pandora-kernel.git
13 years agosfc: Limit filter search depth further for performance hints (i.e. RFS)
Ben Hutchings [Tue, 4 Jan 2011 15:22:36 +0000 (15:22 +0000)]
sfc: Limit filter search depth further for performance hints (i.e. RFS)

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
13 years agonet: RPS: Make hardware-accelerated RFS conditional on NETIF_F_NTUPLE
Ben Hutchings [Tue, 15 Feb 2011 20:32:04 +0000 (20:32 +0000)]
net: RPS: Make hardware-accelerated RFS conditional on NETIF_F_NTUPLE

For testing and debugging purposes it is useful to be able to disable
hardware acceleration of RFS without disabling RFS altogether.  Since
this is a similar feature to 'n-tuple' flow steering through the
ethtool API, test the same feature flag that controls that.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
13 years agosfc: Add TX queues for high-priority traffic
Ben Hutchings [Mon, 10 Jan 2011 21:18:20 +0000 (21:18 +0000)]
sfc: Add TX queues for high-priority traffic

Implement the ndo_setup_tc() operation with 2 traffic classes.

Current Solarstorm controllers do not implement TX queue priority, but
they do allow queues to be 'paced' with an enforced delay between
packets.  Paced and unpaced queues are scheduled in round-robin within
two separate hardware bins (paced queues with a large delay may be
placed into a third bin temporarily, but we won't use that).  If there
are queues in both bins, the TX scheduler will alternate between them.

If we make high-priority queues unpaced and best-effort queues paced,
and high-priority queues are mostly empty, a single high-priority queue
can then instantly take 50% of the packet rate regardless of how many
of the best-effort queues have descriptors outstanding.

We do not actually want an enforced delay between packets on best-
effort queues, so we set the pace value to a reserved value that
actually results in a delay of 0.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
13 years agosfc: Distinguish queue lookup from test for queue existence
Ben Hutchings [Mon, 7 Feb 2011 23:04:38 +0000 (23:04 +0000)]
sfc: Distinguish queue lookup from test for queue existence

efx_channel_get_{rx,tx}_queue() currently return NULL if the channel
isn't used for traffic in that direction.  In most cases this is a
bug, but some callers rely on it as an existence test.

Add existence test functions efx_channel_has_{rx_queue,tx_queues}()
and use them as appropriate.

Change efx_channel_get_{rx,tx}_queue() to assert that the requested
queue exists.

Remove now-redundant initialisation from efx_set_channels().

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
13 years agosfc: Move TX queue core queue mapping into tx.c
Ben Hutchings [Wed, 12 Jan 2011 18:39:40 +0000 (18:39 +0000)]
sfc: Move TX queue core queue mapping into tx.c

efx_hard_start_xmit() needs to implement a mapping which is the
inverse of tx_queue::core_txq.  Move the initialisation of
tx_queue::core_txq next to efx_hard_start_xmit() to make the
connection more obvious.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
13 years agonet: Adjust TX queue kobjects if number of queues changes during unregister
Ben Hutchings [Tue, 15 Feb 2011 19:39:21 +0000 (19:39 +0000)]
net: Adjust TX queue kobjects if number of queues changes during unregister

If the root qdisc for a net device is mqprio, and the driver's
ndo_setup_tc() operation dynamically adds and remvoes TX queues,
netif_set_real_num_tx_queues() will be called during device
unregistration to remove the extra TX queues when the qdisc is
destroyed.  Currently this causes the corresponding kobjects
to be leaked, and the device's reference count never drops to 0.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
13 years agosch_mqprio: Always set num_tc to 0 in mqprio_destroy()
Ben Hutchings [Mon, 14 Feb 2011 19:02:23 +0000 (19:02 +0000)]
sch_mqprio: Always set num_tc to 0 in mqprio_destroy()

All the cleanup code in mqprio_destroy() is currently conditional on
priv->qdiscs being non-null, but that condition should only apply to
the per-queue qdisc cleanup.  We should always set the number of
traffic classes back to 0 here.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
13 years agoMerge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
David S. Miller [Wed, 9 Feb 2011 01:19:01 +0000 (17:19 -0800)]
Merge branch 'master' of /linux/kernel/git/davem/net-2.6

Conflicts:
drivers/net/e1000e/netdev.c

13 years agopch_can: fix module reload issue with MSI
Tomoya [Mon, 7 Feb 2011 23:29:03 +0000 (23:29 +0000)]
pch_can: fix module reload issue with MSI

Currently, in case reload pch_can,
pch_can not to be able to catch interrupt.

The cause is bus-master is not set in pch_can.
Thus, add enabling bus-master processing.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agopch_can: fix rmmod issue
Tomoya [Mon, 7 Feb 2011 23:29:02 +0000 (23:29 +0000)]
pch_can: fix rmmod issue

Currently, when rmmod pch_can, kernel failure occurs.
The cause is pci_iounmap executed before pch_can_reset.
Thus pci_iounmap moves after pch_can_reset.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agopch_can: fix 800k comms issue
Tomoya [Mon, 7 Feb 2011 23:29:01 +0000 (23:29 +0000)]
pch_can: fix 800k comms issue

Currently, 800k comms fails since prop_seg set zero.
(EG20T PCH CAN register of prop_seg must be set more than 1)
To prevent prop_seg set to zero, change tseg2_min 1 to 2.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: Kill NETEVENT_PMTU_UPDATE.
David S. Miller [Wed, 9 Feb 2011 00:17:55 +0000 (16:17 -0800)]
net: Kill NETEVENT_PMTU_UPDATE.

Nobody actually does anything in response to the event,
so just kill it off.

Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: Remove bogus barrier() in dst_allfrag().
David S. Miller [Tue, 8 Feb 2011 23:33:22 +0000 (15:33 -0800)]
net: Remove bogus barrier() in dst_allfrag().

I simply missed this one when modifying the other dst
metric interfaces earlier.

Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: Fix lockdep regression caused by initializing netdev queues too early.
David S. Miller [Tue, 8 Feb 2011 23:02:50 +0000 (15:02 -0800)]
net: Fix lockdep regression caused by initializing netdev queues too early.

In commit aa9421041128abb4d269ee1dc502ff65fb3b7d69 ("net: init ingress
queue") we moved the allocation and lock initialization of the queues
into alloc_netdev_mq() since register_netdevice() is way too late.

The problem is that dev->type is not setup until the setup()
callback is invoked by alloc_netdev_mq(), and the dev->type is
what determines the lockdep class to use for the locks in the
queues.

Fix this by doing the queue allocation after the setup() callback
runs.

This is safe because the setup() callback is not allowed to make any
state changes that need to be undone on error (memory allocations,
etc.).  It may, however, make state changes that are undone by
free_netdev() (such as netif_napi_add(), which is done by the
ipoib driver's setup routine).

The previous code also leaked a reference to the &init_net namespace
object on RX/TX queue allocation failures.

Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet/caif: Fix dangling list pointer in freed object on error.
David S. Miller [Tue, 8 Feb 2011 22:31:31 +0000 (14:31 -0800)]
net/caif: Fix dangling list pointer in freed object on error.

rtnl_link_ops->setup(), and the "setup" callback passed to alloc_netdev*(),
cannot make state changes which need to be undone on failure.  There is
no cleanup mechanism available at this point.

So we have to add the caif private instance to the global list once we
are sure that register_netdev() has succedded in ->newlink().

Otherwise, if register_netdev() fails, the caller will invoke free_netdev()
and we will have a reference to freed up memory on the chnl_net_list.

Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoipsec: allow to align IPv4 AH on 32 bits
Nicolas Dichtel [Wed, 2 Feb 2011 06:29:02 +0000 (06:29 +0000)]
ipsec: allow to align IPv4 AH on 32 bits

The Linux IPv4 AH stack aligns the AH header on a 64 bit boundary
(like in IPv6). This is not RFC compliant (see RFC4302, Section
3.3.3.2.1), it should be aligned on 32 bits.

For most of the authentication algorithms, the ICV size is 96 bits.
The AH header alignment on 32 or 64 bits gives the same results.

However for SHA-256-128 for instance, the wrong 64 bit alignment results
in adding useless padding in IPv4 AH, which is forbidden by the RFC.

To avoid breaking backward compatibility, we use a new flag
(XFRM_STATE_ALIGN4) do change original behavior.

Initial patch from Dang Hongwu <hongwu.dang@6wind.com> and
Christophe Gouault <christophe.gouault@6wind.com>.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoUSB CDC NCM errata updates for cdc_ncm host driver
Alexey Orishko [Mon, 7 Feb 2011 09:45:10 +0000 (09:45 +0000)]
USB CDC NCM errata updates for cdc_ncm host driver

Specification links:
- CDC NCM errata link:
  http://www.usb.org/developers/devclass_docs/NCM10_012011.zip
- CDC and WMC errata link:
  http://www.usb.org/developers/devclass_docs/CDC1.2_WMC1.1_012011.zip

Changes:
- driver updated to match cdc.h header with errata changes
- added support for USB_CDC_SET_NTB_INPUT_SIZE control request with
  8 byte length
- fixes to comply with specification: send only control requests supported by
  device, set number of datagrams for IN direction, connection speed structure
  update, etc.
- packet loss fixed for tx direction; misleading flag renamed.
- adjusted hard_mtu value.

Signed-off-by: Alexey Orishko <alexey.orishko@stericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoCDC NCM errata updates for cdc.h
Alexey Orishko [Mon, 7 Feb 2011 09:45:09 +0000 (09:45 +0000)]
CDC NCM errata updates for cdc.h

Changes are based on the following documents:
- CDC NCM errata:
http://www.usb.org/developers/devclass_docs/NCM10_012011.zip
- CDC and WMC errata link:
http://www.usb.org/developers/devclass_docs/CDC1.2_WMC1.1_012011.zip

Signed-off-by: Alexey Orishko <alexey.orishko@stericsson.com>
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
David S. Miller [Tue, 8 Feb 2011 21:52:31 +0000 (13:52 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next-2.6

13 years agoMerge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/jkirsher/net-2.6
David S. Miller [Tue, 8 Feb 2011 20:16:52 +0000 (12:16 -0800)]
Merge branch 'master' of /linux/kernel/git/jkirsher/net-2.6

13 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
David S. Miller [Tue, 8 Feb 2011 20:03:54 +0000 (12:03 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-2.6

13 years agoixgbe: update version string
Don Skidmore [Wed, 26 Jan 2011 06:04:37 +0000 (06:04 +0000)]
ixgbe: update version string

This will synchronize the version string with that of the latest source
forge driver which shares its functionality.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: cleanup variable initialization
Don Skidmore [Wed, 26 Jan 2011 06:04:17 +0000 (06:04 +0000)]
ixgbe: cleanup variable initialization

The ixgbe_fcoe_ddp_get function wasn't initializing one of its variables
and this was producing compiler warnings.  This patch cleans that up.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Stephen Ko <stephen.s.ko@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: limit VF access to network traffic
Alexander Duyck [Wed, 19 Jan 2011 18:33:05 +0000 (18:33 +0000)]
ixgbe: limit VF access to network traffic

This change fixes VM pool allocation issues based on MAC address filtering,
as well as limits the scope of VF access to promiscuous mode.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: fix for 82599 erratum on Header Splitting
Don Skidmore [Tue, 18 Jan 2011 22:53:47 +0000 (22:53 +0000)]
ixgbe: fix for 82599 erratum on Header Splitting

We have found a hardware erratum on 82599 hardware that can lead to
unpredictable behavior when Header Splitting mode is enabled.  So
we are no longer enabling this feature on affected hardware.

Please see the 82599 Specification Update for more information.

CC: stable@kernel.org
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Stephen Ko <stephen.s.ko@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: fix variable set but not used warnings by gcc 4.6
Emil Tantilov [Wed, 5 Jan 2011 07:09:41 +0000 (07:09 +0000)]
ixgbe: fix variable set but not used warnings by gcc 4.6

Caught with gcc 4.6 -Wunused-but-set-variable

Remove unused napi_vectors variable.

Fix the use of reset_bit in ixgbe_reset_hw_X540()

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Stephen Ko <stephen.s.ko@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoe1000: add support for Marvell Alaska M88E1118R PHY
Florian Fainelli [Mon, 24 Jan 2011 14:48:03 +0000 (14:48 +0000)]
e1000: add support for Marvell Alaska M88E1118R PHY

This patch adds support for Marvell Alask M88E188R PHY chips. Support for
other M88* PHYs is already there, so there is nothing more to add than its
PHY id.

CC: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Florian Fainelli <ffainelli@freebox.fr>
Acked-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoe1000e: tx_timeout should not increment for non-hang events
Jesse Brandeburg [Wed, 2 Feb 2011 10:19:55 +0000 (10:19 +0000)]
e1000e: tx_timeout should not increment for non-hang events

Currently the driver increments the tx_timeout counter (an error counter)
when simply resetting the part with outstanding transmit work pending.
This is an unnecessary count of an error, when all we should be doing is
just resetting the part and discarding the transmits.  With this change the
only increment of tx_timeout is when the stack calls the watchdog reset
function due to a true Tx timeout.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Reviewed-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoMerge branch 'batman-adv/merge' of git://git.open-mesh.org/ecsv/linux-merge
David S. Miller [Tue, 8 Feb 2011 03:54:14 +0000 (19:54 -0800)]
Merge branch 'batman-adv/merge' of git://git.open-mesh.org/ecsv/linux-merge

13 years agobatman-adv: Linearize fragment packets before merge
Sven Eckelmann [Sun, 6 Feb 2011 23:26:43 +0000 (23:26 +0000)]
batman-adv: Linearize fragment packets before merge

We access the data inside the skbs of two fragments directly using memmove
during the merge. The data of the skb could span over multiple skb pages. An
direct access without knowledge about the pages would lead to an invalid memory
access.

