
Update the rest to match the draft.
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Use the added dev_alloc_name() call to create tunnel device name,
rather than iterate in a hand-made loop with an artificial limit.
Thanks Patrick for noticing this.
[ The way this works is, when the device is actually registered,
the generic code noticed the '%' in the name and invokes
dev_alloc_name() to fully resolve the name. -DaveM ]
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add missing structure kernel-doc descriptions to sock.h & skbuff.h
to fix kernel-doc warnings.
(I think that Stephen H. sent a similar patch, but I can't find it.
I just want to kill the warnings, with either patch.)
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Al Viro spotted a bogus use of u64 on the input sequence number which
is big-endian. This patch fixes it by giving the input sequence number
its own member in the xfrm_skb_cb structure.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes unused declaration of dflt_rt_lookup() method in
include/net/ndisc.h
Signed-off-by: Rami Rosen <ramirose@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch changes current use of: init_timer(), add_timer()
and del_timer() to setup_timer() with mod_timer(), which
should be safer anyway.
Reported-by: Jann Traschewski <jann@gmx.de>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
According to one of Jann's OOPS reports it looks like
BUG_ON(timer_pending(timer)) triggers during add_timer()
in ax25_start_t1timer(). This patch changes current use
of: init_timer(), add_timer() and del_timer() to
setup_timer() with mod_timer(), which should be safer
anyway.
Reported-by: Jann Traschewski <jann@gmx.de>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
All these static inlines are unused:
in_own_zone 1 (net/tipc/addr.h)
msg_dataoctet 1 (net/tipc/msg.h)
msg_direct 1 (include/net/tipc/tipc_msg.h)
msg_options 1 (include/net/tipc/tipc_msg.h)
tipc_nmap_get 1 (net/tipc/bcast.h)
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes some unused definitions and one method typedef
declaration (f_pnode)
in include/net/ip6_fib.h, as they are not used in the kernel.
Signed-off-by: Rami Rosen <ramirose@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove IP6_RT_PRIO_FW and IP6_RT_FLOW_MASK definitions in
include/net/ip6_route.h, as they are not used in the kernel.
Signed-off-by: Rami Rosen <ramirose@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ->move operation has two bugs:
- It is called with the same extension as source and destination,
so it doesn't update the new extension.
- The address of the old extension is calculated incorrectly,
instead of (void *)ct->ext + ct->ext->offset[i] it uses
ct->ext + ct->ext->offset[i].
Fixes a crash on x86_64 reported by Chuck Ebbert <cebbert@redhat.com>
and Thomas Woerner <twoerner@redhat.com>.
Tested-by: Thomas Woerner <twoerner@redhat.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This merges the mux.c (including the connection interface) with trans_fd
in preparation for transport API changes. Ultimately, trans_fd will need
to be rewritten to clean it up and simplify the implementation, but this
reorganization is viewed as the first step.
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
GDM gets unhappy if /var/gdm doesn't have the sticky bit set. This patch adds
support for the sticky bit in much the same way setuid/setgid is supported.
With this patch, I can launch X from a v9fs rootfs (although I quickly run out
of fds in the server once gnome starts up).
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Acked-by: Eric Van Hensbergen <ericvh@gmail.com>
This replaces the console-based virto client with a block-based
client using a single request queue.
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
Add a new transport function which allows a cut-thru directly to
the transport instead of processing request through the mux if the
cut-thru exists.
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
Add a new set of configuration functions to the NetLabel/LSM API so that
LSMs can perform their own configuration of the NetLabel subsystem without
relying on assistance from userspace.
Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: Casey Schaufler <casey@schaufler-ca.com>
Reviewed-by: James Morris <jmorris@namei.org>
Cc: Chris Wright <chrisw@sous-sol.org>
Cc: Stephen Smalley <sds@tycho.nsa.gov>
Cc: Casey Schaufler <casey@schaufler-ca.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I was notified by Randy Stewart that lksctp claims to be
"the reference implementation". First of all, "the
refrence implementation" was the original implementation
of SCTP in usersapce written ty Randy and a few others.
Second, after looking at the definiton of 'reference implementation',
we don't really meet the requirements.
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
The port offset calculations depend on the protocol family, but, as
Adrian noticed, I broke this logic with the commit
5ee31fc1ec
[INET]: Consolidate inet(6)_hash_connect.
