]> err.no Git - linux-2.6/log
linux-2.6
16 years ago[NET]: Make netns cleanup to run in a separate queue
Benjamin Thery [Tue, 20 Nov 2007 07:18:16 +0000 (23:18 -0800)]
[NET]: Make netns cleanup to run in a separate queue

This patch adds a separate workqueue for cleaning up a network
namespace. If we use the keventd workqueue to execute cleanup_net(),
there is a problem to unregister devices in IPv6. Indeed the code
that cleans up also schedule work in keventd: as long as cleanup_net()
hasn't return, dst_gc_task() cannot run and as long as dst_gc_task() has
not run, there are still some references pending on the net devices and
cleanup_net() can not unregister and exit the keventd workqueue.

Signed-off-by: Benjamin Thery <benjamin.thery@bull.net>
Signed-off-by: Daniel Lezcano <dlezcano@fr.ibm.com>
Acked-by: Denis V. Lunev <den@openvz.org>
Acked-By: Kirill Korotaev <dev@sw.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPVS]: Relax the module get/put in ip_vs_app.c
Pavel Emelyanov [Tue, 20 Nov 2007 06:52:41 +0000 (22:52 -0800)]
[IPVS]: Relax the module get/put in ip_vs_app.c

Both try_module_get/module_put already handle the module == NULL
case, so no need in manual checking.

This patch fits both net-2.6 and net-2.6.25.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Acked-by: Simon Horman <horms@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TUN]: Use iov_length()
Akinobu Mita [Tue, 20 Nov 2007 06:46:51 +0000 (22:46 -0800)]
[TUN]: Use iov_length()

Use iov_length() instead of tun's homemade iov_total().

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET] net/core/request_sock.c: Remove unused exports.
Adrian Bunk [Tue, 20 Nov 2007 06:45:20 +0000 (22:45 -0800)]
[NET] net/core/request_sock.c: Remove unused exports.

This patch removes the following unused EXPORT_SYMBOL's:
- reqsk_queue_alloc
- __reqsk_queue_destroy
- reqsk_queue_destroy

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[PATCH] IPV4 : Move ip route cache flush (secret_rebuild) from softirq to workqueue
Eric Dumazet [Tue, 20 Nov 2007 06:43:37 +0000 (22:43 -0800)]
[PATCH] IPV4 : Move ip route cache flush (secret_rebuild) from softirq to workqueue

Every 600 seconds (ip_rt_secret_interval), a softirq flush of the
whole ip route cache is triggered. On loaded machines, this can starve
softirq for many seconds and can eventually crash.

This patch moves this flush to a workqueue context, using the worker
we intoduced in commit 39c90ece7565f5c47110c2fa77409d7a9478bd5b (IPV4:
Convert rt_check_expire() from softirq processing to workqueue.)

Also, immediate flushes (echo 0 >/proc/sys/net/ipv4/route/flush) are
using rt_do_flush() helper function, wich take attention to
rescheduling.

Next step will be to handle delayed flushes
("echo -1 >/proc/sys/net/ipv4/route/flush" or "ip route flush cache")

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[RAW]: Consolidate proc interface.
Pavel Emelyanov [Tue, 20 Nov 2007 06:38:33 +0000 (22:38 -0800)]
[RAW]: Consolidate proc interface.

Both ipv6/raw.c and ipv4/raw.c use the seq files to walk
through the raw sockets hash and show them.

The "walking" code is rather huge, but is identical in both
cases. The difference is the hash table to walk over and
the protocol family to check (this was not in the first
virsion of the patch, which was noticed by YOSHIFUJI)

Make the ->open store the needed hash table and the family
on the allocated raw_iter_state and make the start/next/stop
callbacks work with it.

This removes most of the code.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[RAW]: Consolidate proto->unhash callback
Pavel Emelyanov [Tue, 20 Nov 2007 06:37:58 +0000 (22:37 -0800)]
[RAW]: Consolidate proto->unhash callback

Same as the ->hash one, this is easily consolidated.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[RAW]: Consolidate proto->hash callback
Pavel Emelyanov [Tue, 20 Nov 2007 06:37:24 +0000 (22:37 -0800)]
[RAW]: Consolidate proto->hash callback

Having the raw_hashinfo it's easy to consolidate the
raw[46]_hash functions.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[RAW]: Introduce raw_hashinfo structure
Pavel Emelyanov [Tue, 20 Nov 2007 06:36:45 +0000 (22:36 -0800)]
[RAW]: Introduce raw_hashinfo structure

The ipv4/raw.c and ipv6/raw.c contain many common code (most
of which is proc interface) which can be consolidated.

Most of the places to consolidate deal with the raw sockets
hashtable, so introduce a struct raw_hashinfo which describes
the raw sockets hash.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPv6] RAW: Compact the API for the kernel
Pavel Emelyanov [Tue, 20 Nov 2007 06:35:57 +0000 (22:35 -0800)]
[IPv6] RAW: Compact the API for the kernel

Same as in the previous patch for ipv4, compact the
API and hide hash table and rwlock inside the raw.c
file.

Plus fix some "bad" places from checkpatch.pl point
of view (assignments inside if()).

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPv4] RAW: Compact the API for the kernel
Pavel Emelyanov [Tue, 20 Nov 2007 06:35:07 +0000 (22:35 -0800)]
[IPv4] RAW: Compact the API for the kernel

The raw sockets functions are explicitly used from
inside the kernel in two places:

1. in ip_local_deliver_finish to intercept skb-s
2. in icmp_error

For this purposes many functions and even data structures,
that are naturally internal for raw protocol, are exported.

Compact the API to two functions and hide all the other
(including hash table and rwlock) inside the net/ipv4/raw.c

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Consolidate net namespace related proc files creation.
Denis V. Lunev [Tue, 20 Nov 2007 06:31:54 +0000 (22:31 -0800)]
[NET]: Consolidate net namespace related proc files creation.

Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Make AF_UNIX per network namespace safe [v2]
Denis V. Lunev [Tue, 20 Nov 2007 06:29:30 +0000 (22:29 -0800)]
[NET]: Make AF_UNIX per network namespace safe [v2]

Because of the global nature of garbage collection, and because of the
cost of per namespace hash tables unix_socket_table has been kept
global.  With a filter added on lookups so we don't see sockets from
the wrong namespace.

Currently I don't fold the namesapce into the hash so multiple
namespaces using the same socket name will be guaranteed a hash
collision.

