Here is a patch which fixes an issue observed when using TCP over IPv6
and AH from IPsec.
When a connection gets closed the 4-way method and the last ACK from
the server gets dropped, the subsequent FINs from the client do not
get ACKed because tcp_v6_send_response does not set the transport
header pointer. This causes ah6_output to try to allocate a lot of
memory, which typically fails, so the ACKs never make it out of the
stack.
I have reproduced the problem on kernel 2.6.7, but after looking at
the latest kernel it seems the problem is still there.
Signed-off-by: Cosmin Ratiu <cratiu@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This fixed a lockdep warning which appeared when doing stress
memory tests over NFS:
inconsistent {RECLAIM_FS-ON-W} -> {IN-RECLAIM_FS-W} usage.
page reclaim => nfs_writepage => tcp_sendmsg => lock sk_lock
mount_root => nfs_root_data => tcp_close => lock sk_lock =>
tcp_send_fin => alloc_skb_fclone => page reclaim
David raised a concern that if the allocation fails in tcp_send_fin(), and it's
GFP_ATOMIC, we are going to yield() (which sleeps) and loop endlessly waiting
for the allocation to succeed.
But fact is, the original GFP_KERNEL also sleeps. GFP_ATOMIC+yield() looks
weird, but it is no worse the implicit sleep inside GFP_KERNEL. Both could
loop endlessly under memory pressure.
CC: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
CC: David S. Miller <davem@davemloft.net>
CC: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Christoph Lameter pointed out that packet drops at qdisc level where not
accounted in SNMP counters. Only if application sets IP_RECVERR, drops
are reported to user (-ENOBUFS errors) and SNMP counters updated.
IP_RECVERR is used to enable extended reliable error message passing,
but these are not needed to update system wide SNMP stats.
This patch changes things a bit to allow SNMP counters to be updated,
regardless of IP_RECVERR being set or not on the socket.
Example after an UDP tx flood
# netstat -s
...
IP:
1487048 outgoing packets dropped
...
Udp:
...
SndbufErrors: 1487048
send() syscalls, do however still return an OK status, to not
break applications.
Note : send() manual page explicitly says for -ENOBUFS error :
"The output queue for a network interface was full.
This generally indicates that the interface has stopped sending,
but may be caused by transient congestion.
(Normally, this does not occur in Linux. Packets are just silently
dropped when a device queue overflows.) "
This is not true for IP_RECVERR enabled sockets : a send() syscall
that hit a qdisc drop returns an ENOBUFS error.
Many thanks to Christoph, David, and last but not least, Alexey !
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
All instances of file_operations should be const.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The function block inet_connect_sock_af_ops contains no data
make it constant.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
qdisc drops should be notified to IP_RECVERR enabled sockets, as done in IPV4.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
These tables are never modified at runtime. Move to read-only
section.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
struct net::ipv6.ip6_dst_ops is separatedly dynamically allocated,
but there is no fundamental reason for it. Embed it directly into
struct netns_ipv6.
For that:
* move struct dst_ops into separate header to fix circular dependencies
I honestly tried not to, it's pretty impossible to do other way
* drop dynamical allocation, allocate together with netns
For a change, remove struct dst_ops::dst_net, it's deducible
by using container_of() given dst_ops pointer.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When processing a received IPv6 Router Advertisement, the kernel
creates or updates an IPv6 Neighbor Cache entry for the sender --
but presently this does not occur if IPv6 forwarding is enabled
(net.ipv6.conf.*.forwarding = 1), or if IPv6 Router Advertisements
are not accepted (net.ipv6.conf.*.accept_ra = 0), because in these
cases processing of the Router Advertisement has already halted.
This patch allows the Neighbor Cache to be updated in these cases,
while still avoiding any modification to routes or link parameters.
This continues to satisfy RFC 4861, since any entry created in the
Neighbor Cache as the result of a received Router Advertisement is
still placed in the STALE state.
Signed-off-by: David Ward <david.ward@ll.mit.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
if tunnel parameters have frag_off set to IP_DF, pmtudisc on the ipv4 link
will be performed by deriving the mtu from the ipv4 link and setting the
DF-Flag of the encapsulating IPv4 Header. If fragmentation is needed on the
way, the IPv4 pmtu gets adjusted, the ipv6 package will be resent eventually,
using the new and lower mtu and everyone is happy.
