Arnaldo and I agreed it could be applied now, because I have other
pending patches depending on this one (Thank you Arnaldo)
(The other important patch moves skc_refcnt in a separate cache line,
so that the SMP/NUMA performance doesnt suffer from cache line ping pongs)
1) First some performance data :
--------------------------------
tcp_v4_rcv() wastes a *lot* of time in __inet_lookup_established()
The most time critical code is :
sk_for_each(sk, node, &head->chain) {
if (INET_MATCH(sk, acookie, saddr, daddr, ports, dif))
goto hit; /* You sunk my battleship! */
}
The sk_for_each() does use prefetch() hints but only the begining of
"struct sock" is prefetched.
As INET_MATCH first comparison uses inet_sk(__sk)->daddr, wich is far
away from the begining of "struct sock", it has to bring into CPU
cache cold cache line. Each iteration has to use at least 2 cache
lines.
This can be problematic if some chains are very long.
2) The goal
-----------
The idea I had is to change things so that INET_MATCH() may return
FALSE in 99% of cases only using the data already in the CPU cache,
using one cache line per iteration.
3) Description of the patch
---------------------------
Adds a new 'unsigned int skc_hash' field in 'struct sock_common',
filling a 32 bits hole on 64 bits platform.
struct sock_common {
unsigned short skc_family;
volatile unsigned char skc_state;
unsigned char skc_reuse;
int skc_bound_dev_if;
struct hlist_node skc_node;
struct hlist_node skc_bind_node;
atomic_t skc_refcnt;
+ unsigned int skc_hash;
struct proto *skc_prot;
};
Store in this 32 bits field the full hash, not masked by (ehash_size -
1) Using this full hash as the first comparison done in INET_MATCH
permits us immediatly skip the element without touching a second cache
line in case of a miss.
Suppress the sk_hashent/tw_hashent fields since skc_hash (aliased to
sk_hash and tw_hash) already contains the slot number if we mask with
(ehash_size - 1)
File include/net/inet_hashtables.h
64 bits platforms :
#define INET_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
(((__sk)->sk_hash == (__hash))
((*((__u64 *)&(inet_sk(__sk)->daddr)))== (__cookie)) && \
((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports)) && \
(!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
32bits platforms:
#define TCP_IPV4_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
(((__sk)->sk_hash == (__hash)) && \
(inet_sk(__sk)->daddr == (__saddr)) && \
(inet_sk(__sk)->rcv_saddr == (__daddr)) && \
(!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
- Adds a prefetch(head->chain.first) in
__inet_lookup_established()/__tcp_v4_check_established() and
__inet6_lookup_established()/__tcp_v6_check_established() and
__dccp_v4_check_established() to bring into cache the first element of the
list, before the {read|write}_lock(&head->lock);
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Clean up timer initialization by introducing DEFINE_TIMER a'la
DEFINE_SPINLOCK. Build and boot-tested on x86. A similar patch has been
been in the -RT tree for some time.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
# grep -r 'netif_carrier_o[nf]' linux-2.6.12 | wc -l
246
# size vmlinux.org vmlinux.carrier
text data bss dec hex filename
4339634 1054414 259296 5653344 564360 vmlinux.org
4337710 1054414 259296 5651420 563bdc vmlinux.carrier
And this ain't an allyesconfig kernel!
Signed-off-by: David S. Miller <davem@davemloft.net>
qdisc_create_dflt() is missing to destroy the newly allocated
default qdisc if the initialization fails resulting in leaks
of all kinds. The only caller in mainline which may trigger
this bug is sch_tbf.c in tbf_create_dflt_qdisc().
Note: qdisc_create_dflt() doesn't fulfill the official locking
requirements of qdisc_destroy() but since the qdisc could
never be seen by the outside world this doesn't matter
and it can stay as-is until the locking of pkt_sched
is cleaned up.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
More unusable TCF_META_* match types that need to get eliminated
before 2.6.13 goes out the door.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Thomas Graf <tgraf@suug.ch>
It won't exist any longer when we shrink the SKB in 2.6.14,
and we should kill this off before anyone in userspace starts
using it.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Thomas Graf <tgraf@suug.ch>
The current call to __qdisc_dequeue_head leads to a branch
misprediction for every loop iteration, the fact that the
most common priority is 2 makes this even worse. This issue
has been brought up by Eric Dumazet <dada1@cosmosbay.com>
but unlike his solution which was to manually unroll the loop,
this approach preserves the possibility to increase the number
of bands at compile time.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the protocol specific config options out to the specific protocols.
With this change net/Kconfig now starts to become readable and serve as a
good basis for further re-structuring.