Changes from v1:
- fixed unix_seq_open

Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Make AF_PACKET handle multiple network namespaces
Denis V. Lunev [Tue, 20 Nov 2007 06:28:35 +0000 (22:28 -0800)]
[NET]: Make AF_PACKET handle multiple network namespaces

This is done by making packet_sklist_lock and packet_sklist per
network namespace and adding an additional filter condition on
received packets to ensure they came from the proper network
namespace.

Changes from v1:
- prohibit to call inet_dgram_ops.ioctl in other than init_net

Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Make the netlink methods in rtnetlink handle multiple network namespaces
Eric W. Biederman [Tue, 20 Nov 2007 06:27:40 +0000 (22:27 -0800)]
[NET]: Make the netlink methods in rtnetlink handle multiple network namespaces

After the previous prep work this just consists of removing checks
limiting the code to work in the initial network namespace, and
updating rtmsg_ifinfo so we can generate events for devices in
something other then the initial network namespace.

Referring to network other network devices like the IFLA_LINK
and IFLA_MASTER attributes do, gets interesting if those network
devices happen to be in other network namespaces.  Currently
ifindex numbers are allocated globally so I have taken the path
of least resistance and not still report the information even
though the devices they are talking about are invisible.

If applications start getting confused or when ifindex
numbers become local to the network namespace we may need
to do something different in the future.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Denis V. Lunev <den@openz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Make rtnetlink infrastructure network namespace aware (v3)
Denis V. Lunev [Tue, 20 Nov 2007 06:26:51 +0000 (22:26 -0800)]
[NET]: Make rtnetlink infrastructure network namespace aware (v3)

After this patch none of the netlink callback support anything
except the initial network namespace but the rtnetlink infrastructure
now handles multiple network namespaces.

Changes from v2:
- IPv6 addrlabel processing

Changes from v1:
- no need for special rtnl_unlock handling
- fixed IPv6 ndisc

Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Modify all rtnetlink methods to only work in the initial namespace (v2)
Denis V. Lunev [Fri, 30 Nov 2007 13:21:31 +0000 (00:21 +1100)]
[NET]: Modify all rtnetlink methods to only work in the initial namespace (v2)

Before I can enable rtnetlink to work in all network namespaces I need
to be certain that something won't break.  So this patch deliberately
disables all of the rtnletlink methods in everything except the
initial network namespace.  After the methods have been audited this
extra check can be disabled.

Changes from v1:
- added IPv6 addrlabel protection

Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
16 years ago[MACVLAN]: Allow setting mac address while device is up
Patrick McHardy [Tue, 20 Nov 2007 06:00:42 +0000 (22:00 -0800)]
[MACVLAN]: Allow setting mac address while device is up

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[MACVLAN]: Remove unnecessary IFF_UP check
Patrick McHardy [Tue, 20 Nov 2007 06:00:00 +0000 (22:00 -0800)]
[MACVLAN]: Remove unnecessary IFF_UP check

Only devices that are UP are in the hash, so macvlan_broadcast() doesn't
need to check for IFF_UP.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPCONFIG]: Mark vendor_class_identifier as __initdata.
David S. Miller [Tue, 20 Nov 2007 05:56:16 +0000 (21:56 -0800)]
[IPCONFIG]: Mark vendor_class_identifier as __initdata.

Based upon a suggestion by Francois Romieu.

Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPVS]: Create synced connections with their real state
Rumen G. Bogdanovski [Tue, 20 Nov 2007 05:53:27 +0000 (21:53 -0800)]
[IPVS]: Create synced connections with their real state

With this patch the synced connections are created with their real state,
which can be changed on the next synchronizations if necessary. This way
on fail-over all the connections will be treated according to their actual
state, causing no scheduling problems (the active and the nonactive
connections have different weights in the schedulers).
The backwards compatibility is preserved and the existing tools will show
the true connection states even on the backup director.

Signed-off-by: Rumen G. Bogdanovski <rumen@voicecho.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPVS]: Flag synced connections and expose them in proc
Rumen G. Bogdanovski [Tue, 20 Nov 2007 05:52:42 +0000 (21:52 -0800)]
[IPVS]: Flag synced connections and expose them in proc

This patch labels the sync-created connections with IP_VS_CONN_F_SYNC
flag and creates /proc/net/ip_vs_conn_sync to enable monitoring of the
origin of the connections, if they are local or created by the
synchronization.

Signed-off-by: Rumen G. Bogdanovski <rumen@voicecho.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoieee80211: Add IEEE80211_MAX_FRAME_LEN to linux/ieee80211.h
Michael Wu [Tue, 30 Oct 2007 20:50:05 +0000 (16:50 -0400)]
ieee80211: Add IEEE80211_MAX_FRAME_LEN to linux/ieee80211.h

This patch adds IEEE80211_MAX_FRAME_LEN which is useful for drivers trying
to determine how much to allocate for their RX buffers.

It also updates the comment on IEEE80211_MAX_DATA_LEN based on revisions
in 802.11e.

IEEE80211_MAX_FRAG_THRESHOLD and IEEE80211_MAX_RTS_THRESHOLD are also
revised due to the new maximum frame size.

Signed-off-by: Michael Wu <flamingice@sourmilk.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agomac80211: Accept auto txpower setting
Mattias Nissler [Wed, 24 Oct 2007 21:30:36 +0000 (23:30 +0200)]
mac80211: Accept auto txpower setting

This changes the SIWTXPOWER ioctl to also accept a txpower setting of
"automatic". Since mac80211 currently cannot tell drivers to automatically
adjust tx power, we select the tx power level of the current channel. While
this is kind of a hack, it certainly saves some iwconfig users from headaches.

Signed-off-by: Mattias Nissler <mattias.nissler@gmx.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NETPOLL]: Don't need rx_flags.
Stephen Hemminger [Tue, 20 Nov 2007 03:37:09 +0000 (19:37 -0800)]
[NETPOLL]: Don't need rx_flags.