If the frag_off parameter is unset, the mtu for the tunnel will be derived
from the tunnel device or the ipv6 pmtu, which might be higher than the ipv4
pmtu. In that case we must allow the fragmentation of the IPv4 packet because
the IPv6 mtu wouldn't 'learn' from the adjusted IPv4 pmtu, resulting in
frequent icmp_frag_needed and package loss on the IPv6 layer.
This patch allows fragmentation when tunnel was created with parameter
nopmtudisc, like in ipip/gre tunnels.
Signed-off-by: Sascha Hlusiak <contact@saschahlusiak.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 63d9950b08
(ipv6: Make v4-mapped bindings consistent with IPv4)
changes behavior of inet6_bind() for v4-mapped addresses so it should
behave the same way as inet_bind().
During this change setting of err to -EADDRNOTAVAIL got lost:
af_inet.c:469 inet_bind()
err = -EADDRNOTAVAIL;
if (!sysctl_ip_nonlocal_bind &&
!(inet->freebind || inet->transparent) &&
addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
chk_addr_ret != RTN_LOCAL &&
chk_addr_ret != RTN_MULTICAST &&
chk_addr_ret != RTN_BROADCAST)
goto out;
af_inet6.c:463 inet6_bind()
if (addr_type == IPV6_ADDR_MAPPED) {
int chk_addr_ret;
/* Binding to v4-mapped address on a v6-only socket
* makes no sense
*/
if (np->ipv6only) {
err = -EINVAL;
goto out;
}
/* Reproduce AF_INET checks to make the bindings consitant */
v4addr = addr->sin6_addr.s6_addr32[3];
chk_addr_ret = inet_addr_type(net, v4addr);
if (!sysctl_ip_nonlocal_bind &&
!(inet->freebind || inet->transparent) &&
v4addr != htonl(INADDR_ANY) &&
chk_addr_ret != RTN_LOCAL &&
chk_addr_ret != RTN_MULTICAST &&
chk_addr_ret != RTN_BROADCAST)
goto out;
} else {
Signed-off-by Bruno Prémont <bonbons@linux-vserver.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch addresses:
* assigning -1 to np->tclass as it is currently done is not very meaningful,
since it turns into 0xff;
* RFC 3542, 6.5 allows -1 for clearing the sticky IPV6_TCLASS option
and specifies -1 to mean "use kernel default":
- RFC 2460, 7. requires that the default traffic class must be zero for
all 8 bits,
- this is consistent with RFC 2474, 4.1 which recommends a default PHB of 0,
in combination with a value of the ECN field of "non-ECT" (RFC 3168, 5.).
This patch changes the meaning of -1 from assigning 255 to mean the RFC 2460
default, which at the same time allows to satisfy clearing the sticky TCLASS
option as per RFC 3542, 6.5.
(When passing -1 as ancillary data, the fallback remains np->tclass, which
has either been set via socket options, or contains the default value.)
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
This replaces assignments of the type "int on LHS" = "u8 on RHS" with
simpler code. The LHS can express all of the unsigned right hand side
values, hence the assigned value can not be negative.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
If an interface has multiple addresses, the current message for DAD
failure isn't really helpful, so this patch adds the address itself to
the printk.
Signed-off-by: Jens Rosenboom <jens@mcbone.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
String literals are constant, and usually, we can also tag the array
of pointers const too, moving it to the .rodata section.
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
This renames away a variable clash:
* ipv6_table[] is declared as a static global table;
* ipv6_sysctl_net_init() uses ipv6_table to refer/destroy dynamic memory;
* ipv6_sysctl_net_exit() also uses ipv6_table for the same purpose;
* both the two last functions call kfree() on ipv6_table.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Choose saner defaults for xfrm[4|6] gc_thresh values on init
Currently, the xfrm[4|6] code has hard-coded initial gc_thresh values
(set to 1024). Given that the ipv4 and ipv6 routing caches are sized
dynamically at boot time, the static selections can be non-sensical.
This patch dynamically selects an appropriate gc threshold based on
the corresponding main routing table size, using the assumption that
we should in the worst case be able to handle as many connections as
the routing table can.
