Get rid of custom locking that was using wait queue, lock, and atomic
to basically build a queued mutex.  Use RCU for read side.
Signed-off-by: NStephen Hemminger <shemminger@vyatta.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Convert AF_PACKET to use RCU, eliminating one more reader/writer lock.

There is no need for a real sk_del_node_init_rcu(), because sk_del_node_init
is doing the equivalent thing to hlst_del_init_rcu already; but added
some comments to try and make that obvious.
Signed-off-by: NStephen Hemminger <shemminger@vyatta.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

To see the effect make sure you have an empty SPD.
On window1 "ip xfrm mon" and on window2 issue "ip xfrm policy flush"
You get prompt back in window2 and you see the flush event on window1.
With this fix, you still get prompt on window1 but no event on window2.

Thanks to Alexey Dobriyan for finding a bug in earlier version
when using pfkey to do the flushing.
Signed-off-by: NJamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

To see the effect make sure you have an empty SAD.
On window1 "ip xfrm mon" and on window2 issue "ip xfrm state flush"
You get prompt back in window2 and you see the flush event on window1.
With this fix, you still get prompt on window1 but no event on window2.

Thanks to Alexey Dobriyan for finding a bug in earlier version
when using pfkey to do the flushing.
Signed-off-by: NJamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

RFC 2367 says flushing behavior should be:
1) user space -> kernel: flush
2) kernel: flush
3) kernel -> user space: flush event to ALL listeners

This is not realistic today in the presence of selinux policies
which may reject the flush etc. So we make the sequence become:
1) user space -> kernel: flush
2) kernel: flush
3) kernel -> user space: flush response to originater from #1
4) if there were no errors then:
kernel -> user space: flush event to ALL listeners
Signed-off-by: NJamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

commit 3bc38712 (handle NF_STOP and unknown verdicts in
nf_reinject) was a partial fix to packet leaks.

If user asks NF_STOLEN status, we must free the skb as well.
Reported-by: NAfi Gjermund <afigjermund@gmail.com>
Signed-off-by: NEric DUmazet <eric.dumazet@gmail.com>
Signed-off-by: NPatrick McHardy <kaber@trash.net>

This patch fixes a bug that triggers an assertion if you create
a conntrack entry with a helper and netfilter debugging is enabled.
Basically, we hit the assertion because the confirmation flag is
set before the conntrack extensions are added. To fix this, we
move the extension addition before the aforementioned flag is
set.

This patch also removes the possibility of setting a helper for
existing conntracks. This operation would also trigger the
assertion since we are not allowed to add new extensions for
existing conntracks. We know noone that could benefit from
this operation sanely.

Thanks to Eric Dumazet for initial posting a preliminary patch
to address this issue.
Reported-by: NDavid Ramblewski <David.Ramblewski@atosorigin.com>
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NPatrick McHardy <kaber@trash.net>

XFRMINHDRERROR counter is ambigous when validating forwarding
path. It makes it tricky to debug when you have both in and fwd
validation.
Signed-off-by: NJamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch enables fast retransmissions after one dupACK for
TCP if the stream is identified as thin. This will reduce
latencies for thin streams that are not able to trigger fast
retransmissions due to high packet interarrival time. This
mechanism is only active if enabled by iocontrol or syscontrol
and the stream is identified as thin.
Signed-off-by: NAndreas Petlund <apetlund@simula.no>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch will make TCP use only linear timeouts if the
stream is thin. This will help to avoid the very high latencies
that thin stream suffer because of exponential backoff. This
mechanism is only active if enabled by iocontrol or syscontrol
and the stream is identified as thin. A maximum of 6 linear
timeouts is tried before exponential backoff is resumed.
Signed-off-by: NAndreas Petlund <apetlund@simula.no>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Make remaining netlink policies as const.
Fixup coding style where needed.
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Dunno, what was the idea, it wasn't used for a long time.
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