The rx_flags variable is redundant. Turning rx on/off is done
via setting the rx_np pointer.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NETPOLL]: Kill NETPOLL_RX_DROP, set but never tested.
Stephen Hemminger [Tue, 20 Nov 2007 03:24:52 +0000 (19:24 -0800)]
[NETPOLL]: Kill NETPOLL_RX_DROP, set but never tested.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NETPOLL]: no need to store local_mac
Stephen Hemminger [Tue, 20 Nov 2007 03:23:29 +0000 (19:23 -0800)]
[NETPOLL]: no need to store local_mac

The local_mac is managed by the network device, no need to keep a
spare copy and all the management problems that could cause.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NETPOLL]: netpoll_poll() cleanup
Stephen Hemminger [Tue, 20 Nov 2007 03:18:11 +0000 (19:18 -0800)]
[NETPOLL]: netpoll_poll() cleanup

Restructure code slightly to improve readability:
  * dereference device once
  * change obvious while() loop
  * let poll_napi() handle null list itself

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NETPOLL]: Use skb_queue_purge().
Stephen Hemminger [Tue, 20 Nov 2007 03:15:03 +0000 (19:15 -0800)]
[NETPOLL]: Use skb_queue_purge().

Use standard routine for flushing queue.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: Correct DSACK check placing
Ilpo Järvinen [Sat, 17 Nov 2007 00:17:05 +0000 (16:17 -0800)]
[TCP]: Correct DSACK check placing

Previously one of the in-block skip branches was missing it.

Also, drop it from tail-fully-processed case because the next
iteration will do exactly the same thing, i.e., process the
SACK block that contains the DSACK information.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[CAN]: Add documentation
Oliver Hartkopp [Sat, 17 Nov 2007 00:09:28 +0000 (16:09 -0800)]
[CAN]: Add documentation

This patch adds documentation for the PF_CAN protocol family.

Signed-off-by: Oliver Hartkopp <oliver.hartkopp@volkswagen.de>
Signed-off-by: Urs Thuermann <urs.thuermann@volkswagen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[CAN]: Add maintainer entries
Oliver Hartkopp [Sat, 17 Nov 2007 00:07:41 +0000 (16:07 -0800)]
[CAN]: Add maintainer entries

This patch adds entries in the CREDITS and MAINTAINERS file for CAN.

Signed-off-by: Oliver Hartkopp <oliver.hartkopp@volkswagen.de>
Signed-off-by: Urs Thuermann <urs.thuermann@volkswagen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[CAN]: Add missing Kbuild entries
Oliver Hartkopp [Fri, 28 Dec 2007 00:51:46 +0000 (16:51 -0800)]
[CAN]: Add missing Kbuild entries

This patch adds the missing Kbuild entries and the missing Kbuild file
in include/linux/can for the CAN subsystem.

Signed-off-by: Oliver Hartkopp <oliver@hartkopp.net>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[CAN]: Fix plain integer definitions in userspace header.
Oliver Hartkopp [Fri, 28 Dec 2007 00:50:06 +0000 (16:50 -0800)]
[CAN]: Fix plain integer definitions in userspace header.

This patch fixes the use of plain integers instead of __u32 in a struct
that is visible from kernel space and user space.

Thanks to Sam Ravnborg for pointing out the wrong plain int usage.

Signed-off-by: Oliver Hartkopp <oliver@hartkopp.net>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[CAN]: Add virtual CAN netdevice driver
Oliver Hartkopp [Fri, 16 Nov 2007 23:56:08 +0000 (15:56 -0800)]
[CAN]: Add virtual CAN netdevice driver

This patch adds the virtual CAN bus (vcan) network driver.
The vcan device is just a loopback device for CAN frames, no
real CAN hardware is involved.

Signed-off-by: Oliver Hartkopp <oliver.hartkopp@volkswagen.de>
Signed-off-by: Urs Thuermann <urs.thuermann@volkswagen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[CAN]: Add broadcast manager (bcm) protocol
Oliver Hartkopp [Fri, 16 Nov 2007 23:53:52 +0000 (15:53 -0800)]
[CAN]: Add broadcast manager (bcm) protocol

This patch adds the CAN broadcast manager (bcm) protocol.

Signed-off-by: Oliver Hartkopp <oliver.hartkopp@volkswagen.de>
Signed-off-by: Urs Thuermann <urs.thuermann@volkswagen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[CAN]: Add raw protocol
Oliver Hartkopp [Fri, 16 Nov 2007 23:53:09 +0000 (15:53 -0800)]
[CAN]: Add raw protocol

This patch adds the CAN raw protocol.

Signed-off-by: Oliver Hartkopp <oliver.hartkopp@volkswagen.de>
Signed-off-by: Urs Thuermann <urs.thuermann@volkswagen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[CAN]: Add PF_CAN core module
Oliver Hartkopp [Fri, 16 Nov 2007 23:52:17 +0000 (15:52 -0800)]
[CAN]: Add PF_CAN core module

This patch adds the CAN core functionality but no protocols or drivers.
No protocol implementations are included here.  They come as separate
patches.  Protocol numbers are already in include/linux/can.h.

Signed-off-by: Oliver Hartkopp <oliver.hartkopp@volkswagen.de>
Signed-off-by: Urs Thuermann <urs.thuermann@volkswagen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[CAN]: Allocate protocol numbers for PF_CAN
Oliver Hartkopp [Sun, 16 Dec 2007 23:59:24 +0000 (15:59 -0800)]
[CAN]: Allocate protocol numbers for PF_CAN

This patch adds a protocol/address family number, ARP hardware type,
ethernet packet type, and a line discipline number for the SocketCAN
implementation.

Signed-off-by: Oliver Hartkopp <oliver.hartkopp@volkswagen.de>
Signed-off-by: Urs Thuermann <urs.thuermann@volkswagen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: NET_CLS_ROUTE : convert ip_rt_acct to per_cpu variables
Eric Dumazet [Fri, 16 Nov 2007 11:32:10 +0000 (03:32 -0800)]
[NET]: NET_CLS_ROUTE : convert ip_rt_acct to per_cpu variables

ip_rt_acct needs 4096 bytes per cpu to perform some accounting.
It is actually allocated as a single huge array [4096*NR_CPUS]
(rounded up to a power of two)

Converting it to a per cpu variable is wanted to :
 - Save space on machines were num_possible_cpus() < NR_CPUS
 - Better NUMA placement (each cpu gets memory on its node)

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: Rewrite SACK block processing & sack_recv_cache use
Ilpo Järvinen [Fri, 16 Nov 2007 03:50:37 +0000 (19:50 -0800)]
[TCP]: Rewrite SACK block processing & sack_recv_cache use

Key points of this patch are:

  - In case new SACK information is advance only type, no skb
    processing below previously discovered highest point is done
  - Optimize cases below highest point too since there's no need
    to always go up to highest point (which is very likely still
    present in that SACK), this is not entirely true though
    because I'm dropping the fastpath_skb_hint which could
    previously optimize those cases even better. Whether that's
    significant, I'm not too sure.

