platform/kernel/kernel-mfld-blackbay.git
14 years agotcp: fix a timewait refcnt race
Eric Dumazet [Thu, 3 Dec 2009 00:49:01 +0000 (00:49 +0000)]
tcp: fix a timewait refcnt race

After TCP RCU conversion, tw->tw_refcnt should not be set to 1 in
inet_twsk_alloc(). It allows a RCU reader to get this timewait socket,
while we not yet stabilized it.

Only choice we have is to set tw_refcnt to 0 in inet_twsk_alloc(),
then atomic_add() it later, once everything is done.

Location of this atomic_add() is tricky, because we dont want another
writer to find this timewait in ehash, while tw_refcnt is still zero !

Thanks to Kapil Dakhane tests and reports.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotcp: connect() race with timewait reuse
Eric Dumazet [Wed, 2 Dec 2009 22:31:19 +0000 (22:31 +0000)]
tcp: connect() race with timewait reuse

Its currently possible that several threads issuing a connect() find
the same timewait socket and try to reuse it, leading to list
corruptions.

Condition for bug is that these threads bound their socket on same
address/port of to-be-find timewait socket, and connected to same
target. (SO_REUSEADDR needed)

To fix this problem, we could unhash timewait socket while holding
ehash lock, to make sure lookups/changes will be serialized. Only
first thread finds the timewait socket, other ones find the
established socket and return an EADDRNOTAVAIL error.

This second version takes into account Evgeniy's review and makes sure
inet_twsk_put() is called outside of locked sections.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: Add support for ethtool self test
Suresh R [Fri, 4 Dec 2009 00:15:52 +0000 (16:15 -0800)]
be2net: Add support for ethtool self test

This patch adds support for ethtool selftest.

From: Suresh R <sureshr@serverengines.com>
Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotcp: diag: Dont report negative values for rx queue
Eric Dumazet [Fri, 4 Dec 2009 00:06:13 +0000 (16:06 -0800)]
tcp: diag: Dont report negative values for rx queue

Both netlink and /proc/net/tcp interfaces can report transient
negative values for rx queue.

ss ->
State   Recv-Q Send-Q  Local Address:Port  Peer Address:Port
ESTAB   -6     6       127.0.0.1:45956     127.0.0.1:3333

netstat ->
tcp   4294967290      6 127.0.0.1:37784  127.0.0.1:3333 ESTABLISHED

This is because we dont lock socket while computing
tp->rcv_nxt - tp->copied_seq,
and another CPU can update copied_seq before rcv_next in RX path.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Fix ks8851 snl NULL pointer dereference Oops
Tristram Ha [Thu, 3 Dec 2009 11:06:42 +0000 (11:06 +0000)]
net: Fix ks8851 snl NULL pointer dereference Oops

This fixes the NULL pointer dereference Oops in the ks8851 snl network
driver during transmission.  All socket buffers from the queue are
processed inside the loop, but new workqueues have been scheduled to
run.

Signed-off-by: Tristram Ha <Tristram.Ha@micrel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonetdevice: provide common routine for macvlan and vlan operstate management
Patrick Mullaney [Thu, 3 Dec 2009 23:59:22 +0000 (15:59 -0800)]
netdevice: provide common routine for macvlan and vlan operstate management

Provide common routine for the transition of operational state for a leaf
device during a root device transition.

Signed-off-by: Patrick Mullaney <pmullaney@novell.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobnx2: Refine VPD logic.
Michael Chan [Thu, 3 Dec 2009 23:58:00 +0000 (15:58 -0800)]
bnx2: Refine VPD logic.

- Replace magic values with constants
- Simplify length calculation and fix a bug

Based on valuable feedback from Ben Hutchings <bhutchings@solarflare.com>

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoixgbe: use EIAM to automask MSI-X
Jesse Brandeburg [Thu, 3 Dec 2009 11:33:54 +0000 (11:33 +0000)]
ixgbe: use EIAM to automask MSI-X

when disabling interrupts, driver was writing with IO, this is no
necessary because on ixgbe parts the hardware can "oneshot"
disable and clear the interrupt.  So on 82598/82599 use of EIAM
should avoid one posted write per interrupt when in MSI-X mode.
This should improve performance and seems to in my limited
testing, reduce CPU utilization VERY slightly.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoixgbe: performance tweaks
Jesse Brandeburg [Thu, 3 Dec 2009 11:33:29 +0000 (11:33 +0000)]
ixgbe: performance tweaks

drop variables that had cache lines modified in simultaneous hot paths.
keep some variables modified on hot paths but make their storage per queue.
cache align DMA data buffer start addresses.
cache align (padding) some structures that end within a cacheline.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoixgbe: change default ring size
Jesse Brandeburg [Thu, 3 Dec 2009 11:33:07 +0000 (11:33 +0000)]
ixgbe: change default ring size

decrease the memory utilization of the tx / rx queue allocation
by changing the default ring size to 512 (from 1024).  At
1024 rx entries of 2KB each (from 4kB slab) with 16 queues
ixgbe was using 64 MB of memory per port, which is not
necessary.

Users can still change queue lengths with ethtool -k.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoixgbe: select FCoE Tx queue in ndo_select_queue
Yi Zou [Thu, 3 Dec 2009 11:32:44 +0000 (11:32 +0000)]
ixgbe: select FCoE Tx queue in ndo_select_queue

This removes the Tx queue selection for FCoE traffic from ixgbe_xmit_frame()
and does it in the ndo_select_queue() call, moving all Tx queue selection
into a single routine.

Signed-off-by: Yi Zou <yi.zou@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoixgbe: use known user priority for FCoE when DCB is enabled
Yi Zou [Thu, 3 Dec 2009 11:32:22 +0000 (11:32 +0000)]
ixgbe: use known user priority for FCoE when DCB is enabled

Store the user priority for FCoE and use it directly for outgoing
FCoE traffic when DCB is enabled.

Signed-off-by: Yi Zou <yi.zou@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agousbnet & cdc-ether: Autosuspend for online devices
Oliver Neukum [Thu, 3 Dec 2009 23:31:18 +0000 (15:31 -0800)]
usbnet & cdc-ether: Autosuspend for online devices

Using remote wakeup and delayed transmission to allow
online device to go into usb autosuspend.
Minimal alternate support for devices that don't support
remote wakeup.