ICMP6 MIB statistics was per-netns for quite a time.
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The lock used in unix_state_lock() is a spin_lock not reader-writer.
Signed-off-by: NStephen Hemminger <shemminger@vyatta.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As reported by Randy Dunlap <randy.dunlap@oracle.com>, compilation
of nf_defrag_ipv4 fails with:

include/net/netfilter/nf_conntrack.h:94: error: field 'ct_general' has incomplete type
include/net/netfilter/nf_conntrack.h:178: error: 'const struct sk_buff' has no member named 'nfct'
include/net/netfilter/nf_conntrack.h:185: error: implicit declaration of function 'nf_conntrack_put'
include/net/netfilter/nf_conntrack.h:294: error: 'const struct sk_buff' has no member named 'nfct'
net/ipv4/netfilter/nf_defrag_ipv4.c:45: error: 'struct sk_buff' has no member named 'nfct'
net/ipv4/netfilter/nf_defrag_ipv4.c:46: error: 'struct sk_buff' has no member named 'nfct'

net/nf_conntrack.h must not be included with NF_CONNTRACK=n, add a
few #ifdefs. Long term the header file should be fixed to be usable
even with NF_CONNTRACK=n.
Tested-by: NRandy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: NPatrick McHardy <kaber@trash.net>

Enhance IPVS to load balance SCTP transport protocol packets. This is done
based on the SCTP rfc 4960. All possible control chunks have been taken
care. The state machine used in this code looks some what lengthy. I tried
to make the state machine easy to understand.
Signed-off-by: NVenkata Mohan Reddy Koppula <mohanreddykv@gmail.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NPatrick McHardy <kaber@trash.net>

Only used for writing, so convert to spinlock
Signed-off-by: NStephen Hemminger <shemminger@vyatta.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Export sk_attach_filter/sk_detach_filter routines,
so that tun module can use them.
Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Traffic (tcp) doesnot start on a vlan interface when gro is enabled.
Even the tcp handshake was not taking place.
This is because, the eth_type_trans call before the netif_receive_skb
in napi_gro_finish() resets the skb->dev to napi->dev from the previously
set vlan netdev interface. This causes the ip_route_input to drop the
incoming packet considering it as a packet coming from a martian source.

I could repro this on 2.6.32.7 (stable) and 2.6.33-rc7.
With this fix, the traffic starts and the test runs fine on both vlan
and non-vlan interfaces.

CC: Herbert Xu <herbert@gondor.apana.org.au>
CC: Patrick McHardy <kaber@trash.net>
Signed-off-by: NAjit Khaparde <ajitk@serverengines.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As reported by Alexey Dobriyan:

--------------------
setkey now takes several seconds to run this simple script
and it spits "recv: Resource temporarily unavailable" messages.

#!/usr/sbin/setkey -f
flush;
spdflush;

add A B ipcomp 44 -m tunnel -C deflate;
add B A ipcomp 45 -m tunnel -C deflate;

spdadd A B any -P in ipsec
        ipcomp/tunnel/192.168.1.2-192.168.1.3/use;
spdadd B A any -P out ipsec
        ipcomp/tunnel/192.168.1.3-192.168.1.2/use;
--------------------

Obviously applications want the events even when the table
is empty.  So we cannot make this behavioral change.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The n-tuple list should be flushed if and only if the ETH_RESET_FILTER
flag is set and the driver is able to reset filtering/flow direction
hardware without also resetting a component whose flag is not set.
This test is best left to the driver.
Signed-off-by: NBen Hutchings <bhutchings@solarflare.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

kasprintf() makes code smaller.
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 2eff25c1
(netfilter: xt_hashlimit: fix race condition and simplify locking)
added a mutex deadlock :
htable_create() is called with hashlimit_mutex already locked
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The net being checked there is dev_net(dev) and thus this if
is always false.

Fits both net and net-next trees.
Signed-off-by: NPavel Emelyanov <xemul@openvz.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

call_rcu() will unconditionally reinitialize RCU head anyway.
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Acked-by: NPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add __percpu sparse annotations to net.

These annotations are to make sparse consider percpu variables to be
in a different address space and warn if accessed without going
through percpu accessors.  This patch doesn't affect normal builds.

