linux-2.6
15 years agoAX.25: Fix sysctl registration if !CONFIG_AX25_DAMA_SLAVE
Ralf Baechle [Wed, 6 Aug 2008 01:46:57 +0000 (18:46 -0700)] 
AX.25: Fix sysctl registration if !CONFIG_AX25_DAMA_SLAVE

Since 49ffcf8f99e8d33ec8afb450956804af518fd788 ("sysctl: update
sysctl_check_table") setting struct ctl_table.procname = NULL does no
longer work as it used to the way the AX.25 code is expecting it to
resulting in the AX.25 sysctl registration code to break if
CONFIG_AX25_DAMA_SLAVE was not set as in some distribution kernels.
Kernel releases from 2.6.24 are affected.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agopktgen: mac count
Robert Olsson [Wed, 6 Aug 2008 01:45:05 +0000 (18:45 -0700)] 
pktgen: mac count

dst_mac_count and src_mac_count patch from Eneas Hunguana
We have sent one mac address to much.

Signed-off-by: Robert Olsson <robert.olsson@its.uu.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agopktgen: random flow
Robert Olsson [Wed, 6 Aug 2008 01:44:26 +0000 (18:44 -0700)] 
pktgen: random flow

Random flow generation has not worked. This fixes it.

Signed-off-by: Robert Olsson <robert.olsson@its.uu.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agobridge: Eliminate unnecessary forward delay
Stephen Hemminger [Wed, 6 Aug 2008 01:42:51 +0000 (18:42 -0700)] 
bridge: Eliminate unnecessary forward delay

From: Stephen Hemminger <shemminger@vyatta.com>

Based upon original patch by Herbert Xu, which contained
the following problem description:

--------------------
When the forward delay is set to zero, we still delay the setting
of the forwarding state by one or possibly two timers depending
on whether STP is enabled.  This could either turn out to be
instantaneous, or horribly slow depending on the load of the
machine.

As there is nothing preventing us from enabling forwarding straight
away, this patch eliminates this potential delay by executing the
code directly if the forward delay is zero.

The effect of this problem is that immediately after the carrier
comes on a port, the bridge will drop all packets received from
that port until it enters forwarding mode, thus causing unnecessary
packet loss.

Note that this patch doesn't fully remove the delay due to the
link watcher.  We should also check the carrier state when we
are about to drop an incoming packet because the port is disabled.
But that's for another patch.
--------------------

This version of the fix takes a different approach, in that
it just does the state change directly.

Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoMerge branch 'no-ath9k' of git://git.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Tue, 5 Aug 2008 08:28:35 +0000 (01:28 -0700)] 
Merge branch 'no-ath9k' of git://git./linux/kernel/git/linville/wireless-2.6

15 years agobridge: fix compile warning in net/bridge/br_netfilter.c
Rami Rosen [Tue, 5 Aug 2008 08:21:22 +0000 (01:21 -0700)] 
bridge: fix compile warning in net/bridge/br_netfilter.c

This patch fixes the following warning due to incompatible pointer
assignment:

net/bridge/br_netfilter.c: In function 'br_netfilter_rtable_init':
net/bridge/br_netfilter.c:116: warning: assignment from incompatible
pointer type

This warning is due to commit 4adf0af6818f3ea52421dc0bae836cfaf20ef72a
from July 30 (send correct MTU value in PMTU (revised)).

Signed-off-by: Rami Rosen <ramirose@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoipv4: remove unused field in struct flowi (include/net/flow.h).
Rami Rosen [Tue, 5 Aug 2008 08:19:50 +0000 (01:19 -0700)] 
ipv4: remove unused field in struct flowi (include/net/flow.h).

This patch removes an unused field (flags) from struct flowi; it seems
that this "flags" field was used once in the past for multipath
routing with FLOWI_FLAG_MULTIPATHOLDROUTE flag (which does no longer
exist); however, the "flags" field of struct flowi is not used
anymore.

Signed-off-by: Rami Rosen <ramirose@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agotg3: Fix 'scheduling while atomic' errors
Matt Carlson [Tue, 5 Aug 2008 06:17:34 +0000 (23:17 -0700)] 
tg3: Fix 'scheduling while atomic' errors

This patch fixes the 'scheduling while atomic' errors introduced by
commit 12dac0756d357325b107fe6ec24921ec38661839 ("tg3: adapt tg3 to
use reworked PCI PM code").

The first hunk of the patch removes an unnecessary
tg3_set_power_state() call.  The chip will already be in the D0 state
either due to a chip reset or through a previous call to
tg3_set_power_state().

The second hunk of the patch moves the tg3_set_power_state() call
outside the critical section guarded by tg3_full_lock() and
tg3_full_unlock() functions.  The power state of the device is and
should be outside the lock's domain and all other
tg3_set_power_state() calls support this.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: Kill plain NET_XMIT_BYPASS.
David S. Miller [Tue, 5 Aug 2008 06:04:08 +0000 (23:04 -0700)] 
net: Kill plain NET_XMIT_BYPASS.

dst_input() was doing something completely absurd, looping
on skb->dst->input() if NET_XMIT_BYPASS was seen, but these
functions never return such an error.

And as a result plain ole' NET_XMIT_BYPASS has no more
references and can be completely killed off.

Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet_sched: Add qdisc __NET_XMIT_BYPASS flag
Jarek Poplawski [Tue, 5 Aug 2008 05:39:11 +0000 (22:39 -0700)] 
net_sched: Add qdisc __NET_XMIT_BYPASS flag

Patrick McHardy <kaber@trash.net> noticed that it would be nice to
handle NET_XMIT_BYPASS by NET_XMIT_SUCCESS with an internal qdisc flag
__NET_XMIT_BYPASS and to remove the mapping from dev_queue_xmit().

David Miller <davem@davemloft.net> spotted a serious bug in the first
version of this patch.

Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet_sched: Add qdisc __NET_XMIT_STOLEN flag
Jarek Poplawski [Tue, 5 Aug 2008 05:31:03 +0000 (22:31 -0700)] 
net_sched: Add qdisc __NET_XMIT_STOLEN flag

Patrick McHardy <kaber@trash.net> noticed:
"The other problem that affects all qdiscs supporting actions is
TC_ACT_QUEUED/TC_ACT_STOLEN getting mapped to NET_XMIT_SUCCESS
even though the packet is not queued, corrupting upper qdiscs'
qlen counters."

and later explained:
"The reason why it translates it at all seems to be to not increase
the drops counter. Within a single qdisc this could be avoided by
other means easily, upper qdiscs would still increase the counter
when we return anything besides NET_XMIT_SUCCESS though.