Return this logic back, by passing the port offset directly into the
consolidated function.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Noticed-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move ipv6_icmp_sysctl_init and ipv6_route_sysctl_init into the right
ifdef section otherwise that does not compile when CONFIG_SYSCTL=yes
and CONFIG_PROC_FS=no
Signed-off-by: Daniel Lezcano <dlezcano@fr.ibm.com>
Acked-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
compile error building without CONFIG_FS_PROC:
net/ipv4/fib_frontend.c: In function 'fib_net_init':
net/ipv4/fib_frontend.c:1032: error: implicit declaration of function 'fib_proc_
init'
net/ipv4/fib_frontend.c: In function 'fib_net_exit':
net/ipv4/fib_frontend.c:1047: error: implicit declaration of function 'fib_proc_
exit'
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This way we can remove TCP and DCCP specific versions of
sk->sk_prot->get_port: both v4 and v6 use inet_csk_get_port
sk->sk_prot->hash: inet_hash is directly used, only v6 need
a specific version to deal with mapped sockets
sk->sk_prot->unhash: both v4 and v6 use inet_hash directly
struct inet_connection_sock_af_ops also gets a new member, bind_conflict, so
that inet_csk_get_port can find the per family routine.
Now only the lookup routines receive as a parameter a struct inet_hashtable.
With this we further reuse code, reducing the difference among INET transport
protocols.
Eventually work has to be done on UDP and SCTP to make them share this
infrastructure and get as a bonus inet_diag interfaces so that iproute can be
used with these protocols.
net-2.6/net/ipv4/inet_hashtables.c:
struct proto | +8
struct inet_connection_sock_af_ops | +8
2 structs changed
__inet_hash_nolisten | +18
__inet_hash | -210
inet_put_port | +8
inet_bind_bucket_create | +1
__inet_hash_connect | -8
5 functions changed, 27 bytes added, 218 bytes removed, diff: -191
net-2.6/net/core/sock.c:
proto_seq_show | +3
1 function changed, 3 bytes added, diff: +3
net-2.6/net/ipv4/inet_connection_sock.c:
inet_csk_get_port | +15
1 function changed, 15 bytes added, diff: +15
net-2.6/net/ipv4/tcp.c:
tcp_set_state | -7
1 function changed, 7 bytes removed, diff: -7
net-2.6/net/ipv4/tcp_ipv4.c:
tcp_v4_get_port | -31
tcp_v4_hash | -48
tcp_v4_destroy_sock | -7
tcp_v4_syn_recv_sock | -2
tcp_unhash | -179
5 functions changed, 267 bytes removed, diff: -267
net-2.6/net/ipv6/inet6_hashtables.c:
__inet6_hash | +8
1 function changed, 8 bytes added, diff: +8
net-2.6/net/ipv4/inet_hashtables.c:
inet_unhash | +190
inet_hash | +242
2 functions changed, 432 bytes added, diff: +432
vmlinux:
16 functions changed, 485 bytes added, 492 bytes removed, diff: -7
/home/acme/git/net-2.6/net/ipv6/tcp_ipv6.c:
tcp_v6_get_port | -31
tcp_v6_hash | -7
tcp_v6_syn_recv_sock | -9
3 functions changed, 47 bytes removed, diff: -47
/home/acme/git/net-2.6/net/dccp/proto.c:
dccp_destroy_sock | -7
dccp_unhash | -179
dccp_hash | -49
dccp_set_state | -7
dccp_done | +1
5 functions changed, 1 bytes added, 242 bytes removed, diff: -241
/home/acme/git/net-2.6/net/dccp/ipv4.c:
dccp_v4_get_port | -31
dccp_v4_request_recv_sock | -2
2 functions changed, 33 bytes removed, diff: -33
/home/acme/git/net-2.6/net/dccp/ipv6.c:
dccp_v6_get_port | -31
dccp_v6_hash | -7
dccp_v6_request_recv_sock | +5
3 functions changed, 5 bytes added, 38 bytes removed, diff: -33
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The namespace is not available in the fib_sync_down_addr, add it as a
parameter.
Looking up a device by the pointer to it is OK. Looking up using a
result from fib_trie/fib_hash table lookup is also safe. No need to
fix that at all. So, just fix lookup by address and insertion to the
hash table path.
Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is required to make fib_info lookups namespace aware. In the
other case initial namespace devices are marked as dead in the local
routing table during other namespace stop.
Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
fib_sync_down can be called with an address and with a device. In
reality it is called either with address OR with a device. The
codepath inside is completely different, so lets separate it into two
calls for these two cases.
Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current ip route cache implementation is not suited to large caches.
We can consume a lot of CPU when cache must be invalidated, since we
currently need to evict all cache entries, and this eviction is
sometimes asynchronous. min_delay & max_delay can somewhat control this
asynchronism behavior, but whole thing is a kludge, regularly triggering
infamous soft lockup messages. When entries are still in use, this also
consumes a lot of ram, filling dst_garbage.list.
A better scheme is to use a generation identifier on each entry,
so that cache invalidation can be performed by changing the table
identifier, without having to scan all entries.
No more delayed flushing, no more stalling when secret_interval expires.
Invalidated entries will then be freed at GC time (controled by
ip_rt_gc_timeout or stress), or when an invalidated entry is found
in a chain when an insert is done.
Thus we keep a normal equilibrium.
This patch :
- renames rt_hash_rnd to rt_genid (and makes it an atomic_t)
- Adds a new rt_genid field to 'struct rtable' (filling a hole on 64bit)
- Checks entry->rt_genid at appropriate places :
Add a net argument to inet6_lookup and propagate it further.
Actually, this is tcp-v6 implementation of what was done for
tcp-v4 sockets in a previous patch.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a net argument to inet_lookup and propagate it further
into lookup calls. Plus tune the __inet_check_established.
The dccp and inet_diag, which use that lookup functions
pass the init_net into them.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This tags the inet_bind_bucket struct with net pointer,
initializes it during creation and makes a filtering
during lookup.
A better hashfn, that takes the net into account is to
be done in the future, but currently all bind buckets
with similar port will be in one hash chain.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
These two functions are the same except for what they call
to "check_established" and "hash" for a socket.
This saves half-a-kilo for ipv4 and ipv6.
add/remove: 1/0 grow/shrink: 1/4 up/down: 582/-1128 (-546)
function old new delta
__inet_hash_connect - 577 +577
arp_ignore 108 113 +5
static.hint 8 4 -4
rt_worker_func 376 372 -4
inet6_hash_connect 584 25 -559
inet_hash_connect 586 25 -561
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Constify a few data tables use const qualifiers on variables where
possible in the nf_*_proto_tcp sources.
Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename all "conntrack" variables to "ct" for more consistency and
avoiding some overly long lines.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Reorder struct nf_conntrack_l4proto so all members used during packet
processing are in the same cacheline.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
nf_ct_tuple_src_equal() and nf_ct_tuple_dst_equal() both compare the protocol
numbers. Unfortunately gcc doesn't optimize out the second comparison, so
remove it and prefix both functions with __ to indicate that they should not
be used directly.
Saves another 16 byte of text in __nf_conntrack_find() on x86_64:
nf_conntrack_tuple_taken | -20 # 320 -> 300, size inlines: 181 -> 161
__nf_conntrack_find | -16 # 267 -> 251, size inlines: 127 -> 115
__nf_conntrack_confirm | -40 # 875 -> 835, size inlines: 570 -> 537
3 functions changed, 76 bytes removed
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ignoring specific entries in __nf_conntrack_find() is only needed by NAT
for nf_conntrack_tuple_taken(). Remove it from __nf_conntrack_find()
and make nf_conntrack_tuple_taken() search the hash itself.
Saves 54 bytes of text in the hotpath on x86_64:
__nf_conntrack_find | -54 # 321 -> 267, # inlines: 3 -> 2, size inlines: 181 -> 127
nf_conntrack_tuple_taken | +305 # 15 -> 320, lexblocks: 0 -> 3, # inlines: 0 -> 3, size inlines: 0 -> 181
nf_conntrack_find_get | -2 # 90 -> 88
3 functions changed, 305 bytes added, 56 bytes removed, diff: +249
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
With the RCU conversion only write_lock usages of nf_conntrack_lock are
left (except one read_lock that should actually use write_lock in the
H.323 helper). Switch to a spinlock.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use RCU for expectation hash. This doesn't buy much for conntrack
runtime performance, but allows to reduce the use of nf_conntrack_lock
for /proc and nf_netlink_conntrack.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