For ipv4, the maximum route cache size is 16 * the number of hash
buckets in the route cache. Given that xfrm4 starts garbage
collection at the gc_thresh and prevents new allocations at 2 *
gc_thresh, we set gc_thresh to half the maximum route cache size.
For ipv6, its a bit trickier. there is no maximum route cache size,
but the ipv6 dst_ops gc_thresh is statically set to 1024. It seems
sane to select a simmilar gc_thresh for the xfrm6 code that is half
the number of hash buckets in the v6 route cache times 16 (like the v4
code does).
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Export garbage collector thresholds for xfrm[4|6]_dst_ops
Had a problem reported to me recently in which a high volume of ipsec
connections on a system began reporting ENOBUFS for new connections
eventually.
It seemed that after about 2000 connections we started being unable to
create more. A quick look revealed that the xfrm code used a dst_ops
structure that limited the gc_thresh value to 1024, and always
dropped route cache entries after 2x the gc_thresh.
It seems the most direct solution is to export the gc_thresh values in
the xfrm[4|6] dst_ops as sysctls, like the main routing table does, so
that higher volumes of connections can be supported. This patch has
been tested and allows the reporter to increase their ipsec connection
volume successfully.
Reported-by: Joe Nall <joe@nall.com>
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
ipv4/xfrm4_policy.c | 18 ++++++++++++++++++
ipv6/xfrm6_policy.c | 18 ++++++++++++++++++
2 files changed, 36 insertions(+)
Signed-off-by: David S. Miller <davem@davemloft.net>
The local variable 'idev' shadows the function argument 'idev' to
ip6_mc_add_src(). Fixed by removing the local declaration, as pmc->idev
should be identical with 'idev' passed as argument.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: David L Stevens <dlstevens@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the TCP connection handshake completes on the passive
side, a variety of state must be set up in the "child" sock,
including the key if MD5 authentication is being used. Fix TCP
for both address families to label the key with the peer's
destination address, rather than the address from the listening
sock, which is usually the wildcard.
Reported-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: John Dykstra <john.dykstra1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix MD5 signature checking so that an IPv4 active open
to an IPv6 socket can succeed. In particular, use the
correct address family's signature generation function
for the SYN/ACK.
Reported-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: John Dykstra <john.dykstra1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- validate and forward GSO UDP/IPv6 packets from untrusted sources.
- do software UFO if the outgoing device doesn't support UFO.
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
- move ipv6_select_ident() inline function to ipv6.h and remove the unused
skb argument
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
- fix gso_size setting for ipv6 fragment to be a multiple of 8 bytes.
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
- add HW checksum support for outgoing large UDP/IPv6 packets destined for
a UFO enabled device.
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
The sit module makes use of skb->dst in it's xmit function, so since
93f154b594 ("net: release dst entry in dev_hard_start_xmit()") sit
tunnels are broken, because the flag IFF_XMIT_DST_RELEASE is not
unset.
This patch unsets that flag for sit devices to fix this
regression.
Signed-off-by: Sascha Hlusiak <contact@saschahlusiak.de>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After commit 2b85a34e91
(net: No more expensive sock_hold()/sock_put() on each tx)
we do not take any more references on sk->sk_refcnt on outgoing packets.
I forgot to delete two __sock_put() from ip_push_pending_frames()
and ip6_push_pending_frames().
Reported-by: Emil S Tantilov <emils.tantilov@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Tested-by: Emil S Tantilov <emils.tantilov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The routine ipv6_rcv() uses magic number 0 for a return when it drops a
packet. This corresponds to NET_RX_SUCCESS, which is obviously
incorrect. Correct this by using NET_RX_DROP instead.
ps. It isn't exactly clear who the IPv6 maintainers are, apologies if
I've missed any.
Signed-off-by: Mark Smith <markzzzsmith@yahoo.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is the result of an automatic spatch transformation to convert
all ndo_start_xmit() return values of 0 to NETDEV_TX_OK.
Some occurences are missed by the automatic conversion, those will be
handled in a seperate patch.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
There's a bug in addrconf_prefix_rcv() where it won't update the
preferred lifetime of an IPv6 address if the current valid lifetime
of the address is less than 2 hours (the minimum value in the RA).
