Part of a performance problem with ip_tables is that memory allocation
is not NUMA aware, but 'only' SMP aware (ie each CPU normally touch
separate cache lines)
Even with small iptables rules, the cost of this misplacement can be
high on common workloads. Instead of using one vmalloc() area
(located in the node of the iptables process), we now allocate an area
for each possible CPU, using vmalloc_node() so that memory should be
allocated in the CPU's node if possible.
Port to arp_tables and ip6_tables by Harald Welte.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace existing BIC version 1.1 with new version 2.0.
The main change is to replace the window growth function
with a cubic function as described in:
http://www.csc.ncsu.edu/faculty/rhee/export/bitcp/cubic-paper.pdf
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The latest BICTCP patch at:
http://www.csc.ncsu.edu:8080/faculty/rhee/export/bitcp/index_files/Page546.htm
disables the low_utilization feature of BICTCP because it doesn't work
in some cases. This patch removes it.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
As noticed by Phil Oester, the GRE NAT protocol helper is initialized
before the NAT core, which makes registration fail.
Change the linking order to make NAT be initialized first.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The skb_postpull_rcsum introduced a bug to the checksum modification.
Although the length pulled is offset bytes, the origin of the pulling
is the GRE header, not the IP header.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a TFTP client is SNATed so that the port is also changed, the
port is never changed back for the expected connection.
Signed-off-by: Marcus Sundberg <marcus@ingate.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
We have to store the congestion control timestamp on the SKB before we
clone it, not after. Else we get no timestamping information at all.
tcp_transmit_skb() has been reworked so that we can do the timestamp
still in one spot, instead of at all the call sites.
Problem discovered, and initial fix, from Tom Young
<tyo@ee.unimelb.edu.au>.
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove unneeded call to tcp_vegas_rtt_calc. The more accurate
microsecond value has already been registered prior to calling
tcp_vegas_cong_avoid.
Signed-off-by: Thomas Young <tyo@ee.mu.oz.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the resetting of rtt measurements to inside the once per RTT
block of code.
Signed-off-by: Thomas Young <tyo@ee.mu.oz.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
NFA_NEST calls NFA_PUT which jumps to nfattr_failure if the skb has no
room left. We call read_unlock_bh at nfattr_failure for the NFA_PUT inside
the locked section, so move NFA_NEST inside the locked section too.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Should have been marked EXPERIMENTAL from the beginning, as the current
bunch of fixes show.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
ip_conntrack_flush() used to be part of ip_conntrack_cleanup(), which needs
to drop _all_ references on module unload. Table flushed using ctnetlink
just needs to clean the table and doesn't need to flush the event cache or
wait for any references attached to skbs. Move everything but pure table
flushing back to ip_conntrack_cleanup().
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
ip_nat_initialized() takes enum ip_nat_manip_type as it's second argument,
not a hook number.
Noticed and initial patch by Marcus Sundberg <marcus@ingate.com>.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is a coding error in inet_create that causes it to always return
ESOCKTNOSUPPORT. It should return EPROTONOSUPPORT when there are
protocols registered for a given socket type but none of them match
the requested protocol.
This is based on a patch by Jayachandran C.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
From: David Stevens <dlstevens@us.ibm.com>
As explained at:
http://www.cs.ucsb.edu/~krishna/igmp_dos/
With IGMP version 1 and 2 it is possible to inject a unicast
report to a client which will make it ignore multicast
reports sent later by the router.
The fix is to only accept the report if is was sent to a
multicast or unicast address.
Signed-off-by: David S. Miller <davem@davemloft.net>
The receive path for fib_lookup netlink messages is lacking sanity
checks for header and payload and is thus vulnerable to malformed
netlink messages causing illegal memory references.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Around jiffies wrap time (i.e. within first 5 mins after boot), recent
match rules which contain both --seconds and --hitcount arguments
experience false matches.
This is because the last_pkts array is filled with zeros on creation, and
when comparing 'now' to 0 (+ --seconds argument), time_before_eq thinks it
has found a hit.
