Commit Graph

13772 Commits

Author SHA1 Message Date
Eric Dumazet
1240d1373c ipv4: udp: Optimise multicast reception
UDP multicast rx path is a bit complex and can hold a spinlock
for a long time.

Using a small (32 or 64 entries) stack of socket pointers can help
to perform expensive operations (skb_clone(), udp_queue_rcv_skb())
outside of the lock, in most cases.

It's also a base for a future RCU conversion of multicast recption.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Lucian Adrian Grijincu <lgrijincu@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-08 20:53:08 -08:00
Eric Dumazet
fddc17defa ipv6: udp: optimize unicast RX path
We first locate the (local port) hash chain head
If few sockets are in this chain, we proceed with previous lookup algo.

If too many sockets are listed, we take a look at the secondary
(port, address) hash chain.

We choose the shortest chain and proceed with a RCU lookup on the elected chain.

But, if we chose (port, address) chain, and fail to find a socket on given address,
 we must try another lookup on (port, in6addr_any) chain to find sockets not bound
to a particular IP.

-> No extra cost for typical setups, where the first lookup will probabbly
be performed.

RCU lookups everywhere, we dont acquire spinlock.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-08 20:53:07 -08:00
Eric Dumazet
5051ebd275 ipv4: udp: optimize unicast RX path
We first locate the (local port) hash chain head
If few sockets are in this chain, we proceed with previous lookup algo.

If too many sockets are listed, we take a look at the secondary
(port, address) hash chain we added in previous patch.

We choose the shortest chain and proceed with a RCU lookup on the elected chain.

But, if we chose (port, address) chain, and fail to find a socket on given address,
 we must try another lookup on (port, INADDR_ANY) chain to find socket not bound
to a particular IP.

-> No extra cost for typical setups, where the first lookup will probabbly
be performed.

RCU lookups everywhere, we dont acquire spinlock.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-08 20:53:07 -08:00
Eric Dumazet
512615b6b8 udp: secondary hash on (local port, local address)
Extends udp_table to contain a secondary hash table.

socket anchor for this second hash is free, because UDP
doesnt use skc_bind_node : We define an union to hold
both skc_bind_node & a new hlist_nulls_node udp_portaddr_node

udp_lib_get_port() inserts sockets into second hash chain
(additional cost of one atomic op)

udp_lib_unhash() deletes socket from second hash chain
(additional cost of one atomic op)

Note : No spinlock lockdep annotation is needed, because
lock for the secondary hash chain is always get after
lock for primary hash chain.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-08 20:53:06 -08:00
Eric Dumazet
d4cada4ae1 udp: split sk_hash into two u16 hashes
Union sk_hash with two u16 hashes for udp (no extra memory taken)

One 16 bits hash on (local port) value (the previous udp 'hash')

One 16 bits hash on (local address, local port) values, initialized
but not yet used. This second hash is using jenkin hash for better
distribution.

Because the 'port' is xored later, a partial hash is performed
on local address + net_hash_mix(net)

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-08 20:53:05 -08:00
Eric Dumazet
fdcc8aa953 udp: add a counter into udp_hslot
Adds a counter in udp_hslot to keep an accurate count
of sockets present in chain.

This will permit to upcoming UDP lookup algo to chose
the shortest chain when secondary hash is added.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-08 20:53:04 -08:00
Stephen Rothwell
415ce61aef net/appletalk: using compat_ptr needs inclusion of linux/compat.h
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-08 20:41:03 -08:00
Eric W. Biederman
81adee47df net: Support specifying the network namespace upon device creation.
There is no good reason to not support userspace specifying the
network namespace during device creation, and it makes it easier
to create a network device and pass it to a child network namespace
with a well known name.

We have to be careful to ensure that the target network namespace
for the new device exists through the life of the call.  To keep
that logic clear I have factored out the network namespace grabbing
logic into rtnl_link_get_net.

In addtion we need to continue to pass the source network namespace
to the rtnl_link_ops.newlink method so that we can find the base
device source network namespace.