Signed-off-by: Oliver Neukum <oliver@neukum.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocdc-ether: Implement reset_resume()
Oliver Neukum [Thu, 3 Dec 2009 11:41:07 +0000 (11:41 +0000)]
cdc-ether: Implement reset_resume()

Normal resume can do double duty as reset_resume() for this driver
as it keeps no state in the device

Signed-off-by: Oliver Neukum <oliver@neukum.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/holtmann/bluet...
David S. Miller [Thu, 3 Dec 2009 21:51:02 +0000 (13:51 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/holtmann/bluetooth-next-2.6

14 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/kaber/nf-next-2.6
David S. Miller [Thu, 3 Dec 2009 21:23:12 +0000 (13:23 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/kaber/nf-next-2.6

14 years agobnx2: Update version to 2.0.3.
Michael Chan [Thu, 3 Dec 2009 09:46:35 +0000 (09:46 +0000)]
bnx2: Update version to 2.0.3.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobnx2: Read firmware version from VPD.
Michael Chan [Thu, 3 Dec 2009 09:46:34 +0000 (09:46 +0000)]
bnx2: Read firmware version from VPD.

And display it through ethtool -i.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobnx2: Print warning when unable to allocate the full SKB/page ring.
Michael Chan [Thu, 3 Dec 2009 09:46:33 +0000 (09:46 +0000)]
bnx2: Print warning when unable to allocate the full SKB/page ring.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobnx2: Dump some state during tx timeout.
Michael Chan [Thu, 3 Dec 2009 09:46:32 +0000 (09:46 +0000)]
bnx2: Dump some state during tx timeout.

To help debug the problem.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobnx2: Protect tx timeout reset with rtnl_lock().
Michael Chan [Thu, 3 Dec 2009 09:46:31 +0000 (09:46 +0000)]
bnx2: Protect tx timeout reset with rtnl_lock().

To prevent race conditions with other reset events.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotg3: Update version to 3.105
Matt Carlson [Thu, 3 Dec 2009 08:36:25 +0000 (08:36 +0000)]
tg3: Update version to 3.105

This patch updates the tg3 version to 3.105.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotg3: Clean tg3_init_one()
Matt Carlson [Thu, 3 Dec 2009 08:36:24 +0000 (08:36 +0000)]
tg3: Clean tg3_init_one()

Tg3's firmware is no longer loaded from tg3_init_one().  Remove the
error path that releases the firmware back to the system.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotg3: Use pci_read_vpd() instead of private method
Matt Carlson [Thu, 3 Dec 2009 08:36:23 +0000 (08:36 +0000)]
tg3: Use pci_read_vpd() instead of private method

If the driver needs to use PCI configuration space to read the VPD data,
the accesses could collide with the pci_read_vpd() function.  This patch
converts the driver to use pci_read_vpd() to avoid any potential
collisions.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotg3: Add some VPD preprocessor constants
Matt Carlson [Thu, 3 Dec 2009 08:36:22 +0000 (08:36 +0000)]
tg3: Add some VPD preprocessor constants

This patch cleans up the VPD code by creating preprocessor definitions
and using them in the place of hardcoded constants.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotg3: Add 57765 asic rev
Matt Carlson [Thu, 3 Dec 2009 08:36:21 +0000 (08:36 +0000)]
tg3: Add 57765 asic rev

This patch adds the 57765 asic revision.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotg3: Make TSS enable independent of MSI-X enable
Matt Carlson [Thu, 3 Dec 2009 08:36:20 +0000 (08:36 +0000)]
tg3: Make TSS enable independent of MSI-X enable

The 57765 asic rev has MSI-X capability, but does not support TSS.  This
patch changes the tx paths so that TSS is explicitly mentioned, rather
than implied through the ENABLE_MSIX flag.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agodrivers/net: Move && and || to end of previous line
Joe Perches [Thu, 3 Dec 2009 07:58:21 +0000 (07:58 +0000)]
drivers/net: Move && and || to end of previous line

Only files where David Miller is the primary git-signer.
wireless, wimax, ixgbe, etc are not modified.

Compile tested x86 allyesconfig only
Not all files compiled (not x86 compatible)

Added a few > 80 column lines, which I ignored.
Existing checkpatch complaints ignored.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: Bump up the driver version number
Ajit Khaparde [Thu, 3 Dec 2009 06:19:08 +0000 (06:19 +0000)]
be2net: Bump up the driver version number

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: Support for WoL using magic packet after suspend.
Ajit Khaparde [Thu, 3 Dec 2009 06:16:59 +0000 (06:16 +0000)]
be2net: Support for WoL using magic packet after suspend.

Add support for WOL using Magic Packet after suspend is done.

Signed-off-by: Sarveshwar Bandi <sarveshwarb@serverengines.com>
Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: Changes to print fw command opcode when command is failed by controller.
Ajit Khaparde [Thu, 3 Dec 2009 06:12:06 +0000 (06:12 +0000)]
be2net: Changes to print fw command opcode when command is failed by controller.

When a firmware command fails, only the failure codes are printed.
It is difficult to co-relate this to the actual command that has failed.
These changes will now print the command code that has failed.

Signed-off-by: Sarveshwar Bandi <sarveshwarb@serverengines.com>
Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Batch inet_twsk_purge
Eric W. Biederman [Thu, 3 Dec 2009 02:29:09 +0000 (02:29 +0000)]
net: Batch inet_twsk_purge

This function walks the whole hashtable so there is no point in
passing it a network namespace.  Instead I purge all timewait
sockets from dead network namespaces that I find.  If the namespace
is one of the once I am trying to purge I am guaranteed no new timewait
sockets can be formed so this will get them all.  If the namespace
is one I am not acting for it might form a few more but I will
call inet_twsk_purge again and  shortly to get rid of them.  In
any even if the network namespace is dead timewait sockets are
useless.

Move the calls of inet_twsk_purge into batch_exit routines so
that if I am killing a bunch of namespaces at once I will just
call inet_twsk_purge once and save a lot of redundant unnecessary
work.

My simple 4k network namespace exit test the cleanup time dropped from
roughly 8.2s to 1.6s.  While the time spent running inet_twsk_purge fell
to about 2ms.  1ms for ipv4 and 1ms for ipv6.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Use rcu lookups in inet_twsk_purge.
Eric W. Biederman [Thu, 3 Dec 2009 02:29:08 +0000 (02:29 +0000)]
net: Use rcu lookups in inet_twsk_purge.

While we are looking up entries to free there is no reason to take
the lock in inet_twsk_purge.  We have to drop locks and restart
occassionally anyway so adding a few more in case we get on the
wrong list because of a timewait move is no big deal.  At the
same time not taking the lock for long periods of time is much
more polite to the rest of the users of the hash table.

