openwrt/staging/blogic.git
12 years agomwifiex: improve scan delay logic during Tx traffic
Amitkumar Karwar [Sat, 4 Aug 2012 01:06:04 +0000 (18:06 -0700)]
mwifiex: improve scan delay logic during Tx traffic

Earlier scan command was postponed by 20msec multiple times to give
preference to Tx traffic until we find empty wmm queue.

There is a corner case in which wmm queue becomes empty immediately
after processing the packet(before 20msec) and there may be next
packet coming after some time. In this case we should not resume scan
operation.

We will use new flag to check Tx traffic and resume scan operation
only if there is no traffic for 200msec.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agomwifiex: update max_chan_per_scan correctly for SSID scan
Amitkumar Karwar [Sat, 4 Aug 2012 01:06:03 +0000 (18:06 -0700)]
mwifiex: update max_chan_per_scan correctly for SSID scan

As per recent patch "658f37b mwifiex: scan less channels..."
less channels are scanned per scan command in associated state.

Default number of channels per scan command for normal scan are
already 4, but those are 14 for SSID specific scan operation.

This code change in this patch is required for SSID specific scan.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agomwifiex: do le_to_cpu conversion for Rx packet header elements
Amitkumar Karwar [Sat, 4 Aug 2012 01:06:02 +0000 (18:06 -0700)]
mwifiex: do le_to_cpu conversion for Rx packet header elements

We do similar conversion for Tx packet header but it was missing
for Rx packet header. Without this fix, Rx packet header won't
be correctly parsed on big endian platform.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agomwifiex: correction in MSDU padding logic
Yogesh Ashok Powar [Sat, 4 Aug 2012 01:06:01 +0000 (18:06 -0700)]
mwifiex: correction in MSDU padding logic

Padding arithmetic will always work for MSDUs provided first MSDU
ends on 4-byte boundary. Fixing it by making sure that all MSDU ends
on 4-byte boundary.

Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agomwifiex: remove extra padding to AMSDU
Yogesh Ashok Powar [Sat, 4 Aug 2012 01:06:00 +0000 (18:06 -0700)]
mwifiex: remove extra padding to AMSDU

Since commit: fb3c19bc9687d972b83faf366b38ac807eca8f25, adding
extra padding to AMSDU is redundant since same is being performed at
mwifiex_11n_aggregate_pkt after forming the AMSDU packet. Fixing it
by removing it.

Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agomwifiex: copy MSDU subframes correctly
Yogesh Ashok Powar [Sat, 4 Aug 2012 01:05:59 +0000 (18:05 -0700)]
mwifiex: copy MSDU subframes correctly

All MSDU subframes, except the first one in AMSDU, were being written
wrongly at the location of first MSDU frame. Fixing that by copying
the MSDU at skb->tail of AMSDU.

Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agoath9k: fix indentation in ar9003_hw_set_power_per_rate_table
Gabor Juhos [Thu, 2 Aug 2012 14:00:52 +0000 (16:00 +0200)]
ath9k: fix indentation in ar9003_hw_set_power_per_rate_table

The current indentation is off by one tab.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agoath9k: optimize power level initialization for CTL_[25]GHT20
Gabor Juhos [Thu, 2 Aug 2012 14:00:51 +0000 (16:00 +0200)]
ath9k: optimize power level initialization for CTL_[25]GHT20

The first part of the power array is initialized in a loop
and the last two values are initialized separately. Extend
the loop to cover the last two items, and remove the separate
initialization.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agocarl9170: Add support for NL80211_IFTYPE_MESH_POINT interfaces
Javier Lopez [Fri, 27 Jul 2012 18:27:25 +0000 (11:27 -0700)]
carl9170: Add support for NL80211_IFTYPE_MESH_POINT interfaces

This patch contains following modifications:

- Add mesh capabilities on fw.c to permit creation of mesh
  interfaces using this driver.

- Modify carl9170_set_operating_mode, to use AP-style beaconing
  with mesh interfaces.

- Allow beacon updates for NL80211_IFTYPE_MESH_POINT type in
  carl9170_handle_command_response.