Currently it will provide skipping by walking. Combined with
RB-tree, all skipping would become fast too regardless of window
size (can be done incrementally later).

Previously a number of cases in TCP SACK processing fails to
take advantage of costly stored information in sack_recv_cache,
most importantly, expected events such as cumulative ACK and new
hole ACKs. Processing on such ACKs result in rather long walks
building up latencies (which easily gets nasty when window is
huge). Those latencies are often completely unnecessary
compared with the amount of _new_ information received, usually
for cumulative ACK there's no new information at all, yet TCP
walks whole queue unnecessary potentially taking a number of
costly cache misses on the way, etc.!

Since the inclusion of highest_sack, there's a lot information
that is very likely redundant (SACK fastpath hint stuff,
fackets_out, highest_sack), though there's no ultimate guarantee
that they'll remain the same whole the time (in all unearthly
scenarios). Take advantage of this knowledge here and drop
fastpath hint and use direct access to highest SACKed skb as
a replacement.

Effectively "special cased" fastpath is dropped. This change
adds some complexity to introduce better coveraged "fastpath",
though the added complexity should make TCP behave more cache
friendly.

The current ACK's SACK blocks are compared against each cached
block individially and only ranges that are new are then scanned
by the high constant walk. For other parts of write queue, even
when in previously known part of the SACK blocks, a faster skip
function is used (if necessary at all). In addition, whenever
possible, TCP fast-forwards to highest_sack skb that was made
available by an earlier patch. In typical case, no other things
but this fast-forward and mandatory markings after that occur
making the access pattern quite similar to the former fastpath
"special case".

DSACKs are special case that must always be walked.

The local to recv_sack_cache copying could be more intelligent
w.r.t DSACKs which are likely to be there only once but that
is left to a separate patch.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: Earlier SACK block verification & simplify access to them
Ilpo Järvinen [Fri, 16 Nov 2007 03:49:47 +0000 (19:49 -0800)]
[TCP]: Earlier SACK block verification & simplify access to them

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: Create tcp_sacktag_one().
Ilpo Järvinen [Fri, 16 Nov 2007 03:44:56 +0000 (19:44 -0800)]
[TCP]: Create tcp_sacktag_one().

Worker function that implements the main logic of
the inner-most loop of tcp_sacktag_write_queue().

Idea was originally presented by David S. Miller.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: Prior_fackets can be replaced by highest_sack seq
Ilpo Järvinen [Fri, 16 Nov 2007 03:43:56 +0000 (19:43 -0800)]
[TCP]: Prior_fackets can be replaced by highest_sack seq

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: Make lost retrans detection more self-contained
Ilpo Järvinen [Fri, 16 Nov 2007 03:42:54 +0000 (19:42 -0800)]
[TCP]: Make lost retrans detection more self-contained

Highest_sack_end_seq is no longer calculated in the loop,
thus it can be pushed to the worker function altogether
making that function independent of the sacktag.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: Convert highest_sack to sk_buff to allow direct access
Ilpo Järvinen [Fri, 16 Nov 2007 03:41:46 +0000 (19:41 -0800)]
[TCP]: Convert highest_sack to sk_buff to allow direct access

It is going to replace the sack fastpath hint quite soon... :-)

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: non-FACK SACK follows conservative SACK loss recovery
Ilpo Järvinen [Fri, 16 Nov 2007 03:39:31 +0000 (19:39 -0800)]
[TCP]: non-FACK SACK follows conservative SACK loss recovery

Many assumptions that are true when no reordering or other
strange events happen are not a part of the RFC3517. FACK
implementation is based on such assumptions. Previously (before
the rewrite) the non-FACK SACK was basically doing fast rexmit
and then it times out all skbs when first cumulative ACK arrives,
which cannot really be called SACK based recovery :-).

RFC3517 SACK disables these things:
- Per SKB timeouts & head timeout entry to recovery
- Marking at least one skb while in recovery (RFC3517 does this
  only for the fast retransmission but not for the other skbs
  when cumulative ACKs arrive in the recovery)
- Sacktag's loss detection flavors B and C (see comment before
  tcp_sacktag_write_queue)

This does not implement the "last resort" rule 3 of NextSeg, which
allows retransmissions also when not enough SACK blocks have yet
arrived above a segment for IsLost to return true [RFC3517].

The implementation differs from RFC3517 in these points:
- Rate-halving is used instead of FlightSize / 2
- Instead of using dupACKs to trigger the recovery, the number
  of SACK blocks is used as FACK does with SACK blocks+holes
  (which provides more accurate number). It seems that the
  difference can affect negatively only if the receiver does not
  generate SACK blocks at all even though it claimed to be
  SACK-capable.
- Dupthresh is not a constant one. Dynamical adjustments include
  both holes and sacked segments (equal to what FACK has) due to
  complexity involved in determining the number sacked blocks
  between highest_sack and the reordered segment. Thus it's will
  be an over-estimate.

Implementation note:

tcp_clean_rtx_queue doesn't need a lost_cnt tweak because head
skb at that point cannot be SACKED_ACKED (nor would such
situation last for long enough to cause problems).

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: Extend reordering detection to cover CA_Loss partially
Ilpo Järvinen [Fri, 16 Nov 2007 03:35:11 +0000 (19:35 -0800)]
[TCP]: Extend reordering detection to cover CA_Loss partially

This implements more accurately what is stated in sacktag's
overall comment:

  "Both of these heuristics are not used in Loss state, when
   we cannot account for retransmits accurately."

When CA_Loss state is entered, the state changer ensures that
undo_marker is only set if no TCPCB_RETRANS skbs were found,
thus having non-zero undo_marker in CA_Loss basically tells
that the R-bits still accurately reflect the current state
of TCP.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: Move !in_sack test earlier in sacktag & reorganize if()s
Ilpo Järvinen [Fri, 16 Nov 2007 03:33:31 +0000 (19:33 -0800)]
[TCP]: Move !in_sack test earlier in sacktag & reorganize if()s

All intermediate conditions include it already, make them
simpler as well.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Move sock_valbool_flag to socket.c
Pavel Emelyanov [Thu, 15 Nov 2007 11:03:19 +0000 (03:03 -0800)]
[NET]: Move sock_valbool_flag to socket.c

The sock_valbool_flag() helper is used in setsockopt to
set or reset some flag on the sock. This helper is required
in the net/socket.c only, so move it there.