This means we need a new NET_XMIT return value to indicate this to
the upper qdiscs. So I'd suggest to introduce NET_XMIT_STOLEN,
return that to upper qdiscs and translate it to NET_XMIT_SUCCESS
in dev_queue_xmit, similar to NET_XMIT_BYPASS."

David Miller <davem@davemloft.net> noticed:
"Maybe these NET_XMIT_* values being passed around should be a set of
bits. They could be composed of base meanings, combined with specific
attributes.

So you could say "NET_XMIT_DROP | __NET_XMIT_NO_DROP_COUNT"

The attributes get masked out by the top-level ->enqueue() caller,
such that the base meanings are the only thing that make their
way up into the stack. If it's only about communication within the
qdisc tree, let's simply code it that way."

This patch is trying to realize these ideas.

Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoiwl3945: fix merge mistake for packet injection
Zhu Yi [Mon, 4 Aug 2008 08:00:48 +0000 (16:00 +0800)] 
iwl3945: fix merge mistake for packet injection

We should allow packets transmission in monitor mode for 3945. The
patch fixes a merge error with 2.6.26 kernel.

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: grap nic access before accessing periphery registers
Tomas Winkler [Mon, 4 Aug 2008 08:00:47 +0000 (16:00 +0800)] 
iwlwifi: grap nic access before accessing periphery registers

We need to grap nic access before accessing periphery registers.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: decrement rx skb counter in scan abort handler
Emmanuel Grumbach [Mon, 4 Aug 2008 08:00:46 +0000 (16:00 +0800)] 
iwlwifi: decrement rx skb counter in scan abort handler

This patch decrements rx skb counter in scan abort handler.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: fix unhandled interrupt when HW rfkill is on
Tomas Winkler [Mon, 4 Aug 2008 08:00:43 +0000 (16:00 +0800)] 
iwlwifi: fix unhandled interrupt when HW rfkill is on

This patch fixes unhandled interrupt when HW rfkill is on during
devices start up. The behavior changes, now open is successful even
when rfkill is on. This is to align with the situation when rfkill
is set on after opening.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: implement iwl5000_calc_rssi
Tomas Winkler [Mon, 4 Aug 2008 08:00:42 +0000 (16:00 +0800)] 
iwlwifi: implement iwl5000_calc_rssi

This patch implements rssi calculation for 5000 HW.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: memory allocation optimization
Gregory Greenman [Mon, 4 Aug 2008 08:00:40 +0000 (16:00 +0800)] 
iwlwifi: memory allocation optimization

This patch optimizes memory allocation. The cmd member of
iwl_tx_queue was allocated previously as a continuous block
of memory. This patch allocates separate memory chunks for each command
and maps/unmaps these chunks in the run time.

Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: HW bug fixes
Tomas Winkler [Mon, 4 Aug 2008 08:00:39 +0000 (16:00 +0800)] 
iwlwifi: HW bug fixes

This patch adds few HW bug fixes.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agop54: Fix potential concurrent access to private data
Larry Finger [Sun, 3 Aug 2008 22:58:36 +0000 (17:58 -0500)] 
p54: Fix potential concurrent access to private data

Experience with the rtl8187 driver has shown that mac80211 can make
calls to the config callback routine in rapid succession. This patch
creates a mutex that protects the private data in several of the routines
called by mac80211.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Disable link tuning in rt2500usb
Ivo van Doorn [Sun, 3 Aug 2008 21:36:01 +0000 (23:36 +0200)] 
rt2x00: Disable link tuning in rt2500usb

In the legacy rt2570 driver the link tuner was never
really called. And now the reason has finally become
apparent: It breaks TX capabilities

As soon as the device has been associated all following
TX frames will be queued in the hardware and never transmitted
to the air. Disabling sections of the link tuner did not
have the expected result, but completely disabling the
link tuner did have the right result (Both of my rt2570 devices
came back to life).

This should fix Fedora bug: 411481

v2: Fix typos

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: Don't use buffer allocated on the stack for led names
Sven Wegener [Fri, 1 Aug 2008 19:57:16 +0000 (21:57 +0200)] 
iwlwifi: Don't use buffer allocated on the stack for led names

Having the buffer on the stack and even re-using it for all led devices is bad.
Not being able to resolve the name member of the led device structure to a
meaningful value leads to confusion during ad-hoc debugging and potential
breakage in the future, if we ever decide to access the name member outside of
the registration function. Move the buffer to our private per led device
structures so that it is accessible after registration.

A quick grep didn't yield any occurence of using the led device name parameter
outside of the led device registration function, so currently we should already
be safe for normal operation.

Signed-off-by: Sven Wegener <sven.wegener@stealer.net>
Cc: Richard Purdie <rpurdie@rpsys.net>
Acked-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: rename 4965 to AGN
Tomas Winkler [Mon, 21 Jul 2008 15:54:42 +0000 (18:54 +0300)] 
iwlwifi: rename 4965 to AGN

This patch renames driver name from 4965 to AGN
The driver supports both 4965AGN and 5000AGN family

The driver's original module name iwl4965.ko remains as an alias

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: fix checkpatch.pl errors
Tomas Winkler [Sun, 20 Jul 2008 23:40:14 +0000 (02:40 +0300)] 
iwlwifi: fix checkpatch.pl errors

This patch fixes errors reported by checkpatch in iwlwifi drivers

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: rename iwl4965-base.c to iwl-agn.c
Tomas Winkler [Sun, 20 Jul 2008 23:40:13 +0000 (02:40 +0300)] 
iwlwifi: rename iwl4965-base.c to iwl-agn.c

This patch renames iwl4965-base.c to iwl-agn.c

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: automatic IBSS channel selection
Daniel Drake [Sat, 19 Jul 2008 22:31:17 +0000 (23:31 +0100)] 
mac80211: automatic IBSS channel selection

When joining an ad-hoc network, the user is currently required to specify
the channel. The network will not be joined otherwise, unless it happens
to be sitting on the currently active channel.

