Like many other places, we have to check that the array index is
within allowed limits, or otherwise, a kernel oops and other nastiness
can ensue when we access memory beyond the end of the array.
[ 5954.115381] BUG: unable to handle kernel paging request at 0000004000000000
[ 5954.120014] IP: __find_logger+0x6f/0xa0
[ 5954.123979] nf_log_bind_pf+0x2b/0x70
[ 5954.123979] nfulnl_recv_config+0xc0/0x4a0 [nfnetlink_log]
[ 5954.123979] nfnetlink_rcv_msg+0x12c/0x1b0 [nfnetlink]
...
The problem goes back to v2.6.30-rc1~1372~1342~31 where nf_log_bind
was decoupled from nf_log_register.
Reported-by: Miguel Di Ciurcio Filho <miguel.filho@gmail.com>,
via irc.freenode.net/#netfilter
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Fix dst_lock usage in __ip_vs_update_dest. We need
_bh locking because destination is updated in user context.
Can cause lockups on frequent destination updates.
Problem reported by Simon Kirby. Bug was introduced
in 2.6.37 from the "ipvs: changes for local real server"
change.
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Hans Schillstrom <hans@schillstrom.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
This patch fixes the out of sync scenarios while in SYN_RECV state.
Quoting Jozsef, what it happens if we are out of sync if the
following:
> > b. conntrack entry is outdated, new SYN received
> > - (b1) we ignore it but save the initialization data from it
> > - (b2) when the reply SYN/ACK receives and it matches the saved data,
> > we pick up the new connection
This is what it should happen if we are in SYN_RECV state. Initially,
the SYN packet hits b1, thus we save data from it. But the SYN/ACK
packet is considered a retransmission given that we're in SYN_RECV
state. Therefore, we never hit b2 and we don't get in sync. To fix
this, we ignore SYN/ACK if we are in SYN_RECV. If the previous packet
was a SYN, then we enter the ignore case that get us in sync.
This patch helps a lot to conntrackd in stress scenarios (assumming a
client that generates lots of small TCP connections). During the failover,
consider that the new primary has injected one outdated flow in SYN_RECV
state (this is likely to happen if the conntrack event rate is high
because the backup will be a bit delayed from the primary). With the
current code, if the client starts a new fresh connection that matches
the tuple, the SYN packet will be ignored without updating the state
tracking, and the SYN+ACK in reply will blocked as it will not pass
checkings III or IV (since all state tracking in the original direction
is not initialized because of the SYN packet was ignored and the ignore
case that get us in sync is not applied).
I posted a couple of patches before this one. Changli Gao spotted
a simpler way to fix this problem. This patch implements his idea.
Cc: Changli Gao <xiaosuo@gmail.com>
Cc: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
lc and wlc use the same formula, but lblc and lblcr use another one. There
is no reason for using two different formulas for the lc variants.
The formula used by lc is used by all the lc variants in this patch.
Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Acked-by: Wensong Zhang <wensong@linux-vs.org>
Signed-off-by: Simon Horman <horms@verge.net.au>
Assigning a socket in timewait state to skb->sk can trigger
kernel oops, e.g. in nfnetlink_log, which does:
if (skb->sk) {
read_lock_bh(&skb->sk->sk_callback_lock);
if (skb->sk->sk_socket && skb->sk->sk_socket->file) ...
in the timewait case, accessing sk->sk_callback_lock and sk->sk_socket
is invalid.
Either all of these spots will need to add a test for sk->sk_state != TCP_TIME_WAIT,
or xt_TPROXY must not assign a timewait socket to skb->sk.
This does the latter.
If a TW socket is found, assign the tproxy nfmark, but skip the skb->sk assignment,
thus mimicking behaviour of a '-m socket .. -j MARK/ACCEPT' re-routing rule.
The 'SYN to TW socket' case is left unchanged -- we try to redirect to the
listener socket.
Cc: Balazs Scheidler <bazsi@balabit.hu>
Cc: KOVACS Krisztian <hidden@balabit.hu>
Signed-off-by: Florian Westphal <fwestphal@astaro.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
When IP_VS schedulers do not find a destination, they output a terse
"WLC: no destination available" message through kernel syslog, which I
can not only make sense of because syslog puts them in a logfile
together with keepalived checker results.
This patch makes the output a bit more informative, by telling you which
virtual service failed to find a destination.
Example output:
kernel: [1539214.552233] IPVS: wlc: TCP 192.168.8.30:22 - no destination available
kernel: [1539299.674418] IPVS: wlc: FWM 22 0x00000016 - no destination available
I have tested the code for IPv4 and FWM services, as you can see from
the example; I do not have an IPv6 setup to test the third code path
with.
To avoid code duplication, I put a new function ip_vs_scheduler_err()
into ip_vs_sched.c, and use that from the schedulers instead of calling
IP_VS_ERR_RL directly.
Signed-off-by: Patrick Schaaf <netdev@bof.de>
Signed-off-by: Simon Horman <horms@verge.net.au>
Remove code that should not be called anymore.