Signed-off-by: Eric W. Biederman <ebiederm@aristanetworks.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
2009-11-08 00:53:51 -08:00
Joe Perches
f7a3a1d8af appletalk/ddp.c: Neaten checksum function
atalk_sum_partial can now use the rol16 function in bitops.h

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-08 00:43:19 -08:00
Eric Dumazet
fd5c002761 ipv6: avoid dev_hold()/dev_put() in rawv6_bind()
Using RCU helps not touching device refcount in rawv6_bind()

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-08 00:43:18 -08:00
Eric Dumazet
e0d087af72 rtnetlink: Cleanups
Pure cleanups patch

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-07 01:26:17 -08:00
Arnd Bergmann
91774904fb net/x25: push BKL usage into x25_proto
The x25 driver uses lock_kernel() implicitly through
its proto_ops wrapper. The makes the usage explicit
in order to get rid of that wrapper and to better document
the usage of the BKL.

The next step should be to get rid of the usage of the BKL
in x25 entirely, which requires understanding what data
structures need serialized accesses.

Cc: Henner Eisen <eis@baty.hanse.de>
Cc: David S. Miller <davem@davemloft.net>
Cc: linux-x25@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-07 00:46:40 -08:00
Arnd Bergmann
58a9d73202 net/irda: push BKL into proto_ops
The irda driver uses the BKL implicitly in its protocol
operations. Replace the wrapped proto_ops with explicit
lock_kernel() calls makes the usage more obvious and
shrinks the size of the object code.

The calls t lock_kernel() should eventually all be replaced
by other serialization methods, which requires finding out

The calls t lock_kernel() should eventually all be replaced
by other serialization methods, which requires finding out
which data actually needs protection.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-07 00:46:39 -08:00
Arnd Bergmann
83927ba069 net/ipx: push down BKL into a ipx_dgram_ops
Making the BKL usage explicit in ipx makes it more
obvious where it is used, reduces code size and helps
getting rid of the BKL in common code.

I did not analyse how to kill lock_kernel from ipx
entirely, this will involve either proving that it's not
needed, or replacing with a proper mutex or spinlock,
after finding out which data structures are protected
by the lock.

Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: David S. Miller <davem@davemloft.net>
Cc: Stephen Hemminger <shemminger@vyatta.com>
Cc: netdev@vger.kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-07 00:46:39 -08:00
Arnd Bergmann
ecced8ba87 net/appletalk: push down BKL into a atalk_dgram_ops
Making the BKL usage explicit in appletalk makes it more
obvious where it is used, reduces code size and helps
getting rid of the BKL in common code.

I did not analyse how to kill lock_kernel from appletalk
entirely, this will involve either proving that it's not
needed, or replacing with a proper mutex or spinlock,
after finding out which data structures are protected
by the lock.

Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: David S. Miller <davem@davemloft.net>
Cc: Stephen Hemminger <shemminger@vyatta.com>
Cc: netdev@vger.kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-07 00:46:37 -08:00
Thomas Gleixner
d3bcfefaca net: Replace old style lock initializer
SPIN_LOCK_UNLOCKED is deprecated. Use DEFINE_SPINLOCK instead.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-07 00:46:34 -08:00
Arnd Bergmann
9177efd399 net, compat_ioctl: handle more ioctls correctly
The MII ioctls and SIOCSIFNAME need to go through ifsioc conversion,
which they never did so far. Some others are not implemented in the
native path, so we can just return -EINVAL directly.

Add IFSLAVE ioctls to the EINVAL list and move it to the end to
optimize the code path for the common case.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-06 23:11:15 -08:00
Arnd Bergmann
6b96018b28 compat: move sockios handling to net/socket.c
This removes the original socket compat_ioctl code
from fs/compat_ioctl.c and converts the code from the copy
in net/socket.c into a single function. We add a few cycles
of runtime to compat_sock_ioctl() with the long switch()
statement, but gain some cycles in return by simplifying
the call chain to get there.

Due to better inlining, save 1.5kb of object size in the
process, and enable further savings:

before:
   text    data     bss     dec     hex filename
  13540   18008    2080   33628    835c obj/fs/compat_ioctl.o
  14565     636      40   15241    3b89 obj/net/socket.o

after:
   text    data     bss     dec     hex filename
   8916   15176    2080   26172    663c obj/fs/compat_ioctl.o
  20725     636      40   21401    5399 obj/net/socket.o

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-06 23:10:54 -08:00
Arnd Bergmann
2066022177 appletalk: handle SIOCATALKDIFADDR compat ioctl
We must not have a compat ioctl handler for SIOCATALKDIFADDR
in common code, because the same number is used in other protocols
with different data structures.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-06 23:01:14 -08:00
Arnd Bergmann
7a229387d3 net: copy socket ioctl code to net/socket.h
This makes an identical copy of the socket compat_ioctl code
from fs/compat_ioctl.c to net/socket.c, as a preparation
for moving the functionality in a way that can be easily
reviewed.