This patch implements automatic channel selection when the user has not
locked the interface onto a specific channel.

Signed-off-by: Daniel Drake <dsd@gentoo.org>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: use dtim_period from association, and set listen_interval
Tomas Winkler [Sat, 19 Jul 2008 01:41:24 +0000 (04:41 +0300)] 
iwlwifi: use dtim_period from association, and set listen_interval

This patch uses dtim_period from association, and sets the listen_interval.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: RS small compile warnings without CONFIG_IWLWIFI_DEBUG
Denis V. Lunev [Sat, 19 Jul 2008 01:04:18 +0000 (04:04 +0300)] 
iwlwifi: RS small compile warnings without CONFIG_IWLWIFI_DEBUG

iwl-agn-rs.c: In function 'rs_clear':
iwl-agn-rs.c:2405: warning: unused variable 'priv

Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: bug fix in AGG flow - cast const to ULL
Emmanuel Grumbach [Fri, 18 Jul 2008 05:53:09 +0000 (13:53 +0800)] 
iwlwifi: bug fix in AGG flow - cast const to ULL

This patch fixes a bug in AGG flow:
u64 bitmap = 0;
bitmap |= 1 << 32 results to be 0xffffffff80000000.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: set led register in disassociation
Esti Kummer [Fri, 18 Jul 2008 05:53:07 +0000 (13:53 +0800)] 
iwlwifi: set led register in disassociation

This patch sets the led register in disassociation flow according to
rf-kill state : off - in case of rf_kill, on - otherwise.

Signed-off-by: Esti Kummer <ester.kummer@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: kill iwl4965_fill_rs_info
Tomas Winkler [Fri, 18 Jul 2008 05:53:06 +0000 (13:53 +0800)] 
iwlwifi: kill iwl4965_fill_rs_info

iwl4965_fill_rs_info was used in sysfs. This info is already present
in iwl-agn-rs debugfs.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: rename iwl-4695-rs to iwl-agn-rs
Tomas Winkler [Fri, 18 Jul 2008 05:53:05 +0000 (13:53 +0800)] 
iwlwifi: rename iwl-4695-rs to iwl-agn-rs

This patch renames iwl-4965-rs to iwl-agn-rs as it provides
rate scale capability for all AGN capable iwlwifi drivers.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: move iwl4965_set_pwr_src to iwl4965-base.c
Tomas Winkler [Fri, 18 Jul 2008 05:53:04 +0000 (13:53 +0800)] 
iwlwifi: move iwl4965_set_pwr_src to iwl4965-base.c

This patch moves iwl4965_set_pwr_src to iwl4965-base.c.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: move beacon handling to iwl4965-base.c
Tomas Winkler [Fri, 18 Jul 2008 05:53:03 +0000 (13:53 +0800)] 
iwlwifi: move beacon handling to iwl4965-base.c

This patch concentrates becaon handling in iwl4965-base.c.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: move iwl4965_mac_ampdu_action to iwl4965-base.c
Tomas Winkler [Fri, 18 Jul 2008 05:53:02 +0000 (13:53 +0800)] 
iwlwifi: move iwl4965_mac_ampdu_action to iwl4965-base.c

This patch moves iwl4965_mac_ampdu_action to iwl4965-base.c.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: make listen_interval be limited by low level driver
Tomas Winkler [Fri, 18 Jul 2008 05:53:00 +0000 (13:53 +0800)] 
mac80211: make listen_interval be limited by low level driver

This patch makes possible for a driver to specify maximal listen interval
The possibility for user to configure listen interval is not implemented
yet, currently the maximum provided by the driver or 1 is used.
Mac80211 uses config handler to set listen interval for to the driver.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: pass dtim_period to low level driver
Emmanuel Grumbach [Fri, 18 Jul 2008 05:52:59 +0000 (13:52 +0800)] 
mac80211: pass dtim_period to low level driver

This patch adds the dtim_period in ieee80211_bss_conf, this allows the low
level driver to know the dtim_period, and to plan power save accordingly.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: corrects power_level in sysfs
Esti Kummer [Fri, 18 Jul 2008 05:52:58 +0000 (13:52 +0800)] 
iwlwifi: corrects power_level in sysfs

This patch corrects power_level in sysfs.

Signed-off-by: Esti Kummer <ester.kummer@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: add power save to 5000 HW
Mohamed Abbas [Fri, 18 Jul 2008 05:52:57 +0000 (13:52 +0800)] 
iwlwifi: add power save to 5000 HW

This patch adds support for power save for 5000 HW.

Signed-off-by: Mohamed Abbas <mohamed.abbas@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agonet: eliminate refcounting in backlog queue
Stephen Hemminger [Mon, 4 Aug 2008 04:29:57 +0000 (21:29 -0700)] 
net: eliminate refcounting in backlog queue

Avoid the overhead of atomic increment/decrement on each received packet.
This helps performance of non-NAPI devices (like loopback).
Use cleanup function to walk queue on each cpu and clean out any
left over packets.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoipv6: Do not drop packet if skb->local_df is set to true
Wei Yongjun [Mon, 4 Aug 2008 04:15:59 +0000 (21:15 -0700)] 
ipv6: Do not drop packet if skb->local_df is set to true

The old code will drop IPv6 packet if ipfragok is not set, since
ipfragok is obsoleted, will be instead by used skb->local_df, so this
check must be changed to skb->local_df.

This patch fix this problem and not drop packet if skb->local_df is
set to true.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agosctp: Drop ipfargok in sctp_xmit function
Herbert Xu [Mon, 4 Aug 2008 04:15:08 +0000 (21:15 -0700)] 
sctp: Drop ipfargok in sctp_xmit function

The ipfragok flag controls whether the packet may be fragmented
either on the local host on beyond.  The latter is only valid on
IPv4.

In fact, we never want to do the latter even on IPv4 when PMTU is
enabled.  This is because even though we can't fragment packets
within SCTP due to the prtocol's inherent faults, we can still
fragment it at IP layer.  By setting the DF bit we will improve
the PMTU process.

RFC 2960 only says that we SHOULD clear the DF bit in this case,
so we're compliant even if we set the DF bit.  In fact RFC 4960
no longer has this statement.

Once we make this change, we only need to control the local
fragmentation.  There is already a bit in the skb which controls
that, local_df.  So this patch sets that instead of using the
ipfragok argument.

