I was very frustrated about the fact that I have to recompile the kernel
to change the hash size. So, I created this patch.

If IPVS is built-in you can append ip_vs.conn_tab_bits=?? to kernel
command line, or, if you built IPVS as modules, you can add
options ip_vs conn_tab_bits=??.

To keep everything backward compatible, you still can select the size at
compile time, and that will be used as default.

It has been about a year since this patch was originally posted
and subsequently dropped on the basis of insufficient test data.

Mark Bergsma has provided the following test results which seem
to strongly support the need for larger hash table sizes:

We do however run into the same problem with the default setting (212 =
4096 entries), as most of our LVS balancers handle around a million
connections/SLAB entries at any point in time (around 100-150 kpps
load). With only 4096 hash table entries this implies that each entry
consists of a linked list of 256 connections *on average*.

To provide some statistics, I did an oprofile run on an 2.6.31 kernel,
with both the default 4096 table size, and the same kernel recompiled
with IP_VS_CONN_TAB_BITS set to 18 (218 = 262144 entries). I built a
quick test setup with a part of Wikimedia/Wikipedia's live traffic
mirrored by the switch to the test host.

With the default setting, at ~ 120 kpps packet load we saw a typical %si
CPU usage of around 30-35%, and oprofile reported a hot spot in
ip_vs_conn_in_get:

samples  %        image name               app name
symbol name
1719761  42.3741  ip_vs.ko                 ip_vs.ko      ip_vs_conn_in_get
302577    7.4554  bnx2                     bnx2          /bnx2
181984    4.4840  vmlinux                  vmlinux       __ticket_spin_lock
128636    3.1695  vmlinux                  vmlinux       ip_route_input
74345     1.8318  ip_vs.ko                 ip_vs.ko      ip_vs_conn_out_get
68482     1.6874  vmlinux                  vmlinux       mwait_idle

After loading the recompiled kernel with 218 entries, %si CPU usage
dropped in half to around 12-18%, and oprofile looks much healthier,
with only 7% spent in ip_vs_conn_in_get:

samples  %        image name               app name
symbol name
265641   14.4616  bnx2                     bnx2         /bnx2
143251    7.7986  vmlinux                  vmlinux      __ticket_spin_lock
140661    7.6576  ip_vs.ko                 ip_vs.ko     ip_vs_conn_in_get
94364     5.1372  vmlinux                  vmlinux      mwait_idle
86267     4.6964  vmlinux                  vmlinux      ip_route_input

[ horms@verge.net.au: trivial up-port and minor style fixes ]
Signed-off-by: NCatalin(ux) M. BOIE <catab@embedromix.ro>
Cc: Mark Bergsma <mark@wikimedia.org>
Signed-off-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NPatrick McHardy <kaber@trash.net>

This cleanup patch puts struct/union/enum opening braces,
in first line to ease grep games.

struct something
{

becomes :

struct something {
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid 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: NJan Engelhardt <jengelh@medozas.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since pr_err and friends are used instead of printk there is no point
in keeping IP_VS_ERR and friends.  Furthermore make use of '__func__'
instead of hard coded function names.
Signed-off-by: NHannes Eder <heder@google.com>
Acked-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

While being at it cleanup whitespace.
Signed-off-by: NHannes Eder <heder@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Base versions handle constant folding now.  For headers exposed to
userspace, we must only expose the __ prefixed versions.
Signed-off-by: NHarvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The first argument to csum_partial is const void *
casts to char/u8 * are not necessary
Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Remove the 'supports_ipv6' scheduler flag since all schedulers now
support IPv6.
Signed-off-by: NJulius Volz <julius.volz@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Using NIPQUAD() with NIPQUAD_FMT, %d.%d.%d.%d or %u.%u.%u.%u
can be replaced with %pI4
Signed-off-by: NHarvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NHarvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NHarvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

__FUNCTION__ is gcc-specific, use __func__
Signed-off-by: NHarvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

inet_iif() in inet_sock.h requires route.h. Since users of inet_iif()
usually require other route.h functionality anyway this patch moves
inet_iif() to route.h.
Signed-off-by: NKOVACS Krisztian <hidden@sch.bme.hu>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Instead of duplicating the fields, integrate a user stats structure into
the kernel stats structure. This is more robust when the members are
changed, because they are now automatically kept in sync.
Signed-off-by: NSven Wegener <sven.wegener@stealer.net>
Reviewed-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Instead of checking the value in include/net/ip_vs.h, we can just
restrict the range in our Kconfig file. This will prevent values outside
of the range early.
Signed-off-by: NSven Wegener <sven.wegener@stealer.net>
Reviewed-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Adjust various debug outputs to use the new *_BUF macro variants for
correct output of v4/v6 addresses.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Convert functions for looking up destinations (real servers) to support
IPv6 services/dests.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Add xmit functions for IPv6. Also add the already needed __ip_vs_get_out_rt_v6()
to ip_vs_core.c. Bind the new xmit functions to v6 connections.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Extend functions for getting/creating connections and connection
templates for IPv6 support and fix the callers.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Extend protocol DNAT/SNAT and state handlers to work with IPv6. Also
change/introduce new checksumming helper functions for this.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Add 'af' arguments to conn_schedule(), conn_in_get(), conn_out_get() and
csum_check() function pointers in struct ip_vs_protocol. Extend the
respective functions for TCP, UDP, AH and ESP and adjust the callers.