In my test configuration of killing 4k network namespaces
this change causes 4k back to back runs of inet_twsk_purge on an
empty hash table to go from roughly 20.7s to 3.3s, and the total
time to destroy 4k network namespaces goes from roughly 44s to
3.3s.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Allow fib_rule_unregister to batch
Eric W. Biederman [Thu, 3 Dec 2009 20:22:55 +0000 (12:22 -0800)]
net: Allow fib_rule_unregister to batch

Refactor the code so fib_rules_register always takes a template instead
of the actual fib_rules_ops structure that will be used.  This is
required for network namespace support so 2 out of the 3 callers already
do this, it allows the error handling to be made common, and it allows
fib_rules_unregister to free the template for hte caller.

Modify fib_rules_unregister to use call_rcu instead of syncrhonize_rcu
to allw multiple namespaces to be cleaned up in the same rcu grace
period.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonetns: Add an explicit rcu_barrier to unregister_pernet_{device|subsys}
Eric W. Biederman [Thu, 3 Dec 2009 02:29:06 +0000 (02:29 +0000)]
netns: Add an explicit rcu_barrier to unregister_pernet_{device|subsys}

This allows namespace exit methods to batch work that comes requires an
rcu barrier using call_rcu without having to treat the
unregister_pernet_operations cases specially.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Allow xfrm_user_net_exit to batch efficiently.
Eric W. Biederman [Thu, 3 Dec 2009 02:29:05 +0000 (02:29 +0000)]
net: Allow xfrm_user_net_exit to batch efficiently.

xfrm.nlsk is provided by the xfrm_user module and is access via rcu from
other parts of the xfrm code.  Add xfrm.nlsk_stash a copy of xfrm.nlsk that
will never be set to NULL.  This allows the synchronize_net and
netlink_kernel_release to be deferred until a whole batch of xfrm.nlsk sockets
have been set to NULL.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Move network device exit batching
Eric W. Biederman [Thu, 3 Dec 2009 02:29:04 +0000 (02:29 +0000)]
net: Move network device exit batching

Move network device exit batching from a special case in
net_namespace.c to using common mechanisms in dev.c

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Add support for batching network namespace cleanups
Eric W. Biederman [Thu, 3 Dec 2009 02:29:03 +0000 (02:29 +0000)]
net: Add support for batching network namespace cleanups

- Add exit_list to struct net to support building lists of network
  namespaces to cleanup.

- Add exit_batch to pernet_operations to allow running operations only
  once during a network namespace exit.  Instead of once per network
  namespace.

- Factor opt ops_exit_list and ops_exit_free so the logic with cleanup
  up a network namespace does not need to be duplicated.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoipv4 05/05: add sysctl to accept packets with local source addresses
Patrick McHardy [Thu, 3 Dec 2009 01:25:58 +0000 (01:25 +0000)]
ipv4 05/05: add sysctl to accept packets with local source addresses

commit 8ec1e0ebe26087bfc5c0394ada5feb5758014fc8
Author: Patrick McHardy <kaber@trash.net>
Date:   Thu Dec 3 12:16:35 2009 +0100

    ipv4: add sysctl to accept packets with local source addresses

    Change fib_validate_source() to accept packets with a local source address when
    the "accept_local" sysctl is set for the incoming inet device. Combined with the
    previous patches, this allows to communicate between multiple local interfaces
    over the wire.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet 04/05: fib_rules: allow to delete local rule
Patrick McHardy [Thu, 3 Dec 2009 01:25:57 +0000 (01:25 +0000)]
net 04/05: fib_rules: allow to delete local rule

commit d124356ce314fff22a047ea334379d5105b2d834
Author: Patrick McHardy <kaber@trash.net>
Date:   Thu Dec 3 12:16:35 2009 +0100

    net: fib_rules: allow to delete local rule

    Allow to delete the local rule and recreate it with a higher priority. This
    can be used to force packets with a local destination out on the wire instead
    of routing them to loopback. Additionally this patch allows to recreate rules
    with a priority of 0.

    Combined with the previous patch to allow oif classification, a socket can
    be bound to the desired interface and packets routed to the wire like this:

    # move local rule to lower priority
    ip rule add pref 1000 lookup local
    ip rule del pref 0

    # route packets of sockets bound to eth0 to the wire independant
    # of the destination address
    ip rule add pref 100 oif eth0 lookup 100
    ip route add default dev eth0 table 100

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet 03/05: fib_rules: add oif classification
Patrick McHardy [Thu, 3 Dec 2009 01:25:56 +0000 (01:25 +0000)]
net 03/05: fib_rules: add oif classification

commit 68144d350f4f6c348659c825cde6a82b34c27a91
Author: Patrick McHardy <kaber@trash.net>
Date:   Thu Dec 3 12:05:25 2009 +0100

    net: fib_rules: add oif classification

    Support routing table lookup based on the flow's oif. This is useful to
    classify packets originating from sockets bound to interfaces differently.

    The route cache already includes the oif and needs no changes.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet 02/05: fib_rules: rename ifindex/ifname/FRA_IFNAME to iifindex/iifname/FRA_IIFNAME
Patrick McHardy [Thu, 3 Dec 2009 01:25:54 +0000 (01:25 +0000)]
net 02/05: fib_rules: rename ifindex/ifname/FRA_IFNAME to iifindex/iifname/FRA_IIFNAME

commit 229e77eec406ad68662f18e49fda8b5d366768c5
Author: Patrick McHardy <kaber@trash.net>
Date:   Thu Dec 3 12:05:23 2009 +0100

    net: fib_rules: rename ifindex/ifname/FRA_IFNAME to iifindex/iifname/FRA_IIFNAME

    The next patch will add oif classification, rename interface related members
    and attributes to reflect that they're used for iif classification.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet 01/05: fib_rules: rearrange struct fib_rule
Patrick McHardy [Thu, 3 Dec 2009 01:25:53 +0000 (01:25 +0000)]
net 01/05: fib_rules: rearrange struct fib_rule

commit b8952893d5d86f69c4e499d191b98c6658f64b0f
Author: Patrick McHardy <kaber@trash.net>
Date:   Thu Dec 3 12:05:22 2009 +0100

    net: fib_rules: rearrange struct fib_rule

    The ifname member is only used to resolve interface names and is not needed
    during rule lookups. The target and ctarget members however are used during
    rule lookups and are currently located in a second cacheline.

