firefly-linux-kernel-4.4.55.git
15 years agoiwlagn: improve rate scale table search
Mohamed Abbas [Fri, 8 May 2009 20:44:39 +0000 (13:44 -0700)]
iwlagn: improve rate scale table search

iwlagn rate scaling will periodically search other rate scale
tables to switch to the best table regarding performance. In the past
the number of search tables were 3. Every time the rate scale algorithm
goes through these available tables in will stay in current table for
some time before start searching again. Recent driver support more
feature and antenna, so we have more tables to search. This patch make
sure we go through all available tables.

Signed-off-by: Mohamed Abbas <mohamed.abbas@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: clean up PS code
Johannes Berg [Fri, 8 May 2009 20:44:38 +0000 (13:44 -0700)]
iwlwifi: clean up PS code

This removes all the dead code that tries to adjust the power
saving level based on the system AC state (inacceptable policy
in the kernel) or based on overtemp conditions (unused).

Also, pass _all_ policy wrt. enabling PS to mac80211, since
we do not use the power_disabled internally I now use that to
mirror the mac80211 CONF_PS setting. When mac80211 turns off
CONF_PS we follow suit. This means that the user power level
(which can currently only be set from sysfs) is not touched
for mac80211 powersave changes.

This means no "association status" checks are necessary since
mac80211 will not allow power save to be enabled when not
associated.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Acked-by: Mohamed Abbas <mohamed.abbas@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: fix PS disable status race
Johannes Berg [Fri, 8 May 2009 20:44:37 +0000 (13:44 -0700)]
iwlwifi: fix PS disable status race

iwlwifi internally needs to keep track of whether PS
is enabled in the firmware or not. To do this, it keeps
a bit in the status flags, called STATUS_POWER_PMI.

The code to set this bit looks as follows:

static int iwl_set_power(struct iwl_priv *priv, void *cmd)
{
return iwl_send_cmd_pdu_async(priv, POWER_TABLE_CMD,
      sizeof(struct iwl_powertable_cmd),
      cmd, NULL);
}

int iwl_power_update_mode(...)
{
[...]
if (final_mode != IWL_POWER_MODE_CAM)
set_bit(STATUS_POWER_PMI, &priv->status);

iwl_update_power_cmd(priv, &cmd, final_mode);
cmd.keep_alive_beacons = 0;

if (final_mode == IWL_POWER_INDEX_5)
cmd.flags |= IWL_POWER_FAST_PD;

ret = iwl_set_power(priv, &cmd);

if (final_mode == IWL_POWER_MODE_CAM)
clear_bit(STATUS_POWER_PMI, &priv->status);
else
set_bit(STATUS_POWER_PMI, &priv->status);

if (priv->cfg->ops->lib->update_chain_flags && update_chains)
priv->cfg->ops->lib->update_chain_flags(priv);
[...]
}

Now, this bit really needs to track what the _firmware_
thinks, not what the driver thinks. Therefore, there is
a race condition here -- the driver sets the bit before
it knows that the async command sent to the card in the
iwl_set_power function has been processed. As a result,
the call to update_chain_flags() may think that the card
has been woken up (PMI bit cleared) while in reality it
hasn't processed the async POWER_TABLE_CMD yet.

This leads to bugs -- any commands the update_chain_flags
function sends can get stuck and subsequent commands also
fail.

The fix is almost trivial: since there's no reason to send
an async command here (in fact, there almost never should
be since many mac80211 callbacks can sleep) just make the
function wait for the card to process the command and then
return and clear the PMI bit.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Acked-by: Mohamed Abbas <mohamed.abbas@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: do proper hw restart
Johannes Berg [Fri, 8 May 2009 20:44:36 +0000 (13:44 -0700)]
iwlwifi: do proper hw restart

When the microcode fails for any reason, ask mac80211 to
recover instead of trying ourselves and failing at it.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: remove ieee80211_ht_bss_info
Johannes Berg [Fri, 8 May 2009 19:21:41 +0000 (21:21 +0200)]
mac80211: remove ieee80211_ht_bss_info

This struct is no longer used (and hasn't been for a while).

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: properly track HT operation_mode
Johannes Berg [Fri, 8 May 2009 19:21:06 +0000 (21:21 +0200)]
mac80211: properly track HT operation_mode

When we disassociate, we set the channel to non-HT which
obviously invalidates any ht_operation_mode setting. But
when we then associate with the next AP again, we might
still have the ht_operation_mode from the previous AP
cached and fail to configure the hardware with the new
(but unchanged) operation mode. This patch fixes it by
separately tracking whether our cache is valid.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: move HT operation mode BSS info
Johannes Berg [Fri, 8 May 2009 18:47:39 +0000 (20:47 +0200)]
mac80211: move HT operation mode BSS info

There really is no need to have a separate struct for a
single variable. The fact that it exists is due to the
code legacy, but we can remove that now. Very simple.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Fix chipset detection for rt73usb
Ivo van Doorn [Fri, 8 May 2009 16:30:43 +0000 (18:30 +0200)]
rt2x00: Fix chipset detection for rt73usb

The lower 4 bytes of the chipset revision must contain
a non-zero value. This bug was introduced by
"rt2x00: Simplify rt2x00_check_rev".

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: improve scan timing
Johannes Berg [Fri, 8 May 2009 12:12:21 +0000 (14:12 +0200)]
mac80211: improve scan timing

The call to ieee80211_hw_config() is supposed to apply changes
synchronously, so once it returns the parameters are applied to
the hardware. Thus, there really is no need to delay the probing
by the channel switch time again since the channel switch has
already happened once we get to this code.

Additionally, there is no need to wait for a NAV update (probe
delay) when the channel is passively scanned. Remove that extra
time too.

This cuts scanning time from over 7 seconds to under 4 on ar9170,
which is due to the number of channels scanned and ar9170's switch
time being advertised as 135ms (my test now indicates it is about
77ms with the current driver, but the difference might also be due
to using a different machine with different USB controllers).

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: MFP - Drop unprotected Action frames prior key setup
Jouni Malinen [Fri, 8 May 2009 09:36:03 +0000 (12:36 +0300)]
mac80211: MFP - Drop unprotected Action frames prior key setup

When management frame protection (IEEE 802.11w) is used, unprotected
Robust Action frames are not allowed prior to key configuration.
However, unprotected Deauthentication and Disassociation frames are
allowed at that point, but not after key configuration.