The changes in the callers need to be somewhat extensive, since they now
need to pass a filled out struct ip_vs_iphdr * to the modified functions
instead of a struct iphdr *.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Add 'supports_ipv6' flag to struct ip_vs_scheduler to indicate whether a
scheduler supports IPv6. Set the flag to 1 in schedulers that work with
IPv6, 0 otherwise. This flag is checked in a later patch while trying to
add a service with a specific scheduler. Adjust debug in v6-supporting
schedulers to work with both address families.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Add support for selecting services based on their address family to
ip_vs_service_get() and adjust the callers.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Add extended internal versions of struct ip_vs_service_user and struct
ip_vs_dest_user (the originals can't be modified as they are part
of the old sockopt interface). Adjust ip_vs_ctl.c to work with the new
data structures and add some minor AF-awareness.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Add some debugging macros that allow conditional output of either v4 or v6
addresses, depending on an 'af' parameter. This is done by creating a
temporary string buffer in an outer debug macro and writing addresses'
string representations into it from another macro which can only be used
when inside the outer one.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Add a struct ip_vs_iphdr for easier handling of common v4 and v6 header
fields in the same code path. ip_vs_fill_iphdr() helps to fill this struct
from an IPv4 or IPv6 header. Add further helper functions for copying and
comparing addresses.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Introduce new 'af' fields into IPVS data structures for specifying an
entry's address family. Convert IP addresses to be of type union
nf_inet_addr.
Signed-off-by: NJulius Volz <juliusv@google.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>

Commit 8ab19ea3 ("ipvs: Fix possible deadlock
in estimator code") fixed a deadlock condition, but that condition can only
happen during unload of IPVS, because during normal operation there is at least
our global stats structure in the estimator list. The mod_timer() and
del_timer_sync() calls are actually initialization and cleanup code in
disguise. Let's make it explicit and move them to their own init and cleanup
function.
Signed-off-by: NSven Wegener <sven.wegener@stealer.net>
Signed-off-by: NSimon Horman <horms@verge.net.au>

There's no reason for dynamically allocating an estimator object for every
stats object. Directly embed an estimator object into every stats object and
switch to using the kernel-provided list implementation. This makes the code
much simpler and faster, as we do not need to traverse the list of all
estimators to find the one belonging to a stats object. There's no need to use
an rwlock, as we only have one reader. Also reorder the members of the
estimator structure slightly to avoid padding overhead. This can't be done
with the stats object as the members are currently copied to our user space
object via memcpy() and changing it would break ABI.
Signed-off-by: NSven Wegener <sven.wegener@stealer.net>
Acked-by: NSimon Horman <horms@verge.net.au>

Signed-off-by: NSven Wegener <sven.wegener@stealer.net>
Acked-by: NSimon Horman <horms@verge.net.au>

Signed-off-by: NSven Wegener <sven.wegener@stealer.net>
Acked-by: NSimon Horman <horms@verge.net.au>

Current versions of ipvsadm include "/usr/src/linux/include/net/ip_vs.h"
directly. This file also contains kernel-only definitions. Normally, public
definitions should live in include/linux, so this patch moves the
definitions shared with userspace to a new file, "include/linux/ip_vs.h".

This also removes the unused NFC_IPVS_PROPERTY bitmask, which was once
used to point into skb->nfcache.

To make old ipvsadms still compile with this, the old header file includes
the new one.

Thanks to Dave Miller and Horms for noting/adding the missing Kbuild entry
for the new header file.
Signed-off-by: NJulius Volz <juliusv@google.com>
Acked-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

	Fixes bug http://bugzilla.kernel.org/show_bug.cgi?id=10556
where conn templates with protocol=IPPROTO_IP can oops backup box.

        Result from ip_vs_proto_get() should be checked because
protocol value can be invalid or unsupported in backup. But
for valid message we should not fail for templates which use
IPPROTO_IP. Also, add checks to validate message limits and
connection state. Show state NONE for templates using IPPROTO_IP.
Signed-off-by: NJulian Anastasov <ja@ssi.bg>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The feature of ipvs ctls is that the net/ipv4/vs path
is common for core ipvs ctls and for two schedulers,
so I make it exported and re-use it in modules.