    Move ifname further to the end to make sure both target and ctarget are
    located in the same cacheline as other members used during rule lookups.

    The layout on 64 bit changes from:

    struct fib_rule {
     ...
            u32                        table;                /*    56     4 */
            u8                         action;               /*    60     1 */

            /* XXX 3 bytes hole, try to pack */

            /* --- cacheline 1 boundary (64 bytes) --- */
            u32                        target;               /*    64     4 */

            /* XXX 4 bytes hole, try to pack */

            struct fib_rule *          ctarget;              /*    72     8 */
            struct rcu_head            rcu;                  /*    80    16 */
            struct net *               fr_net;               /*    96     8 */
    };

    to:

    struct fib_rule {
     ...
            u32                        table;                /*    40     4 */
            u8                         action;               /*    44     1 */

            /* XXX 3 bytes hole, try to pack */

            u32                        target;               /*    48     4 */

            /* XXX 4 bytes hole, try to pack */

            struct fib_rule *          ctarget;              /*    56     8 */
            /* --- cacheline 1 boundary (64 bytes) --- */
            char                       ifname[16];           /*    64    16 */
            struct rcu_head            rcu;                  /*    80    16 */
            struct net *               fr_net;               /*    96     8 */

    };

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoBluetooth: Declare MODULE_FIRMWARE for Marvell SDIO driver
Ben Hutchings [Sat, 7 Nov 2009 21:41:18 +0000 (21:41 +0000)]
Bluetooth: Declare MODULE_FIRMWARE for Marvell SDIO driver

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Add RFCOMM option to use L2CAP ERTM mode
Marcel Holtmann [Mon, 5 Oct 2009 10:23:48 +0000 (12:23 +0200)]
Bluetooth: Add RFCOMM option to use L2CAP ERTM mode

By default the RFCOMM layer would still use L2CAP basic mode. For testing
purposes this option enables RFCOMM to select enhanced retransmission
mode.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Add L2CAP option for max transmit value
Marcel Holtmann [Mon, 5 Oct 2009 09:35:43 +0000 (11:35 +0200)]
Bluetooth: Add L2CAP option for max transmit value

For testing purposes it is important to modify the max transmit value.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Enable auto sleep mode for btmrvl driver
Amitkumar Karwar [Wed, 30 Sep 2009 18:18:30 +0000 (11:18 -0700)]
Bluetooth: Enable auto sleep mode for btmrvl driver

The auto sleep mode for btmrvl driver is not enabled by default.
This patch enables auto sleep mode when card is probed.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Fix 'SendRRorRNR' to send the ReqSeq value
Gustavo F. Padovan [Sat, 3 Oct 2009 05:34:40 +0000 (02:34 -0300)]
Bluetooth: Fix 'SendRRorRNR' to send the ReqSeq value

SendRRorRNR needs to acknowledge received I-frames (actually every packet
needs to acknowledge received I-frames by sending the proper packet
sequence number), so ReqSeq is set to the next I-frame number sequence to
be pulled by the reassembly function.
SendRRorRNR tells the remote side about local busy conditions, it sends
a Receiver Ready frame if local busy is false or a Receiver Not Ready
if local busy is true.
ReqSeq is the packet's field to send the number of the acknowledged
packets.

Signed-off-by: Gustavo F. Padovan <gustavo@las.ic.unicamp.br>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Implement RejActioned flag
Gustavo F. Padovan [Sat, 3 Oct 2009 05:34:39 +0000 (02:34 -0300)]
Bluetooth: Implement RejActioned flag

RejActioned is used to prevent retransmission when a entity is on the
WAIT_F state, i.e., waiting for a frame with F-bit set due local busy
condition or a expired retransmission timer. (When these two events raise
they send a frame with the Poll bit set and enters in the WAIT_F state to
wait for a frame with the Final bit set.)
The local entity doesn't send I-frames(the data frames) until the receipt
of a frame with F-bit set. When that happens it also set RejActioned to false.
RejActioned is a mandatory feature of ERTM spec.

Signed-off-by: Gustavo F. Padovan <gustavo@las.ic.unicamp.br>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Fix sending ReqSeq on I-frames
Gustavo F. Padovan [Sat, 3 Oct 2009 05:34:38 +0000 (02:34 -0300)]
Bluetooth: Fix sending ReqSeq on I-frames

As specified by ERTM spec an ERTM channel can acknowledge received
I-frames(the data frames) by sending an I-frame with the proper ReqSeq
value (i.e. ReqSeq is set to BufferSeq).  Until now we aren't setting the
ReqSeq value on I-frame control bits. That way we can save sending
S-frames(Supervise frames) only to acknowledge receipt of I-frames. It
is very helpful to the full-duplex channel.
ReqSeq is the packet sequence number sent in an acknowledgement frame to
acknowledge receipt of frames up to (ReqSeq - 1).
BufferSeq controls the receiver buffer, it is used to delay
acknowledgement of new frames to not cause buffer overflow. BufferSeq
value is not increased until frames are pulled by reassembly function.

Signed-off-by: Gustavo F. Padovan <gustavo@las.ic.unicamp.br>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Fix unset of SrejActioned flag
Gustavo F. Padovan [Sat, 3 Oct 2009 05:34:37 +0000 (02:34 -0300)]
Bluetooth: Fix unset of SrejActioned flag

SrejActioned  is a flag that when set prevents local side to retransmit a
I-frame(the data frame) already retransmitted. The local entity can
retransmit again only when it receives a SREJ frame with the F-bit set.
SREJ frame - Selective Reject frame  - is sent when an entity wants the
retransmission of a specific I-frame that was lost or corrupted.
This bug can put ERTM in an unknown state once the entity can't
retransmit.
A frame with the Final bit set is expected when the local side sends a
frame with the Poll bit set due to a local busy condition or a
retransmission timer expired. (Receipt of P-bit shall always be replied by
a frame with the F-bit set).
pi->conn_state keeps informations about many ERTM flags including
SrejActioned.

Signed-off-by: Gustavo F. Padovan <gustavo@las.ic.unicamp.br>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Initialize variables and timers for both channel's sides
Gustavo F. Padovan [Sat, 3 Oct 2009 05:34:36 +0000 (02:34 -0300)]
Bluetooth: Initialize variables and timers for both channel's sides

Fix ERTM's full-duplex channel to work as specified by ERTM spec. ERTM
needs to handle state vars, timers and counters to send and receive
I-frames(the data frames), i.e., for both sides of data communication.
We initialize all of them to the default values here.
Full-duplex channel is a mandatory feature of ERTM spec.