Make ieee80211_drop_unencrypted() handle the special cases for MFP by
separating the basic Data frame case from Management frame processing
and handle the Management frames only if MFP has been negotiated. In
addition, do not use sdata->drop_unencrypted for Management frames
since the decision on whether to accept the frame depends on the key
being configured.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: Drop unencrypted frames based on key setup
Jouni Malinen [Fri, 8 May 2009 09:34:10 +0000 (12:34 +0300)]
mac80211: Drop unencrypted frames based on key setup

When using nl80211, we do not have a mechanism to set
sdata->drop_unencrypted. Currently, this breaks code that is supposed
to drop unencrypted frames when protection is expected since
ieee80211_rx_h_decrypt() is optimized to not set rx->key when the
frame is not protected.

This patch modifies ieee80211_rx_h_decrypt() to set rx->key for all
frames and only skip decryption if the frame is not protected. This
allows ieee80211_drop_unencrypted() to correctly drop frames even if
drop_unencrypted is not set.

The changes here are not enough to handle all cases, though. Additional
patches will be needed to implement proper IEEE 802.1X PAE for station
mode (currently, this is only used for AP mode) and some additional
rules are needed for MFP to drop unprotected Robust Action frames prior
to having PTK and IGTK configured.

In theory, the unprotected frames could and should be dropped in
ieee80211_rx_h_decrypt(). However, due to the special case with EAPOL
frames that have to be allowed to be received unprotected even when
keys are set, it is simpler to only set rx->key and allow the
ieee80211_frame_allowed() function to handle the actual dropping of
data frames after 802.11->802.3 header conversion. In addition,
unprotected robust management frames are dropped before they are
processed.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: make iwl_set_rate static
Johannes Berg [Fri, 8 May 2009 17:31:42 +0000 (19:31 +0200)]
iwlwifi: make iwl_set_rate static

It's not needed outside iwl-core.c

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Acked-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agocfg80211: fix wext iw_freq parsing
Johannes Berg [Fri, 8 May 2009 07:42:33 +0000 (09:42 +0200)]
cfg80211: fix wext iw_freq parsing

The function to parse a struct iw_freq has a stupid bug,
it returns NULL when the channel cannot be found at all,
but NULL is supposed to mean "auto". Fix this by checking
the return value of ieee80211_get_channel() and returning
ERR_PTR(-EINVAL) if it returned NULL (channel not found).

This fixes an issue where you could say (in IBSS mode)
iwconfig wlan0 channel 21
and it would use channel 1 instead because that's the
first available channel with IBSS allowed (which is what
the "auto" setting uses).

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: set default QoS values according to spec
Johannes Berg [Thu, 7 May 2009 14:16:24 +0000 (16:16 +0200)]
mac80211: set default QoS values according to spec

We've never really cared about the default QoS (WMM) values, but
we really should if the AP doesn't send any. This patch makes
mac80211 use the default values according to 802.11-2007, and
additionally syncs the default values when we disassociate so
whatever the last AP said gets "unconfigured".

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: fix scan channel race
Johannes Berg [Thu, 7 May 2009 12:23:01 +0000 (14:23 +0200)]
mac80211: fix scan channel race

When a software scan starts, it first sets sw_scanning, but
leaves the scan_channel "unset" (it currently actually gets
initialised to a default). Now, when something else tries
to (re)configure the hardware in the window between these two
events (after sw_scanning = true, but before scan_channel is
set), the current code switches to the (unset!) scan_channel.
This causes trouble, especially when switching bands and
sending frames on the wrong channel.

To work around this, leave scan_channel initialised to NULL
and use it to determine whether or not a switch to a different
channel should occur (and also use the same condition to check
whether to adjust power for scan or not).

Additionally, avoid reconfiguring the hardware completely when
recalculating idle resulted in no changes, this was the problem
that originally led us to discover the race condition in the
first place, which was helpfully bisected by Pavel. This part
of the patch should not be necessary with the other fixes, but
not calling the ieee80211_hw_config function when we know it to
be unnecessary is certainly a correct thing to do.

Unfortunately, this patch cannot and does not fix the race
condition completely, but due to the way the scan code is
structured it makes the particular problem Pavel discovered
(race while changing channel at the same time as transmitting
frames) go away. To fix it completely, more work especially
with locking configuration is needed.

Bisected-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agowireless: WL12XX should depend on GENERIC_HARDIRQS
Geert Uytterhoeven [Wed, 6 May 2009 20:35:58 +0000 (22:35 +0200)]
wireless: WL12XX should depend on GENERIC_HARDIRQS

m68k allmodconfig:
| drivers/net/wireless/wl12xx/main.c: In function 'wl12xx_probe':
| drivers/net/wireless/wl12xx/main.c:1273: error: implicit declaration of function 'set_irq_type'
| make[1]: *** [drivers/net/wireless/wl12xx/main.o] Error 1

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Kalle Valo <kalle.valo@nokia.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agonl80211 : Add support for configuring MFP
Jouni Malinen [Wed, 6 May 2009 19:09:37 +0000 (22:09 +0300)]
nl80211 : Add support for configuring MFP

NL80211_CMD_ASSOCIATE request must be able to indicate whether
management frame protection (IEEE 802.11w) is being used. mac80211 was
able to use MFP in client mode only with WEXT, but the new
NL80211_ATTR_USE_MFP attribute will allow this to be done with
nl80211, too.

Since we are currently using nl80211 for MFP only with drivers that
use user space SME, only MFP disabled and required values are
used. However, the NL80211_ATTR_USE_MFP attribute is an enum that can
be extended with MFP optional in the future, if that is needed with
some drivers (e.g., if the RSN IE is generated by the driver).

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoRemove duplicate slow protocol define in bond_3ad.h
Richard Genoud [Sat, 9 May 2009 06:59:16 +0000 (06:59 +0000)]
Remove duplicate slow protocol define in bond_3ad.h

ETH_P_SLOW is already defined in include/linux/if_ether.h.
There's no need to define BOND_ETH_P_LACPDU in drivers/net/bonding/bond_3ad.h

Signed-off-by: Richard Genoud <richard.genoud@gmail.com>
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: check retval of dev_addr_init()
Jiri Pirko [Fri, 8 May 2009 13:30:17 +0000 (13:30 +0000)]
net: check retval of dev_addr_init()

Add missed checking of dev_addr_init return value in alloc_netdev_mq.