The code is hidden inside of #if 0 and gets activated in the
patch that will make it work.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-06 23:00:29 -08:00
David S. Miller
10d626f4f4 Merge branch 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/lowpan/lowpan 2009-11-06 17:57:51 -08:00
David S. Miller
62d83681e5 Merge branch 'linux-2.6.33.y' of git://git.kernel.org/pub/scm/linux/kernel/git/inaky/wimax 2009-11-06 05:01:54 -08:00
Dmitry Eremin-Solenikov
bb1cafb8fc ieee802154: add support for creation/removal of logic interfaces
Add support for two more NL802154 commands: ADD_IFACE and DEL_IFACE,
thus allowing creation and removal of logic WPAN interfaces on the top
of wpan-phy.

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
2009-11-06 14:32:24 +03:00
Dmitry Eremin-Solenikov
0a868b26c8 ieee802154: add PHY_NAME to LIST_IFACE command results
Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
2009-11-06 14:32:22 +03:00
Dmitry Eremin-Solenikov
339b4ca5f6 ieee802154: add two nl802154 helpers
Add two nl802154 helpers: one for starting a reply message, one for sending.

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
2009-11-06 14:32:21 +03:00
Dmitry Eremin-Solenikov
1eaa9d03d3 ieee802154: add LIST_PHY command support
Add nl802154 command to get information about PHY's present in
the system.

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
2009-11-06 14:31:22 +03:00
Dmitry Eremin-Solenikov
78fe738d1a ieee802154: split away MAC commands implementation
Move all mac-related stuff to separate file so that ieee802154/netlink.c
contains only generic code.

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
2009-11-06 14:31:20 +03:00
Dmitry Eremin-Solenikov
cb6b376357 ieee802154: merge nl802154 and wpan-class in single module
There is no real need to have ieee802154 interfaces separate
into several small modules, as neither of them has it's own use.

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
2009-11-06 14:31:18 +03:00
Dmitry Eremin-Solenikov
e9cf356c0c wpan-phy: follow usual patter of devices registration
Follow the usual pattern of devices registration by adding new function
(wpan_phy_set_dev) that sets child->parent relationship and removing
parent argument from wpan_phy_register call.

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
2009-11-06 14:29:50 +03:00
Dmitry Eremin-Solenikov
a0b4a738e0 wpan-phy: allow specifying a per-page channel mask
IEEE 802.15.4-2006 defines channel pages that hold channels (max 32 pages,
27 channels per page). Allow the driver to specify supported channels
on pages, other than the first one.

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
2009-11-06 14:23:41 +03:00
Dmitry Eremin-Solenikov
375bb0e04b wpan-phy: use snprintf to limit the amount of chars written
Use snprintf to limit the amount of chars put in the buffer for attr -> show.

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
2009-11-06 14:13:45 +03:00
Dmitry Eremin-Solenikov
37eb0edc84 wpan-phy: init channel/page fields
Set page to zero (for compatibility w/ devices supporting only first page).
Also init channel by default to -1 to disallow transfers for non-initialised
devices.

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
2009-11-06 14:13:22 +03:00
Dmitry Eremin-Solenikov
1c889f4db6 wpan-phy: add wpan-phy iteration functions
Add API to iterate over the wpan-phy instances.

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
2009-11-06 14:12:24 +03:00
David S. Miller
230f9bb701 Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Conflicts:
	drivers/net/usb/cdc_ether.c

All CDC ethernet devices of type USB_CLASS_COMM need to use
'&mbm_info'.

Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-06 00:55:55 -08:00
Eric Dumazet
887e671f32 decnet: netdevice refcount leak
While working on device refcount stuff, I found a device refcount leak
through DECNET.
This nasty bug can be used to hold refcounts on any !DECNET netdevice.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-06 00:50:39 -08:00
Jozsef Kadlecsik
f9dd09c7f7 netfilter: nf_nat: fix NAT issue in 2.6.30.4+
Vitezslav Samel discovered that since 2.6.30.4+ active FTP can not work
over NAT. The "cause" of the problem was a fix of unacknowledged data
detection with NAT (commit a3a9f79e36).
However, actually, that fix uncovered a long standing bug in TCP conntrack:
when NAT was enabled, we simply updated the max of the right edge of
the segments we have seen (td_end), by the offset NAT produced with
changing IP/port in the data. However, we did not update the other parameter
(td_maxend) which is affected by the NAT offset. Thus that could drift
away from the correct value and thus resulted breaking active FTP.