The macro and type tricks around snmp stats make things a bit
interesting.  DEFINE/DECLARE_SNMP_STAT() macros mark the target field
as __percpu and SNMP_UPD_PO_STATS() macro is updated accordingly.  All
snmp_mib_*() users which used to cast the argument to (void **) are
updated to cast it to (void __percpu **).
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NDavid S. Miller <davem@davemloft.net>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: Vlad Yasevich <vladislav.yasevich@hp.com>
Cc: netdev@vger.kernel.org
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Eric's version fixed it for pfkey. This one is for xfrm user.
I thought about amortizing those two get_acqseq()s but it seems
reasonable to have two of these sequence spaces for the two different
interfaces.

cheers,
jamal
commit d5168d5addbc999c94aacda8f28a4a173756a72b
Author: Jamal Hadi Salim <hadi@cyberus.ca>
Date:   Tue Feb 16 06:51:22 2010 -0500

    xfrm: avoid spinlock in get_acqseq() used by xfrm user

    This is in the same spirit as commit 28aecb9d
    by Eric Dumazet.
    Use atomic_inc_return() in get_acqseq() to avoid taking a spinlock
Signed-off-by: NJamal Hadi Salim <hadi@cyberus.ca>
Acked-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Stop computing the number of neighbour table settings we have by
counting the number of binary sysctls.  This behaviour was silly
and meant that we could not add another neighbour table setting
without also adding another binary sysctl.

Don't pass the binary sysctl path for neighour table entries
into neigh_sysctl_register.  These parameters are no longer
used and so are just dead code.
Signed-off-by: NEric W. Biederman <ebiederm@xmission.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Stop using the binary sysctl enumeartion in sysctl.h as an index into
a per interface array.  This leads to unnecessary binary sysctl number
allocation, and a fragility in data structure and implementation
because of unnecessary coupling.
Signed-off-by: NEric W. Biederman <ebiederm@xmission.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Same stuff as in ip_gre patch: receive hook can be called before netns
setup is done, oopsing in net_generic().
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

GRE protocol receive hook can be called right after protocol addition is done.
If netns stuff is not yet initialized, we're going to oops in
net_generic().

This is remotely oopsable if ip_gre is compiled as module and packet
comes at unfortunate moment of module loading.
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

xfrm_state_clone calls kfree instead of xfrm_state_put to free
a failed state.  Depending on the state of the failed state, it
can cause leaks to things like module references.

All states should be freed by xfrm_state_put past the point of
xfrm_init_state.
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When ipcomp_tunnel_attach fails we will call ipcomp_destroy twice.
This may lead to double-frees on certain structures.

As there is no reason to explicitly call ipcomp_destroy, this patch
removes it from ipcomp*.c and lets the standard xfrm_state destruction
take place.

This is based on the discovery and patch by Alexey Dobriyan.
Tested-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Looks like an oversight in GRO design.
Signed-off-by: NStephen Hemminger <shemminger@vyatta.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add the required handlers to convert 32 bit
ebtables mark match and match target structs to 64bit layout.
Signed-off-by: NFlorian Westphal <fwestphal@astaro.com>

ebt_limit structure is larger on 64 bit systems due
to "long" type used in the (kernel-only) data section.

Setting .compatsize is enough in this case, these values
have no meaning in userspace.
Signed-off-by: NFlorian Westphal <fwestphal@astaro.com>

ebtables can be compiled to perform userspace-side padding of
structures. In that case, all the structures are already in the
'native' format expected by the kernel.

This tries to determine what format the userspace program is
using.

For most set/getsockopts, this can be done by checking
the len argument for sizeof(compat_ebt_replace) and
re-trying the native handler on error.

In case of EBT_SO_GET_ENTRIES, the native handler is tried first,
it will error out early when checking the *len argument
(the compat version has to defer this check until after
 iterating over the kernel data set once, to adjust for all
 the structure size differences).

As this would cause error printks, remove those as well, as
recommended by Bart de Schuymer.
Signed-off-by: NFlorian Westphal <fw@strlen.de>

Main code for 32 bit userland ebtables binary with 64 bit kernels
support.

Tested on x86_64 kernel only, using 64bit ebtables binary
for output comparision.

At least ebt_mark, m_mark and ebt_limit need CONFIG_COMPAT hooks, too.

remaining problem:

The ebtables userland makefile has:
ifeq ($(shell uname -m),sparc64)
	CFLAGS+=-DEBT_MIN_ALIGN=8 -DKERNEL_64_USERSPACE_32
endif

struct ebt_replace, ebt_entry_match etc. then contain userland-side
padding, i.e.  even if we are called from a 32 bit userland, the
structures may already be in the right format.

This problem is addressed in a follow-up patch.
Signed-off-by: NFlorian Westphal <fwestphal@astaro.com>

allows to call do_update_counters() from upcoming CONFIG_COMPAT
code instead of copy&pasting the same code.
Signed-off-by: NFlorian Westphal <fw@strlen.de>