Signed-off-by: Jiri Pirko <jpirko@redhat.com>
 net/core/dev.c |   15 ++++++++++++---
 1 files changed, 12 insertions(+), 3 deletions(-)
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agodecnet: Use data ready call back, rather than hand coding it
Steven Whitehouse [Fri, 8 May 2009 06:30:25 +0000 (06:30 +0000)]
decnet: Use data ready call back, rather than hand coding it

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: trivial register access cleanup
Dhananjay Phadke [Fri, 8 May 2009 22:02:30 +0000 (22:02 +0000)]
netxen: trivial register access cleanup

Remove unnecessary length parameter since it's always 4 bytes.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: disable msi-x for NC512m adapter
Dhananjay Phadke [Fri, 8 May 2009 22:02:29 +0000 (22:02 +0000)]
netxen: disable msi-x for NC512m adapter

This board doesn't suppot msi-x well due to msi-x table
mapping (hardware) issue.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: fix irq tear down and msix leak.
Dhananjay Phadke [Fri, 8 May 2009 22:02:28 +0000 (22:02 +0000)]
netxen: fix irq tear down and msix leak.

o Fix the order of irq and hardware context teardown.
  Also synchronize the interrupt in dev close() before
  releasing tx buffers.

o Fix possible msi-x vector leak if available vectors are
  less than requested.

o Request multiple msix vectors only if hardware supports
  multiple rx queues.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: cache msi register offset
Dhananjay Phadke [Fri, 8 May 2009 22:02:27 +0000 (22:02 +0000)]
netxen: cache msi register offset

Store msi target status register offset in adapter struct.
This avoids contention on msi_tgt_status table from interrupt
hadlers of different pci function.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoNetwork Drop Monitor: Fix skb_kill_datagram
John Dykstra [Fri, 8 May 2009 21:57:01 +0000 (14:57 -0700)]
Network Drop Monitor: Fix skb_kill_datagram

Commit ead2ceb0ec9f85cff19c43b5cdb2f8a054484431 ("Network Drop Monitor:
Adding kfree_skb_clean for non-drops and modifying end-of-line points
for skbs") established new conventions for identifying dropped packets.

Align skb_kill_datagram() with these conventions so that packets that
get dropped just before the copy to userspace are properly tracked.

Signed-off-by: John Dykstra <john.dykstra1@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: do not return negative number of received packages
Stanislaw Gruszka [Tue, 5 May 2009 23:22:12 +0000 (23:22 +0000)]
bnx2x: do not return negative number of received packages

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Acked-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2x: free workqueue when driver fail to register
Stanislaw Gruszka [Tue, 5 May 2009 23:22:01 +0000 (23:22 +0000)]
bnx2x: free workqueue when driver fail to register

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Acked-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
David S. Miller [Fri, 8 May 2009 19:46:17 +0000 (12:46 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next-2.6

15 years agoMerge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
David S. Miller [Fri, 8 May 2009 09:48:30 +0000 (02:48 -0700)]
Merge branch 'master' of /linux/kernel/git/davem/net-2.6

Conflicts:
include/net/tcp.h

15 years agonet: remove stale reference to fastroute from Kconfig help text
Ashish Karkare [Thu, 7 May 2009 23:31:01 +0000 (16:31 -0700)]
net: remove stale reference to fastroute from Kconfig help text

Signed-off-by: Ashish Karkare <akarkare@marvell.com>
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosmsc911x: fix calculation of res_size for ioremap
Matthias Ludwig [Thu, 7 May 2009 22:00:12 +0000 (15:00 -0700)]
smsc911x: fix calculation of res_size for ioremap

fix size of remaped iomem, which is 1 byte to small
(e.g. mappes only 0xff bytes instead of 0x100)

Signed-off-by: Matthias Ludwig <mludwig@ultratronik.de>
Acked-by: Steve Glendinning <steve.glendinning@smsc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoe1000/e1000e/igb/ixgb: don't txhang after link down
Jesse Brandeburg [Thu, 7 May 2009 11:07:35 +0000 (11:07 +0000)]
e1000/e1000e/igb/ixgb: don't txhang after link down

after the recent changes to wired drivers to use only
netif_carrier_off the driver can have outstanding tx work to
complete that will never complete once link is down.  Since the
intel hardware will hold this tx work forever, the driver
notices a tx timeout condition internally and might try
to instigate printk and reset of the part with a
netif_stop_queue, which doesn't work because link is down.

Don't bother arming to tx hang detection when link is down.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: Typecase '1' for 64 bit shift
Nelson, Shannon [Thu, 7 May 2009 10:40:15 +0000 (10:40 +0000)]
ixgbe: Typecase '1' for 64 bit shift

Make sure we don't get any sign-extend issues when we shift a 1
into bit 31.

Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: fix failing to log fan failures
Don Skidmore [Thu, 7 May 2009 10:39:54 +0000 (10:39 +0000)]
ixgbe: fix failing to log fan failures

We weren't logging the 82598AT fan failure if it occurred before (ixgbe_open)
as we hadn't sent up to catch the interrupt that event caused.

This patch checks for this failure in:
ixgbe_probe - So we can log the failure asap. We check right after we
set up the adapter->flags, which is when we know that we have a fan.
ixgbe_up_complete - To catch failures that may have happened between probe
and when we set up the interrupt that would normally detect the fan failure.

To enable all of this we need to initialize the adapter flag with
IXGBE_FLAG_FAN_FAIL_CAPABLE when the NIC contained a fan.

Signed-off-by: Don Skidmore <donald.c.skidmore@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>
15 years agoixgbe: Enable L2 header split in 82599
Yi Zou [Thu, 7 May 2009 10:39:35 +0000 (10:39 +0000)]
ixgbe: Enable L2 header split in 82599

This enables L2 header split when packet split is enabled for 82599.

Signed-off-by: Yi Zou <yi.zou@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: set queue0 for srrctl configuration correctly for DCB
Alexander Duyck [Thu, 7 May 2009 10:39:16 +0000 (10:39 +0000)]
ixgbe: set queue0 for srrctl configuration correctly for DCB

The current configuration is not setting queue 0 correctly for DCB
configurations.  As a result unconfigured queues are being used to setup
the SRRCTL register rx buffer len sizes.

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>
15 years agoixgbe: always set header length in SRRCTL
Alexander Duyck [Thu, 7 May 2009 10:38:56 +0000 (10:38 +0000)]
ixgbe: always set header length in SRRCTL

As per the documentation for 82599 in order to support hardware RSC the
header size must be set.  This is only currently done for packet split
mode.  This patch sets the header buffer length for all modes.