The only complication is that there isn't a struct sock object
per transport, so for IPv4 we have to resort to changing the
pmtudisc field for every packet.  This should be safe though
as the protocol is single-threaded.

Note that after this patch we can remove ipfragok from the rest
of the stack too.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoipv6: Fix the return value of Set Hop-by-Hop options header with NULL data pointer
Yang Hongyang [Mon, 4 Aug 2008 01:16:15 +0000 (18:16 -0700)] 
ipv6: Fix the return value of Set Hop-by-Hop options header with NULL data pointer

When Set Hop-by-Hop options header with NULL data
pointer and optlen is not zero use setsockopt(),
the kernel successfully return 0 instead of
return error EINVAL or EFAULT.

This patch fix the problem.

Signed-off-by: Yang Hongyang <yanghy@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agoipv6: syncookies: free reqsk on xfrm_lookup error
Florian Westphal [Mon, 4 Aug 2008 01:13:44 +0000 (18:13 -0700)] 
ipv6: syncookies: free reqsk on xfrm_lookup error

cookie_v6_check() did not call reqsk_free() if xfrm_lookup() fails,
leaking the request sock.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: Add missing extra2 parameter for ip_default_ttl sysctl
Sven Wegener [Sun, 3 Aug 2008 21:06:44 +0000 (14:06 -0700)] 
net: Add missing extra2 parameter for ip_default_ttl sysctl

Commit 76e6ebfb40a2455c18234dcb0f9df37533215461 ("netns: add namespace
parameter to rt_cache_flush") acceses the extra2 parameter of the
ip_default_ttl ctl_table, but it is never set to a meaningful
value. When e84f84f276473dcc673f360e8ff3203148bdf0e2 ("netns: place
rt_genid into struct net") is applied, we'll oops in
rt_cache_invalidate(). Set extra2 to init_net, to avoid that.

Reported-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Sven Wegener <sven.wegener@stealer.net>
Tested-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Acked-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: use software GSO for SG+CSUM capable netdevices
Lennert Buytenhek [Sun, 3 Aug 2008 08:23:10 +0000 (01:23 -0700)] 
net: use software GSO for SG+CSUM capable netdevices

If a netdevice does not support hardware GSO, allowing the stack to
use GSO anyway and then splitting the GSO skb into MSS-sized pieces
as it is handed to the netdevice for transmitting is likely still
a win as far as throughput and/or CPU usage are concerned, since it
reduces the number of trips through the output path.

This patch enables the use of GSO on any netdevice that supports SG.
If a GSO skb is then sent to a netdevice that supports SG but does not
support hardware GSO, net/core/dev.c:dev_hard_start_xmit() will take
care of doing the necessary GSO segmentation in software.

Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: fix missing pneigh entries in the neighbor seq_file code
Chris Larson [Sun, 3 Aug 2008 08:10:55 +0000 (01:10 -0700)] 
net: fix missing pneigh entries in the neighbor seq_file code

When pneigh entries exist, but the user's read buffer isn't sufficient to
hold them all, one of the pneigh entries will be missing from the results.

In neigh_get_idx_any, the number of elements which neigh_get_idx
encountered is not correctly subtracted from the position number before
the call to pneigh_get_idx.  neigh_get_idx reduces the position by 1 for
each call to neigh_get_next, but it does not reduce it by one for the
first element (neigh_get_first). The patch alters the neigh_get_idx and
pneigh_get_idx functions to subtract one from pos, for the first element,
when pos is non-zero.

Signed-off-by: Chris Larson <clarson@mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agonet: in the first call to neigh_seq_next, call neigh_get_first, not neigh_get_idx.
Chris Larson [Sun, 3 Aug 2008 08:02:41 +0000 (01:02 -0700)] 
net: in the first call to neigh_seq_next, call neigh_get_first, not neigh_get_idx.

neigh_seq_next won't be called both with *pos > 0 && v ==
SEQ_START_TOKEN, so there's no point calling neigh_get_idx when we're
on the start token, just call neigh_get_first directly.

Signed-off-by: Chris Larson <clarson@mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agopkt_sched: Make sure RTNL is held in qdisc_root_lock().
David S. Miller [Sun, 3 Aug 2008 06:27:37 +0000 (23:27 -0700)] 
pkt_sched: Make sure RTNL is held in qdisc_root_lock().

It is the only legal environment in which this can be
used.

Add some commentary explaining the situation.

Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agomac80211: Use queue_lock() in ieee80211_ht_agg_queue_remove().
David S. Miller [Sun, 3 Aug 2008 06:25:50 +0000 (23:25 -0700)] 
mac80211: Use queue_lock() in ieee80211_ht_agg_queue_remove().

qdisc_root_lock() is only %100 safe to use when the RTNL
semaphore is held.

Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agopkt_sched: Use qdisc_lock() on already sampled root qdisc.
David S. Miller [Sun, 3 Aug 2008 03:02:43 +0000 (20:02 -0700)] 
pkt_sched: Use qdisc_lock() on already sampled root qdisc.

Based upon a bug report by Jeff Kirsher.

Don't use qdisc_root_lock() in these cases as the root
qdisc could have been changed, and we'd thus lock the
wrong object.

Tested by Emil S Tantilov who confirms that this seems
to fix the problem.

Signed-off-by: David S. Miller <davem@davemloft.net>
15 years agort2x00: Fix compile warning
Ivo van Doorn [Sat, 2 Aug 2008 08:31:09 +0000 (01:31 -0700)] 
rt2x00: Fix compile warning

rt2x00usb_vendor_request_large_buff is write-only, so it is
safe to make the argument a const.

Fixes compile warning:
drivers/net/wireless/rt2x00/rt73usb.c: In function 'rt73usb_load_firmware':
drivers/net/wireless/rt2x00/rt73usb.c:916: warning: passing argument 5 of 'rt2x00usb_vendor_request_large_buff' discards qualifiers from pointer target typ

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.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 [Sat, 2 Aug 2008 05:08:51 +0000 (22:08 -0700)] 
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-2.6

15 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6
Linus Torvalds [Fri, 1 Aug 2008 21:59:11 +0000 (14:59 -0700)] 
Merge branch 'release' of git://git./linux/kernel/git/aegl/linux-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6:
  [IA64] Move include/asm-ia64 to arch/ia64/include/asm

15 years agoembedded: fix vc_translate operator precedence
Tim Bird [Fri, 1 Aug 2008 21:05:50 +0000 (14:05 -0700)] 
embedded: fix vc_translate operator precedence

This fixes a bug in operator precedence in the newly introduced vc_translate
macro.  Without this fix, the translation of some characters on the
kernel console is garbled.