- Add NL80211_IFTYPE_MESH_POINT case on carl9170_op_add_interfaces to
  support mesh/ap/sta virtual interface combinations.

Signed-off-by: Javier Lopez <jlopex@cozybit.com>
Acked-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agob43: N-PHY: add overriding RF control for rev7+
Rafał Miłecki [Thu, 26 Jul 2012 08:25:09 +0000 (10:25 +0200)]
b43: N-PHY: add overriding RF control for rev7+

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agob43: be more user friendly with PHY info
Rafał Miłecki [Thu, 26 Jul 2012 06:16:01 +0000 (08:16 +0200)]
b43: be more user friendly with PHY info

First of all, use PHY names instead of magic numbers. It should make
configuring kernel easier in case of not enabled PHY type support.
Secondly, always print info about PHY. This is really basic info about
hardware and quite important for the support level.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agob43: N-PHY: init 0x2057 radio
Rafał Miłecki [Wed, 25 Jul 2012 22:07:39 +0000 (00:07 +0200)]
b43: N-PHY: init 0x2057 radio

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agob43: N-PHY: finish 0x2056 radio setup
Rafał Miłecki [Wed, 25 Jul 2012 22:07:38 +0000 (00:07 +0200)]
b43: N-PHY: finish 0x2056 radio setup

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agob43: add helper waiting for given value in radio reg
Rafał Miłecki [Wed, 25 Jul 2012 22:07:37 +0000 (00:07 +0200)]
b43: add helper waiting for given value in radio reg

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agob43: rename host flags defines
Rafał Miłecki [Wed, 25 Jul 2012 14:58:38 +0000 (16:58 +0200)]
b43: rename host flags defines

There are more than 3 registers on new hardware. Host flags handling has
to be rewritten, as we can't use u128 type to handle all 5 regs.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agop54pci: convert driver to use asynchronous firmware loading
Larry Finger [Tue, 24 Jul 2012 19:16:42 +0000 (21:16 +0200)]
p54pci: convert driver to use asynchronous firmware loading

Drivers that load firmware from their probe routine have problems with the
latest versions of udev as they get timeouts while waiting for user
space to start. The problem is fixed by using request_firmware_nowait()
and delaying the start of mac80211 until the firmware is loaded.

To prevent the possibility of the driver being unloaded while the firmware
loading callback is still active, a completion queue entry is used.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agob43: N-PHY: add PHY rev7+ workarounds
Rafał Miłecki [Tue, 24 Jul 2012 17:18:20 +0000 (19:18 +0200)]
b43: N-PHY: add PHY rev7+ workarounds

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agob43: N-PHY: update workarounds
Rafał Miłecki [Tue, 24 Jul 2012 17:18:19 +0000 (19:18 +0200)]
b43: N-PHY: update workarounds

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agobrcmsmac: use ieee80211_channel_to_frequency
Johannes Berg [Tue, 24 Jul 2012 15:35:57 +0000 (17:35 +0200)]
brcmsmac: use ieee80211_channel_to_frequency

Instead of hard-coding almost the same functionality,
just use ieee80211_channel_to_frequency() directly.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
12 years agoipv4: Properly purge netdev references on uncached routes.
David S. Miller [Tue, 31 Jul 2012 22:06:50 +0000 (15:06 -0700)]
ipv4: Properly purge netdev references on uncached routes.

When a device is unregistered, we have to purge all of the
references to it that may exist in the entire system.

If a route is uncached, we currently have no way of accomplishing
this.

So create a global list that is scanned when a network device goes
down.  This mirrors the logic in net/core/dst.c's dst_ifdown().

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv4: Cache routes in nexthop exception entries.
David S. Miller [Tue, 31 Jul 2012 22:02:02 +0000 (15:02 -0700)]
ipv4: Cache routes in nexthop exception entries.

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv4: percpu nh_rth_output cache
Eric Dumazet [Tue, 31 Jul 2012 05:45:30 +0000 (05:45 +0000)]
ipv4: percpu nh_rth_output cache

Input path is mostly run under RCU and doesnt touch dst refcnt

But output path on forwarding or UDP workloads hits
badly dst refcount, and we have lot of false sharing, for example
in ipv4_mtu() when reading rt->rt_pmtu

Using a percpu cache for nh_rth_output gives a nice performance
increase at a small cost.