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>
15 years agotcp: tcp_prequeue() can use keyed wakeups
Eric Dumazet [Thu, 7 May 2009 07:20:39 +0000 (07:20 +0000)]
tcp: tcp_prequeue() can use keyed wakeups

We can avoid waking up tasks not interested in receive notifications,
using wake_up_interruptible_poll() instead of wake_up_interruptible()

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agotcp: tcp_prequeue() cleanup
Eric Dumazet [Thu, 7 May 2009 07:08:38 +0000 (07:08 +0000)]
tcp: tcp_prequeue() cleanup

Small cleanup patch to reduce line lengths, before a change in
tcp_prequeue().

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: Add missing rculist.h include to netdevice.h
David S. Miller [Wed, 6 May 2009 23:52:51 +0000 (16:52 -0700)]
net: Add missing rculist.h include to netdevice.h

Otherwise list_for_each_entry_rcu() et al. aren't visible
and we get build failures in some configurations.

Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: Make inet_twsk_put similar to sock_put
Arnaldo Carvalho de Melo [Wed, 6 May 2009 23:50:52 +0000 (16:50 -0700)]
net: Make inet_twsk_put similar to sock_put

By separating the freeing code from the refcounting decrementing.
Probably reducing icache pressure when we still have reference counts to
go.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: update skb_recycle_check() for hardware timestamping changes
Lennert Buytenhek [Wed, 6 May 2009 23:49:18 +0000 (16:49 -0700)]
net: update skb_recycle_check() for hardware timestamping changes

Commit ac45f602ee3d1b6f326f68bc0c2591ceebf05ba4 ("net: infrastructure
for hardware time stamping") added two skb initialization actions to
__alloc_skb(), which need to be added to skb_recycle_check() as well.

Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: Patrick Ohly <patrick.ohly@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobnx2: Fix panic in bnx2_poll_work().
Michael Chan [Wed, 6 May 2009 23:46:47 +0000 (16:46 -0700)]
bnx2: Fix panic in bnx2_poll_work().

Add barrier() to bnx2_get_hw_{tx|rx}_cons() to fix this issue:

http://bugzilla.kernel.org/show_bug.cgi?id=12698

This issue was reported by multiple i386 users.  Without barrier(),
the compiled code looks like the following where %eax contains the
address of the tx_cons or rx_cons in the DMA status block.  The
status block contents can change between the cmpb and the movzwl
instruction.  The driver would crash if the value was not 0xff during
the cmpb instruction, but changed to 0xff during the movzwl
instruction.

6828: 80 38 ff              cmpb   $0xff,(%eax)
682b: 0f b7 10              movzwl (%eax),%edx

With the added barrier(), the compiled code now looks correct:

683d: 0f b7 10              movzwl (%eax),%edx
6840: 0f b6 c2              movzbl %dl,%eax
6843: 3d ff 00 00 00        cmp    $0xff,%eax

Thanks to Pascal de Bruijn <pmjdebruijn@pcode.nl> for reporting the
problem and Holger Noefer <hnoefer@pironet-ndh.com> for patiently
testing test patches for us.

Also updated version to 2.0.1.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet-sched: fix bfifo default limit
Patrick McHardy [Wed, 6 May 2009 23:45:07 +0000 (16:45 -0700)]
net-sched: fix bfifo default limit

When no limit is given, the bfifo uses a default of tx_queue_len * mtu.
Packets handled by qdiscs include the link layer header, so this should
be taken into account, similar to what other qdiscs do.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoigb: resolve panic on shutdown when SR-IOV is enabled
Alexander Duyck [Wed, 6 May 2009 23:43:48 +0000 (16:43 -0700)]
igb: resolve panic on shutdown when SR-IOV is enabled

The setup_rctl call was making a call into the ring structure after it had
been freed.  This was causing a panic on shutdown.  This call wasn't
necessary since it is possible to get the needed index from
adapter->vfs_allocated_count.

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>
15 years agoMerge branch 'linux-2.6.30.y' of git://git.kernel.org/pub/scm/linux/kernel/git/inaky...
David S. Miller [Wed, 6 May 2009 23:42:19 +0000 (16:42 -0700)]
Merge branch 'linux-2.6.30.y' of git://git./linux/kernel/git/inaky/wimax

15 years agoixgbe: Proposed PARCH PCIE legacy I/O port free intel 10Gb NIC driver
gouji-new [Wed, 6 May 2009 10:44:45 +0000 (10:44 +0000)]
ixgbe: Proposed PARCH PCIE legacy I/O port free intel 10Gb NIC driver

Traditionally Intel based NIC drivers request I/O port even though it
doesn't need that really.

Intel PCIE 10Gb driver (ixgbe) also requests I/O port but it doesn't
need it either.

This is a little inconvenient situation because sometimes we have to
handle those cards on the slots where any I/O space is not attached.

So we made pach which makes ixgbe driver legacy I/O port free.

Signed-off-by: Masayuki Gouji <gouji.masayuki@jp.fujitsu.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: Return PCI_ERS_RESULT_DISCONNECT when bus is disabled
Breno Leitao [Wed, 6 May 2009 10:44:26 +0000 (10:44 +0000)]
ixgbe: Return PCI_ERS_RESULT_DISCONNECT when bus is disabled

According to the "PCI Error Recovery" document, if after a recovery,
the bus is disabled, the error_detected function should return
PCI_ERS_RESULT_DISCONNECT. Actually ixgbe error_detected function is
always returning PCI_ERS_RESULT_NEED_RESET, even if the bus is in failure.
This patch just check if the bus is disabled and then returns
PCI_ERS_RESULT_DISCONNET.

Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoixgbe: remove ixgbe_napi_add/del_all references
Alexander Duyck [Wed, 6 May 2009 10:44:07 +0000 (10:44 +0000)]
ixgbe: remove ixgbe_napi_add/del_all references

There were still some references to napi_add/del_all left after the dynamic
vector allocation patch.  This patch removes those references since the
ixgbe_napi_add/del_all calls are no longer needed as the napi struct is
added when the vector is created, and deleted when the vector is freed.

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>
15 years agoixgbe: skb_record_rx_queue should record rx queue instead of vector
Alexander Duyck [Wed, 6 May 2009 10:43:47 +0000 (10:43 +0000)]
ixgbe: skb_record_rx_queue should record rx queue instead of vector

currently ixgbe_receive_skb is passing the vector index to
skb_record_rx_queue instead of the queue index.  This patch changes that so
that the ring index is passed instead.