Besides, patch two places in sys_setsockopt() that repeat
this helper functionality manually.

Since this is not a bugfix, but a trivial cleanup, I
prepared this patch against net-2.6.25, but it also
applies (with a single offset) to the latest net-2.6.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Use sockfd_lookup_light in the rest of the net/socket.c
Pavel Emelyanov [Thu, 15 Nov 2007 00:01:43 +0000 (16:01 -0800)]
[NET]: Use sockfd_lookup_light in the rest of the net/socket.c

Some time ago a sockfd_lookup_light was introduced and
most of the socket.c file was patched to use it. However
two routines were left - sys_sendto and sys_recvfrom.

Patch them as well, since this helper does exactly what
these two need.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPV4] ipconfig: Implement DHCP Class-identifier
Rainer Jochem [Wed, 14 Nov 2007 10:18:39 +0000 (02:18 -0800)]
[IPV4] ipconfig: Implement DHCP Class-identifier

From : Rainer Jochem <rainer.jochem@mpi-sb.mpg.de>

Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Move Qdisc_class_ops and Qdisc_ops in appropriate sections.
Eric Dumazet [Wed, 14 Nov 2007 09:44:41 +0000 (01:44 -0800)]
[NET]: Move Qdisc_class_ops and Qdisc_ops in appropriate sections.

Qdisc_class_ops are const, and Qdisc_ops are mostly read.

Using "const" and "__read_mostly" qualifiers helps to reduce false
sharing.

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPV6] ADDRCONF: Support RFC3484 configurable address selection policy table.
YOSHIFUJI Hideaki [Wed, 14 Nov 2007 06:56:23 +0000 (15:56 +0900)]
[IPV6] ADDRCONF: Support RFC3484 configurable address selection policy table.

Policy table is implemented as an RCU linear list since we do not expect
large list nor frequent updates.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPV6] ADDRCONF: Allow address selection policy with ifindex.
YOSHIFUJI Hideaki [Wed, 14 Nov 2007 06:56:15 +0000 (15:56 +0900)]
[IPV6] ADDRCONF: Allow address selection policy with ifindex.

This patch allows ifindex to be a key for address selection policy table.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPV6] ADDRCONF: Rename ipv6_saddr_label() to ipv6_addr_label().
YOSHIFUJI Hideaki [Wed, 14 Nov 2007 06:55:29 +0000 (15:55 +0900)]
[IPV6] ADDRCONF: Rename ipv6_saddr_label() to ipv6_addr_label().

This patch renames ipv6_saddr_label() to ipv6_addr_label() because
address label is used for both of source address and destination
address.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Kill afinfo->nf_post_routing
David S. Miller [Wed, 14 Nov 2007 09:57:47 +0000 (01:57 -0800)]
[IPSEC]: Kill afinfo->nf_post_routing

After changeset:

[NETFILTER]: Introduce NF_INET_ hook values

It always evaluates to NF_INET_POST_ROUTING.

Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NETFILTER]: Introduce NF_INET_ hook values
Patrick McHardy [Tue, 20 Nov 2007 02:53:30 +0000 (18:53 -0800)]
[NETFILTER]: Introduce NF_INET_ hook values

The IPv4 and IPv6 hook values are identical, yet some code tries to figure
out the "correct" value by looking at the address family. Introduce NF_INET_*
values for both IPv4 and IPv6. The old values are kept in a #ifndef __KERNEL__
section for userspace compatibility.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Add async resume support on input
Herbert Xu [Tue, 20 Nov 2007 02:50:17 +0000 (18:50 -0800)]
[IPSEC]: Add async resume support on input

This patch adds support for async resumptions on input.  To do so, the
transform would return -EINPROGRESS and subsequently invoke the
function xfrm_input_resume to resume processing.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Remove nhoff from xfrm_input
Herbert Xu [Tue, 20 Nov 2007 02:47:58 +0000 (18:47 -0800)]
[IPSEC]: Remove nhoff from xfrm_input

The nhoff field isn't actually necessary in xfrm_input.  For tunnel
mode transforms we now throw away the output IP header so it makes no
sense to fill in the nexthdr field.  For transport mode we can now let
the function transport_finish do the setting and it knows where the
nexthdr field is.

The only other thing that needs the nexthdr field to be set is the
header extraction code.  However, we can simply move the protocol
extraction out of the generic header extraction.

We want to minimise the amount of info we have to carry around between
transforms as this simplifies the resumption process for async crypto.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Make x->lastused an unsigned long
Herbert Xu [Wed, 14 Nov 2007 05:47:08 +0000 (21:47 -0800)]
[IPSEC]: Make x->lastused an unsigned long

Currently x->lastused is u64 which means that it cannot be
read/written atomically on all architectures.  David Miller observed
that the value stored in it is only an unsigned long which is always
atomic.

So based on his suggestion this patch changes the internal
representation from u64 to unsigned long while the user-interface
still refers to it as u64.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Move state lock into x->type->input
Herbert Xu [Wed, 14 Nov 2007 05:45:58 +0000 (21:45 -0800)]
[IPSEC]: Move state lock into x->type->input

This patch releases the lock on the state before calling
x->type->input.  It also adds the lock to the spots where they're
currently needed.

Most of those places (all except mip6) are expected to disappear with
async crypto.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Move integrity stat collection into xfrm_input
Herbert Xu [Sun, 16 Dec 2007 23:55:02 +0000 (15:55 -0800)]
[IPSEC]: Move integrity stat collection into xfrm_input

Similar to the moving out of the replay processing on the output, this
patch moves the integrity stat collectin from x->type->input into
xfrm_input.

This would eventually allow transforms such as AH/ESP to be lockless.

The error value EBADMSG (currently unused in the crypto layer) is used
to indicate a failed integrity check.  In future this error can be
directly returned by the crypto layer once we switch to aead
algorithms.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Store xfrm states in security path directly
Herbert Xu [Wed, 14 Nov 2007 05:44:55 +0000 (21:44 -0800)]
[IPSEC]: Store xfrm states in security path directly

As it is xfrm_input first collects a list of xfrm states on the stack
before storing them in the packet's security path just before it
returns.  For async crypto, this construction presents an obstacle
since we may need to leave the loop after each transform.