Signed-off-by: Sven Eckelmann <sven@narfation.org>
[lindner_marek@yahoo.de: Move return from function to the end]
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
13 years agox25: possible skb leak on bad facilities
andrew hendry [Mon, 7 Feb 2011 00:08:15 +0000 (00:08 +0000)]
x25: possible skb leak on bad facilities

Originally x25_parse_facilities returned
-1 for an error
 0 meaning 0 length facilities
>0 the length of the facilities parsed.

5ef41308f94dc ("x25: Prevent crashing when parsing bad X.25 facilities") introduced more
error checking in x25_parse_facilities however used 0 to indicate bad parsing
a6331d6f9a429 ("memory corruption in X.25 facilities parsing") followed this further for
DTE facilities, again using 0 for bad parsing.

The meaning of 0 got confused in the callers.
If the facilities are messed up we can't determine where the data starts.
So patch makes all parsing errors return -1 and ensures callers close and don't use the skb further.

Reported-by: Andy Whitcroft <apw@canonical.com>
Signed-off-by: Andrew Hendry <andrew.hendry@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agomac80211: remove unneeded check
Dan Carpenter [Mon, 7 Feb 2011 19:03:35 +0000 (22:03 +0300)]
mac80211: remove unneeded check

"ap" is the address of sdata->u.ap so it can never be NULL here.  Also
we dereferenced it on the previous line.  I removed the check.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoath9k: Update comments for not parsing DTIM period
Mohammed Shafi Shajakhan [Mon, 7 Feb 2011 15:38:28 +0000 (21:08 +0530)]
ath9k: Update comments for not parsing DTIM period

Add few comments for not parsing DTIM period from mac80211

Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomac80211: Update comments on radiotap MCS index
Mohammed Shafi Shajakhan [Mon, 7 Feb 2011 04:40:04 +0000 (10:10 +0530)]
mac80211: Update comments on radiotap MCS index

mac80211 now supports passing MCS index to radiotap, so update the
comments regarding this

Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomac80211: as a 4-addr station, do not receive packets for other stations
Felix Fietkau [Sat, 5 Feb 2011 22:48:37 +0000 (23:48 +0100)]
mac80211: as a 4-addr station, do not receive packets for other stations

Since 4-addr frames completely override the source address which will
make it into the converted 802.3 frames, receiving frames for other
4-addr stations will confuse the bridging code.

To be able to handle traffic for all connected devices, the bridge
code will automatically turn on promiscuous mode, which triggers
this problem.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Reported-by: Steve Brown <sbrown@cortland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomac80211: Make some mlme timers module paramaters.
Ben Greear [Fri, 4 Feb 2011 23:30:24 +0000 (15:30 -0800)]
mac80211: Make some mlme timers module paramaters.

This allows users to tune the connection-loss algorithms
to be more or less lenient.  In particular, larger
null-func retries helps when using lots of virtual
stations on a loaded network.

Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoath5k: Fix fast channel switching
Nick Kossifidis [Thu, 3 Feb 2011 23:41:02 +0000 (01:41 +0200)]
ath5k: Fix fast channel switching

Fast channel change fixes:

a) Always set OFDM timings
b) Don't re-activate PHY
c) Enable only NF calibration, not AGC

Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomac80211: fix the skb cloned check in the tx path
Felix Fietkau [Mon, 7 Feb 2011 11:05:00 +0000 (12:05 +0100)]
mac80211: fix the skb cloned check in the tx path

Using skb_header_cloned to check if it's safe to write to the skb is not
enough - mac80211 also touches the tailroom of the skb.
Initially this check was only used to increase a counter, however this
commit changed the code to also skip skb data reallocation if no extra
head/tailroom was needed:

commit 4cd06a344db752f513437138953af191cbe9a691
mac80211: skip unnecessary pskb_expand_head calls

It added a regression at least with iwl3945, which is fixed by this patch.

Reported-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Tested-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: Re-enable RF_KILL interrupt when down
Don Fry [Sun, 6 Feb 2011 17:29:45 +0000 (09:29 -0800)]
iwlagn: Re-enable RF_KILL interrupt when down

With commit 554d1d027b19265c4aa3f718b3126d2b86e09a08 only one RF_KILL
interrupt will be seen by the driver when the interface is down.

Re-enable the interrupt when it occurs to see all transitions.

Signed-off-by: Don Fry <donald.h.fry@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agossb-pcmcia: Fix parsing of invariants tuples
Michael Büsch [Fri, 4 Feb 2011 22:34:45 +0000 (23:34 +0100)]
ssb-pcmcia: Fix parsing of invariants tuples

This fixes parsing of the device invariants (MAC address)
for PCMCIA SSB devices.

ssb_pcmcia_do_get_invariants expects an iv pointer as data
argument.

Tested-by: dylan cristiani <d.cristiani@idem-tech.it>
Signed-off-by: Michael Buesch <mb@bu3sch.de>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoenic: Update MAINTAINERS
Vasanthy Kolluri [Fri, 4 Feb 2011 16:17:26 +0000 (16:17 +0000)]
enic: Update MAINTAINERS

Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoenic: Clean up: Remove support for an older version of hardware
Vasanthy Kolluri [Fri, 4 Feb 2011 16:17:21 +0000 (16:17 +0000)]
enic: Clean up: Remove support for an older version of hardware

Remove support for an older version (A1) of hardware

Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoenic: Bug Fix: Reorder firmware devcmds - CMD_INIT and CMD_IG_VLAN_REWRITE_MODE
Vasanthy Kolluri [Fri, 4 Feb 2011 16:17:16 +0000 (16:17 +0000)]
enic: Bug Fix: Reorder firmware devcmds - CMD_INIT and CMD_IG_VLAN_REWRITE_MODE

Firmware requires CMD_IG_VLAN_REWRITE_MODE be issued before a CMD_INIT.

Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoenic: Bug Fix: Fix return values of enic_add/del_station_addr routines
Vasanthy Kolluri [Fri, 4 Feb 2011 16:17:10 +0000 (16:17 +0000)]
enic: Bug Fix: Fix return values of enic_add/del_station_addr routines

Fix enic_add/del_station_addr routines to return appropriate error code when an invalid address is added or deleted.

Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoenic: Clean up: Organize devcmd wrapper routines
Vasanthy Kolluri [Fri, 4 Feb 2011 16:17:05 +0000 (16:17 +0000)]
enic: Clean up: Organize devcmd wrapper routines

Organize the wrapper routines for firmware devcmds into a separate file.

Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoenic: Decouple mac address registration and deregistration from port profile set...
Roopa Prabhu [Fri, 4 Feb 2011 12:57:16 +0000 (12:57 +0000)]
enic: Decouple mac address registration and deregistration from port profile set operation

This patch removes VM mac address registration and deregistration code during
port profile set operation. We can delay mac address registration until
enic_open.

Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobnx2x: Proper netdev->ndo_set_rx_mode() implementation.
Vladislav Zolotarov [Sun, 6 Feb 2011 19:25:41 +0000 (11:25 -0800)]
bnx2x: Proper netdev->ndo_set_rx_mode() implementation.

Completed the bnx2x_set_rx_mode() to a proper netdev->ndo_set_rx_mode
implementation:
 - Added a missing configuration of a unicast MAC addresses list.
 - Changed bp->dma_lock from being a mutex to a spinlock as long as it's taken
under netdev->addr_list_lock now.

Signed-off-by: Vladislav Zolotarov <vladz@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobnx2x: Duplication in promisc mode
Vladislav Zolotarov [Sun, 6 Feb 2011 19:21:49 +0000 (11:21 -0800)]
bnx2x: Duplication in promisc mode

Prevent packets duplication for frames targeting FCoE L2 ring:
packets were arriving to stack from both L2 RSS and from FCoE
L2 in a promiscuous mode.

Configure FCoE L2 ring to DROP_ALL rx mode, when interface is
configured to PROMISC, and to accept only unicast frames, when
interface is configured to ALL_MULTI.

Signed-off-by: Vladislav Zolotarov <vladz@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobnx2x: MTU for FCoE L2 ring
Vladislav Zolotarov [Sun, 6 Feb 2011 19:21:02 +0000 (11:21 -0800)]
bnx2x: MTU for FCoE L2 ring

Always configure an FCoE L2 ring with a mini-jumbo MTU size (2500).
To do that we had to move the rx_buf_size parameter from per
function level to a per ring level.

Signed-off-by: Vladislav Zolotarov <vladz@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoiwlwifi: fix compiling error with different configuration
Wey-Yi Guy [Fri, 4 Feb 2011 18:46:41 +0000 (10:46 -0800)]
iwlwifi: fix compiling error with different configuration

When .config has different configuration, it might fail to compile
iwlwifi. fix it

Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
13 years agoiwlwifi: remove unnecessary locking
Johannes Berg [Fri, 4 Feb 2011 14:57:45 +0000 (06:57 -0800)]
iwlwifi: remove unnecessary locking

This code, and the places that set the variable
is_internal_short_scan and the vif pointers are
all protected by the mutex, there's no point in
locking the spinlock here as well (any more).

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
13 years agotcp: Add reference to initial CWND ietf draft.
David S. Miller [Sun, 6 Feb 2011 02:13:45 +0000 (18:13 -0800)]
tcp: Add reference to initial CWND ietf draft.

Suggested by Alexander Zimmermann

Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agosis900: Fix mem leak in sis900_rx error path
Jesper Juhl [Sat, 5 Feb 2011 10:41:53 +0000 (10:41 +0000)]
sis900: Fix mem leak in sis900_rx error path

Fix memory leak in error path of sis900_rx(). If we don't do this we'll
leak the skb we dev_alloc_skb()'ed just a few lines above when the
variable goes out of scope.

Signed-off-by: Jesper Juhl <jj@chaosbits.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobenet: Avoid potential null deref in be_cmd_get_seeprom_data()
Ajit Khaparde [Fri, 4 Feb 2011 17:18:28 +0000 (17:18 +0000)]
benet: Avoid potential null deref in be_cmd_get_seeprom_data()

Found by: Jesper Juhl <jj@chaosbits.net>

Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoinetpeer: Move ICMP rate limiting state into inet_peer entries.
David S. Miller [Fri, 4 Feb 2011 23:55:25 +0000 (15:55 -0800)]
inetpeer: Move ICMP rate limiting state into inet_peer entries.

Like metrics, the ICMP rate limiting bits are cached state about
a destination.  So move it into the inet_peer entries.

If an inet_peer cannot be bound (the reason is memory allocation
failure or similar), the policy is to allow.

Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoipv4: Don't miss existing cached metrics in new routes.
David S. Miller [Fri, 4 Feb 2011 22:37:30 +0000 (14:37 -0800)]
ipv4: Don't miss existing cached metrics in new routes.

Always lookup to see if we have an existing inetpeer entry for
a route.  Let FLOWI_FLAG_PRECOW_METRICS merely influence the
"create" argument to rt_bind_peer().

Also, call rt_bind_peer() unconditionally since it is not
possible for rt->peer to be non-NULL at this point.

Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
David S. Miller [Fri, 4 Feb 2011 22:28:58 +0000 (14:28 -0800)]
Merge branch 'master' of /linux/kernel/git/davem/net-2.6

13 years agomac80211: Optimize scans on current operating channel.
Ben Greear [Fri, 4 Feb 2011 19:54:17 +0000 (11:54 -0800)]
mac80211: Optimize scans on current operating channel.

This should decrease un-necessary flushes, on/off channel work,
and channel changes in cases where the only scanned channel is
the current operating channel.

* Removes SCAN_OFF_CHANNEL flag, uses SDATA_STATE_OFFCHANNEL
  and is-scanning flags instead.

* Add helper method to determine if we are currently configured
  for the operating channel.

* Do no blindly go off/on channel in work.c  Instead, only call
  appropriate on/off code when we really need to change channels.
  Always enable offchannel-ps mode when starting work,
  and disable it when we are done.

* Consolidate ieee80211_offchannel_stop_station and
  ieee80211_offchannel_stop_beaconing, call it
  ieee80211_offchannel_stop_vifs instead.

* Accept non-beacon frames when scanning on operating channel.

* Scan state machine optimized to minimize on/off channel
  transitions.  Also, when going on-channel, go ahead and
  re-enable beaconing.  We're going to be there for 200ms,
  so seems like some useful beaconing could happen.
  Always enable offchannel-ps mode when starting software
  scan, and disable it when we are done.

* Grab local->mtx earlier in __ieee80211_scan_completed_finish
  so that we are protected when calling hw_config(), etc.

* Pass probe-responses up the stack if scanning on local
  channel, so that mlme can take a look.

Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoath9k: add additional checks for the baseband hang detection
Felix Fietkau [Fri, 4 Feb 2011 19:09:25 +0000 (20:09 +0100)]
ath9k: add additional checks for the baseband hang detection

Since even with the latest changes the false positive issue of the baseband
hang check is not fully solved yet, additional checks are needed.
If the baseband hang occurs, the rx_clear signal will be stuck to high, so
we can use the cycle counters to confirm it.
With this patch, a hardware reset is only triggered if the baseband hang
check returned true three times in a row, with a beacon interval between
each check and if the busy time was also 99% or more during the check
intervals.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomac80211: do not send duplicate data frames to the cooked monitor interface
Felix Fietkau [Fri, 4 Feb 2011 18:20:08 +0000 (19:20 +0100)]
mac80211: do not send duplicate data frames to the cooked monitor interface

I can't think of a valid use case for this aside from debugging (which can
also be done with a real monitor interface), and dropping these frames saves
some precious CPU cycles.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoath: Fix clearing of secondary key cache entry for TKIP
Jouni Malinen [Fri, 4 Feb 2011 11:51:28 +0000 (13:51 +0200)]
ath: Fix clearing of secondary key cache entry for TKIP

All register writes to the key cache have to be done in pairs. However,
the clearing of a separate MIC entry with hardware revisions that use
combined MIC key layout did not do that with one of the registers. Add
the matching register write to the following register to make the KEY4
register write actually complete.