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>
15 years agoixgbe: make q_vectors dynamic to reduce netdev size
Alexander Duyck [Wed, 6 May 2009 10:43:28 +0000 (10:43 +0000)]
ixgbe: make q_vectors dynamic to reduce netdev size

Currently the q_vectors are being allocated statically inside of the
adapter struct.  This increases the overall size of the adapter struct when
we can easily allocate the vectors dynamically.  This patch changes that
behavior so that the q_vectors are allocated dynamically and the napi
structures are automatically allocated inside of the q_vectors as needed.

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>
15 years agopci: update 82576 sr-iov quirk with latest device IDs
Alexander Duyck [Wed, 6 May 2009 10:25:42 +0000 (10:25 +0000)]
pci: update 82576 sr-iov quirk with latest device IDs

The current quirk doesn't include all 82576 device IDs.  This update
resolves that.

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>
15 years agoigb/igbvf: set rx csum always enabled in hw, disable via sw
Alexander Duyck [Wed, 6 May 2009 10:25:23 +0000 (10:25 +0000)]
igb/igbvf: set rx csum always enabled in hw, disable via sw

An issue was found in which rx checksum could not be enabled without
resetting the interface.  The issue was the hardware enable was not being
done via ethtool.  To resolve this issue and prevent conflicts with VF
configuration we will leave the feature always enabled in hardware, and
then in software we will choose to ignore the results via a sw flag.

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>
15 years agoigbvf: cleanup flags and allow for rx checksum to be disabled
Alexander Duyck [Wed, 6 May 2009 10:25:01 +0000 (10:25 +0000)]
igbvf: cleanup flags and allow for rx checksum to be disabled

This patch cleans up a number of unused or unneeded feature flags.  As a
result of these changes the user should now be able to enable or disable rx
checksumming via ethtool.

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>
15 years agomv643xx_eth: only unmask RX and TX_END interrupts for available queues
Saeed Bishara [Wed, 6 May 2009 03:02:01 +0000 (03:02 +0000)]
mv643xx_eth: only unmask RX and TX_END interrupts for available queues

It is not a good idea to blindly unmask the RX and TX_END interrupts
for all eight queues on all mv643xx_eth hardware, since some variations
of the hardware have less than eight transmit/receive queues, and the
RX/TX_END interrupts for the queues they don't have can be in use by
other interrupt sources.

Signed-off-by: Saeed Bishara <saeed@marvell.com>
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agomv643xx_eth: rework receive skb cache alignment
Lennert Buytenhek [Wed, 6 May 2009 03:01:22 +0000 (03:01 +0000)]
mv643xx_eth: rework receive skb cache alignment

On the platforms that mv643xx_eth is used on, the manual skb->data
alignment logic in mv643xx_eth can be simplified, as the only case we
need to handle is where NET_SKB_PAD is not a multiple of the cache
line size.  If this is the case, the extra padding we need can be
computed at compile time, while if NET_SKB_PAD _is_ a multiple of
the cache line size, the code can be optimised out entirely.

Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agomv643xx_eth: minor register definition cleanup
Lennert Buytenhek [Wed, 6 May 2009 03:01:17 +0000 (03:01 +0000)]
mv643xx_eth: minor register definition cleanup

Move the definitions for the SDMA and port serial configuration
register values to where all the other register definitions live,
and expand the shifts to 32 bit constants.

Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agomv643xx_eth: count handling a link event as one unit of napi work
Lennert Buytenhek [Wed, 6 May 2009 03:01:10 +0000 (03:01 +0000)]
mv643xx_eth: count handling a link event as one unit of napi work

Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agomv643xx_eth: fix variable shadowing sparse warning
Lennert Buytenhek [Wed, 6 May 2009 03:00:59 +0000 (03:00 +0000)]
mv643xx_eth: fix variable shadowing sparse warning

Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: fix bonding support
Dhananjay Phadke [Tue, 5 May 2009 19:05:08 +0000 (19:05 +0000)]
netxen: fix bonding support

o Pause traffic during mac addr change.
o Enable setting mac address for NX3031.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: fix mac list management
Dhananjay Phadke [Tue, 5 May 2009 19:05:07 +0000 (19:05 +0000)]
netxen: fix mac list management

o use standard linked list api for mac addr list management
  in NX3031.
o release mac addresses in firmware in dev close().

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonetxen: fix race in tx ring acccess
Dhananjay Phadke [Tue, 5 May 2009 19:05:06 +0000 (19:05 +0000)]
netxen: fix race in tx ring acccess

Fix the distance check between tx ring producer and consumer that
could lead to tx ring wrap around.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agowimax: oops: wimax_dev_add() is the only one that can initialize the state
Inaky Perez-Gonzalez [Sat, 2 May 2009 09:30:28 +0000 (02:30 -0700)]
wimax: oops: wimax_dev_add() is the only one that can initialize the state

When a new wimax_dev is created, it's state has to be __WIMAX_ST_NULL
until wimax_dev_add() is succesfully called. This allows calls into
the stack that happen before said time to be rejected.

Until now, the state was being set (by mistake) to UNINITIALIZED,
which was allowing calls such as wimax_report_rfkill_hw() to go
through even when a call to wimax_dev_add() had failed; that was
causing an oops when touching uninitialized data.

This situation is normal when the device starts reporting state before
the whole initialization has been completed. It just has to be dealt
with.

Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
15 years agowimax: fix oops if netlink fails to add attribute
Inaky Perez-Gonzalez [Tue, 31 Mar 2009 00:50:17 +0000 (17:50 -0700)]
wimax: fix oops if netlink fails to add attribute

When sending a message to user space using wimax_msg(), if nla_put()
fails, correctly interpret the return code from wimax_msg_alloc() as
an err ptr and return the error code instead of crashing (as it is
assuming than non-NULL means the pointer is ok).

Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
15 years agomac80211: Update SA Query transaction id length
Jouni Malinen [Wed, 6 May 2009 17:34:10 +0000 (20:34 +0300)]
mac80211: Update SA Query transaction id length

IEEE 802.11w/D8.0 changed the length of the SA Query transaction
identifier from 16 to 2 octets.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: remove redundant AR9285 checks
Gabor Juhos [Wed, 6 May 2009 07:47:30 +0000 (09:47 +0200)]
ath9k: remove redundant AR9285 checks

The AR_SREV_9285_1[12]_OR_LATER macros already contains the
AR_SREV_9285 check.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: make private driver rate tables const
Luis R. Rodriguez [Wed, 6 May 2009 06:20:00 +0000 (02:20 -0400)]
ath9k: make private driver rate tables const

On x86 this allows us to do the following small savings:
shave off 23 % off of the module's data, and
shave off 6  % off of the module's text.