In fact, it's much easier to just skip the stack completely and always
store to the security path.  This is proven by the fact that this
patch actually shrinks the code.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Merge most of the input path
Herbert Xu [Wed, 14 Nov 2007 05:44:23 +0000 (21:44 -0800)]
[IPSEC]: Merge most of the input path

As part of the work on asynchronous cryptographic operations, we need
to be able to resume from the spot where they occur.  As such, it
helps if we isolate them to one spot.

This patch moves most of the remaining family-specific processing into
the common input code.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Add async resume support on output
Herbert Xu [Wed, 14 Nov 2007 05:43:43 +0000 (21:43 -0800)]
[IPSEC]: Add async resume support on output

This patch adds support for async resumptions on output.  To do so,
the transform would return -EINPROGRESS and subsequently invoke the
function xfrm_output_resume to resume processing.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Merge most of the output path
Herbert Xu [Wed, 14 Nov 2007 05:43:11 +0000 (21:43 -0800)]
[IPSEC]: Merge most of the output path

As part of the work on asynchrnous cryptographic operations, we need
to be able to resume from the spot where they occur.  As such, it
helps if we isolate them to one spot.

This patch moves most of the remaining family-specific processing into
the common output code.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPV6]: Add ip6_local_out
Herbert Xu [Sat, 12 Jan 2008 03:15:08 +0000 (19:15 -0800)]
[IPV6]: Add ip6_local_out

Most callers of the LOCAL_OUT chain will set the IP packet length
before doing so.  They also share the same output function dst_output.

This patch creates a new function called ip6_local_out which does all
of that and converts the appropriate users over to it.

Apart from removing duplicate code, it will also help in merging the
IPsec output path.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPV4]: Add ip_local_out
Herbert Xu [Sat, 12 Jan 2008 03:14:00 +0000 (19:14 -0800)]
[IPV4]: Add ip_local_out

Most callers of the LOCAL_OUT chain will set the IP packet length and
header checksum before doing so.  They also share the same output
function dst_output.

This patch creates a new function called ip_local_out which does all
of that and converts the appropriate users over to it.

Apart from removing duplicate code, it will also help in merging the
IPsec output path once the same thing is done for IPv6.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Separate inner/outer mode processing on input
Herbert Xu [Wed, 14 Nov 2007 05:41:28 +0000 (21:41 -0800)]
[IPSEC]: Separate inner/outer mode processing on input

With inter-family transforms the inner mode differs from the outer
mode.  Attempting to handle both sides from the same function means
that it needs to handle both IPv4 and IPv6 which creates duplication
and confusion.

This patch separates the two parts on the input path so that each
function deals with one family only.

In particular, the functions xfrm4_extract_inut/xfrm6_extract_inut
moves the pertinent fields from the IPv4/IPv6 IP headers into a
neutral format stored in skb->cb.  This is then used by the inner mode
input functions to modify the inner IP header.  In this way the input
function no longer has to know about the outer address family.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Separate inner/outer mode processing on output
Herbert Xu [Wed, 14 Nov 2007 05:40:52 +0000 (21:40 -0800)]
[IPSEC]: Separate inner/outer mode processing on output

With inter-family transforms the inner mode differs from the outer
mode.  Attempting to handle both sides from the same function means
that it needs to handle both IPv4 and IPv6 which creates duplication
and confusion.

This patch separates the two parts on the output path so that each
function deals with one family only.

In particular, the functions xfrm4_extract_output/xfrm6_extract_output
moves the pertinent fields from the IPv4/IPv6 IP headers into a
neutral format stored in skb->cb.  This is then used by the outer mode
output functions to write the outer IP header.  In this way the output
function no longer has to know about the inner address family.

Since the extract functions are only called by tunnel modes (the only
modes that can support inter-family transforms), I've also moved the
xfrm*_tunnel_check_size calls into them.  This allows the correct ICMP
message to be sent as opposed to now where you might call icmp_send
with an IPv6 packet and vice versa.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[INET]: Give outer DSCP directly to ip*_copy_dscp
Herbert Xu [Wed, 14 Nov 2007 05:40:13 +0000 (21:40 -0800)]
[INET]: Give outer DSCP directly to ip*_copy_dscp

This patch changes the prototype of ipv4_copy_dscp and ipv6_copy_dscp so
that they directly take the outer DSCP rather than the outer IP header.
This will help us to unify the code for inter-family tunnels.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Move x->outer_mode->output out of locked section
Herbert Xu [Wed, 14 Nov 2007 05:39:38 +0000 (21:39 -0800)]
[IPSEC]: Move x->outer_mode->output out of locked section

RO mode is the only one that requires a locked output function.  So
it's easier to move the lock into that function rather than requiring
everyone else to run under the lock.

In particular, this allows us to move the size check into the output
function without causing a potential dead-lock should the ICMP error
somehow hit the same SA on transmission.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Forbid BEET + ipcomp for now
Herbert Xu [Wed, 14 Nov 2007 05:39:08 +0000 (21:39 -0800)]
[IPSEC]: Forbid BEET + ipcomp for now

While BEET can theoretically work with IPComp the current code can't
do that because it tries to construct a BEET mode tunnel type which
doesn't (and cannot) exist.  In fact as it is it won't even attach a
tunnel object at all for BEET which is bogus.

To support this fully we'd also need to change the policy checks on
input to recognise a plain tunnel as a legal variant of an optional
BEET transform.

This patch simply fails such constructions for now.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Merge common code into xfrm_bundle_create
Herbert Xu [Tue, 11 Dec 2007 17:32:34 +0000 (09:32 -0800)]
[IPSEC]: Merge common code into xfrm_bundle_create

Half of the code in xfrm4_bundle_create and xfrm6_bundle_create are
common.  This patch extracts that logic and puts it into
xfrm_bundle_create.  The rest of it are then accessed through afinfo.

As a result this fixes the problem with inter-family transforms where
we treat every xfrm dst in the bundle as if it belongs to the top
family.

This patch also fixes a long-standing error-path bug where we may free
the xfrm states twice.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Move flow construction into xfrm_dst_lookup
Herbert Xu [Wed, 14 Nov 2007 05:37:28 +0000 (21:37 -0800)]
[IPSEC]: Move flow construction into xfrm_dst_lookup

This patch moves the flow construction from the callers of
xfrm_dst_lookup into that function.  It also changes xfrm_dst_lookup
so that it takes an xfrm state as its argument instead of explicit
addresses.