24 udpflood test on my 24 cpu machine (dummy0 output device)
(each process sends 1.000.000 udp frames, 24 processes are started)

before : 5.24 s
after : 2.06 s
For reference, time on linux-3.5 : 6.60 s

Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv4: Restore old dst_free() behavior.
Eric Dumazet [Tue, 31 Jul 2012 01:08:23 +0000 (01:08 +0000)]
ipv4: Restore old dst_free() behavior.

commit 404e0a8b6a55 (net: ipv4: fix RCU races on dst refcounts) tried
to solve a race but added a problem at device/fib dismantle time :

We really want to call dst_free() as soon as possible, even if sockets
still have dst in their cache.
dst_release() calls in free_fib_info_rcu() are not welcomed.

Root of the problem was that now we also cache output routes (in
nh_rth_output), we must use call_rcu() instead of call_rcu_bh() in
rt_free(), because output route lookups are done in process context.

Based on feedback and initial patch from David Miller (adding another
call_rcu_bh() call in fib, but it appears it was not the right fix)

I left the inet_sk_rx_dst_set() helper and added __rcu attributes
to nh_rth_output and nh_rth_input to better document what is going on in
this code.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agobridge: make port attributes const
stephen hemminger [Mon, 30 Jul 2012 08:55:49 +0000 (08:55 +0000)]
bridge: make port attributes const

Simple table that can be marked const.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv4: remove rt_cache_rebuild_count
Eric Dumazet [Mon, 30 Jul 2012 07:14:29 +0000 (07:14 +0000)]
ipv4: remove rt_cache_rebuild_count

After IP route cache removal, rt_cache_rebuild_count is no longer
used.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: ipv4: fix RCU races on dst refcounts
Eric Dumazet [Sun, 29 Jul 2012 23:20:37 +0000 (23:20 +0000)]
net: ipv4: fix RCU races on dst refcounts

commit c6cffba4ffa2 (ipv4: Fix input route performance regression.)
added various fatal races with dst refcounts.

crashes happen on tcp workloads if routes are added/deleted at the same
time.

The dst_free() calls from free_fib_info_rcu() are clearly racy.

We need instead regular dst refcounting (dst_release()) and make
sure dst_release() is aware of RCU grace periods :

Add DST_RCU_FREE flag so that dst_release() respects an RCU grace period
before dst destruction for cached dst

Introduce a new inet_sk_rx_dst_set() helper, using atomic_inc_not_zero()
to make sure we dont increase a zero refcount (On a dst currently
waiting an rcu grace period before destruction)

rt_cache_route() must take a reference on the new cached route, and
release it if was not able to install it.

With this patch, my machines survive various benchmarks.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: TCP early demux cleanup
Eric Dumazet [Sun, 29 Jul 2012 21:06:13 +0000 (21:06 +0000)]
net: TCP early demux cleanup

early_demux() handlers should be called in RCU context, and as we
use skb_dst_set_noref(skb, dst), caller must not exit from RCU context
before dst use (skb_dst(skb)) or release (skb_drop(dst))

Therefore, rcu_read_lock()/rcu_read_unlock() pairs around
->early_demux() are confusing and not needed :

Protocol handlers are already in an RCU read lock section.
(__netif_receive_skb() does the rcu_read_lock() )

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotun: Fix formatting.
David S. Miller [Mon, 30 Jul 2012 21:52:48 +0000 (14:52 -0700)]
tun: Fix formatting.