This is mostly a fix for a theoretical issue since the incorrect entry
that could potentially be left behind in the key cache would not match
with received frames. Anyway, better make this code clean the entry
correctly using paired register writes.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomac80211: do not restart ps timer during scan or offchannel
Rajkumar Manoharan [Wed, 2 Feb 2011 17:27:53 +0000 (22:57 +0530)]
mac80211: do not restart ps timer during scan or offchannel

While leaving oper channel, STA informs sleep state to AP to
stop sending data. Till sending ack for the nullfunc, AP
continues to send the data to STA which restarts ps_timer that
is causing unnecessary nullfunc exchange on timer expiry
when the STA was already moved to offchannel. So don't restart ps_timer
on data reception during scan. This issue was identified by
the following warning.

WARNING: at net/mac80211/tx.c:661 invoke_tx_handlers+0xf07/0x1330 [mac80211]
wlan0: Dropped data frame as no usable bitrate found while scanning and
associated. Target station: 00:03:7f:0b:a6:1b on 5 GHz band
Call Trace:
  [<ffffffffa0413ba7>] invoke_tx_handlers+0xf07/0x1330 [mac80211]
  [<ffffffffa0414056>] ieee80211_tx+0x86/0x2c0 [mac80211]
  [<ffffffffa0414345>] ieee80211_xmit+0xb5/0x1d0 [mac80211]
  [<ffffffffa04037e0>] ieee80211_dynamic_ps_enable_work+0x0/0xb0 [mac80211]
  [<ffffffffa04158cf>] ieee80211_tx_skb+0x4f/0x60 [mac80211]
  [<ffffffffa04026e6>] ieee80211_send_nullfunc+0x46/0x60 [mac80211]
  [<ffffffffa0403885>] ieee80211_dynamic_ps_enable_work+0xa5/0xb0 [mac80211]

Reviewed-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: add useful debug output
Jussi Kivilinna [Mon, 31 Jan 2011 18:50:31 +0000 (20:50 +0200)]
zd1211rw: add useful debug output

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: enable NL80211_IFTYPE_AP
Jussi Kivilinna [Mon, 31 Jan 2011 18:50:21 +0000 (20:50 +0200)]
zd1211rw: enable NL80211_IFTYPE_AP

It should be safe to enable AP-mode now.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: reset rx urbs after idle period of 30 seconds
Jussi Kivilinna [Mon, 31 Jan 2011 18:50:12 +0000 (20:50 +0200)]
zd1211rw: reset rx urbs after idle period of 30 seconds

RX appears to freeze while idle. Resetting rx-urbs appears to be enough to fix
this. Do reset 30 seconds after last rx.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: reset device when CR_BCN_FIFO_SEMAPHORE freezes in beacon setup
Jussi Kivilinna [Mon, 31 Jan 2011 18:50:02 +0000 (20:50 +0200)]
zd1211rw: reset device when CR_BCN_FIFO_SEMAPHORE freezes in beacon setup

When driver fails to acquire device semaphore lock, device usually
freezes soon afterwards. So failing to acquire lock indicates us that
not everything is going right in device/fw. So reset device when
this happens.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: add TX watchdog and device resetting
Jussi Kivilinna [Mon, 31 Jan 2011 18:49:52 +0000 (20:49 +0200)]
zd1211rw: add TX watchdog and device resetting

When doing transfers at high speed for long time, tx queue can freeze. So add
tx watchdog. TX-watchdog checks for locked tx-urbs and reset hardware when
such is detected. Merely unlinking urb was not enough, device have to be
reseted. Hw settings are restored so that any open link will stay on after
reset.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: collect driver settings and add function to restore theim
Jussi Kivilinna [Mon, 31 Jan 2011 18:49:43 +0000 (20:49 +0200)]
zd1211rw: collect driver settings and add function to restore theim

We need HW hard reset later in patchset to reset device after TX-stall.
Collect all settings that we have set to driver for later reset and
add restore function.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: lower hw command timeouts
Jussi Kivilinna [Mon, 31 Jan 2011 18:49:33 +0000 (20:49 +0200)]
zd1211rw: lower hw command timeouts

Device command timeouts are set up very high (1 sec) and this causes
AP beacon to lock up for long for example. Checking timeouts on device
it's easy to see that 1 sec timeout is not needed, when device fails
to response longer timeout doesn't help:

[  473.074419] zd1211rw 1-1:1.0: print_times() Read times:
[  473.175163] zd1211rw 1-1:1.0: print_time()     0 - 10 msec: 1506
[  473.176429] zd1211rw 1-1:1.0: print_time()    11 - 50 msec: 0
[  473.177955] zd1211rw 1-1:1.0: print_time()   51 - 100 msec: 0
[  473.180703] zd1211rw 1-1:1.0: print_time()  101 - 250 msec: 0
[  473.182101] zd1211rw 1-1:1.0: print_time()  251 - 500 msec: 0
[  473.183221] zd1211rw 1-1:1.0: print_time() 500 - 1000 msec: 20
[  473.184381] zd1211rw 1-1:1.0: print_time() 1000 - ... msec: 18

Also vendor driver doesn't use this long timeout. Therefore change
timeout to 50msec.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: change interrupt URB buffer to DMA buffer
Jussi Kivilinna [Mon, 31 Jan 2011 18:49:24 +0000 (20:49 +0200)]
zd1211rw: change interrupt URB buffer to DMA buffer

As might lower beacon update CPU usage.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: use stack and preallocated memory for small cmd-buffers
Jussi Kivilinna [Mon, 31 Jan 2011 18:49:14 +0000 (20:49 +0200)]
zd1211rw: use stack and preallocated memory for small cmd-buffers

Use stack for allocing small < 64 byte arrays in zd_chip.c and preallocated
buffer in zd_usb.c. This might lower CPU usage for beacon setup.

v2:
 - Do not use stack buffers in zd_usb.c as they would be used for urb
   transfer_buffer.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: batch beacon config commands together
Jussi Kivilinna [Mon, 31 Jan 2011 18:49:05 +0000 (20:49 +0200)]
zd1211rw: batch beacon config commands together

Beacon config function writes beacon to hw one write per byte. This is very
slow (usually taking more than 100ms to finish) and causes high CPU usage
when in AP-mode (kworker at ~50% on Intel Atom N270). By batching commands
together zd_mac_config_beacon() runtime can be lowered to 1/5th and lower
CPU usage to saner levels (<10% on Atom).