This patch was copied to the e-mail list previously for testing.  Now,
all reports confirm that it works, so this is an official post for
application.

Signed-off-by: Tim Bird <tim.bird@am.sony.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
15 years agoRemove EXPORTS of follow_page & zap_page_range
Jack Steiner [Fri, 1 Aug 2008 20:08:15 +0000 (15:08 -0500)] 
Remove EXPORTS of follow_page & zap_page_range

Delete 2 EXPORTs that were accidentally sent upstream.

Signed-off-by: Jack Steiner <steiner@sgi.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodual license ftrace.txt
Steven Rostedt [Thu, 31 Jul 2008 16:40:52 +0000 (12:40 -0400)] 
dual license ftrace.txt

I asked legal about the licensing of ftrace.txt, and they told me that,
unless the Documentation directory is specifically set up to handle non
GPL licenses (which it does not appear to be), then it would be best to
put ftrace.txt under the GPL.

This patch adds a dual license to ftrace.txt such that it is under both
the FDL and the GPL.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoFRV: Wire up new system calls
David Howells [Thu, 31 Jul 2008 12:46:33 +0000 (13:46 +0100)] 
FRV: Wire up new system calls

Wire up for FRV the system calls that were added in the last merge window.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMN10300: Wire up new system calls
David Howells [Thu, 31 Jul 2008 12:01:30 +0000 (13:01 +0100)] 
MN10300: Wire up new system calls

Wire up system calls added in the last merge window for the MN10300 arch.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agotry harder to load tty ldisc driver
Eugeniy Meshcheryakov [Thu, 31 Jul 2008 09:03:19 +0000 (10:03 +0100)] 
try harder to load tty ldisc driver

Currently function tty_ldisc_get() tries to load an ldisc driver module
only when tty_ldisc_try_get() returns -EAGAIN. This happens only if
module is being unloaded. If ldisc module is not loaded
tty_ldisc_try_get() returns -EINVAL and this case is not handled in
tty_ldisc_get(), so request_module() is not called.

Attached patch fixes this by calling request_module() if
tty_ldisc_try_get() returned any error code.

I discovered this when my UMTS modem stopped working with 2.6.27-rc1
because module ppp_async was not loaded.

Signed-off-by: Eugeniy Meshcheryakov <eugen@debian.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'kvm-updates-2.6.27' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 1 Aug 2008 19:48:16 +0000 (12:48 -0700)] 
Merge branch 'kvm-updates-2.6.27' of git://git./linux/kernel/git/avi/kvm

* 'kvm-updates-2.6.27' of git://git.kernel.org/pub/scm/linux/kernel/git/avi/kvm:
  KVM: s390: Fix kvm on IBM System z10
  KVM: Advertise synchronized mmu support to userspace
  KVM: Synchronize guest physical memory map to host virtual memory map
  KVM: Allow browsing memslots with mmu_lock
  KVM: Allow reading aliases with mmu_lock

15 years agoMerge branch 'zero-len' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik...
Linus Torvalds [Fri, 1 Aug 2008 19:47:53 +0000 (12:47 -0700)] 
Merge branch 'zero-len' of git://git./linux/kernel/git/jgarzik/misc-2.6

* 'zero-len' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/misc-2.6:
  drivers/media, include/media: delete zero-length files

15 years agomm/hugetlb: don't crash when HPAGE_SHIFT is 0
Benjamin Herrenschmidt [Thu, 31 Jul 2008 07:07:30 +0000 (00:07 -0700)] 
mm/hugetlb: don't crash when HPAGE_SHIFT is 0

Some platform decide whether they support huge pages at boot time.  On
these, such as powerpc, HPAGE_SHIFT is a variable, not a constant, and is
set to 0 when there is no such support.

The patches to introduce multiple huge pages support broke that causing
the kernel to crash at boot time on machines such as POWER3 which lack
support for multiple page sizes.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMissing symbol prefix on vmlinux.lds.h
Yoshinori Sato [Thu, 31 Jul 2008 07:07:29 +0000 (00:07 -0700)] 
Missing symbol prefix on vmlinux.lds.h

ARCH=h8300:

init/main.c:781: undefined reference to `___early_initcall_end'

Same problem have
__start___bug_table
__stop___bug_table
__tracedata_start
__tracedata_end
__per_cpu_start
__per_cpu_end

When defining a symbol in vmlinux.lds, use the VMLINUX_SYMBOL macro.
VMLINUX_SYMBOL adds a prefix charactor.

You can't just use straight symbol names in common header files as they
dont take into consideration weird arch-specific ABI conventions.  in the
case of Blackfin/h8300, the ABI dictates that any C-visible symbols have
an underscore prefixed to them.  Thus all symbols in vmlinux.lds.h need to
be wrapped in VMLINUX_SYMBOL() so that each arch can put hide this magic
in their own files.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Cc: "Mike Frysinger" <vapier.adi@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoPNP: fix formatting of dbg_pnp_show_resources() output
Bjorn Helgaas [Thu, 31 Jul 2008 07:07:27 +0000 (00:07 -0700)] 
PNP: fix formatting of dbg_pnp_show_resources() output

Each resource should be printed on its own line, so start snprintf'ing
at the beginning of the buffer every time through the loop.

Also, use scnprintf() rather than snprintf() when building up the
buffer to print.  scnprintf() returns the number of characters actually
written into the buffer (not including the trailing NULL).

snprintf() returns the number of characters that *would be* written,
assuming everything would fit in the buffer.  That's nice if we want to
resize the buffer to make sure everything fits, but in this case, I
just want to keep from overflowing the buffer, and it's OK if the
output is truncated.