Below patch adds a break if the packet value is zero. This has the
unfortunate side effect of causing mismatches if a packet was received
when jiffies really was equal to zero. The odds of that happening are
slim compared to the problems caused by not adding the break however.
Plus, the author used this same method just below, so it is "good enough".
This fixes netfilter bugs #383 and #395.
Signed-off-by: Phil Oester <kernel@linuxace.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mounting NFS file systems after a (warm) reboot could take a long time if
firewalling and connection tracking was enabled.
The reason is that the NFS clients tends to use the same ports (800 and
counting down). Now on reboot, the server would still have a TCB for an
existing TCP connection client:800 -> server:2049. The client sends a
SYN from port 800 to server:2049, which elicits an ACK from the server.
The firewall on the client drops the ACK because (from its point of
view) the connection is still in half-open state, and it expects to see
a SYNACK.
The client will eventually time out after several minutes.
The following patch corrects this, by accepting ACKs on half open
connections as well.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch contains the following cleanups:
- make needlessly global code static
- ip_conntrack_core.c: ip_conntrack_flush() -> ip_conntrack_flush(void)
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch makes two needlessly global functions static.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
the patch below marks various variables const in net/; the goal is to
move them to the .rodata section so that they can't false-share
cachelines with things that get written to, as well as potentially
helping gcc a bit with optimisations. (these were found using a gcc
patch to warn about such variables)
Signed-off-by: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The tcp_ehash hash table gets too big on systems with really big memory.
It is worse on systems with pages larger than 4KB. It wastes memory that
could be better used. It also makes the netstat command slow because reading
/proc/net/tcp and /proc/net/tcp6 needs to go through the full hash table.
The default value should not be larger for larger page sizes. It seems
that the effect of page size is an unintended error dating back a long
time. I also wonder if the default value really should be a larger
fraction of memory for systems with more memory. While systems with
really big ram can afford more space for hash tables, it is not clear to
me that they benefit from increasing the allocation ratio for this table.
The amount of memory allocated is determined by net/ipv4/tcp.c:tcp_init and
mm/page_alloc.c:alloc_large_system_hash.
tcp_init calls alloc_large_system_hash passing parameters-
bucketsize=sizeof(struct tcp_ehash_bucket)
numentries=thash_entries
scale=(num_physpages >= 128 * 1024) ? (25-PAGE_SHIFT) : (27-PAGE_SHIFT)
limit=0
On i386, PAGE_SHIFT is 12 for a page size of 4K
On ia64, PAGE_SHIFT defaults to 14 for a page size of 16K
The num_physpages test above makes the allocation take a larger fraction
of the total memory on systems with larger memory. The threshold size
for a i386 system is 512MB. For an ia64 system with 16KB pages the
threshold is 2GB.
For smaller memory systems-
On i386, scale = (27 - 12) = 15
On ia64, scale = (27 - 14) = 13
For larger memory systems-
On i386, scale = (25 - 12) = 13
On ia64, scale = (25 - 14) = 11
For the rest of this discussion, I'll just track the larger memory case.
The default behavior has numentries=thash_entries=0, so the allocated
size is determined by either scale or by the default limit of 1/16 of
total memory.
In alloc_large_system_hash-
| numentries = (flags & HASH_HIGHMEM) ? nr_all_pages : nr_kernel_pages;
| numentries += (1UL << (20 - PAGE_SHIFT)) - 1;
| numentries >>= 20 - PAGE_SHIFT;
| numentries <<= 20 - PAGE_SHIFT;
At this point, numentries is pages for all of memory, rounded up to the
nearest megabyte boundary.
| /* limit to 1 bucket per 2^scale bytes of low memory */
| if (scale > PAGE_SHIFT)
| numentries >>= (scale - PAGE_SHIFT);
| else
| numentries <<= (PAGE_SHIFT - scale);
On i386, numentries >>= (13 - 12), so numentries is 1/8196 of
bytes of total memory.