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: add beacon watchdog and setting HW beacon more failsafe
Jussi Kivilinna [Mon, 31 Jan 2011 18:48:55 +0000 (20:48 +0200)]
zd1211rw: add beacon watchdog and setting HW beacon more failsafe

When doing tx/rx at high packet rate (for example simply using ping -f),
device starts to fail to respond to control messages. On non-AP modes
this only causes problems for LED updating code but when we are running
in AP-mode we are writing new beacon to HW usually every 100ms. Now if
control message fails in HW beacon setup, device lock is kept locked
and beacon data partially written. This can and usually does cause:

 1. HW beacon setup fail now on, as driver cannot acquire device lock.
 2. Beacon-done interrupt stop working as device has incomplete beacon.

Therefore make zd_mac_config_beacon() always try to release device lock
and add beacon watchdog to restart beaconing when stall is detected.

Also fix zd_mac_config_beacon() try acquiring device lock for max 500ms,
as what old code appeared to be trying to do using loop and msleep(1).

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomac80211: fix race between next beacon dtim and ieee80211_get_buffered_bc
Christian Lamparter [Mon, 31 Jan 2011 18:48:44 +0000 (20:48 +0200)]
mac80211: fix race between next beacon dtim and ieee80211_get_buffered_bc

On review of 'zd1211rw: implement beacon fetching and handling
ieee80211_get_buffered_bc()', Christian Lamparter noted that [1]:

   Since zd_beacon_done also uploads the next beacon so long in advance,
   there could be an equally long race between the outdated state of the
   next beacon's DTIM broadcast traffic indicator (802.11-2007 7.3.2.6)
   which -in your case- was uploaded almost a beacon interval ago and
   the xmit of ieee80211_get_buffered_bc *now*.

   The dtim bc/mc bit might be not set, when a mc/bc arrived after the
   beacon was uploaded, but before the "beacon done event" from the
   hardware. So, dozing stations don't expect the broadcast traffic
   and of course, they might miss it completely.

   It's probably better to fix this in mac80211 (see the attached hack).

[1] http://marc.info/?l=linux-wireless&m=129435041117256&w=2

CC: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: implement beacon fetching and handling ieee80211_get_buffered_bc()
Jussi Kivilinna [Mon, 31 Jan 2011 18:48:35 +0000 (20:48 +0200)]
zd1211rw: implement beacon fetching and handling ieee80211_get_buffered_bc()

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: let zd_set_beacon_interval() set dtim_period and add AP-beacon flag
Jussi Kivilinna [Mon, 31 Jan 2011 18:48:25 +0000 (20:48 +0200)]
zd1211rw: let zd_set_beacon_interval() set dtim_period and add AP-beacon flag

Add support for AP-mode beacon. Also disable beacon when interface is set
down as otherwise hw will keep flooding NEXT_BCN interrupts.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: fix ack_pending in filter_ack causing tx-packet ordering problem on monitor
Jussi Kivilinna [Mon, 31 Jan 2011 18:48:16 +0000 (20:48 +0200)]
zd1211rw: fix ack_pending in filter_ack causing tx-packet ordering problem on monitor

For reasons not very clear yet to me, filter_ack leaves matching tx-packet
pending with 'ack_pending'. This causes tx-packet to be passed back to upper
layer after next packet has been transfered and tx-packets might end up
coming come out of monitor interface in wrong order vs. rx.

Because of this when enable AP-mode, hostapd monitor interface would get
packets in wrong order causing problems in WPA association.

So don't use mac->ack_pending when in AP-mode.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: support setting BSSID for AP mode
Jussi Kivilinna [Mon, 31 Jan 2011 18:48:06 +0000 (20:48 +0200)]
zd1211rw: support setting BSSID for AP mode

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: move set_rts_cts_work to bss_info_changed
Jussi Kivilinna [Mon, 31 Jan 2011 18:47:56 +0000 (20:47 +0200)]
zd1211rw: move set_rts_cts_work to bss_info_changed

As bss_info_changed may sleep, we can as well set RTS_CTS register right away.
Keep mac->short_preamble for later use (hw reset).

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: move set_multicast_hash and set_rx_filter from workers to configure_filter
Jussi Kivilinna [Mon, 31 Jan 2011 18:47:46 +0000 (20:47 +0200)]
zd1211rw: move set_multicast_hash and set_rx_filter from workers to configure_filter

Workers not needed anymore since configure_filter may sleep. Keep
mac->multicast_hash for later use (hw reset).

Signed-off-by: Jussi Kivilinna <jussi.kivilina@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: fix beacon interval setup
Jussi Kivilinna [Mon, 31 Jan 2011 18:47:36 +0000 (20:47 +0200)]
zd1211rw: fix beacon interval setup

Vendor driver uses CR_BNC_INTERVAL at various places, one is HW_EnableBeacon()
that combinies beacon interval with BSS-type flag and DTIM value in upper 16bit
of u32. The other one is HW_UpdateBcnInterval() that set_aw_pt_bi()
appears to be based on. HW_UpdateBcnInterval() takes interval argument as u16
and uses that for calculations, set_aw_pt_bi() uses u32 value that has flags
and dtim in upper part. This clearly seems wrong. Also HW_UpdateBcnInterval()
updates only lower 16bit part of CR_BNC_INTERVAL. So make set_aw_pt_bi() do
calculations on only lower u16 part of s->beacon_interval.

Also set 32bit beacon interval register before reading values from device,
as HW_EnableBeacon() on vendor driver does. This is required to make beacon
work on AP-mode, simply reading and then writing updated values is not enough
at least with zd1211b.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: add locking for mac->process_intr
Jussi Kivilinna [Mon, 31 Jan 2011 18:47:27 +0000 (20:47 +0200)]
zd1211rw: add locking for mac->process_intr

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: cancel process_intr work on zd_chip_disable_int()
Jussi Kivilinna [Mon, 31 Jan 2011 18:47:17 +0000 (20:47 +0200)]
zd1211rw: cancel process_intr work on zd_chip_disable_int()

OOPS if worker is running and disconnect() is called (triggered
by unpluging device). Much harder to trigger at this stage but
later when we have AP beacon work in process_intr it happens very
easy.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agozd1211rw: use urb anchors for tx and fix tx-queue disabling
Jussi Kivilinna [Mon, 31 Jan 2011 18:47:08 +0000 (20:47 +0200)]
zd1211rw: use urb anchors for tx and fix tx-queue disabling

When stress testing AP-mode I hit OOPS when unpluging or rmmodding
driver.