We save 456 bytes, for those counting.

$ size ath9k.ko
   text    data     bss     dec     hex filename
 250794    3628    1600  256022   3e816 ath9k.ko
$ size ath9k-old.ko
   text    data     bss     dec     hex filename
 239114   15308    1600  256022   3e816 ath9k-old.ko

$ du -b ath9k.ko
4034244 ath9k.ko
$ du -b ath9k-old.ko
4033788 ath9k-old.ko

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54: call p54_wake_free_queues on every p54_free_skb and p54_rx_frame_sent
Max Filippov [Tue, 5 May 2009 21:47:02 +0000 (01:47 +0400)]
p54: call p54_wake_free_queues on every p54_free_skb and p54_rx_frame_sent

Currently queues are stopped when their length reaches their length limit,
but are restarted only when the size of freed range of packet buffer is
not less than the size of the largest possible packet.

This causes permanent queue stop on radio visibility loss in the middle
of ping series: there is plenty of room in the packet buffer, but it is
never freed more than 3 (size of 'best effort' queue) * 288 (ping packet
plus headers) bytes at once.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: Comment the order of HT RX reorder handler vs. RX handlers
Jouni Malinen [Tue, 5 May 2009 17:35:15 +0000 (20:35 +0300)]
mac80211: Comment the order of HT RX reorder handler vs. RX handlers

We are currently processing block ack reordering as a separate task
before all other RX handlers. In theory, this is wrong since this step
should be done only after duplicate removal (see Figure 6-1 in IEEE
802.11n). However, moving this needs some work and the current
situation is not too bad. Add a comment here so that this small detail
does not get forgotten and who knows, maybe someone has some extra
time to take a look at cleaning this up.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: Add a timeout for frames in the RX reorder buffer
Jouni Malinen [Tue, 5 May 2009 17:35:14 +0000 (20:35 +0300)]
mac80211: Add a timeout for frames in the RX reorder buffer

This patch allows skbs to be released from the RX reorder buffer in
case they have been there for an unexpectedly long time without us
having received the missing frames before them. Previously, these
frames were only released when the reorder window moved and that could
take very long time unless new frames were received constantly (e.g.,
TCP connections could be killed more or less indefinitely).

This situation should not happen very frequently, but it looks like
there are some scenarious that trigger it for some reason. As such,
this should be considered mostly a workaround to speed up recovery
from unexpected siutation that could result in connections hanging for
long periods of time.

The changes here will only check for timeout situation when adding new
RX frames to the reorder buffer. It does not handle all possible
cases, but seems to help for most cases that could result from common
network usage (e.g., TCP retrying at least couple of times). For more
completely coverage, a timer could be used to periodically check
whether there are any frames remaining in the reorder buffer if no new
frames are received.

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: Use a shared function to release frames from RX reorder buf
Jouni Malinen [Tue, 5 May 2009 17:35:13 +0000 (20:35 +0300)]
mac80211: Use a shared function to release frames from RX reorder buf

No need to duplicate the same code in two places (and that would be
three after the followup patch).

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Simplify rt2x00_check_rev
Ivo van Doorn [Tue, 5 May 2009 17:46:08 +0000 (19:46 +0200)]
rt2x00: Simplify rt2x00_check_rev

rt2x00_check_rev() was too specific for rt2500usb and rt73usb,
by adding the mask argument (instead of hardcoding it into
the function itself) we can use the function in rt2800usb as
well.

v2: Fix revision mask for rt2800usb

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: Fix sparse warning for ssid_len on ieee80211_sta_config_auth()
Luis R. Rodriguez [Tue, 5 May 2009 19:48:39 +0000 (15:48 -0400)]
mac80211: Fix sparse warning for ssid_len on ieee80211_sta_config_auth()

net/mac80211/mlme.c:2079:28: warning: symbol 'ssid_len' shadows an earlier one
net/mac80211/mlme.c:2022:12: originally declared here

ssid_len is already being declared and checked above so there is
no need for it again.

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54: correct merge damage from "p54: more SoftLED updates"
John W. Linville [Wed, 6 May 2009 18:16:15 +0000 (14:16 -0400)]
p54: correct merge damage from "p54: more SoftLED updates"

Ooops...

Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoar9170: fix build when !CONFIG_PM
Alexander Beregalov [Mon, 4 May 2009 16:46:25 +0000 (20:46 +0400)]
ar9170: fix build when !CONFIG_PM

Fix this build error when CONFIG_PM is not set:
drivers/net/wireless/ath/ar9170/usb.c: In function 'ar9170_usb_probe':
drivers/net/wireless/ath/ar9170/usb.c:692:
error: 'struct usb_device' has no member named 'reset_resume'

Signed-off-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: report operating frequency rather than current
Johannes Berg [Mon, 4 May 2009 15:52:10 +0000 (17:52 +0200)]
mac80211: report operating frequency rather than current

It's not very helpful to see, in iwconfig, the current frequency
the card is tuned to if that frequency is currently somewhere
across the board because we're scanning. Since we keep track of
the frequency the user wants, display that instead.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agocfg80211: make nl80211_send_mlme_timeout() static
Luis R. Rodriguez [Sat, 2 May 2009 04:34:48 +0000 (00:34 -0400)]
cfg80211: make nl80211_send_mlme_timeout() static

Fixes sparse complaint:

  CHECK   net/wireless/nl80211.c
net/wireless/nl80211.c:3694:6:
warning: symbol 'nl80211_send_mlme_timeout'
was not declared. Should it be static?

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54usb: Fixes compile error with CONFIG_PM=n
Hauke Mehrtens [Fri, 1 May 2009 11:12:36 +0000 (13:12 +0200)]
p54usb: Fixes compile error with CONFIG_PM=n

/drivers/net/wireless/p54/p54usb.c: In function 'p54u_probe':
/drivers/net/wireless/p54/p54usb.c:923: error: 'struct usb_device' has no member named 'reset_resume'

In the struct usb_device the reset_resume attribute is only available
when CONFIG_PM is defined.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoWireless: remove driver_data direct access of struct device
Greg Kroah-Hartman [Fri, 1 May 2009 06:02:47 +0000 (23:02 -0700)]
Wireless: remove driver_data direct access of struct device

In the near future, the driver core is going to not allow direct access
to the driver_data pointer in struct device.  Instead, the functions
dev_get_drvdata() and dev_set_drvdata() should be used.  These functions
have been around since the beginning, so are backwards compatible with
all older kernel versions.