On ia64, numentries <<= (14 - 11), so numentries is 1/2048 of
bytes of total memory.
| log2qty = long_log2(numentries);
|
| do {
| size = bucketsize << log2qty;
bucketsize is 16, so size is 16 times numentries, rounded
down to a power of two.
On i386, size is 1/512 of bytes of total memory.
On ia64, size is 1/128 of bytes of total memory.
For smaller systems the results are
On i386, size is 1/2048 of bytes of total memory.
On ia64, size is 1/512 of bytes of total memory.
The large page effect can be removed by just replacing
the use of PAGE_SHIFT with a constant of 12 in the calls to
alloc_large_system_hash. That makes them more like the other uses of
that function from fs/inode.c and fs/dcache.c
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/netfilter/ip_conntrack_netlink.c: In function 'ctnetlink_dump_table':
net/ipv4/netfilter/ip_conntrack_netlink.c:409: warning: implicit declaration of function 'local_bh_disable'
net/ipv4/netfilter/ip_conntrack_netlink.c:427: warning: implicit declaration of function 'local_bh_enable'
Signed-off-by: Benoit Boissinot <benoit.boissinot@ens-lyon.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove proto == NULL checking since ip_conntrack_[nat_]proto_find_get
always returns a valid pointer.
Fix missing ip_conntrack_proto_put in some paths.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes the problem with promoting aliases when:
a) a single primary and > 1 secondary addresses
b) multiple primary addresses each with at least one secondary address
Based on earlier efforts from Brian Pomerantz <bapper@piratehaven.org>,
Patrick McHardy <kaber@trash.net> and Thomas Graf <tgraf@suug.ch>
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
- IP_NF_CONNTRACK_MARK is bool and depends on only IP_NF_CONNTRACK
which is tristate. If a variable depends on IP_NF_CONNTRACK_MARK and
doesn't care about IP_NF_CONNTRACK, it can be y. This must be avoided.
- IP_NF_CT_ACCT has same problem.
- IP_NF_TARGET_CLUSTERIP also depends on IP_NF_MANGLE.
Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Don't show local table to behave similar to fib_hash.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since we've converted the ftp/irc/tftp helpers to use the new
module_parm_array() some time ago, we ware accidentially using signed data
types - thus preventing those modules from being used on ports >= 32768.
This patch fixes it by using 'ushort' module parameters.
Thanks to Jan Nijs for reporting this bug.
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is a compile error that crept in with the last patch of
TCP patches.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Both of ipq and frag_queue have *next and **prev, and they can be replaced
with hlist. Thanks Arnaldo Carvalho de Melo for the suggestion.
Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Acked-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch unconditionally requires CAP_NET_ADMIN for all nfnetlink
messages. It also removes the per-message cap_required field, since all
existing subsystems use CAP_NET_ADMIN for all their messages anyway.
Patrick McHardy owes me a beer if we ever need to re-introduce this.
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add missing size checks. Thanks Patrick McHardy for the hint.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make gcc-4.x happy. Use size_t instead of int. Thanks to Patrick McHardy
for the hint.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some devices (e.g. Qlogic iSCSI HBA hardware like QLA4010 up to firmware
3.0.0.4) initiates TCP with SYN and PUSH flags set.
The Linux TCP/IP stack deals fine with that, but the connection tracking
code doesn't.
This patch alters TCP connection tracking to accept SYN+PUSH as a valid
flag combination.
Signed-off-by: Vlad Drukker <vlad@storewiz.com>
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use "hints" to speed up the SACK processing. Various forms
of this have been used by TCP developers (Web100, STCP, BIC)
to avoid the 2x linear search of outstanding segments.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a patch for discussion addressing some receive buffer growing issues.
This is partially related to the thread "Possible BUG in IPv4 TCP window
handling..." last week.
Specifically it addresses the problem of an interaction between rcvbuf
moderation (receiver autotuning) and rcv_ssthresh. The problem occurs when
sending small packets to a receiver with a larger MTU. (A very common case I
have is a host with a 1500 byte MTU sending to a host with a 9k MTU.) In
such a case, the rcv_ssthresh code is targeting a window size corresponding
to filling up the current rcvbuf, not taking into account that the new rcvbuf
moderation may increase the rcvbuf size.