This removes any address-specific logic from the callers of
xfrm_dst_lookup which is needed to correctly support inter-family
transforms.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Replace x->type->{local,remote}_addr with flags
Herbert Xu [Wed, 14 Nov 2007 05:36:51 +0000 (21:36 -0800)]
[IPSEC]: Replace x->type->{local,remote}_addr with flags

The functions local_addr and remote_addr are more than what they're
needed for.  The same thing can be done easily with flags on the type
object.  This patch does that and simplifies the wrapper functions in
xfrm6_policy accordingly.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Make sure idev is consistent with dev in xfrm_dst
Herbert Xu [Wed, 14 Nov 2007 05:36:07 +0000 (21:36 -0800)]
[IPSEC]: Make sure idev is consistent with dev in xfrm_dst

Previously we took the device from the bottom route and idev from the
top route.  This is bad because idev may well point to a different
device.  This patch changes it so that we get the idev from the device
directly.

It also makes it an error if either dev or idev is NULL.  This is
consistent with the rest of the routing code which also treats these
cases as errors.

I've removed the err initialisation in xfrm6_policy.c because it
achieves no purpose and hid a bug when an initial version of this
patch neglected to set err to -ENODEV (fortunately the IPv4 version
warned about it).

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Set dst->input to dst_discard
Herbert Xu [Wed, 14 Nov 2007 05:35:32 +0000 (21:35 -0800)]
[IPSEC]: Set dst->input to dst_discard

The input function should never be invoked on IPsec dst objects.  This
is because we don't apply IPsec on input until after we've made the
routing decision.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Only set neighbour on top xfrm dst
Herbert Xu [Wed, 14 Nov 2007 05:35:01 +0000 (21:35 -0800)]
[IPSEC]: Only set neighbour on top xfrm dst

The neighbour field is only used by dst_confirm which only ever happens on
the top-most xfrm dst.  So it's a waste to duplicate for every other xfrm
dst.  This patch moves its setting out of the loop so that only the top one
gets set.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Remove unnecessary inclusion of dst.h
Herbert Xu [Wed, 14 Nov 2007 05:34:34 +0000 (21:34 -0800)]
[NET]: Remove unnecessary inclusion of dst.h

The file net/netevent.h only refers to struct dst_entry * so it
doesn't need to include dst.h.  I've replaced it with a forward
declaration.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Eliminate duplicate copies of dst_discard
Herbert Xu [Wed, 14 Nov 2007 05:34:06 +0000 (21:34 -0800)]
[NET]: Eliminate duplicate copies of dst_discard

We have a number of copies of dst_discard scattered around the place
which all do the same thing, namely free a packet on the input or
output paths.

This patch deletes all of them except dst_discard and points all the
users to it.

The only non-trivial bit is decnet where it returns an error.
However, conceptually this is identical to the blackhole functions
used in IPv4 and IPv6 which do not return errors.  So they should
either all return errors or all return zero.  For now I've stuck with
the majority and picked zero as the return value.

It doesn't really matter in practice since few if any driver would
react differently depending on a zero return value or NET_RX_DROP.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPV6]: Move nfheader_len into rt6_info
Herbert Xu [Wed, 14 Nov 2007 05:33:32 +0000 (21:33 -0800)]
[IPV6]: Move nfheader_len into rt6_info

The dst member nfheader_len is only used by IPv6.  It's also currently
creating a rather ugly alignment hole in struct dst.  Therefore this patch
moves it from there into struct rt6_info.

It also reorders the fields in rt6_info to minimize holes.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPSEC]: Use dst->header_len when resizing on output
Herbert Xu [Wed, 14 Nov 2007 05:33:01 +0000 (21:33 -0800)]
[IPSEC]: Use dst->header_len when resizing on output

Currently we use x->props.header_len when resizing on output.
However, if we're resizing at all we might as well go the whole hog
and do it for the whole dst.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPV6]: Only set nfheader_len for top xfrm dst
Herbert Xu [Wed, 14 Nov 2007 05:32:26 +0000 (21:32 -0800)]
[IPV6]: Only set nfheader_len for top xfrm dst

We only need to set nfheader_len in the top xfrm dst.  This is because
we only ever read the nfheader_len from the top xfrm dst.

It is also easier to count nfheader_len as part of header_len which
then lets us remove the ugly wrapper functions for incrementing and
decrementing header lengths in xfrm6_policy.c.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[NET]: Convert init_timer into setup_timer
Pavel Emelyanov [Thu, 24 Jan 2008 05:20:07 +0000 (21:20 -0800)]
[NET]: Convert init_timer into setup_timer

Many-many code in the kernel initialized the timer->function
and  timer->data together with calling init_timer(timer). There
is already a helper for this. Use it for networking code.

The patch is HUGE, but makes the code 130 lines shorter
(98 insertions(+), 228 deletions(-)).

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPV6]: Add raw6 drops counter.
Wang Chen [Wed, 14 Nov 2007 04:31:14 +0000 (20:31 -0800)]
[IPV6]: Add raw6 drops counter.

Add raw drops counter for IPv6 in /proc/net/raw6 .

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[IPV4]: Add raw drops counter.
Wang Chen [Wed, 14 Nov 2007 04:30:01 +0000 (20:30 -0800)]
[IPV4]: Add raw drops counter.

Add raw drops counter for IPv4 in /proc/net/raw .

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: Make tcp_splice_data_recv() static.
Adrian Bunk [Wed, 7 Nov 2007 07:32:26 +0000 (23:32 -0800)]
[TCP]: Make tcp_splice_data_recv() static.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP] splice: add tcp_splice_read() to IPV6
Jens Axboe [Wed, 7 Nov 2007 07:31:58 +0000 (23:31 -0800)]
[TCP] splice: add tcp_splice_read() to IPV6

Thanks to YOSHIFUJI Hideaki for the hint!

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[TCP]: Splice receive support.
Jens Axboe [Wed, 7 Nov 2007 07:30:13 +0000 (23:30 -0800)]
[TCP]: Splice receive support.