Signed-off-by: Gustavo F. Padovan <gustavo@las.ic.unicamp.br>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Fix handling of BNEP setup connection requests
Vikram Kandukuri [Thu, 3 Dec 2009 09:42:51 +0000 (15:12 +0530)]
Bluetooth: Fix handling of BNEP setup connection requests

According to BNEP test specification the proper response should be sent
for a setup connection request message after the BNEP connection setup
has been completed.

Signed-off-by: Vikram Kandukuri <vikram.kandukuri@atheros.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Unobfuscate tasklet_schedule usage
Marcel Holtmann [Wed, 18 Nov 2009 00:02:54 +0000 (01:02 +0100)]
Bluetooth: Unobfuscate tasklet_schedule usage

The tasklet schedule function helpers are just an obfuscation. So remove
them and call the schedule functions directly.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Turn hci_recv_frame into an exported function
Marcel Holtmann [Tue, 17 Nov 2009 23:40:39 +0000 (00:40 +0100)]
Bluetooth: Turn hci_recv_frame into an exported function

For future simplification it is important that the hci_recv_frame
function is no longer an inline function. So move it into the module
itself and export it.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Return ENETDOWN when interface is down
Marcel Holtmann [Wed, 18 Nov 2009 00:05:00 +0000 (01:05 +0100)]
Bluetooth: Return ENETDOWN when interface is down

Sending commands to a down interface results in a timeout while clearly
it should just return ENETDOWN. When using the ioctls this works fine,
but not when using the HCI sockets sendmsg interface.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Implement raw output support for HIDP layer
Jiri Kosina [Thu, 26 Nov 2009 15:20:56 +0000 (16:20 +0100)]
Bluetooth: Implement raw output support for HIDP layer

Implement raw output callback which is used by hidraw to send raw data to
the underlying device.

Without this patch, the userspace hidraw-based applications can't send
output reports to HID Bluetooth devices.

Reported-and-tested-by: Brian Gunn <bgunn@solekai.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Fix miscdev ops owner for virtual driver
Marcel Holtmann [Thu, 3 Dec 2009 17:07:28 +0000 (18:07 +0100)]
Bluetooth: Fix miscdev ops owner for virtual driver

The /dev/vhci ops don't refer to the module and so it is possible to
unload the module while the file descriptor is in use. This was an
accidental removal after the cleanup.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Remove unused global minor variable
Marcel Holtmann [Thu, 3 Dec 2009 17:05:16 +0000 (18:05 +0100)]
Bluetooth: Remove unused global minor variable

After the removal of the module parameter for setting the minor number,
this variable became unused. So just remove it.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Remove stub ioctl in hci_vhci
Thomas Gleixner [Thu, 15 Oct 2009 20:28:53 +0000 (20:28 +0000)]
Bluetooth: Remove stub ioctl in hci_vhci

Remove the empty ioctl which just returns -EINVAL. vfs_ioctl() will
return -ENOTTY instead, but I doubt that any application will notice
the difference :)

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agoBluetooth: Removal of unused variable in btmrvl driver
Rahul Tank [Tue, 29 Sep 2009 02:43:35 +0000 (19:43 -0700)]
Bluetooth: Removal of unused variable in btmrvl driver

This patch removes unused variable "drvdbg" from btmrvl_debugfs_data
structure.

Signed-off-by: Rahul Tank <rahult@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
14 years agosunhme: fix dma warning in Sun HME
Mikulas Patocka [Thu, 3 Dec 2009 06:26:38 +0000 (22:26 -0800)]
sunhme: fix dma warning in Sun HME

Fix dma-api-checking warnings in Sun HME

Sun HME driver is mapping the first fragment with dma_map_single and subsequent
fragments with dma_map_page. It is unmapping all fragments with dma_unmap_single
and that produces the warning.

This patch changes it so that it unmaps only the first fragment with
dma_unmap_single and subsequent fragments are unmapped with dma_unmap_page.

WARNING: at lib/dma-debug.c:816 check_unmap+0x3ac/0x780()
hme 0000:01:01.1: DMA-API: device driver frees DMA memory with wrong function [device address=0x00000000c1082000] [size=32 bytes] [mapped as page] [unmapped as single]
Modules linked in: nbd sunhme openpromfs sermouse unix
Call Trace:
 [0000000000456910] warn_slowpath_common+0x50/0xa0
 [0000000000571f4c] check_unmap+0x3ac/0x780
 [0000000000572570] debug_dma_unmap_page+0x50/0x60
 [000000001002f5fc] happy_meal_tx+0x11c/0x260 [sunhme]
 [000000001002fc4c] happy_meal_interrupt+0xcc/0xe0 [sunhme]
 [0000000000492d94] handle_fasteoi_irq+0x74/0x100
 [000000000042ac0c] handler_irq+0xcc/0x100
 [0000000000426a54] valid_addr_bitmap_patch+0x14/0x1c0
 [0000000000665de0] _spin_unlock_irqrestore+0x40/0x60
 [0000000000462bb8] mod_timer+0x118/0x1a0
 [00000000005ec254] sk_reset_timer+0x14/0x40
 [0000000000635e4c] tcp_event_new_data_sent+0x8c/0xc0
 [0000000000639374] __tcp_push_pending_frames+0x34/0xc0
---[ end trace 73d5c42c1e9f11c4 ]---
Mapped at:
 [<000000001002f148>] happy_meal_start_xmit+0x308/0x480 [sunhme]
 [<00000000005fc858>] dev_hard_start_xmit+0x318/0x3c0
 [<000000000060fec4>] sch_direct_xmit+0x1a4/0x200
 [<00000000005fced0>] dev_queue_xmit+0x410/0x560
 [<0000000000604a1c>] neigh_resolve_output+0xfc/0x300

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotcp: clear hints to avoid a stale one (nfs only affected?)
Ilpo Järvinen [Thu, 3 Dec 2009 06:24:02 +0000 (22:24 -0800)]
tcp: clear hints to avoid a stale one (nfs only affected?)

Eric Dumazet mentioned in a context of another problem:

"Well, it seems NFS reuses its socket, so maybe we miss some
cleaning as spotted in this old patch"

I've not check under which conditions that actually happens but
if true, we need to make sure we don't accidently leave stale
hints behind when the write queue had to be purged (whether reusing
with NFS can actually happen if purging took place is something I'm
not sure of).