Using snprintf() meant that my "len" could grow to be more than the
the buffer size, which makes "sizeof(buf) - len" negative, which causes
this alarming WARN_ON:
    http://marc.info/?l=linux-kernel&m=121736480005656&w=2

More useful snprintf/scnprintf discussion:
    http://lwn.net/Articles/69419/

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Reported-by: Pete Clements <clem@clem.clem-digital.net>
Cc: Rene Herman <rene.herman@keyaccess.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoRemove newline from the description of module parameters
Niels de Vos [Thu, 31 Jul 2008 07:07:23 +0000 (00:07 -0700)] 
Remove newline from the description of module parameters

Some module parameters with only one line have the '\n' at the end of the
description.  This is not needed nor wanted as after the description the
type (i.e.  int) is followed by a newline.

Some modules contain a multi-line description, these are not affected
by this patch.

Signed-off-by: Niels de Vos <niels.devos@wincor-nixdorf.com>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: John W. Linville <linville@tuxdriver.com>
Cc: Ed L. Cashin <ecashin@coraid.com>
Cc: Dave Airlie <airlied@linux.ie>
Cc: Roland Dreier <rolandd@cisco.com>
Acked-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Fri, 1 Aug 2008 19:41:29 +0000 (12:41 -0700)] 
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  pata_it821x: Driver updates and reworking
  libata.h: replace __FUNCTION__ with __func__
  ata_piix: subsys 106b:00a3 is apple ich8m too
  libata-core: make sure that ata_force_tbl is freed in case of an error
  libata: update atapi disable handling
  pata_via: add VX800 flag; add function for fixing h/w bugs
  pata_ali: misplaced pci_dev_put()

15 years agoMerge branch 'for-linus' of git://oss.sgi.com:8090/xfs/xfs-pull
Linus Torvalds [Fri, 1 Aug 2008 19:39:09 +0000 (12:39 -0700)] 
Merge branch 'for-linus' of git://oss.sgi.com:8090/xfs/xfs-pull

* 'for-linus' of git://oss.sgi.com:8090/xfs/xfs-pull: (64 commits)
  [XFS] Remove vn_revalidate calls in xfs.
  [XFS] Now that xfs_setattr is only used for attributes set from ->setattr
  [XFS] xfs_setattr currently doesn't just handle the attributes set through
  [XFS] fix use after free with external logs or real-time devices
  [XFS] A bug was found in xfs_bmap_add_extent_unwritten_real(). In a
  [XFS] fix compilation without CONFIG_PROC_FS
  [XFS] s/XFS_PURGE_INODE/IRELE/g s/VN_HOLD(XFS_ITOV())/IHOLD()/
  [XFS] fix mount option parsing in remount
  [XFS] Disable queue flag test in barrier check.
  [XFS] streamline init/exit path
  [XFS] Fix up problem when CONFIG_XFS_POSIX_ACL is not set and yet we still
  [XFS] Don't assert if trying to mount with blocksize > pagesize
  [XFS] Don't update mtime on rename source
  [XFS] Allow xfs_bmbt_split() to fallback to the lowspace allocator
  [XFS] Restore the lowspace extent allocator algorithm
  [XFS] use minleft when allocating in xfs_bmbt_split()
  [XFS] attrmulti cleanup
  [XFS] Check for invalid flags in xfs_attrlist_by_handle.
  [XFS] Fix CI lookup in leaf-form directories
  [XFS] Use the generic xattr methods.
  ...

15 years agoiwl3945: Fix statistics in monitor mode
Maxim Levitsky [Fri, 1 Aug 2008 09:54:27 +0000 (12:54 +0300)] 
iwl3945: Fix statistics in monitor mode

iwl3945_rx_reply_rx was sending packets too early to
mac80211, before updating signal strength/quality.
This resulted in garbage power levels.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoprism54 - Use offsetof()
Takashi Iwai [Thu, 31 Jul 2008 17:03:10 +0000 (19:03 +0200)] 
prism54 - Use offsetof()

Use the standard offsetof() macro to fix a compile warning below:
  CC [M]  drivers/net/wireless/prism54/isl_ioctl.o
drivers/net/wireless/prism54/isl_ioctl.c: In function 'prism2_ioctl_set_generic_element':
drivers/net/wireless/prism54/isl_ioctl.c:2658: warning: cast from pointer to integer of different size

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoipw2200 - Fix bad ipw_write8() macro
Takashi Iwai [Thu, 31 Jul 2008 17:02:06 +0000 (19:02 +0200)] 
ipw2200 - Fix bad ipw_write8() macro

ipw_write8() can't be used alone with a loop because of a wrong definition.

  CC [M]  drivers/net/wireless/ipw2200.o
drivers/net/wireless/ipw2200.c: In function 'ipw_ethtool_set_eeprom':
drivers/net/wireless/ipw2200.c:10579: warning: array subscript is above array bounds
drivers/net/wireless/ipw2200.c: In function 'ipw_load':
drivers/net/wireless/ipw2200.c:2663: warning: array subscript is above array bounds

Add missing do {} while (0) to fix them.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agortl8187: Fix lockups due to concurrent access to config routine
Larry Finger [Fri, 1 Aug 2008 00:30:48 +0000 (19:30 -0500)] 
rtl8187: Fix lockups due to concurrent access to config routine

Some users of the RTL8187B have experienced difficulties since commit
49292d56352a6ab90d04c3448dd8b6106dfef2d6 that introduced the power
management wext hooks. This difficulty has not made much sense until
it was realized that it was possible for mac80211 to make a call to the
config routine while that routine was already being executed. On this
device, it is necessary to loopback the TX when changing channels. Unless
this is properly restored, the device will lockup. A mutex now protects
the device state, and the private data in several places.

The problem was found by Herton Ronaldo Krzesinski <herton@mandriva.com.br>,
who also suggested this type of fix.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Acked-by: Herton Ronaldo Krzesinski <herton@mandriva.com.br>
Acked-by: Hin-Tak Leung <htl10@users.sourceforge.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2500pci: restoring missing line
Peter Chubb [Thu, 31 Jul 2008 00:56:34 +0000 (10:56 +1000)] 
rt2500pci: restoring missing line

In kernel version 2.6.26-rc9 my wireless LAN card worked; but in the
released 2.6.26, my RaLink rt2500 card wouldn't associate.

Git-bisect led me to this patch:

61486e0f68d1f8966c09b734566a187d42d65c54
rt2x00: Remove ieee80211_tx_control argument from write_tx_desc()

I believe that there is a problem with that patch --- it
(inadvertantly) removes an extra line of code, that used to set the
DATABYTE_COUNT field.