Cc: John W. Linville <linville@tuxdriver.com>
Cc: linux-wireless@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agowl12xx: correct printk format warnings
Bob Copeland [Fri, 1 May 2009 12:20:07 +0000 (08:20 -0400)]
wl12xx: correct printk format warnings

Fixes warnings:
  drivers/net/wireless/wl12xx/main.c:87: warning: int format, different
    type arg (arg 2)
  drivers/net/wireless/wl12xx/main.c: In function `wl12xx_fetch_nvs':
  drivers/net/wireless/wl12xx/main.c:125: warning: int format, different
    type arg (arg 2)
  drivers/net/wireless/wl12xx/wl1251.c: In function 'wl1251_upload_firmware':
  drivers/net/wireless/wl12xx/wl1251.c:94: warning: int format, different
    type arg (arg 2)
  drivers/net/wireless/wl12xx/wl1251.c:141: warning: int format, different
    type arg (arg 2)

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Acked-by: Kalle Valo <kalle.valo@nokia.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: replace test_and_set_bit by set_bit in clear stations function
Tomas Winkler [Thu, 30 Apr 2009 20:56:29 +0000 (13:56 -0700)]
iwlwifi: replace test_and_set_bit by set_bit in clear stations function

This patch replaces test_and_set_bit by set_bit since the bit is not
tested anyway

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: "is_fat" bit in rate scale match RXON flag
Wey-Yi Guy [Thu, 30 Apr 2009 20:56:28 +0000 (13:56 -0700)]
iwlwifi: "is_fat" bit in rate scale match RXON flag

This patch change the "is_fat" checking in rate scale to use
iwl_is_fat_tx_allowed() to match the sta and RX_ON command setting.

Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Tested-by: Conrad Kostecki <ConiKost@gmx.de>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlagn: disable PS support for iwlagn
Reinette Chatre [Thu, 30 Apr 2009 20:56:27 +0000 (13:56 -0700)]
iwlagn: disable PS support for iwlagn

Some issues in PS prevent us from supporting it reliably.

When 4965 goes to sleep it stores some data in host DRAM, reads it back
when device wakes up. In 4965 there is a problem that the data is not
correct when ucode starts using it upon wakeup.

For all iwlagn devices there is a problem where command is sent when PS is
enabled. At the moment there is a locking problem with priv->lock not being
held and thus not requesting nic access correctly.

We disable PS until these issues have been resolved.

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: Mohamed Abbas <mohamed.abbas@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwl3945: fix lock dependency
Abhijeet Kolekar [Thu, 30 Apr 2009 20:56:26 +0000 (13:56 -0700)]
iwl3945: fix lock dependency

Patch seperates rx_used and rx_free into two
different atomic contexts. We can now avoid using GFP_ATOMIC
for skb allocation and use GFP_KERNEL.

Signed-off-by: Abhijeet Kolekar <abhijeet.kolekar@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoAdd LED support for AR5BXB6 IBM Thinkpad PCIe adapters
Paride Legovini [Thu, 30 Apr 2009 19:55:52 +0000 (15:55 -0400)]
Add LED support for AR5BXB6 IBM Thinkpad PCIe adapters

Add LED support on the IBM ThinkPad 11a/b/g Wireless LAN Mini Express
Adapter (AR5BXB6), found on the IBM/Lenovo Thinkpad X60/T60/Z60 series.

Signed-off-by: Paride Legovini <legovini@spiro.fisica.unipd.it>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: use ctl settings based on current regdomain
Bob Copeland [Thu, 30 Apr 2009 19:55:51 +0000 (15:55 -0400)]
ath5k: use ctl settings based on current regdomain

Update ath5k to use the ctl settings for tx power based on current
regulatory domain.

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Acked-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Add Spur filter support on newer chips
Nick Kossifidis [Thu, 30 Apr 2009 19:55:50 +0000 (15:55 -0400)]
ath5k: Add Spur filter support on newer chips

* Add spur filter support for RF5413 and later chips

Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Implement antenna control
Nick Kossifidis [Thu, 30 Apr 2009 19:55:49 +0000 (15:55 -0400)]
ath5k: Implement antenna control

* Add code to support the various antenna scenarios supported by hw

 * For now hardcode the default scenario (single or dual omnis with
 tx/rx diversity working and tx antenna handled by session -hw keeps
 track on which antenna it got ack from each ap/station and maps each
 ap/station to one of the antennas-).

Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Enable AP mode
Jiri Slaby [Thu, 30 Apr 2009 19:55:48 +0000 (15:55 -0400)]
ath5k: Enable AP mode

After some debuging we were hitting the following bugs so far...

 * Due to huge channel list hostapd couldn't get infos from the driver
 and couldn't set the channel. If we manualy set the channel after
 hostapd starts (by setting channel to 0 -auto), beacons are sent
 but they wont show up on scan because they are malformed (they have
 channel = 0 because hostapd doesn't update the channel info -this is
 probably a hostapd bug so i'm CCing Jouni) and they get dropped. Bob
 fixed this by only allowing standard channels to be registered so
 now hostapd works as expected.

 * Docs (and HAL source) say that we must write 0 on timer0 when
 operating on AP mode to start TSF increment but this seems to
 mess with DBA in many cases and beacon queue never gets started.
 We fixed that on the previous patch.

 We have some more things to deal with...

 * For some reason (hw bug or something else) after restarting hostapd
 a few times, beacon inteval seems to change from 100ms to a sec
 (we get one beacon per sec).

 * We need to set sleep timers on STA mode and enable power saving +
 support PCF.

 ...but i think it's time we enable AP support "officialy" so that
 we can get more feedback from users. I ran ath5k with the mentioned
 patches + hostapd 0.6.8 and AP mode worked fine (it had some less
 throughput on my tests than IBSS but it worked).

Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Beaconing fixes
Nick Kossifidis [Thu, 30 Apr 2009 19:55:47 +0000 (15:55 -0400)]
ath5k: Beaconing fixes

* Write next beacon timer even on AP mode since without this we get
 no beacons + ath9k does it too.  Docs say that we must write 0 on
 this register on AP mode to start TSF increment, we do both to be
 on the safe side.

 * Fix num_tx_pending function, we never read the register :P that's
 why we got all those "beacon queue 7 didn't stop messages".