...At least it compiles.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotcp: sysctl_tcp_cookie_size needs to be exported to modules.
David S. Miller [Thu, 3 Dec 2009 06:18:58 +0000 (22:18 -0800)]
tcp: sysctl_tcp_cookie_size needs to be exported to modules.

Otherwise:

ERROR: "sysctl_tcp_cookie_size" [net/ipv6/ipv6.ko] undefined!
make[1]: *** [__modpost] Error 1

Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotcp: Fix warning on 64-bit.
David S. Miller [Thu, 3 Dec 2009 06:12:04 +0000 (22:12 -0800)]
tcp: Fix warning on 64-bit.

net/ipv4/tcp_output.c: In function â€˜tcp_make_synack’:
net/ipv4/tcp_output.c:2488: warning: cast from pointer to integer of different size

Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Teach vlans to cleanup as a pernet subsystem
Eric W. Biederman [Wed, 2 Dec 2009 13:19:08 +0000 (13:19 +0000)]
net: Teach vlans to cleanup as a pernet subsystem

Take advantage of the fact that an explicit rtnl_kill_links is
unnecessary (and skipping it improves batching), as network namespace
exit calls dellink on all remaining virtual devices, and
rtnl_link_unregister calls dellink on all outstanding devices in that
network namespace.  To do this we need to leave the vlan proc
directories in place until after network device exit time, which is
done by using register_pernet_subsys instead of
register_pernet_device.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoatl1e: Remove non-implementation of ethtool set_msglevel() operation
Ben Hutchings [Tue, 1 Dec 2009 19:10:18 +0000 (19:10 +0000)]
atl1e: Remove non-implementation of ethtool set_msglevel() operation

Unimplemented operations should not silently fail.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoatl1e: Remove redundant definitions of ethtool operations
Ben Hutchings [Tue, 1 Dec 2009 19:10:03 +0000 (19:10 +0000)]
atl1e: Remove redundant definitions of ethtool operations

These functions provide the default behaviour and do not need to be
set in struct ethtool_ops.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoatl1e: Allow TX checksum offload and TSO to be disabled and reenabled
Ben Hutchings [Tue, 1 Dec 2009 19:09:52 +0000 (19:09 +0000)]
atl1e: Allow TX checksum offload and TSO to be disabled and reenabled

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoTCPCT part 1g: Responder Cookie => Initiator
William Allen Simpson [Wed, 2 Dec 2009 18:25:27 +0000 (18:25 +0000)]
TCPCT part 1g: Responder Cookie => Initiator

Parse incoming TCP_COOKIE option(s).

Calculate <SYN,ACK> TCP_COOKIE option.

Send optional <SYN,ACK> data.

This is a significantly revised implementation of an earlier (year-old)
patch that no longer applies cleanly, with permission of the original
author (Adam Langley):

    http://thread.gmane.org/gmane.linux.network/102586

Requires:
   TCPCT part 1a: add request_values parameter for sending SYNACK
   TCPCT part 1b: generate Responder Cookie secret
   TCPCT part 1c: sysctl_tcp_cookie_size, socket option TCP_COOKIE_TRANSACTIONS
   TCPCT part 1d: define TCP cookie option, extend existing struct's
   TCPCT part 1e: implement socket option TCP_COOKIE_TRANSACTIONS
   TCPCT part 1f: Initiator Cookie => Responder

Signed-off-by: William.Allen.Simpson@gmail.com
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoTCPCT part 1f: Initiator Cookie => Responder
William Allen Simpson [Wed, 2 Dec 2009 18:23:05 +0000 (18:23 +0000)]
TCPCT part 1f: Initiator Cookie => Responder

Calculate and format <SYN> TCP_COOKIE option.

This is a significantly revised implementation of an earlier (year-old)
patch that no longer applies cleanly, with permission of the original
author (Adam Langley):

    http://thread.gmane.org/gmane.linux.network/102586

Requires:
   TCPCT part 1c: sysctl_tcp_cookie_size, socket option TCP_COOKIE_TRANSACTIONS
   TCPCT part 1d: define TCP cookie option, extend existing struct's

Signed-off-by: William.Allen.Simpson@gmail.com
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoTCPCT part 1e: implement socket option TCP_COOKIE_TRANSACTIONS
William Allen Simpson [Wed, 2 Dec 2009 18:19:30 +0000 (18:19 +0000)]
TCPCT part 1e: implement socket option TCP_COOKIE_TRANSACTIONS

Provide per socket control of the TCP cookie option and SYN/SYNACK data.

This is a straightforward re-implementation of an earlier (year-old)
patch that no longer applies cleanly, with permission of the original
author (Adam Langley):

    http://thread.gmane.org/gmane.linux.network/102586

The principle difference is using a TCP option to carry the cookie nonce,
instead of a user configured offset in the data.

Allocations have been rearranged to avoid requiring GFP_ATOMIC.

Requires:
   net: TCP_MSS_DEFAULT, TCP_MSS_DESIRED
   TCPCT part 1c: sysctl_tcp_cookie_size, socket option TCP_COOKIE_TRANSACTIONS
   TCPCT part 1d: define TCP cookie option, extend existing struct's

Signed-off-by: William.Allen.Simpson@gmail.com
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoTCPCT part 1d: define TCP cookie option, extend existing struct's
William Allen Simpson [Wed, 2 Dec 2009 18:17:05 +0000 (18:17 +0000)]
TCPCT part 1d: define TCP cookie option, extend existing struct's

Data structures are carefully composed to require minimal additions.
For example, the struct tcp_options_received cookie_plus variable fits
between existing 16-bit and 8-bit variables, requiring no additional
space (taking alignment into consideration).  There are no additions to
tcp_request_sock, and only 1 pointer in tcp_sock.

This is a significantly revised implementation of an earlier (year-old)
patch that no longer applies cleanly, with permission of the original
author (Adam Langley):

    http://thread.gmane.org/gmane.linux.network/102586

The principle difference is using a TCP option to carry the cookie nonce,
instead of a user configured offset in the data.  This is more flexible and
less subject to user configuration error.  Such a cookie option has been
suggested for many years, and is also useful without SYN data, allowing
several related concepts to use the same extension option.

    "Re: SYN floods (was: does history repeat itself?)", September 9, 1996.
    http://www.merit.net/mail.archives/nanog/1996-09/msg00235.html

    "Re: what a new TCP header might look like", May 12, 1998.
    ftp://ftp.isi.edu/end2end/end2end-interest-1998.mail

These functions will also be used in subsequent patches that implement
additional features.

