[IPv4]: FIB configuration using struct fib_config

Introduces struct fib_config replacing the ugly struct kern_rta
prone to ordering issues. Avoids creating faked netlink messages
for auto generated routes or requests via ioctl.

A new interface net/nexthop.h is added to help navigate through
nexthop configuration arrays.

A new struct nl_info will be used to carry the necessary netlink
information to be used for notifications later on.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET/IPV4/IPV6]: Change some sysctl variables to __read_mostly

Change net/core, ipv4 and ipv6 sysctl variables to __read_mostly.

Couldn't actually measure any performance increase while testing (.3%
I consider noise), but seems like the right thing to do.

Signed-off-by: Brian Haley <brian.haley@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[RTNETLINK]: Unexport rtnl socket

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET] link: Convert notifications to use rtnl_notify()

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[WIRELESS]: Convert notifications to use rtnl_notify()

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BRIDGE]: Convert notifications to use rtnl_notify()

Fixes a wrong use of current->pid as netlink pid.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPv6] prefix: Convert prefix notifications to use rtnl_notify()

Fixes a wrong use of current->pid as netlink pid.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPv6] link: Convert link notifications to use rtnl_notify()

Fixes a wrong use of current->pid as netlink pid.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPv6] route: Convert route notifications to use rtnl_notify()

Fixes a wrong use of current->pid as netlink pid.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPv6] address: Convert address notification to use rtnl_notify()

Fixes a wrong use of current->pid as netlink pid.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPv4] route: Convert route notifications to use rtnl_notify()

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPv4] address: Convert address notification to use rtnl_notify()

Adds support for NLM_F_ECHO allowing applications to easly
see which address have been deleted, added, or promoted.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[DECNET]: Convert DECnet notifications to use rtnl_notify()

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NEIGH]: Convert neighbour notifications ot use rtnl_notify()

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET] fib_rules: Convert fib rule notification to use rtnl_notify()

Adds support for NLM_F_ECHO to simplify the process of identifying
inserted rules with an auto generated priority.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[RTNETLINK]: Add rtnetlink notification interface

Adds rtnl_notify() to send rtnetlink notification messages and
rtnl_set_sk_err() to report notification errors as socket
errors in order to indicate the need of a resync due to loss
of events.

nlmsg_report() is added to properly document the meaning of
NLM_F_ECHO.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETLINK]: Add notification message sending interface

Adds nlmsg_notify() implementing proper notification logic. The
message is multicasted to all listeners in the group. The
applications the requests orignates from can request a unicast
back report in which case said socket will be excluded from the
multicast to avoid duplicated notifications.

nlmsg_multicast() is extended to take allocation flags to
allow notification in atomic contexts.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[RTNETLINK]: Use rtnl_unicast() for rtnetlink unicasts

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6] udp: Fix type in previous change.

UDPv6 stats are UDP6_foo not UDP_foo.

Signed-off-by: David S. Miller <davem@davemloft.net>

[ARCNET]: SoHard PCI support

Add support for a SoHard PCI ARCnet card.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IRDA]: Replace hard-coded dev_self[] array sizes with ARRAY_SIZE()

Several IR drivers used "for (i = 0; i < 4; i++)" to walk their
dev_self[] table.  Better to use ARRAY_SIZE().  And fix ali-ircc so it
won't run off the end if we find too many adapters.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SELINUX]: security/selinux/hooks.c: Make 4 functions static.

This patch makes four needlessly global functions static.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Acked-by: James Morris <jmorris@namei.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: Make unused signal code go away so nobody copies its brokenness

This code is wrong on so many levels, please lose it so it isn't
replicated anywhere else.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Add UDP_MIB_{SND,RCV}BUFERRORS handling.

Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: add the UdpSndbufErrors and UdpRcvbufErrors MIBs

Signed-off-by: Martin Bligh <mbligh@google.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>

[DECNET]: cleanups

- make the following needlessly global functions static:
  - dn_fib.c: dn_fib_sync_down()
  - dn_fib.c: dn_fib_sync_up()
  - dn_rules.c: dn_fib_rule_action()
- remove the following unneeded prototype:
  - dn_fib.c: dn_cache_dump()

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: Constify firmware structs

Constify largish areas of firmware data in Tigon3 ethernet driver.

non-const:

lsmod:
tg3                   101404  0

objdump -x:
.rodata 000003e8
.data 00004a0c

ls -l:
-rw-r--r-- 1 root root 114404 2006-08-19 21:36 drivers/net/tg3.ko

const:

lsmod:
tg3                   101404  0

objdump -x:
.rodata 000042c8
.data 00000b4c

ls -l:
-rw-r--r-- 1 root root 114532 2006-08-19 21:06 drivers/net/tg3.ko

Signed-off-by: Andreas Mohr <andi@lisas.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6] ip6_fib.c: make code static

Make the following needlessly global code static:
- fib6_walker_lock
- struct fib6_walker_list
- fib6_walk_continue()
- fib6_walk()

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[DECNET] Fix to decnet rules compare function

Here is a fix to the DECnet rules compare function where we used 32bit
values rather than 16bit values. Spotted by Patrick McHardy.

Signed-off-by: Steven Whitehouse <steve@chygwyn.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[DECNET] Fix to multiple tables routing

Here is a fix to Patrick McHardy's increase number of routing tables
patch for DECnet. I did just test this and it appears to be working
fine with this patch.

Signed-off-by: Steven Whitehouse <steve@chygwyn.com>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET] netdev: Check name length

Some improvements to robust name interface.  These API's are safe
now by convention, but it is worth providing some safety checks
against future bugs.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[HTB]: rbtree cleanup

Add code to initialize rb tree nodes, and check for double deletion.
This is not a real fix, but I can make it trap sometimes and may
be a bandaid for: http://bugzilla.kernel.org/show_bug.cgi?id=6681

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[HTB]: Use hlist for hash lists.

Use hlist instead of list for the hash list. This saves
space, and we can check for double delete better.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[HTB]: Lindent

Code was a mess in terms of indentation.  Run through Lindent
script, and cleanup the damage. Also, don't use, vim magic
comment, and substitute inline for __inline__.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[HTB]: HTB_HYSTERESIS cleanup

Change the conditional compilation around HTB_HYSTERSIS
since code was splitting mid expression.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[HTB]: Remove lock macro.

Get rid of the macro's being used to obscure the locking.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[HTB]: Remove broken debug code.

The HTB network scheduler had debug code that wouldn't compile
and confused and obfuscated the code, remove it.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Increate RT_TABLE_MAX to 2^32

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[DECNET]: Increase number of possible routing tables to 2^32

Increase the number of possible routing tables to 2^32 by replacing the
fixed sized array of pointers by a hash table and replacing iterations
over all possible table IDs by hash table walking.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Increase number of possible routing tables to 2^32

Increase number of possible routing tables to 2^32 by replacing iterations
over all possible table IDs by hash table walking.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Increase number of possible routing tables to 2^32

Increase the number of possible routing tables to 2^32 by replacing the
fixed sized array of pointers by a hash table and replacing iterations
over all possible table IDs by hash table walking.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Introduce RTA_TABLE/FRA_TABLE attributes

Introduce RTA_TABLE route attribute and FRA_TABLE routing rule attribute
to hold 32 bit routing table IDs. Usespace compatibility is provided by
continuing to accept and send the rtm_table field, but because of its
limited size it can only carry the low 8 bits of the table ID. This
implies that if larger IDs are used, _all_ userspace programs using them
need to use RTA_TABLE.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Use u32 for routing table IDs

Use u32 for routing table IDs in net/ipv4 and net/decnet in preparation of
support for a larger number of routing tables. net/ipv6 already uses u32
everywhere and needs no further changes. No functional changes are made by
this patch.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NEIGHBOUR]: Use ALIGN() macro.

Rather than opencoding the mask, it looks better to use ALIGN()
macro from kernel.h.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Kill double initialization in sock_alloc_inode.

No need to set ei->socket.flags to zero twice.

Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Remove unnecessary config.h includes from net/

config.h is automatically included by kbuild these days.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: sock_register interface changes

The sock_register() doesn't change the family, so the protocols can
define it read-only.  No caller ever checks return value from
sock_unregister()

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: socket family using RCU

Replace the gross custom locking done in socket code for net_family[]
with simple RCU usage. Some reordering necessary to avoid sleep issues
with sock_alloc.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: drop unused elements from net_proto_family

Three values in net_proto_family are defined but never used.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET] socket: code style cleanup

Make socket.c conform to current style:
* run through Lindent
* get rid of unneeded casts
* split assignment and comparsion where possible

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SUNRPC]: Remove the unnecessary check for highmem in xs_sendpages().

Just call kernel_sendpage() directly.

Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[DECNET]: Convert rwlock to spinlock

As per Stephen Hemminger's recent patch to ipv4/fib_semantics.c this
is the same change but for DECnet.

Signed-off-by: Steven Whitehouse <steve@chygwyn.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[DECNET]: Covert rules to use generic code

This patch converts the DECnet rules code to use the generic
rules system created by Thomas Graf <tgraf@suug.ch>.

Signed-off-by: Steven Whitehouse <steve@chygwyn.com>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Use network-order dport for all visible inet_lookup_*

Right now most inet_lookup_* functions take a host-order hnum instead
of a network-order dport because that's how it is represented
internally.

This means that users of these functions have to be careful about
using the right byte-order.  To add more confusion, inet_lookup takes
a network-order dport unlike all other functions.

So this patch changes all visible inet_lookup functions to take a
dport and move all dport->hnum conversion inside them.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4] fib: convert reader/writer to spinlock

Ther is no point in using a more expensive reader/writer lock
for a low contention lock like the fib_info_lock. The only
reader case is in handling route redirects.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Make sure fib6_rule_lookup doesn't return NULL

The callers of fib6_rule_lookup don't expect it to return NULL,
therefore it must return ip6_null_entry whenever fib_rule_lookup fails.

Signed-off-by: Ville Nuorvala <vnuorval@tcs.hut.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Uninline inet_lookup_listener

By modern standards this function is way too big to be inlined.  It's
even bigger than __inet_lookup_listener :)

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[INET]: Remove is_setbyuser patch

The value is_setbyuser from struct ip_options is never used and set
only one time (http://linux-net.osdl.org/index.php/TODO#IPV4).
This little patch removes it from the kernel source.

Signed-off-by: Louis Nyffenegger <louis.nyffenegger@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Kill fib4_rules_clean().

As noted by Adrian Bunk this function is totally unused.

Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Protect RTM_GETRULE table entry with IPV6_MULTIPLE_TABLES ifdef

This is how IPv4 handles this case too.

Based upon a patch from Andrew Morton.

Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Make code static.

This patch makes needlessly global code static.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SUNRPC]: Update to use in-kernel sockets API.

Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Acked-by: James Morris <jmorris@namei.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Round out in-kernel sockets API

This patch implements wrapper functions that provide a convenient way
to access the sockets API for in-kernel users like sunrpc, cifs &
ocfs2 etc and any future users.

Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Acked-by: James Morris <jmorris@namei.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NEIGH]: Move netlink neighbour table bits to linux/neighbour.h

rtnetlink_rcv_msg() is not longer required to parse attributes
for the neighbour tables layer, remove dependency on obsolete and
buggy rta_buf.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NEIGH]: Convert neighbour table dumping to new netlink api

Also fixes skipping of already dumped neighbours.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NEIGH]: Convert neighbour table modification to new netlink api

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NEIGH]: Move netlink neighbour bits to linux/neighbour.h

Moves netlink neighbour bits to linux/neighbour.h. Also
moves bits to be exported to userspace from net/neighbour.h
to linux/neighbour.h and removes __KERNEL__ guards, userspace
is not supposed to be using it.

rtnetlink_rcv_msg() is not longer required to parse attributes
for the neighbour layer, remove dependency on obsolete and
buggy rta_buf.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NEIGH]: Convert neighbour dumping to new netlink api

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NEIGH]: Convert neighbour addition to new netlink api

Fixes:
    Return EAFNOSUPPORT if no table matches the specified
    address family.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NEIGH]: Convert neighbour deletion to new netlink api