Two other .c files required linux/sysctl.h to make the
extern declaration of this path compile well.
Signed-off-by: NPavel Emelyanov <xemul@openvz.org>
Acked-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

[IPVS]: Remove declaration of unimplemented method and remove unused definition from include/net/ip_vs.h

In include/net/ip_vs.h:
- The ip_vs_secure_tcp_set() method is not implemented anywhere.
- IP_VS_APP_TYPE_FTP is an unused definition.
Signed-off-by: NRami Rosen <ramirose@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Switch the remaining IPVS sysctl entries over to to use CTL_UNNUMBERED,
I stronly doubt that anyone is using the sys_sysctl interface to
these variables.
Signed-off-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

sysctl table check failed: /net/ipv4/vs/lblc_expiration .3.5.21.19 Missing strategy
[...]
sysctl table check failed: /net/ipv4/vs/lblcr_expiration .3.5.21.20 Missing strategy

Switch these entried over to use CTL_UNNUMBERED as clearly
the sys_syscal portion wasn't working.

This is along the same lines as Christian Borntraeger's patch that fixes
up entries with no stratergy in net/ipv4/ipvs/ip_vs_ctl.c
Signed-off-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Running the latest git code I get the following messages during boot:
sysctl table check failed: /net/ipv4/vs/drop_entry .3.5.21.4 Missing strategy
[...]		  
sysctl table check failed: /net/ipv4/vs/drop_packet .3.5.21.5 Missing strategy
[...]
sysctl table check failed: /net/ipv4/vs/secure_tcp .3.5.21.6 Missing strategy
[...]
sysctl table check failed: /net/ipv4/vs/sync_threshold .3.5.21.24 Missing strategy

I removed the binary sysctl handler for those messages and also removed
the definitions in ip_vs.h. The alternative would be to implement a 
proper strategy handler, but syscall sysctl is deprecated.

There are other sysctl definitions that are commented out or work with 
the default sysctl_data strategy. I did not touch these. 
Signed-off-by: NChristian Borntraeger <borntraeger@de.ibm.com>
Acked-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch makes the master daemon to sync the connection when it is about
to close.  This makes the connections on the backup to close or timeout
according their state.  Before the sync was performed only if the
connection is in ESTABLISHED state which always made the connections to
timeout in the hard coded 3 minutes. However the Andy Gospodarek's patch
([IPVS]: use proper timeout instead of fixed value) effectively did nothing
more than increasing this to 15 minutes (Established state timeout).  So
this patch makes use of proper timeout since it syncs the connections on
status changes to FIN_WAIT (2min timeout) and CLOSE (10sec timeout).
However if the backup misses CLOSE hopefully it did not miss FIN_WAIT.
Otherwise we will just have to wait for the ESTABLISHED state timeout. As
it is without this patch.  This way the number of the hanging connections
on the backup is kept to minimum. And very few of them will be left to
timeout with a long timeout.

This is important if we want to make use of the fix for the real server
overcommit on master/backup fail-over.
Signed-off-by: NRumen G. Bogdanovski <rumen@voicecho.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch fixes the problem with node overload on director fail-over.
Given the scenario: 2 nodes each accepting 3 connections at a time and 2
directors, director failover occurs when the nodes are fully loaded (6
connections to the cluster) in this case the new director will assign
another 6 connections to the cluster, If the same real servers exist
there.

The problem turned to be in not binding the inherited connections to
the real servers (destinations) on the backup director. Therefore:
"ipvsadm -l" reports 0 connections:
root@test2:~# ipvsadm -l
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  test2.local:5999 wlc
  -> node473.local:5999           Route   1000   0          0
  -> node484.local:5999           Route   1000   0          0

while "ipvs -lnc" is right
root@test2:~# ipvsadm -lnc
IPVS connection entries
pro expire state       source             virtual            destination
TCP 14:56  ESTABLISHED 192.168.0.10:39164 192.168.0.222:5999
192.168.0.51:5999
TCP 14:59  ESTABLISHED 192.168.0.10:39165 192.168.0.222:5999
192.168.0.52:5999

So the patch I am sending fixes the problem by binding the received
connections to the appropriate service on the backup director, if it
exists, else the connection will be handled the old way. So if the
master and the backup directors are synchronized in terms of real
services there will be no problem with server over-committing since
new connections will not be created on the nonexistent real services
on the backup. However if the service is created later on the backup,
the binding will be performed when the next connection update is
received. With this patch the inherited connections will show as
inactive on the backup:

root@test2:~# ipvsadm -l
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  test2.local:5999 wlc
  -> node473.local:5999           Route   1000   0          1
  -> node484.local:5999           Route   1000   0          1

rumen@test2:~$ cat /proc/net/ip_vs
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port Forward Weight ActiveConn InActConn
TCP  C0A800DE:176F wlc
  -> C0A80033:176F      Route   1000   0          1
  -> C0A80032:176F      Route   1000   0          1

Regards,
Rumen Bogdanovski
Acked-by: NJulian Anastasov <ja@ssi.bg>
Signed-off-by: NRumen G. Bogdanovski <rumen@voicecho.com>
Signed-off-by: NSimon Horman <horms@verge.net.au>