The patch below fixes the issue by *not* updating the conntrack parameters
from NAT, but instead taking into account the NAT offsets in conntrack in a
consistent way. (Updating from NAT would be more harder and expensive because
it'd need to re-calculate parameters we already calculated in conntrack.)

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-06 00:43:42 -08:00
David S. Miller
000ba2e43f net: Fix build warning in sock_bindtodevice().
net/core/sock.c: In function 'sock_setsockopt':
net/core/sock.c:396: warning: 'index' may be used uninitialized in this function
net/core/sock.c:396: note: 'index' was declared here

GCC can't see that all paths initialize index, so just
set it to the default (0) and eliminate the specific
code block that handles the null device name string.

Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 22:37:11 -08:00
Eric Dumazet
baac856454 pktgen: tx_bytes might be slightly wrong
cur_pkt_size can be changed in proc fs while pktgen is running,
we better use a private field to get precise tx-bytes counter.

Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 22:34:27 -08:00
Eric Dumazet
bf8e56bfc4 net: sock_bindtodevice() RCU-ification
Avoid dev_hold()/dev_put() in sock_bindtodevice()

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 22:34:24 -08:00
Eric Dumazet
69df9d5993 ip_frag: dont touch device refcount
When sending fragmentation expiration ICMP V4/V6 messages,
we can avoid touching device refcount, thanks to RCU

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 22:34:22 -08:00
Eric Dumazet
bd27a8750c net_cls: Use __dev_get_by_index()
We hold RTNL in tc_dump_tfilter(), we can avoid dev_hold()/dev_put()

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 22:34:21 -08:00
Eric Dumazet
40c9c31e38 sctp: ipv6: avoid touching device refcount
Avoid touching device refcount in sctp/ipv6, thanks to RCU

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 22:34:21 -08:00
Eric Dumazet
122ec6ffca netlabel: remove dev_put() calls
Use dev_get_by_name_rcu() to avoid dev_put() calls,
in sections already inside a rcu_read_lock()

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 22:34:18 -08:00
Eric Dumazet
31ef30c760 bridge: remove dev_put() in add_del_if()
add_del_if() is called with RTNL, we can use __dev_get_by_index()
instead of [dev_get_by_index() + dev_put()]

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 22:34:16 -08:00
Eric Paris
c84b3268da net: check kern before calling security subsystem
Before calling capable(CAP_NET_RAW) check if this operations is on behalf
of the kernel or on behalf of userspace.  Do not do the security check if
it is on behalf of the kernel.

Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 22:18:18 -08:00
Eric Paris
3f378b6844 net: pass kern to net_proto_family create function
The generic __sock_create function has a kern argument which allows the
security system to make decisions based on if a socket is being created by
the kernel or by userspace.  This patch passes that flag to the
net_proto_family specific create function, so it can do the same thing.

Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 22:18:14 -08:00
Eric Paris
13f18aa05f net: drop capability from protocol definitions
struct can_proto had a capability field which wasn't ever used.  It is
dropped entirely.

struct inet_protosw had a capability field which can be more clearly
expressed in the code by just checking if sock->type = SOCK_RAW.

Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 21:40:17 -08:00
Eric Dumazet
b4ec824021 rose: device refcount leak
While hunting dev_put() for net-next-2.6, I found a device refcount
leak in ROSE, ioctl(SIOCADDRT) error path.

Fix is to not touch device refcount, as we hold RTNL

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 20:56:07 -08:00
Stephen Hemminger
1056bd5167 bridge: prevent bridging wrong device
The bridge code assumes ethernet addressing, so be more strict in
the what is allowed. This showed up when GRE had a bug and was not
using correct address format.

Add some more comments for increased clarity.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-05 20:46:52 -08:00
Gilad Ben-Yossef
6a2a2d6bf8 tcp: Use defaults when no route options are available
Trying to parse the option of a SYN packet that we have
no route entry for should just use global wide defaults
for route entry options.

Signed-off-by: Gilad Ben-Yossef <gilad@codefidence.com>
Tested-by: Valdis.Kletnieks@vt.edu
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-04 23:24:15 -08:00