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet/tun: fix ioctl() based info leaks
Mathias Krause [Sun, 29 Jul 2012 19:45:14 +0000 (19:45 +0000)]
net/tun: fix ioctl() based info leaks

The tun module leaks up to 36 bytes of memory by not fully initializing
a structure located on the stack that gets copied to user memory by the
TUNGETIFF and SIOCGIFHWADDR ioctl()s.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotg3: Update version to 3.124
Michael Chan [Sun, 29 Jul 2012 19:15:45 +0000 (19:15 +0000)]
tg3: Update version to 3.124

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotg3: Fix race condition in tg3_get_stats64()
Michael Chan [Sun, 29 Jul 2012 19:15:44 +0000 (19:15 +0000)]
tg3: Fix race condition in tg3_get_stats64()

Spinlock should be taken before checking for tp->hw_stats.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotg3: Add New 5719 Read DMA workaround
Michael Chan [Sun, 29 Jul 2012 19:15:43 +0000 (19:15 +0000)]
tg3: Add New 5719 Read DMA workaround

After Power-on-reset, the 5719's TX DMA length registers may contain
uninitialized values and cause TX DMA to stall.  Check for invalid
values and set a register bit to flush the TX channels.  The bit
needs to be turned off after the DMA channels have been flushed.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotg3: Fix Read DMA workaround for 5719 A0.
Michael Chan [Sun, 29 Jul 2012 19:15:42 +0000 (19:15 +0000)]
tg3: Fix Read DMA workaround for 5719 A0.

The workaround was mis-applied to all 5719 and 5720 chips.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotg3: Request APE_LOCK_PHY before PHY access
Michael Chan [Sun, 29 Jul 2012 19:15:41 +0000 (19:15 +0000)]
tg3: Request APE_LOCK_PHY before PHY access

to prevent PHY access conflict with APE firmware.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv6: fix incorrect route 'expires' value passed to userspace
Li Wei [Sun, 29 Jul 2012 16:01:30 +0000 (16:01 +0000)]
ipv6: fix incorrect route 'expires' value passed to userspace

When userspace use RTM_GETROUTE to dump route table, with an already
expired route entry, we always got an 'expires' value(2147157)
calculated base on INT_MAX.

The reason of this problem is in the following satement:
rt->dst.expires - jiffies < INT_MAX
gcc promoted the type of both sides of '<' to unsigned long, thus
a small negative value would be considered greater than INT_MAX.

With the help of Eric Dumazet, do the out of bound checks in
rtnl_put_cacheinfo(), _after_ conversion to clock_t.

Signed-off-by: Li Wei <lw@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agomISDN: Bugfix only few bytes are transfered on a connection
Karsten Keil [Sun, 29 Jul 2012 07:15:13 +0000 (07:15 +0000)]
mISDN: Bugfix only few bytes are transfered on a connection

The test for the fillempty condition was wrong in one place.
Changed the variable to the right boolean type.

Signed-off-by: Karsten Keil <keil@b1-systems.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoseeq: use PTR_RET at init_module of driver
Devendra Naga [Sun, 29 Jul 2012 03:28:47 +0000 (03:28 +0000)]
seeq: use PTR_RET at init_module of driver

the driver sees wether the dev_seeq pointer is having a error that can be
read by using the PTR_ERR, and returns it at error case, other wise 0 at
success case.

the PTR_RET does the same thing, and use PTR_RET instead of redoing the
code of PTR_RET

Signed-off-by: Devendra Naga <develkernel412222@gmail.com>
Acked-by: David Howells <dhowells@redhat.com>
Reviewed-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agobnx2x: remove cast around the kmalloc in bnx2x_prev_mark_path
Devendra Naga [Sun, 29 Jul 2012 03:19:23 +0000 (03:19 +0000)]
bnx2x: remove cast around the kmalloc in bnx2x_prev_mark_path

casting the void pointer is redundant (Documentation/CodingStyle)

Signed-off-by: Devendra Naga <develkernel412222@gmail.com>
Acked-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv4: clean up put_child
Lin Ming [Sun, 29 Jul 2012 02:00:03 +0000 (02:00 +0000)]
ipv4: clean up put_child

The first parameter struct trie *t is not used anymore.
Remove it.

Signed-off-by: Lin Ming <mlin@ss.pku.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv4: fix debug info in tnode_new
Lin Ming [Sun, 29 Jul 2012 01:19:55 +0000 (01:19 +0000)]
ipv4: fix debug info in tnode_new

It should print size of struct rt_trie_node * allocated instead of size
of struct rt_trie_node.