 * Put full prioriy on beacon queue, lock all queues with lower
 priority using the arblock and also bypass any arblock by seting
 the arblock ignore flag.

 * For the CAB queue (do we need this thing ?, it seems crap) since
 it's supposed to fire up after each beacon (we don't use it on driver
 part, ath9k/MadWiFi does), don't make it DBA gated but instead make
 it fire after each beacon by using the beacon sent gated flag.

 * Increase bmiss threshold to 10, that's what we used on MadWiFi for
 a long time. Also when we have pending frames on the beacon queue (we
 got a beacon that didn't make it on the air) it's more likely that
 the beacon queue never started, probably due to faulty DBA setting,
 so change that "beacon queue didn't stop" message.

 Tested this with AP mode and IBSS mode and seems to work fine ;-)

Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Put remaining EEPROM data on ee struct
Nick Kossifidis [Thu, 30 Apr 2009 19:55:46 +0000 (15:55 -0400)]
ath5k: Put remaining EEPROM data on ee struct

* Put remaining EEPROM information on ee struct and remove is_hb63
 function.

 Now we also have rfkill stuff available.

Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Read Spur channels from EEPROM
Nick Kossifidis [Thu, 30 Apr 2009 19:55:45 +0000 (15:55 -0400)]
ath5k: Read Spur channels from EEPROM

* Read Spur channel information from EEPROM and use default channels
 for RF5413 compatible chips that don't have this info on EEPROM.

Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Allow user/driver to set txpower
Nick Kossifidis [Thu, 30 Apr 2009 19:55:44 +0000 (15:55 -0400)]
ath5k: Allow user/driver to set txpower

* Now that we have regulatory control enable the driver to set
 txpower on hw

 * Also use txpower table offset so that we can match
 power range set by user/driver with indices on power table.

 Tested 2 different cards (a CM9 and an RF5112-based ubnt) and got
 the same output using a remote machine to measure per-packet rssi
 (conected the cards using attenuators). I also switched between
 various tx power levels and i saw an equal power change on the remote
 machine (so txpower changes as expected) and verified that we have
 the same output on each rate.

Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Synchronize initialization with rt2870 driver
Ivo van Doorn [Tue, 28 Apr 2009 18:14:58 +0000 (20:14 +0200)]
rt2x00: Synchronize initialization with rt2870 driver

Ralink released a new rt2870 driver, these are the obvious
differences I could find. It doesn't same to make my device
work better, but neither does it seem to regress...

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Add new USB ID for rt2800usb
Ivo van Doorn [Mon, 27 Apr 2009 21:58:31 +0000 (23:58 +0200)]
rt2x00: Add new USB ID for rt2800usb

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agowl12xx: add driver
Kalle Valo [Wed, 29 Apr 2009 20:33:31 +0000 (23:33 +0300)]
wl12xx: add driver

wl12xx is a driver for TI wl1251 802.11 chipset designed for embedded
devices, supporting both SDIO and SPI busses. Currently the driver
supports only SPI. Adding support 1253 (the 5 GHz version) should be
relatively easy. More information here:

http://focus.ti.com/general/docs/wtbu/wtbuproductcontent.tsp?contentId=4711&navigationId=12494&templateId=6123

(Collapsed original sequence of pre-merge patches into single commit for
initial merge. -- JWL)

Signed-off-by: Kalle Valo <kalle.valo@nokia.com>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Implement support for rt2800usb
Ivo van Doorn [Sun, 26 Apr 2009 13:47:48 +0000 (15:47 +0200)]
rt2x00: Implement support for rt2800usb

Add support for the rt2800usb chipset.

Current problems:
 * Cannot scan 11n AP's
 * No TX during first minute after association
 * Broken Hardware encryption

Includes various patches from Mattias, Felix, Xose and Axel.

Signed-off-by: Mattias Nissler <mattias.nissler@gmx.de>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Xose Vazquez Perez <xose.vazquez@gmail.com>
Signed-off-by: Axel Kollhofer <rain_maker@root-forum.org>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Style fix for interval defines
Alban Browaeys [Wed, 29 Apr 2009 20:00:23 +0000 (22:00 +0200)]
rt2x00: Style fix for interval defines

Extra parenthesis are not needed in these 2 cases,
all other defines in rt2x00 are done without parenthesis
so just fixup these 2 cases.

Signed-off-by: Alban Browaeys <prahal@yahoo.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath9k: uninline ath9k_io{read,write}32 routines
Gabor Juhos [Wed, 29 Apr 2009 11:01:58 +0000 (13:01 +0200)]
ath9k: uninline ath9k_io{read,write}32 routines

The spin_lock handling uses lots of instructions on some archs.
With this patch the size of the ath9k module will be significantly
smaller.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Acked-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: tell driver when idle
Johannes Berg [Wed, 29 Apr 2009 10:26:17 +0000 (12:26 +0200)]
mac80211: tell driver when idle

When we aren't doing anything in mac80211, we can turn off
much of the hardware, depending on the driver/hw. Not doing
anything, aka being idle, means:

 * no monitor interfaces
 * no AP/mesh/wds interfaces
 * any station interfaces are in DISABLED state
 * any IBSS interfaces aren't trying to be in a network
 * we aren't trying to scan

By creating a new function that verifies these conditions and calling
it at strategic points where the states of those conditions change,
we can easily make mac80211 tell the driver when we are idle to save
power.

Additionally, this fixes a small quirk where a recalculated powersave
state is passed to the driver even if the hardware is about to stopped
completely.

This patch intentionally doesn't touch radio_enabled because that is
currently implemented to be a soft rfkill which is inappropriate here
when we need to be able to wake up with low latency.

One thing I'm not entirely sure about is this:

  phy0: device no longer idle - in use
  wlan0: direct probe to AP 00:11:24:91:07:4d try 1
  wlan0 direct probe responded
  wlan0: authenticate with AP 00:11:24:91:07:4d
  wlan0: authenticated
> phy0: device now idle
> phy0: device no longer idle - in use
  wlan0: associate with AP 00:11:24:91:07:4d
  wlan0: RX AssocResp from 00:11:24:91:07:4d (capab=0x401 status=0 aid=1)
  wlan0: associated

Is it appropriate to go into idle state for a short time when we have
just authenticated, but not associated yet? This happens only with the
userspace SME, because we cannot really know how long it will wait
before asking us to associate. Would going idle after a short timeout
be more appropriate? We may need to revisit this, depending on what
happens.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>