For example, If I send a router advertisement with a prefix that
has valid lifetime = preferred lifetime = 2 hours we'll build
this address:
3: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qlen 1000
inet6 2001:1890:1109:a20:217:8ff:fe7d:4718/64 scope global dynamic
valid_lft 7175sec preferred_lft 7175sec
If I then send the same prefix with valid lifetime = preferred
lifetime = 0 it will be ignored since the minimum valid lifetime
is 2 hours:
3: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qlen 1000
inet6 2001:1890:1109:a20:217:8ff:fe7d:4718/64 scope global dynamic
valid_lft 7161sec preferred_lft 7161sec
But according to RFC 4862 we should always reset the preferred lifetime
even if the valid lifetime is invalid, which would cause the address
to immediately get deprecated. So with this patch we'd see this:
5: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qlen 1000
inet6 2001:1890:1109:a20:21f:29ff:fe5a:ef04/64 scope global deprecated dynamic
valid_lft 7163sec preferred_lft 0sec
The comment winds-up being 5x the size of the code to fix the problem.
Update the preferred lifetime of IPv6 addresses derived from a prefix
info option in a router advertisement even if the valid lifetime in
the option is invalid, as specified in RFC 4862 Section 5.5.3e. Fixes
an issue where an address will not immediately become deprecated.
Reported by Jens Rosenboom.
Signed-off-by: Brian Haley <brian.haley@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The SCTP pushed the skb above the sctp chunk header, so the
check of pskb_may_pull(skb, nh + offset + 1 - skb->data) in
_decode_session6() will never return 0 and the ports decode
of sctp will always fail. (nh + offset + 1 - skb->data < 0)
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
As transparent proxying looks up the socket early and assigns
it to the skb for later processing, we must drop any existing
socket ownership prior to that in order to distinguish between
the case where tproxy is active and where it is not.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ipv6 module uses rcu_call() thus it should use rcu_barrier() on
module unload.
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Avoid showing wrong high values when the preferred lifetime of an address
is expired.
Signed-off-by: Jens Rosenboom <me@jayr.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change all the code that deals directly with ICMPv6 type and code
values to use u8 instead of a signed int as that's the actual data
type.
Signed-off-by: Brian Haley <brian.haley@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 2b85a34e91
(net: No more expensive sock_hold()/sock_put() on each tx)
changed initial sk_wmem_alloc value.
We need to take into account this offset when reporting
sk_wmem_alloc to user, in PROC_FS files or various
ioctls (SIOCOUTQ/TIOCOUTQ)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
IPv4:
- make PIM register vifs netns local
- set the netns when a PIM register vif is created
- make PIM available in all network namespaces (if CONFIG_IP_PIMSM_V2)
by adding the protocol handler when multicast routing is initialized
IPv6:
- make PIM register vifs netns local
- make PIM available in all network namespaces (if CONFIG_IPV6_PIMSM_V2)
by adding the protocol handler when multicast routing is initialized
Signed-off-by: Tom Goff <thomas.goff@boeing.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
.ko is normally not included in Kconfig help, make it consistent.
Signed-off-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
One of the problem with sock memory accounting is it uses
a pair of sock_hold()/sock_put() for each transmitted packet.
This slows down bidirectional flows because the receive path
also needs to take a refcount on socket and might use a different
cpu than transmit path or transmit completion path. So these
two atomic operations also trigger cache line bounces.
We can see this in tx or tx/rx workloads (media gateways for example),
where sock_wfree() can be in top five functions in profiles.
We use this sock_hold()/sock_put() so that sock freeing
is delayed until all tx packets are completed.
As we also update sk_wmem_alloc, we could offset sk_wmem_alloc
by one unit at init time, until sk_free() is called.
Once sk_free() is called, we atomic_dec_and_test(sk_wmem_alloc)
to decrement initial offset and atomicaly check if any packets
are in flight.
skb_set_owner_w() doesnt call sock_hold() anymore
sock_wfree() doesnt call sock_put() anymore, but check if sk_wmem_alloc
reached 0 to perform the final freeing.
Drawback is that a skb->truesize error could lead to unfreeable sockets, or
even worse, prematurely calling __sk_free() on a live socket.
Nice speedups on SMP. tbench for example, going from 2691 MB/s to 2711 MB/s
on my 8 cpu dev machine, even if tbench was not really hitting sk_refcnt
contention point. 5 % speedup on a UDP transmit workload (depends
on number of flows), lowering TX completion cpu usage.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