It appears that when tx-queue is disabled, tx-urbs might be left pending.
These can cause ehci to call non-existing tx_urb_complete() (after rmmod)
or uninitialized/reseted private structure (after disconnect()). Add skb
queue for submitted packets and unlink pending urbs on zd_usb_disable_tx().

Part of the problem seems to be usb->free_urb_list that isn't always
working as it should, causing machine freeze when trying to free the list
in zd_usb_disable_tx(). Caching free urbs isn't what other drivers seem
to be doing (usbnet for example) so strip free_usb_list.

Patch makes tx-urb handling saner with use of urb anchors.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoath9k: Fix possible double free of PAPRD skb's
Mohammed Shafi Shajakhan [Fri, 4 Feb 2011 13:08:23 +0000 (18:38 +0530)]
ath9k: Fix possible double free of PAPRD skb's

This patch reverts the following commit
ath9k: remove bfs_paprd_timestamp from struct ath_buf_state

Under high interference/noisy environment conditions where PAPRD frames
fails heavily introduces a possibility of double freeing skb's and causes
kernel panic after some time.This patch reverts back to the original approach
of using paprd_timestamp before freeing the PAPRD frame skb's

[  194.193705] Pid: 0, comm: swapper Tainted: G      D WC
2.6.35-22-generic #33-Ubuntu
[  194.193712] Call Trace:
[  194.193722]  [<c05c6468>] ? printk+0x2d/0x35
[  194.193732]  [<c05c63c3>] panic+0x5a/0xd2
[  194.193741]  [<c05ca3ed>] oops_end+0xcd/0xd0
[  194.193750]  [<c0105f74>] die+0x54/0x80
[  194.193758]  [<c05c9a16>] do_trap+0x96/0xc0
[  194.193837]  [<c0103fb0>] ? do_invalid_op+0x0/0xa0
[  194.193846]  [<c010403b>] do_invalid_op+0x8b/0xa0
[  194.193856]  [<c020bd4c>] ? kfree+0xec/0xf0
[  194.193866]  [<c012ce18>] ? default_spin_lock_flags+0x8/0x10
[  194.193877]  [<c01de47a>] ? free_one_page+0x12a/0x2d0
[  194.193888]  [<c01e04dc>] ? __free_pages+0x1c/0x40
[  194.193897]  [<c05c97a7>] error_code+0x73/0x78
[  194.193906]  [<c020bd4c>] ? kfree+0xec/0xf0
[  194.193915]  [<c04ecdd0>] ? skb_release_data+0x70/0xa0
[  194.193924]  [<c04ecdd0>] skb_release_data+0x70/0xa0
[  194.193933]  [<c04ec997>] __kfree_skb+0x17/0x90
[  194.193941]  [<c04eca31>] consume_skb+0x21/0x40
[  194.193964]  [<f85e0b70>] ieee80211_tx_status+0x760/0x860 [mac80211]
[  194.193979]  [<f85caddf>] ath_tx_complete_buf+0x1bf/0x2c0 [ath9k]
[  194.193988]  [<c05c8b9f>] ? _raw_spin_lock_irqsave+0x2f/0x50
[  194.193997]  [<c04ec40e>] ? skb_queue_tail+0x3e/0x50
[  194.194010]  [<f85cc803>] ath_tx_complete_aggr+0x823/0x940 [ath9k]
[  194.194021]  [<c0108a28>] ? sched_clock+0x8/0x10
[  194.194030]  [<c016bf14>] ? sched_clock_local+0xa4/0x180
[  194.194040]  [<c0139f57>] ? enqueue_sleeper+0x1e7/0x2b0
[  194.194051]  [<c013a194>] ? enqueue_entity+0x174/0x200
[  194.194064]  [<f85ce83d>] ath_tx_edma_tasklet+0x2bd/0x3b0 [ath9k]
[  194.194074]  [<c05c8b9f>] ? _raw_spin_lock_irqsave+0x2f/0x50
[  194.194088]  [<f85c7b9f>] ath9k_tasklet+0x9f/0x190 [ath9k]
[  194.194097]  [<c01505d7>] tasklet_action+0xa7/0xb0
[  194.194107]  [<c015127c>] __do_softirq+0x9c/0x1b0
[  194.194117]  [<c01a7f64>] ? irq_to_desc+0x14/0x20
[  194.194126]  [<c0124fc4>] ? ack_apic_level+0x64/0x1f0
[  194.194136]  [<c01513d5>] do_softirq+0x45/0x50
[  194.194145]  [<c0151545>] irq_exit+0x65/0x70
[  194.194153]  [<c05cf665>] do_IRQ+0x55/0xc0
[  194.194162]  [<c016a6c7>] ? hrtimer_start+0x27/0x30
[  194.194171]  [<c0103630>] common_interrupt+0x30/0x38
[  194.194181]  [<c012c21a>] ? native_safe_halt+0xa/0x10
[  194.194268]  [<c010a2f9>] default_idle+0x49/0xb0
[  194.194277]  [<c0101fcc>] cpu_idle+0x8c/0xd0
[  194.194286]  [<c05b2431>] rest_init+0x71/0x80
[  194.194295]  [<c081981a>] start_kernel+0x36e/0x374
[  194.194305]  [<c08199dd>] ? pass_all_bootoptions+0x0/0xa
[  194.194314]  [<c08190d7>] i386_start_kernel+0xd7/0xdf
[  194.194364] panic occurred, switching back to text console

Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agocarl9170: fix typo in PS code
Christian Lamparter [Thu, 3 Feb 2011 21:22:55 +0000 (22:22 +0100)]
carl9170: fix typo in PS code

This patch fixes a off-by-one bug which bugged
the driver's PS-POLL capability.

Cc: <stable@kernel.org>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agonet: can: janz-ican3: world-writable sysfs termination file
Vasiliy Kulikov [Fri, 4 Feb 2011 02:23:53 +0000 (02:23 +0000)]
net: can: janz-ican3: world-writable sysfs termination file

Don't allow everybody to set terminator via sysfs.

Signed-off-by: Vasiliy Kulikov <segoon@openwall.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: can: at91_can: world-writable sysfs files
Vasiliy Kulikov [Fri, 4 Feb 2011 02:23:50 +0000 (02:23 +0000)]
net: can: at91_can: world-writable sysfs files

Don't allow everybody to write to mb0_id file.

Signed-off-by: Vasiliy Kulikov <segoon@openwall.com>
Acked-by: Kurt Van Dijck <kurt.van.dijck@eia.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMAINTAINERS: update email ids of the be2net driver maintainers.
Ajit Khaparde [Fri, 4 Feb 2011 21:03:35 +0000 (13:03 -0800)]
MAINTAINERS: update email ids of the be2net driver maintainers.

Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobridge: Don't put partly initialized fdb into hash
Pavel Emelyanov [Fri, 4 Feb 2011 21:02:36 +0000 (13:02 -0800)]
bridge: Don't put partly initialized fdb into hash

The fdb_create() puts a new fdb into hash with only addr set. This is
not good, since there are callers, that search the hash w/o the lock
and access all the other its fields.