Now when ip_vs_out handles replies for local clients at
LOCAL_IN hook we do not need to call conn_out_get and
handle_response_icmp from ip_vs_in_icmp* because such
lookups were already performed for the ICMP packet and no
connection was found.
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Fix get_curr_sync_buff to keep buffer for 2 seconds
as intended, not just for the current jiffie. By this way
we will sync more connection structures with single packet.
Signed-off-by: Tinggong Wang <wangtinggong@gmail.com>
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
As noticed by Eric, nf_iterate doesn't use RCU correctly by
accessing the prev pointer of a RCU protected list element when
a verdict of NF_REPEAT is issued.
Fix by jumping backwards to the hook invocation directly instead
of loading the previous list element before continuing the list
iteration.
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
This reverts commit 44bd4de9c2.
I have to revert the early loop termination in connlimit since it generates
problems when an iptables statement does not use -m state --state NEW before
the connlimit match extension.
Signed-off-by: Stefan Berger <stefanb@linux.vnet.ibm.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The patch below introduces an early termination of the loop that is
counting matches. It terminates once the counter has exceeded the
threshold provided by the user. There's no point in continuing the loop
afterwards and looking at other entries.
It plays together with the following code further below:
return (connections > info->limit) ^ info->inverse;
where connections is the result of the counted connection, which in turn
is the matches variable in the loop. So once
-> matches = info->limit + 1
alias -> matches > info->limit
alias -> matches > threshold
we can terminate the loop.
Signed-off-by: Stefan Berger <stefanb@linux.vnet.ibm.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
When SYSCTL and PROC_FS and NETFILTER_NETLINK are not enabled:
net/built-in.o: In function `try_to_load_type':
ip_set_core.c:(.text+0x3ab49): undefined reference to `nfnl_unlock'
ip_set_core.c:(.text+0x3ab4e): undefined reference to `nfnl_lock'
...
Reported-by: Randy Dunlap <randy.dunlap@oracle.com>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The TCP tracking code has a special case that allows to return
NF_REPEAT if we receive a new SYN packet while in TIME_WAIT state.
In this situation, the TCP tracking code destroys the existing
conntrack to start a new clean session.
[DESTROY] tcp 6 src=192.168.0.2 dst=192.168.1.2 sport=38925 dport=8000 src=192.168.1.2 dst=192.168.1.100 sport=8000 dport=38925 [ASSURED]
[NEW] tcp 6 120 SYN_SENT src=192.168.0.2 dst=192.168.1.2 sport=38925 dport=8000 [UNREPLIED] src=192.168.1.2 dst=192.168.1.100 sport=8000 dport=38925
However, this is a problem for the iptables' CT target event filtering
which will not work in this case since the conntrack template will not
be there for the new session. To fix this, we reassign the conntrack
template to the packet if we return NF_REPEAT.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
'!' has higher precedence than '&'. IP_VS_STATE_MASTER is 0x1 so
the original code is equivelent to if (!ipvs->sync_state) ...
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Hans Schillstrom <hans.schillstrom@ericsson.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Use sctp_app_lock instead of tcp_app_lock in the SCTP protocol module.
This appears to be a typo introduced by the netns changes.
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Hans Schillstrom <hans.schillstrom@ericsson.com>
Add a new 'devgroup' match to match on the device group of the
incoming and outgoing network device of a packet.
Signed-off-by: Patrick McHardy <kaber@trash.net>
When a message carries multiple commands and one of them triggers
an error, we have to report to the userspace which one was that.
The line number of the command plays this role and there's an attribute
reserved in the header part of the message to be filled out with the error
line number. In order not to modify the original message received from
the userspace, we construct a new, complete netlink error message and
modifies the attribute there, then send it.
Netlink is notified not to send its ACK/error message.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu
Signed-off-by: Patrick McHardy <kaber@trash.net>
Add a dummy ip_set_get_ip6_port function that unconditionally
returns false for CONFIG_IPV6=n and convert the real function
to ipv6_skip_exthdr() to avoid pulling in the ip6_tables module
when loading ipset.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Don't fall through in the switch statement, otherwise IPv4 headers
are incorrectly parsed again as IPv6 and the return value will always
be 'false'.
Signed-off-by: Patrick McHardy <kaber@trash.net>
ip_vs_sync_cleanup() may be called from ip_vs_init() on error
and thus needs to be accesible from section __init
Reporte-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Hans Schillstrom <hans@schillstrom.com>
Tested-by: Hans Schillstrom <hans@schillstrom.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
This is a rather naieve approach to allowing PVS to compile with
CONFIG_SYSCTL disabled. I am working on a more comprehensive patch which
will remove compilation of all sysctl-related IPVS code when CONFIG_SYSCTL
is disabled.