Requires:
   TCPCT part 1a: add request_values parameter for sending SYNACK
   TCPCT part 1b: generate Responder Cookie secret
   TCPCT part 1c: sysctl_tcp_cookie_size, socket option TCP_COOKIE_TRANSACTIONS

Signed-off-by: William.Allen.Simpson@gmail.com
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoTCPCT part 1c: sysctl_tcp_cookie_size, socket option TCP_COOKIE_TRANSACTIONS
William Allen Simpson [Wed, 2 Dec 2009 18:14:19 +0000 (18:14 +0000)]
TCPCT part 1c: sysctl_tcp_cookie_size, socket option TCP_COOKIE_TRANSACTIONS

Define sysctl (tcp_cookie_size) to turn on and off the cookie option
default globally, instead of a compiled configuration option.

Define per socket option (TCP_COOKIE_TRANSACTIONS) for setting constant
data values, retrieving variable cookie values, and other facilities.

Move inline tcp_clear_options() unchanged from net/tcp.h to linux/tcp.h,
near its corresponding struct tcp_options_received (prior to changes).

This is a straightforward re-implementation of an earlier (year-old)
patch that no longer applies cleanly, with permission of the original
author (Adam Langley):

    http://thread.gmane.org/gmane.linux.network/102586

These functions will also be used in subsequent patches that implement
additional features.

Requires:
   net: TCP_MSS_DEFAULT, TCP_MSS_DESIRED

Signed-off-by: William.Allen.Simpson@gmail.com
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoTCPCT part 1b: generate Responder Cookie secret
William Allen Simpson [Wed, 2 Dec 2009 18:12:09 +0000 (18:12 +0000)]
TCPCT part 1b: generate Responder Cookie secret

Define (missing) hash message size for SHA1.

Define hashing size constants specific to TCP cookies.

Add new function: tcp_cookie_generator().

Maintain global secret values for tcp_cookie_generator().

This is a significantly revised implementation of earlier (15-year-old)
Photuris [RFC-2522] code for the KA9Q cooperative multitasking platform.

Linux RCU technique appears to be well-suited to this application, though
neither of the circular queue items are freed.

These functions will also be used in subsequent patches that implement
additional features.

Signed-off-by: William.Allen.Simpson@gmail.com
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoTCPCT part 1a: add request_values parameter for sending SYNACK
William Allen Simpson [Wed, 2 Dec 2009 18:07:39 +0000 (18:07 +0000)]
TCPCT part 1a: add request_values parameter for sending SYNACK

Add optional function parameters associated with sending SYNACK.
These parameters are not needed after sending SYNACK, and are not
used for retransmission.  Avoids extending struct tcp_request_sock,
and avoids allocating kernel memory.

Also affects DCCP as it uses common struct request_sock_ops,
but this parameter is currently reserved for future use.

Signed-off-by: William.Allen.Simpson@gmail.com
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoMerge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
David S. Miller [Thu, 3 Dec 2009 06:00:34 +0000 (22:00 -0800)]
Merge branch 'master' of /linux/kernel/git/davem/net-2.6

14 years agocnic: Use dma_alloc_coherent().
Michael Chan [Wed, 2 Dec 2009 15:15:39 +0000 (15:15 +0000)]
cnic: Use dma_alloc_coherent().

Replace pci_alloc_consistent() with dma_alloc_coherent() so that
appropriate GFP flags can be used.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocnic: Fix rq_page_table DMA address.
Michael Chan [Wed, 2 Dec 2009 15:15:38 +0000 (15:15 +0000)]
cnic: Fix rq_page_table DMA address.

The old code mistakenly zeroed out the upper 32-bit of the DMA address.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocnic: Fix bogus iSCSI MAC address
Michael Chan [Wed, 2 Dec 2009 15:15:37 +0000 (15:15 +0000)]
cnic: Fix bogus iSCSI MAC address

Fix code to read the proper iSCSI MAC address for bnx2x devices.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocnic: Fix bnx2x ring shutdown.
Michael Chan [Wed, 2 Dec 2009 15:15:36 +0000 (15:15 +0000)]
cnic: Fix bnx2x ring shutdown.

Need to send a HALT command to the firmware to fully shutdown the bnx2x
rings.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocnic: Fix ring I/O address for bnx2x devices.
Michael Chan [Wed, 2 Dec 2009 15:15:35 +0000 (15:15 +0000)]
cnic: Fix ring I/O address for bnx2x devices.

Old code was initializing the ring producers using an incorrect I/O
address.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonetxen: fix failure cases for fw hang recovery
Amit Kumar Salecha [Wed, 2 Dec 2009 15:46:19 +0000 (15:46 +0000)]
netxen: fix failure cases for fw hang recovery

Handle few corner cases in firmware hang detection and recovery:

o Don't mark device state as READY, till handshake with
  firmware is done.
o During probe, if start_firmware fails, restore reference
  count.
o Don't increment refernce count, if start_firmware fails
  during firmware reset.
o Clear __NX_RESETTING bit, incase of fatal error or tempeature
  reaches critical limit so that pci remove() doesn't poll on
  this bit.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonetxen : fix debug tools access for NX2031
Narender Kumar [Wed, 2 Dec 2009 15:46:18 +0000 (15:46 +0000)]
netxen : fix debug tools access for NX2031

Driver maps only some part of the MMIO bar 0 in case of
NX2031. This is sufficient for tx and rx by the driver.
When debug tools need to access resources in
unmapped regions, driver needs to map these on the fly
in order to complete read/write request.

Signed-off-by: Narender Kumar <narender.kumar@qlogic.com>
Signed-off-by: Dhananjay Phadke <dhananjay.phadke@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoe1000e: correct ICH/PCH PHY operations function pointers
Bruce Allan [Wed, 2 Dec 2009 17:03:18 +0000 (17:03 +0000)]
e1000e: correct ICH/PCH PHY operations function pointers

Some function pointers for a few PHY operations (for 82578 and 82567) and
were set incorrectly causing functions to be executed that were accessing
incorrect PHY register offsets for that particular device.  This patch also
moves a few PHY-specific functions from ich8lan.c to the more appropriate
phy.c.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoe1000e: refactor PHY ID detection workaround
Bruce Allan [Wed, 2 Dec 2009 17:02:43 +0000 (17:02 +0000)]
e1000e: refactor PHY ID detection workaround