This patch reinstates that line, and with it my card works again.

Signed-off-by: Peter Chubb <peterc@gelato.unsw.edu.au>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agolibertas: only enable rtap with mesh firmware
Dan Williams [Tue, 29 Jul 2008 17:50:39 +0000 (13:50 -0400)] 
libertas: only enable rtap with mesh firmware

Since only mesh-enabled firmware has the CMD_802_11_MONITOR_MODE on
which the rtap functionality depends, only expose the rtap functionality
when mesh is also available.

Signed-off-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Sequence counter should be protected in irqsave
Ivo van Doorn [Mon, 28 Jul 2008 08:21:16 +0000 (10:21 +0200)] 
rt2x00: Sequence counter should be protected in irqsave

The sequence counter can be accessed in IRQ context,
which means the lock protecting the counter should
be irqsave. To prevent making the entire intf->lock irqsave
without reason, create a new lock which only protects
the sequence counter.

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Fix VGC lower bound initialization
Ivo van Doorn [Mon, 28 Jul 2008 08:20:12 +0000 (10:20 +0200)] 
rt2x00: Fix VGC lower bound initialization

When the EEPROM_BBPTUNE_VGC word is valid, we should
override EEPROM_BBPTUNE_VGCLOWER field with the BBP
value.

And we should _not_ do that when EEPROM_BBPTUNE_R17 is valid.

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: rt61pci needs another millisecond after firmware upload
Ivo van Doorn [Sun, 27 Jul 2008 13:06:50 +0000 (15:06 +0200)] 
rt2x00: rt61pci needs another millisecond after firmware upload

After the hardware has indicated the firmware upload has completed
and the device is ready, we should wait another millisecond to
make sure the device is really ready to continue.

Without this timout, bringing the interface down and up again will
fail due to incorrect register initialization.

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Fix partial antenna configuration
Ivo van Doorn [Sun, 27 Jul 2008 13:06:21 +0000 (15:06 +0200)] 
rt2x00: Fix partial antenna configuration

The if-statement to determine the new TX/RX antenna
configuration was incomplete. It lacks the general
else-clause when the antenna wasn't changed.

This is a correct event, since it can occur when only
one of the antenna's has been changed or when the new
configuration is being forced (like when the interface
has just been added).

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agort2x00: Fix access permissions on debugfs files
Ivo van Doorn [Sun, 27 Jul 2008 13:06:05 +0000 (15:06 +0200)] 
rt2x00: Fix access permissions on debugfs files

Although most rt2x00 debugfs files don't contain
information which could compromise network security,
it is better to set the access permissions to root only.

This will be required when HW crypto is implemented,
because it could be possible to read the HW key from
the registers.

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoRtl8187 PATCH add usb ID for asus wireless link
Andrea Merello [Fri, 25 Jul 2008 17:08:11 +0000 (19:08 +0200)] 
Rtl8187 PATCH add usb ID for asus wireless link

This patch from Davide Cavalca adds a usb ID for an rtl8187L device.

Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoiwlwifi: don't stop queue in the middle of fragmented packet
Tomas Winkler [Thu, 24 Jul 2008 18:33:42 +0000 (21:33 +0300)] 
iwlwifi: don't stop queue in the middle of fragmented packet

This patch avoids stopping queue in the middle of the fragmented packet.
It is required that there will be ~10 (max packet/min fragment) or 16
(4 bits of frag number) free tfds all the time.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: remove obsolete declaration of struct ieee80211_hw_mode
Helmut Schaa [Thu, 24 Jul 2008 16:22:55 +0000 (18:22 +0200)] 
ath5k: remove obsolete declaration of struct ieee80211_hw_mode

Signed-off-by: Helmut Schaa <hschaa@suse.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agomac80211: fix fragmentation kludge
Tomas Winkler [Thu, 24 Jul 2008 15:46:44 +0000 (18:46 +0300)] 
mac80211: fix fragmentation kludge

This patch make mac80211 transmit correctly fragmented packet after
queue was stopped

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoAth5k: mask out unneeded interrupts
Jiri Slaby [Wed, 23 Jul 2008 11:17:34 +0000 (13:17 +0200)] 
Ath5k: mask out unneeded interrupts

Mask out previously demanded interrupt flags because we set
new ones. Don't allow mixing them after switch from sta to
ibss and vice versa.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Cc: Nick Kossifidis <mickflemm@gmail.com>
Cc: Luis R. Rodriguez <mcgrof@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoRFKILL: set the status of the leds on activation.
Dmitry Baryshkov [Tue, 22 Jul 2008 10:21:59 +0000 (14:21 +0400)] 
RFKILL: set the status of the leds on activation.

Provide default activate function to set the state of the led
when the led becomes bound to the trigger

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoRFKILL: allow one to specify led trigger name
Dmitry Baryshkov [Tue, 22 Jul 2008 10:17:37 +0000 (14:17 +0400)] 
RFKILL: allow one to specify led trigger name

Allow the rfkill driver to specify led trigger name.
By default it still defaults to the name of rfkill switch.

Signed-off-by: Dmitry Baryshkov <dbaryshkov@gmail.com>
Acked-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agorfkill: query EV_SW states when rfkill-input (re)?connects to a input device
Henrique de Moraes Holschuh [Thu, 31 Jul 2008 13:53:57 +0000 (10:53 -0300)] 
rfkill: query EV_SW states when rfkill-input (re)?connects to a input device

Every time a new input device that is capable of one of the
rfkill EV_SW events (currently only SW_RFKILL_ALL) is connected to
rfkill-input, we must check the states of the input EV_SW switches
and take action.  Otherwise, we will ignore the initial switch state.

We also need to re-check the states of the EV_SW switches after
a device that was under an exclusive grab is released back to us,
since we got no input events from that device while it was grabbed.