Support for network splice receive.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[SPLICE]: Don't assume regular pages in splice_to_pipe()
Jens Axboe [Wed, 7 Nov 2007 07:29:47 +0000 (23:29 -0800)]
[SPLICE]: Don't assume regular pages in splice_to_pipe()

Allow caller to pass in a release function, there might be
other resources that need releasing as well. Needed for
network receive.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years ago[MAC80211]: Revert unaligned warning removal.
David S. Miller [Fri, 25 Jan 2008 00:57:39 +0000 (16:57 -0800)]
[MAC80211]: Revert unaligned warning removal.

For release Linux removed this warning, but we want it
back for development.

Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog
Linus Torvalds [Mon, 28 Jan 2008 22:00:01 +0000 (09:00 +1100)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog

* git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog:
  [WATCHDOG] constify function pointer tables
  [WATCHDOG] TXx9 watchdog driver
  [WATCHDOG] misc_register patch
  [WATCHDOG] wdt: fix locking

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh-2.6
Linus Torvalds [Mon, 28 Jan 2008 21:52:50 +0000 (08:52 +1100)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh-2.6: (197 commits)
  sh: add spi header and r2d platform data V3
  sh: update r7780rp interrupt code
  sh: remove consistent alloc stuff from the machine vector
  sh: use declared coherent memory for dreamcast pci ethernet adapter
  sh: declared coherent memory support V2
  sh: Add support for SDK7780 board.
  sh: constify function pointer tables
  sh: Kill off -traditional for linker script.
  cdrom: Add support for Sega Dreamcast GD-ROM.
  sh: Kill off hs7751rvoip reference from arch/sh/Kconfig.
  sh: Drop r7780rp_defconfig, use r7780mp_defconfig as kbuild default.
  sh: Kill off dead HS771RVoIP board support.
  sh: r7785rp: Fix up DECLARE_INTC_DESC() arg mismatch.
  sh: r7785rp: Hook up the rest of the HL7785 FPGA IRQ vectors.
  sh: r2d - enable sm501 usb host function
  sh: remove voyagergx
  sh: r2d - add lcd planel timings to sm501 platform data
  sh: Add OHCI and UDC platform devices for SH7720.
  sh: intc - remove default interrupt priority tables
  sh: Correct pte size mismatch for X2 TLB.
  ...

16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid
Linus Torvalds [Mon, 28 Jan 2008 21:52:20 +0000 (08:52 +1100)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid: (24 commits)
  HID: ADS/Tech Radio si470x needs blacklist entry
  HID: Logitech Extreme 3D needs NOGET quirk
  HID: Refactor MS Presenter 8K key mapping
  HID: MS Presenter mapping for PID 0x0701
  HID: Support Samsung IR remote
  HID: fix compilation of hidbp drivers without usbhid
  HID: Blacklist the Gretag-Macbeth Huey display colorimeter
  HID: the `bit' in hidinput_mapping_quirks() is an out parameter
  HID: remove redundant WARN_ON()s in order not to scare users
  HID: force hiddev creation for SONY PS3 controller
  HID: Use hid blacklist in usbmouse/usbkbd
  HID: proper handling of MS 4k and 6k devices
  HID: remove unused variable in quirk event handler
  HID: hid-input quirk for BTC 8193
  HID: separate hid-input event quirks from generic code
  HID: refactor mapping to input subsystem for quirky devices
  HID: Microsoft Wireless Optical Desktop 3.0 quirk
  HID: Add support for Logitech Elite keyboards
  HID: add full support for Genius KB-29E
  HID: fix a potential bug in pointer casting
  ...

16 years agoMerge branch 'for-2.6.25' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Mon, 28 Jan 2008 21:51:56 +0000 (08:51 +1100)]
Merge branch 'for-2.6.25' of git://git.kernel.dk/linux-2.6-block

* 'for-2.6.25' of git://git.kernel.dk/linux-2.6-block:
  block: implement drain buffers
  __bio_clone: don't calculate hw/phys segment counts
  block: allow queue dma_alignment of zero
  blktrace: Add blktrace ioctls to SCSI generic devices

16 years agoMerge branch 'blk-end-request' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Mon, 28 Jan 2008 21:51:32 +0000 (08:51 +1100)]
Merge branch 'blk-end-request' of git://git.kernel.dk/linux-2.6-block

* 'blk-end-request' of git://git.kernel.dk/linux-2.6-block: (30 commits)
  blk_end_request: changing xsysace (take 4)
  blk_end_request: changing ub (take 4)
  blk_end_request: cleanup of request completion (take 4)
  blk_end_request: cleanup 'uptodate' related code (take 4)
  blk_end_request: remove/unexport end_that_request_* (take 4)
  blk_end_request: changing scsi (take 4)
  blk_end_request: add bidi completion interface (take 4)
  blk_end_request: changing ide-cd (take 4)
  blk_end_request: add callback feature (take 4)
  blk_end_request: changing ide normal caller (take 4)
  blk_end_request: changing cpqarray (take 4)
  blk_end_request: changing cciss (take 4)
  blk_end_request: changing ide-scsi (take 4)
  blk_end_request: changing s390 (take 4)
  blk_end_request: changing mmc (take 4)
  blk_end_request: changing i2o_block (take 4)
  blk_end_request: changing viocd (take 4)
  blk_end_request: changing xen-blkfront (take 4)
  blk_end_request: changing viodasd (take 4)
  blk_end_request: changing sx8 (take 4)
  ...

16 years agoMerge branch 'sg' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Mon, 28 Jan 2008 21:51:05 +0000 (08:51 +1100)]
Merge branch 'sg' of git://git.kernel.dk/linux-2.6-block

* 'sg' of git://git.kernel.dk/linux-2.6-block:
  SG: work with the SCSI fixed maximum allocations.
  SG: Convert SCSI to use scatterlist helpers for sg chaining
  SG: Move functions to lib/scatterlist.c and add sg chaining allocator helpers

16 years agoMerge branch 'cfq-ioc-share' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Mon, 28 Jan 2008 21:50:42 +0000 (08:50 +1100)]
Merge branch 'cfq-ioc-share' of git://git.kernel.dk/linux-2.6-block

* 'cfq-ioc-share' of git://git.kernel.dk/linux-2.6-block:
  cfq-iosched: kill some big inlines
  cfq-iosched: relax IOPRIO_CLASS_IDLE restrictions
  kernel: add CLONE_IO to specifically request sharing of IO contexts
  io_context sharing - anticipatory changes
  block: cfq: make the io contect sharing lockless
  io_context sharing - cfq changes
  io context sharing: preliminary support
  ioprio: move io priority from task_struct to io_context