One hunk makes rcv_ssthresh use tcp_rmem[2] as the size target rather than
rcvbuf. The other changes the behavior when it overflows its memory bounds
with in-order data so that it tries to grow rcvbuf (the same as with
out-of-order data).
These changes should help my problem of mixed MTUs, and should also help the
case from last week's thread I think. (In both cases though you still need
tcp_rmem[2] to be set much larger than the TCP window.) One question is if
this is too aggressive at trying to increase rcvbuf if it's under memory
stress.
Orignally-from: John Heffner <jheffner@psc.edu>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is an updated version of the RFC3465 ABC patch originally
for Linux 2.6.11-rc4 by Yee-Ting Li. ABC is a way of counting
bytes ack'd rather than packets when updating congestion control.
The orignal ABC described in the RFC applied to a Reno style
algorithm. For advanced congestion control there is little
change after leaving slow start.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move all the code that does linear TCP slowstart to one
inline function to ease later patch to add ABC support.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Simplify the code that comuputes microsecond rtt estimate used
by TCP Vegas. Move the callback out of the RTT sampler and into
the end of the ack cleanup.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP peformance with TSO over networks with delay is awful.
On a 100Mbit link with 150ms delay, we get 4Mbits/sec with TSO and
50Mbits/sec without TSO.
The problem is with TSO, we intentionally do not keep the maximum
number of packets in flight to fill the window, we hold out to until
we can send a MSS chunk. But, we also don't update the congestion window
unless we have filled, as per RFC2861.
This patch replaces the check for the congestion window being full
with something smarter that accounts for TSO.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Here is the patch that introduces the generic skb_checksum_complete
which also checks for hardware RX checksum faults. If that happens,
it'll call netdev_rx_csum_fault which currently prints out a stack
trace with the device name. In future it can turn off RX checksum.
I've converted every spot under net/ that does RX checksum checks to
use skb_checksum_complete or __skb_checksum_complete with the
exceptions of:
* Those places where checksums are done bit by bit. These will call
netdev_rx_csum_fault directly.
* The following have not been completely checked/converted:
ipmr
ip_vs
netfilter
dccp
This patch is based on patches and suggestions from Stephen Hemminger
and David S. Miller.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Most netlink families make no use of the done() callback, making
it optional gets rid of all unnecessary dummy implementations.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The existing connection tracking subsystem in netfilter can only
handle ipv4. There were basically two choices present to add
connection tracking support for ipv6. We could either duplicate all
of the ipv4 connection tracking code into an ipv6 counterpart, or (the
choice taken by these patches) we could design a generic layer that
could handle both ipv4 and ipv6 and thus requiring only one sub-protocol
(TCP, UDP, etc.) connection tracking helper module to be written.
In fact nf_conntrack is capable of working with any layer 3
protocol.
The existing ipv4 specific conntrack code could also not deal
with the pecularities of doing connection tracking on ipv6,
which is also cured here. For example, these issues include:
1) ICMPv6 handling, which is used for neighbour discovery in
ipv6 thus some messages such as these should not participate
in connection tracking since effectively they are like ARP
messages
2) fragmentation must be handled differently in ipv6, because
the simplistic "defrag, connection track and NAT, refrag"
(which the existing ipv4 connection tracking does) approach simply
isn't feasible in ipv6
3) ipv6 extension header parsing must occur at the correct spots
before and after connection tracking decisions, and there were
no provisions for this in the existing connection tracking
design
4) ipv6 has no need for stateful NAT
The ipv4 specific conntrack layer is kept around, until all of
the ipv4 specific conntrack helpers are ported over to nf_conntrack
and it is feature complete. Once that occurs, the old conntrack
stuff will get placed into the feature-removal-schedule and we will
fully kill it off 6 months later.
Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Eric Dumazet