Signed-off-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Cc: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Update phy calibration functions
Nick Kossifidis [Sun, 20 Jul 2008 03:47:12 +0000 (06:47 +0300)] 
ath5k: Update phy calibration functions

 * Enable I/Q calibration each time we have correction results (we
   were only enabling calibration during reset). If we don't we commit
   the same results each time calibration routine is called.
 * Add some documentation and a TODO on nf calibration
 * Return -EAGAIN on noise floor timeout/failure

Changes-licensed-under: ISC
Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Update channel functions
Nick Kossifidis [Sun, 20 Jul 2008 03:44:43 +0000 (06:44 +0300)] 
ath5k: Update channel functions

 * Add channel function for RF2425 (got this from decompiling binary
   HAL, i have no idea why there is a 5GHz section but i'm looking
   into it)
 * Update RF5112 channel function (also got this from decompiling binary HAL)
 * Set JAPAN setting for channel 14 on all PHY chips

Changes-licensed-under: ISC
Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Add RF2425 initial rfgain values
Nick Kossifidis [Sun, 20 Jul 2008 03:42:47 +0000 (06:42 +0300)] 
ath5k: Add RF2425 initial rfgain values

 * Add initial RF gain settings for RF2425

Changes-licensed-under: ISC
Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Reorder calibration calls during reset and update hw_set_power
Nick Kossifidis [Sun, 20 Jul 2008 03:41:26 +0000 (06:41 +0300)] 
ath5k: Reorder calibration calls during reset and update hw_set_power

 * Update ath5k_hw_reset and add some more documentation about PHY calibration
 * Fix ath5k_hw_set_power to use AR5K_SLEEP_CTL_SLE_ALLOW for Network sleep
 * Preserve sleep duration field while setting AR5K_SLEEP_CTL
   and reduce delays & checks for register's status (got this from
   decompiling & dumps, it works for me but it needs testing)

Changes-licensed-under: ISC
Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Do ADC test during reset
Nick Kossifidis [Sun, 20 Jul 2008 03:38:16 +0000 (06:38 +0300)] 
ath5k: Do ADC test during reset

 * Do an ADC test during reset to match recent regdumps

Changes-licensed-under: ISC
Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Misc hw_reset updates
Nick Kossifidis [Sun, 20 Jul 2008 03:36:52 +0000 (06:36 +0300)] 
ath5k: Misc hw_reset updates

 * Update hw_reset to calculate some of the values we were using as static
 * Increase activation to rx delay

Changes-licensed-under: ISC
Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Misc hw_attach fixes
Nick Kossifidis [Sun, 20 Jul 2008 03:34:39 +0000 (06:34 +0300)] 
ath5k: Misc hw_attach fixes

 * Correctly attach RF2425
 * Update SREV values for Radio chips
 * Update hw_attach to use new SPENDING values
 * Write a bit after POST for some chips

Changes-licensed-under: ISC
Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Restore saved initval after POST
Nick Kossifidis [Sun, 20 Jul 2008 03:32:32 +0000 (06:32 +0300)] 
ath5k: Restore saved initval after POST

 * Restore saved initial value after POST

Changes-licensed-under: ISC
Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoath5k: Update register list
Nick Kossifidis [Wed, 30 Jul 2008 10:18:59 +0000 (13:18 +0300)] 
ath5k: Update register list

* Update list of registers
* Use updated register macros inside hw.c, initvals.c and debug.c

Changes-licensed-under: ISC
Signed-off-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
15 years agoadd reverse dependency of CONFIG_SGI_XP upon CONFIG_SGI_GRU
Dean Nelson [Fri, 1 Aug 2008 14:55:26 +0000 (09:55 -0500)] 
add reverse dependency of CONFIG_SGI_XP upon CONFIG_SGI_GRU

Add a reverse dependency of CONFIG_SGI_XP upon CONFIG_SGI_GRU to Kconfig.

Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoadd dependency of CONFIG_SGI_XP upon CONFIG_NET
Dean Nelson [Fri, 1 Aug 2008 19:19:08 +0000 (14:19 -0500)] 
add dependency of CONFIG_SGI_XP upon CONFIG_NET

Add a dependency of CONFIG_SGI_XP upon CONFIG_NET to Kconfig.

Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agotracehook: fix exit_signal=0 case
Roland McGrath [Thu, 31 Jul 2008 09:04:09 +0000 (02:04 -0700)] 
tracehook: fix exit_signal=0 case

My commit 2b2a1ff64afbadac842bbc58c5166962cf4f7664 introduced a regression
(sorry about that) for the odd case of exit_signal=0 (e.g. clone_flags=0).
This is not a normal use, but it's used by a case in the glibc test suite.

Dying with exit_signal=0 sends no signal, but it's supposed to wake up a
parent's blocked wait*() calls (unlike the delayed_group_leader case).
This fixes tracehook_notify_death() and its caller to distinguish a
"signal 0" wakeup from the delayed_group_leader case (with no wakeup).

Signed-off-by: Roland McGrath <roland@redhat.com>
Tested-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'for-linus' of git://neil.brown.name/md
Linus Torvalds [Fri, 1 Aug 2008 18:56:07 +0000 (11:56 -0700)] 
Merge branch 'for-linus' of git://neil.brown.name/md

* 'for-linus' of git://neil.brown.name/md:
  md: raid10: wake up frozen array
  md: do not count blocked devices as spares
  md: do not progress the resync process if the stripe was blocked
  md: delay notification of 'active_idle' to the recovery thread
  md: fix merge error
  md: move async_tx_issue_pending_all outside spin_lock_irq

15 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfashe...
Linus Torvalds [Fri, 1 Aug 2008 18:54:05 +0000 (11:54 -0700)] 
Merge branch 'upstream-linus' of git://git./linux/kernel/git/mfasheh/ocfs2

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2:
  [PATCH] ocfs2: Release mutex in error handling code
  [PATCH] ocfs2: Fix oops when racing files truncates with writes into an mmap region
  [PATCH 2/2] ocfs2: Fix race between mount and recovery
  [PATCH 1/2] ocfs2: Add counter in struct ocfs2_dinode to track journal replays
  [PATCH] configfs: Convenience macros for attribute definition.
  [PATCH] configfs: Pin configfs subsystems separately from new config_items.
  [PATCH] configfs: Fix open directory making rmdir() fail
  [PATCH] configfs: Lock new directory inodes before removing on cleanup after failure
  [PATCH] configfs: Prevent userspace from creating new entries under attaching directories
  [PATCH] configfs: Fix failing symlink() making rmdir() fail
  [PATCH] configfs: Fix symlink() to a removing item
  [PATCH] configfs: Include linux/err.h in linux/configfs.h