The workaround that detects the correct PHY ID when an initial read of the
PHY ID registers returns an invalid one should retry up to ten times with
a small delay between attempts using a single PHY address and then repeat
using the remaining possible PHY addresses.  Do this instead of trying each
possible PHY address repeating that up to 100 times.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoe1000e: check_polarity function pointers not set for 8257x
Bruce Allan [Wed, 2 Dec 2009 17:02:14 +0000 (17:02 +0000)]
e1000e: check_polarity function pointers not set for 8257x

The function pointers for 8257x devices are not set.  This is not really a
problem now because there is nothing in the driver that references the
pointers for this particular MAC-family (the appropriate functions are
called directly), but the pointers should be set in case they are used in
the future.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoskbuff: remove skb_dma_map/unmap
Alexander Duyck [Wed, 2 Dec 2009 16:49:02 +0000 (16:49 +0000)]
skbuff: remove skb_dma_map/unmap

The two functions skb_dma_map/unmap are unsafe to use as they cause
problems when packets are cloned and sent to multiple devices while a HW
IOMMU is enabled.  Due to this it is best to remove the code so it is not
used by any other network driver maintainters.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotg3: remove use of skb_dma_map/unmap
Alexander Duyck [Wed, 2 Dec 2009 16:48:38 +0000 (16:48 +0000)]
tg3: remove use of skb_dma_map/unmap

Due to the fact that skb_dma_map/unmap do not work correctly when a HW
IOMMU is enabled it has been recommended to go about removing the calls
from the network device drivers.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: Matt Carlson <mcarlson@broadcom.com>
CC: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: remove use of skb_dma_map/unmap
Alexander Duyck [Wed, 2 Dec 2009 16:48:18 +0000 (16:48 +0000)]
be2net: remove use of skb_dma_map/unmap

Due to the fact that skb_dma_map/unmap do not work correctly when a HW
IOMMU is enabled it has been recommended to go about removing the calls
from the network device drivers.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobnx2: remove skb_dma_map/unmap calls from driver
Alexander Duyck [Wed, 2 Dec 2009 16:47:57 +0000 (16:47 +0000)]
bnx2: remove skb_dma_map/unmap calls from driver

Due to the fact that skb_dma_map/unmap do not work correctly when a HW
IOMMU is enabled it has been recommended to go about removing the calls
from the network device drivers.

[ Fix bnx2_free_tx_skbs() ring indexing and use NETDEV_TX_OK return
  code in bnx2_start_xmit() after cleaning up DMA mapping errors. -Mchan ]

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoigbvf: remove skb_dma_map/unmap call from drivers
Alexander Duyck [Wed, 2 Dec 2009 16:47:37 +0000 (16:47 +0000)]
igbvf: remove skb_dma_map/unmap call from drivers

This patch removes the skb_dma_map/unmap calls from the igbvf driver due to
the fact that it does not play well with HW IOMMU when combined with
transmitting cloned skbs.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoigb: remove use of skb_dma_map from driver
Alexander Duyck [Wed, 2 Dec 2009 16:47:18 +0000 (16:47 +0000)]
igb: remove use of skb_dma_map from driver

This change removes skb_dma_map/unmap calls from the igb driver due to the
fact that the call is incompatible with iommu enabled kernels.  In order to
prevent warnings about using the wrong unmap call I have added a
mapped_as_page value to the buffer_info structure to track if the mapped
region is a page or a buffer.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoixgbe: remove skb_dma_map/unmap calls from driver
Alexander Duyck [Wed, 2 Dec 2009 16:46:56 +0000 (16:46 +0000)]
ixgbe: remove skb_dma_map/unmap calls from driver

This patch removes skb_dma_map/unmap calls from the ixgbe driver due to the
fact that the calls don't work with HW IOMMU enabled systems.  The problem
is that multiple mappings will give different results when HW IOMMU is
enabled and the skb_dma_map/unmap calls only have one location to store
mappings.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoixgb: remove use of skb_dma_map from ixgb
Alexander Duyck [Wed, 2 Dec 2009 16:46:31 +0000 (16:46 +0000)]
ixgb: remove use of skb_dma_map from ixgb

skb_dma_map is incompatible with HW iommu due to the fact that multiple
mappings can result in different results each time.  For this reason it is
best to just remove use of these function calls.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoe1000: remove use of skb_dma_map from e1000 driver
Alexander Duyck [Wed, 2 Dec 2009 16:46:00 +0000 (16:46 +0000)]
e1000: remove use of skb_dma_map from e1000 driver

Remove the use of skb_dma_map from the e1000 driver in order to avoid
issues when HW iommu are in use.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoe1000e: remove use of skb_dma_map from e1000e driver
Alexander Duyck [Wed, 2 Dec 2009 16:45:31 +0000 (16:45 +0000)]
e1000e: remove use of skb_dma_map from e1000e driver

In testing we have found that skb_dma_map/unmap is incompatible with HW
IOMMU due to the fact that multiple mappings will return different results.
In order to correct this we need to remove skb_dma_map/unmap calls from the
e1000e driver.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoip: update the description of rp_filter in ip-sysctl.txt
Shan Wei [Wed, 2 Dec 2009 23:39:04 +0000 (15:39 -0800)]
ip: update the description of rp_filter in ip-sysctl.txt

The commit 27fed4175acf81ddd91d9a4ee2fd298981f60295 (ip: fix logic of
reverse path filter sysctl) has changed the logic of rp_filter.  The
document about rp_filter is out of date. Now, setting
conf/all/rp_filte with 0 can also enable source validation.

Update the document according to the commit.

Signed-off-by: Shan Wei <shanwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobnx2: EEH is failing with timeout
Breno Leitao [Thu, 26 Nov 2009 07:31:49 +0000 (07:31 +0000)]
bnx2: EEH is failing with timeout

bnx2 is failing when a PCI error is detected. The error is the
following:

bnx2: Chip not in correct endian mode
bnx2: fw sync timeout, reset code = 404001d

This error was caused because the way pci_restore_state() is working
after commit 4b77b0a2ba27d64f58f16d8d4d48d8319dda36ff ("PCI: Clear
saved_state after the state has been restored").

Signed-off-by: Breno Leitao<leitao@linux.vnet.ibm.com>
Acked-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: Coding style corrections on Sealevel Systems 4021 driver.
Rudy Matela [Wed, 2 Dec 2009 09:26:01 +0000 (01:26 -0800)]
net: Coding style corrections on Sealevel Systems 4021 driver.

Changed braces position on some statements.
Inserted a space between an "if" keyword and a parenthesis.

Signed-off-by: Rudy Matela <rudy.matela@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>