Fixes:
  Return ENOENT if the neighbour is not found (was EINVAL)
  Return EAFNOSUPPORT if no table matches the specified
  address family.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Fix policy routing lookup

When the lookup in a table returns ip6_null_entry the policy routing lookup
returns it instead of continuing in the next table, which effectively means
it only searches the local table.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Fix crash in ip6_del_rt

ip6_null_entry doesn't have rt6i_table set, when trying to delete it the
kernel crashes dereferencing table->tb6_lock.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Fix thinko in rt6_fill_node

This looks like a mistake, the table ID is overwritten again.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_queue: handle GSO packets

Handle GSO packets in nf_queue by segmenting them before queueing to
avoid breaking GSO in case they get mangled.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: Get rid of HW checksum invalidation

Update hardware checksums incrementally to avoid breaking GSO.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Replace CHECKSUM_HW by CHECKSUM_PARTIAL/CHECKSUM_COMPLETE

Replace CHECKSUM_HW by CHECKSUM_PARTIAL (for outgoing packets, whose
checksum still needs to be completed) and CHECKSUM_COMPLETE (for
incoming packets, device supplied full checksum).

Patch originally from Herbert Xu, updated by myself for 2.6.18-rc3.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: netbios conntrack: fix compile

Fix compile breakage caused by move of IFA_F_SECONDARY to new header
file.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Move netlink interface bits to linux/if.h

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Convert link dumping to new netlink api

Transforms netlink code to dump link tables to use the new
netlink api. Makes rtnl_getlink() available regardless of the
availability of the wireless extensions.

Adding copy_rtnl_link_stats() avoids the structural dependency
of struct rtnl_link_stats on struct net_device_stats and thus
avoids troubles later on.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Convert link modification to new netlink api

Transforms do_setlink() into rtnl_setlink() using the new
netlink api. A warning message printed to the console is
added in the event that a change request fails while part
of the change request has been comitted already. The ioctl()
based nature of net devices makes it almost impossible to
move on to atomic netlink operations without obsoleting
some of the functionality.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPv4]: Move interface address bits to linux/if_addr.h

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Convert address dumping to new netlink api

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Convert address deletion to new netlink api

Fixes various unvalidated netlink attributes causing
memory corruptions when left empty by userspace.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Convert address addition to new netlink api

Adds rtm_to_ifaddr() transforming a netlink message to a
struct in_ifaddr. Fixes various unvalidated netlink attributes
causing memory corruptions when left empty by userspace
applications.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETLINK]: Convert core netlink handling to new netlink api

Fixes a theoretical memory and locking leak when the size of
the netlink header would exceed the skb tailroom.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETLINK]: Extend netlink messaging interface

Adds:
 nlmsg_get_pos()                 return current position in message
 nlmsg_trim()                    trim part of message
 nla_reserve_nohdr(skb, len)     reserve room for an attribute w/o hdr
 nla_put_nohdr(skb, len, data)   add attribute w/o hdr
 nla_find_nested()               find attribute in nested attributes

Fixes nlmsg_new() to take allocation flags and consider size.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Use Protocol Independant Policy Routing Rules Framework

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Policy Routing Rules

Adds support for policy routing rules including a new
local table for routes with a local destination.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Protocol Independant Policy Routing Rules Framework

Derived from net/ipv/fib_rules.c

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Multiple Routing Tables

Adds the framework to support multiple IPv6 routing tables.
Currently all automatically generated routes are put into the
same table. This could be changed at a later point after
considering the produced locking overhead.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Remove ndiscs rt6_lock dependency

(Ab)using rt6_lock wouldn't work anymore if rt6_lock is
converted into a per table lock.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NetLabel]: tie NetLabel into the Kconfig system

Modify the net/Kconfig file to enable selecting the NetLabel Kconfig
options.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NetLabel]: SELinux support

Add NetLabel support to the SELinux LSM and modify the
socket_post_create() LSM hook to return an error code.  The most
significant part of this patch is the addition of NetLabel hooks into
the following SELinux LSM hooks:

 * selinux_file_permission()
 * selinux_socket_sendmsg()
 * selinux_socket_post_create()
 * selinux_socket_sock_rcv_skb()
 * selinux_socket_getpeersec_stream()
 * selinux_socket_getpeersec_dgram()
 * selinux_sock_graft()
 * selinux_inet_conn_request()

The basic reasoning behind this patch is that outgoing packets are
"NetLabel'd" by labeling their socket and the NetLabel security
attributes are checked via the additional hook in
selinux_socket_sock_rcv_skb().  NetLabel itself is only a labeling
mechanism, similar to filesystem extended attributes, it is up to the
SELinux enforcement mechanism to perform the actual access checks.