Reported-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Hans Schillstrom <hans@schillstrom.com>
Tested-by: Hans Schillstrom <hans@schillstrom.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Simon Horman <horms@verge.net.au>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Hans Schillstrom <hans@schillstrom.com>
Tested-by: Hans Schillstrom <hans@schillstrom.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Reported-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Hans Schillstrom <hans@schillstrom.com>
Tested-by: Hans Schillstrom <hans@schillstrom.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
net/netfilter/nf_conntrack_netlink.c: In function 'ctnetlink_parse_tuple':
net/netfilter/nf_conntrack_netlink.c:832:11: warning: comparison between 'enum ctattr_tuple' and 'enum ctattr_type'
Use ctattr_type for the 'type' parameter since that's the type of all attributes
passed to this function.
Signed-off-by: Patrick McHardy <kaber@trash.net>
None of the set types need uaccess.h since this is handled centrally
in ip_set_core. Most set types additionally don't need bitops.h and
spinlock.h since they use neither. tcp.h is only needed by those
using before(), udp.h is not needed at all.
Signed-off-by: Patrick McHardy <kaber@trash.net>
For the following rule:
iptables -I PREROUTING -t raw -j CT --ctevents assured
The event delivered looks like the following:
[UPDATE] tcp 6 src=192.168.0.2 dst=192.168.1.2 sport=37041 dport=80 src=192.168.1.2 dst=192.168.1.100 sport=80 dport=37041 [ASSURED]
Note that the TCP protocol state is not included. For that reason
the CT event filtering is not very useful for conntrackd.
To resolve this issue, instead of conditionally setting the CT events
bits based on the ctmask, we always set them and perform the filtering
in the late stage, just before the delivery.
Thus, the event delivered looks like the following:
[UPDATE] tcp 6 432000 ESTABLISHED src=192.168.0.2 dst=192.168.1.2 sport=37041 dport=80 src=192.168.1.2 dst=192.168.1.100 sport=80 dport=37041 [ASSURED]
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The patch adds the combined module of the "SET" target and "set" match
to netfilter. Both the previous and the current revisions are supported.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The module implements the list:set type support in two flavours:
without and with timeout. The sets has two sides: for the userspace,
they store the names of other (non list:set type of) sets: one can add,
delete and test set names. For the kernel, it forms an ordered union of
the member sets: the members sets are tried in order when elements are
added, deleted and tested and the process stops at the first success.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The module implements the hash:net,port type support in four flavours:
for IPv4 and IPv6, both without and with timeout support. The elements
are two dimensional: IPv4/IPv6 network address/prefix and protocol/port
pairs.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The module implements the hash:net type support in four flavours:
for IPv4 and IPv6, both without and with timeout support. The elements
are one dimensional: IPv4/IPv6 network address/prefixes.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The module implements the hash:ip,port,net type support in four flavours:
for IPv4 and IPv6, both without and with timeout support. The elements
are three dimensional: IPv4/IPv6 address, protocol/port and IPv4/IPv6
network address/prefix triples. The different prefixes are searched/matched
from the longest prefix to the shortes one (most specific to least).
In other words the processing time linearly grows with the number of
different prefixes in the set.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The module implements the hash:ip,port,ip type support in four flavours:
for IPv4 and IPv6, both without and with timeout support. The elements
are three dimensional: IPv4/IPv6 address, protocol/port and IPv4/IPv6
address triples.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The module implements the hash:ip,port type support in four flavours:
for IPv4 and IPv6, both without and with timeout support. The elements
are two dimensional: IPv4/IPv6 address and protocol/port pairs. The port
is interpeted for TCP, UPD, ICMP and ICMPv6 (at the latters as type/code
of course).
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The module implements the hash:ip type support in four flavours:
for IPv4 or IPv6, both without and with timeout support.
All the hash types are based on the "array hash" or ahash structure
and functions as a good compromise between minimal memory footprint
and speed. The hashing uses arrays to resolve clashes. The hash table
is resized (doubled) when searching becomes too long. Resizing can be
triggered by userspace add commands only and those are serialized by
the nfnl mutex. During resizing the set is read-locked, so the only
possible concurrent operations are the kernel side readers. Those are
protected by RCU locking.
Because of the four flavours and the other hash types, the functions
are implemented in general forms in the ip_set_ahash.h header file
and the real functions are generated before compiling by macro expansion.
Thus the dereferencing of low-level functions and void pointer arguments
could be avoided: the low-level functions are inlined, the function
arguments are pointers of type-specific structures.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The module implements the bitmap:port type in two flavours, without
and with timeout support to store TCP/UDP ports from a range.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The module implements the bitmap:ip,mac set type in two flavours,
without and with timeout support. In this kind of set one can store
IPv4 address and (source) MAC address pairs. The type supports elements
added without the MAC part filled out: when the first matching from kernel
happens, the MAC part is automatically filled out. The timing out of the
elements stars when an element is complete in the IP,MAC pair.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The module implements the bitmap:ip set type in two flavours, without
and with timeout support. In this kind of set one can store IPv4
addresses (or network addresses) from a given range.
In order not to waste memory, the timeout version does not rely on
the kernel timer for every element to be timed out but on garbage
collection. All set types use this mechanism.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