Applies to current netdev tree.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agor8169: prevent RxFIFO induced loops in the irq handler.
Francois Romieu [Thu, 3 Feb 2011 16:27:52 +0000 (17:27 +0100)]
r8169: prevent RxFIFO induced loops in the irq handler.

While the RxFIFO interruption is masked for most 8168, nothing prevents
it to appear in the irq status word. This is no excuse to crash.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Ivan Vecera <ivecera@redhat.com>
Cc: Hayes <hayeswang@realtek.com>
13 years agor8169: RxFIFO overflow oddities with 8168 chipsets.
Francois Romieu [Thu, 3 Feb 2011 11:02:36 +0000 (12:02 +0100)]
r8169: RxFIFO overflow oddities with 8168 chipsets.

Some experiment-based action to prevent my 8168 chipsets locking-up hard
in the irq handler under load (pktgen ~1Mpps). Apparently a reset is not
always mandatory (is it at all ?).

- RTL_GIGA_MAC_VER_12
- RTL_GIGA_MAC_VER_25
  Missed ~55% packets. Note:
  - this is an old SiS 965L motherboard
  - the 8168 chipset emits (lots of) control frames towards the sender

- RTL_GIGA_MAC_VER_26
  The chipset does not go into a frenzy of mac control pause when it
  crashes yet but it can still be crashed. It needs more work.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Ivan Vecera <ivecera@redhat.com>
Cc: Hayes <hayeswang@realtek.com>
13 years agor8169: use RxFIFO overflow workaround for 8168c chipset.
Ivan Vecera [Thu, 27 Jan 2011 11:24:11 +0000 (12:24 +0100)]
r8169: use RxFIFO overflow workaround for 8168c chipset.

I found that one of the 8168c chipsets (concretely XID 1c4000c0) starts
generating RxFIFO overflow errors. The result is an infinite loop in
interrupt handler as the RxFIFOOver is handled only for ...MAC_VER_11.
With the workaround everything goes fine.

Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Acked-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Hayes <hayeswang@realtek.com>
13 years agobe2net: use device model DMA API
Ivan Vecera [Wed, 2 Feb 2011 08:05:12 +0000 (08:05 +0000)]
be2net: use device model DMA API

Use DMA API as PCI equivalents will be deprecated.

Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Acked-by: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoinclude/net/genetlink.h: Allow genlmsg_cancel to accept a NULL argument
Julia Lawall [Fri, 28 Jan 2011 05:43:40 +0000 (05:43 +0000)]
include/net/genetlink.h: Allow genlmsg_cancel to accept a NULL argument

nlmsg_cancel can accept NULL as its second argument, so for similarity,
this patch extends genlmsg_cancel to be able to accept a NULL second
argument as well.

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: Provide compat support for SIOCGETMIFCNT_IN6 and SIOCGETSGCNT_IN6.
David S. Miller [Fri, 4 Feb 2011 01:59:32 +0000 (17:59 -0800)]
net: Provide compat support for SIOCGETMIFCNT_IN6 and SIOCGETSGCNT_IN6.

Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: Support compat SIOCGETVIFCNT ioctl in ipv4.
David S. Miller [Fri, 4 Feb 2011 01:24:28 +0000 (17:24 -0800)]
net: Support compat SIOCGETVIFCNT ioctl in ipv4.

Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: Fix bug in compat SIOCGETSGCNT handling.
David S. Miller [Fri, 4 Feb 2011 01:21:31 +0000 (17:21 -0800)]
net: Fix bug in compat SIOCGETSGCNT handling.

Commit 709b46e8d90badda1898caea50483c12af178e96 ("net: Add compat
ioctl support for the ipv4 multicast ioctl SIOCGETSGCNT") added the
correct plumbing to handle SIOCGETSGCNT properly.

However, whilst definiting a proper "struct compat_sioc_sg_req" it
isn't actually used in ipmr_compat_ioctl().

Correct this oversight.

Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoniu: Fix races between up/down and get_stats.
David S. Miller [Fri, 4 Feb 2011 00:12:50 +0000 (16:12 -0800)]
niu: Fix races between up/down and get_stats.

As reported by Flavio Leitner, there is no synchronization to protect
NIU's get_stats method from seeing a NULL pointer in either
np->rx_rings or np->tx_rings.  In fact, as far as ->ndo_get_stats
is concerned, these values are set completely asynchronously.

Flavio attempted to fix this using a RW semaphore, which in fact
works most of the time.  However, dev_get_stats() can be invoked
from non-sleepable contexts in some cases, so this fix doesn't
work in all cases.

So instead, control the visibility of the np->{rx,tx}_ring pointers
when the device is being brough up, and use properties of the device
down sequence to our advantage.

In niu_get_stats(), return immediately if netif_running() is false.
The device shutdown sequence first marks the device as not running (by
clearing the __LINK_STATE_START bit), then it performans a
synchronize_rcu() (in dev_deactive_many()), and then finally it
invokes the driver ->ndo_stop() method.

This guarentees that all invocations of niu_get_stats() either see
netif_running() as false, or they see the channel pointers before
->ndo_stop() clears them out.

If netif_running() is true, protect against startup races by loading
the np->{rx,tx}_rings pointer into a local variable, and punting if
it is NULL.  Use ACCESS_ONCE to prevent the compiler from reloading
the pointer on us.

Also, during open, control the order in which the pointers and the
ring counts become visible globally using SMP write memory barriers.
We make sure the np->num_{rx,tx}_rings value is stable and visible
before np->{rx,tx}_rings is.

Such visibility control is not necessary on the niu_free_channels()
side because of the RCU sequencing that happens during device down as
described above.  We are always guarenteed that all niu_get_stats
calls are finished, or will see netif_running() false, by the time
->ndo_stop is invoked.

Reported-by: Flavio Leitner <fleitner@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agomac80211: Add testing functionality for TKIP
Jouni Malinen [Thu, 3 Feb 2011 16:35:19 +0000 (18:35 +0200)]
mac80211: Add testing functionality for TKIP

TKIP countermeasures depend on devices being able to detect Michael
MIC failures on received frames and for stations to report errors to
the AP. In order to test that behavior, it is useful to be able to
send out TKIP frames with incorrect Michael MIC. This testing behavior
has minimal effect on the TX path, so it can be added to mac80211 for
convenient use.

The interface for using this functionality is a file in mac80211
netdev debugfs (tkip_mic_test). Writing a MAC address to the file
makes mac80211 generate a dummy data frame that will be sent out using
invalid Michael MIC value. In AP mode, the address needs to be for one
of the associated stations or ff:ff:ff:ff:ff:ff to use a broadcast
frame. In station mode, the address can be anything, e.g., the current
BSSID. It should be noted that this functionality works correctly only
when associated and using TKIP.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>