In addition to the changes outlined above this patch also includes
some changes to the extended bitmap (ebitmap) and multi-level security
(mls) code to import and export SELinux TE/MLS attributes into and out
of NetLabel.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NetLabel]: CIPSOv4 and Unlabeled packet integration

Add CIPSO/IPv4 and unlabeled packet management to the NetLabel
subsystem.  The CIPSO/IPv4 changes allow the configuration of
CIPSO/IPv4 within the overall NetLabel framework.  The unlabeled
packet changes allows NetLabel to pass unlabeled packets without
error.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NetLabel]: core NetLabel subsystem

Add a new kernel subsystem, NetLabel, to provide explicit packet
labeling services (CIPSO, RIPSO, etc.) to LSM developers.  NetLabel is
designed to work in conjunction with a LSM to intercept and decode
security labels on incoming network packets as well as ensure that
outgoing network packets are labeled according to the security
mechanism employed by the LSM.  The NetLabel subsystem is configured
through a Generic NETLINK interface described in the header files
included in this patch.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NetLabel]: CIPSOv4 engine

Add support for the Commercial IP Security Option (CIPSO) to the IPv4
network stack.  CIPSO has become a de-facto standard for
trusted/labeled networking amongst existing Trusted Operating Systems
such as Trusted Solaris, HP-UX CMW, etc.  This implementation is
designed to be used with the NetLabel subsystem to provide explicit
packet labeling to LSM developers.

The CIPSO/IPv4 packet labeling works by the LSM calling a NetLabel API
function which attaches a CIPSO label (IPv4 option) to a given socket;
this in turn attaches the CIPSO label to every packet leaving the
socket without any extra processing on the outbound side.  On the
inbound side the individual packet's sk_buff is examined through a
call to a NetLabel API function to determine if a CIPSO/IPv4 label is
present and if so the security attributes of the CIPSO label are
returned to the caller of the NetLabel API function.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NetLabel]: core network changes

Changes to the core network stack to support the NetLabel subsystem.  This
includes changes to the IPv4 option handling to support CIPSO labels.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NetLabel]: documentation

Documentation for the NetLabel system, this includes a basic overview
of how NetLabel works, how LSM developers can integrate it into their
favorite LSM, as well as documentation on the CIPSO related sysctl
variables.  Also, due to the difficulty of finding expired IETF
drafts, I am including the IETF CIPSO draft that is the basis of the
NetLabel CIPSO implementation.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[MLSXFRM]: Fix build with SECURITY_NETWORK_XFRM disabled.

The following patch will fix the build problem (encountered by Andrew
Morton) when SECURITY_NETWORK_XFRM is not enabled.

As compared to git-net-selinux_xfrm_decode_session-build-fix.patch in
-mm, this patch sets the return parameter sid to SECSID_NULL in
selinux_xfrm_decode_session() and handles this value in the caller
selinux_inet_conn_request() appropriately.

Signed-off-by: Venkat Yekkirala <vyekkirala@TrustedCS.com>
Acked-by: James Morris <jmorris@namei.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[MLSXFRM]: Auto-labeling of child sockets

This automatically labels the TCP, Unix stream, and dccp child sockets
as well as openreqs to be at the same MLS level as the peer. This will
result in the selection of appropriately labeled IPSec Security
Associations.

This also uses the sock's sid (as opposed to the isec sid) in SELinux
enforcement of secmark in rcv_skb and postroute_last hooks.

Signed-off-by: Venkat Yekkirala <vyekkirala@TrustedCS.com>
Signed-off-by: David S. Miller <davem@davemloft.net>