The menu structure is left almost intact, except that indention is
fixed in most cases. Most visible are the INET changes where several
"depends on INET" are replaced with a single ifdef INET / endif pair.
Several new files were created to accomplish this change - they are
small but serve the purpose that config options are now distributed
out where they belongs.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David 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: David S. Miller <davem@davemloft.net>
Useful in combination with classful qdiscs to drop or
temporary disable certain flows, e.g. one could block
specific ds flows with dsmark.
Unlike the noop qdisc it can be controlled by the user and
statistic accounting is done.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current behaviour is to not report an error if a rate
estimator is created together with a qdisc and the
configuration of the rate estimator is bogus. This leads
to unexpected behaviour because the user is not notified.
New behaviour is to report the error and let the whole
qdisc creation operation fail so the user is able to fix
his mistake.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Adds qdisc_alloc() to share code between qdisc_create()
and qdisc_create_dflt(). Hides the qdisc alignment behind
macros and makes use of them.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Plug holes with padding fields and initialized them to zero.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mostly missing initialization of padding fields of 1 or 2 bytes length,
two instances of uninitialized nlmsgerr->msg of 16 bytes length.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Do not present these confusing new options to the user
unless he picked some facility that makes use of it,
such as NET_EMATCH_TEXT.
Signed-off-by: David S. Miller <davem@davemloft.net>
Removes the skb trimming code which is not needed since we never
touch the skb upon failure. Removes unnecessary initializers,
and simplifies the code a bit.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
prio2list() returns the relevant sk_buff_head for the
band specified by the priority for a given skb.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Removes the skb trimming code which is not needed since we never
touch the skb upon failure. Removes unnecessary includes,
initializers, and simplifies the code a bit.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The simplicity of the fifo qdisc allows several qdisc operations to be
redirected to the relevant queue management function directly. Saves
a lot of code lines and gives the pfifo a byte based backlog.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch converts "unsigned flags" to use more explict types like u16
instead and incrementally introduces NLMSG_NEW().
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes dsmark to do all configuration sanity checks first and
only apply the changes if all of them can be applied without
any errors. Also fixes the weak sanity checks for DSMARK_VALUE
and DSMASK_MASK.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
From: Ralf Baechle <ralf@linux-mips.org>
There are archives of the old list at http://oss.sgi.com/archives/netdev
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is brought to you by the department of applied stupidity.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Adds meta collectors for all socket attributes that make sense
to be filtered upon. Some of them are only useful for debugging
but having them doesn't hurt.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Unused indices which are ignored while walking must still
be counted to avoid dumping the same index twice.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Here is a fixed up version of the reorder feature of netem.
It is the same as the earlier patch plus with the bugfix from Julio merged in.
Has expected backwards compatibility behaviour.
Go ahead and merge this one, the TCP strangeness I was seeing was due
to the reordering bug, and previous version of TSO patch.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Netem works better if there if packets are just queued in the inner discipline
rather than having a separate delayed queue. Change to use the dequeue/requeue
to peek like TBF does.
By doing this potential qlen problems with the old method are avoided. The problems
happened when the netem_run that moved packets from the inner discipline to the nested
discipline failed (because inner queue was full). This happened in dequeue, so the
effective qlen of the netem would be decreased (because of the drop), but there was
no way to keep the outer qdisc (caller of netem dequeue) in sync.
The problem window is still there since this patch doesn't address the issue of
requeue failing in netem_dequeue, but that shouldn't happen since the sequence dequeue/requeue
should always work. Long term correct fix is to implement qdisc->peek in all the qdisc's
to allow for this (needed by several other qdisc's as well).
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Handle duplication of packets in netem by re-inserting at top of qdisc tree.
This avoid problems with qlen accounting with nested qdisc. This recursion
requires no additional locking but will potentially increase stack depth.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Long standing bug.
Policy to repeat an action never worked.
Signed-off-by: J Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix qlen underrun when doing duplication with netem. If netem is used
as leaf discipline, then the parent needs to be tweaked when packets
are duplicated.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Netem currently dumps packets into the queue when timer expires. This
patch makes work by self-clocking (more like TBF). It fixes a bug
when 0 delay is requested (only doing loss or duplication).
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Due to bugs in netem (fixed by later patches), it is possible to get qdisc
qlen to go negative. If this happens the CPU ends up spinning forever
in qdisc_run(). So add a BUG_ON() to trap it.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some network drivers call netif_stop_queue() when detecting loss of
carrier. This leads to packets being queued up at the qdisc level for
an unbound period of time. In order to prevent this effect, the core
networking stack will now cease to queue packets for any device, that
is operationally down (i.e. the queue is flushed and disabled).
Signed-off-by: Tommy S. Christensen <tommy.christensen@tpack.net>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