Signed-off-by: Lin Ming <mlin@ss.pku.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoqlge: Add offload features to vlan interfaces
brenohl@br.ibm.com [Fri, 27 Jul 2012 08:54:52 +0000 (08:54 +0000)]
qlge: Add offload features to vlan interfaces

This patch fills the net_device vlan_features with the proper hardware features,
thus, improving the vlan interface performance.

With the patch applied, I can see around 148% improvement on a TCP_STREAM test,
from 3.5 Gb/s to 8.7 Gb/s. On TCP_RR, I see a 11% improvement, from 18k
to 20. The CPU utilization is almost the same on both cases, from the comparison
above.

Signed-off-by: Breno Leitao <brenohl@br.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Sat, 28 Jul 2012 13:00:39 +0000 (06:00 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:
 "Several bug fixes, some to new features appearing in this merge
  window, some that have been around for a while.

  I have a short list of known problems that need to be sorted out, but
  all of them can be solved easily during the run up to 3.6-final.

  I'll be offline until Sunday afternoon, but nothing need hold up
  3.6-rc1 and the close of the merge window, networking wise, at this
  point.

  1) Fix interface check in ipv4 TCP early demux, from Eric Dumazet.

  2) Fix a long standing bug in TCP DMA to userspace offload that can
     hang applications using MSG_TRUNC, from Jiri Kosina.

  3) Don't allow TCP_USER_TIMEOUT to be negative, from Hangbin Liu.

  4) Don't use GFP_KERNEL under spinlock in kaweth driver, from Dan
     Carpenter"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  tcp: perform DMA to userspace only if there is a task waiting for it
  Revert "openvswitch: potential NULL deref in sample()"
  ipv4: fix TCP early demux
  net: fix rtnetlink IFF_PROMISC and IFF_ALLMULTI handling
  USB: kaweth.c: use GFP_ATOMIC under spin_lock
  tcp: Add TCP_USER_TIMEOUT negative value check
  bcma: add missing iounmap on error path
  bcma: fix regression in interrupt assignment on mips
  mac80211_hwsim: fix possible race condition in usage of info->control.sta & control.vif

12 years agoMerge tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso...
Linus Torvalds [Sat, 28 Jul 2012 03:52:25 +0000 (20:52 -0700)]
Merge tag 'ext4_for_linus' of git://git./linux/kernel/git/tytso/ext4

Pull ext4 updates from Ted Ts'o:
 "The usual collection of bug fixes and optimizations.  Perhaps of
  greatest note is a speed up for parallel, non-allocating DIO writes,
  since we no longer take the i_mutex lock in that case.

  For bug fixes, we fix an incorrect overhead calculation which caused
  slightly incorrect results for df(1) and statfs(2).  We also fixed
  bugs in the metadata checksum feature."

* tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (23 commits)
  ext4: undo ext4_calc_metadata_amount if we fail to claim space
  ext4: don't let i_reserved_meta_blocks go negative
  ext4: fix hole punch failure when depth is greater than 0
  ext4: remove unnecessary argument from __ext4_handle_dirty_metadata()
  ext4: weed out ext4_write_super
  ext4: remove unnecessary superblock dirtying
  ext4: convert last user of ext4_mark_super_dirty() to ext4_handle_dirty_super()
  ext4: remove useless marking of superblock dirty
  ext4: fix ext4 mismerge back in January
  ext4: remove dynamic array size in ext4_chksum()
  ext4: remove unused variable in ext4_update_super()
  ext4: make quota as first class supported feature
  ext4: don't take the i_mutex lock when doing DIO overwrites
  ext4: add a new nolock flag in ext4_map_blocks
  ext4: split ext4_file_write into buffered IO and direct IO
  ext4: remove an unused statement in ext4_mb_get_buddy_page_lock()
  ext4: fix out-of-date comments in extents.c
  ext4: use s_csum_seed instead of i_csum_seed for xattr block
  ext4: use proper csum calculation in ext4_rename
  ext4: fix overhead calculation used by ext4_statfs()
  ...

12 years agoMerge branch 'for-linus' of git://git.linaro.org/people/rmk/linux-arm
Linus Torvalds [Fri, 27 Jul 2012 22:14:26 +0000 (15:14 -0700)]
Merge branch 'for-linus' of git://git.linaro.org/people/rmk/linux-arm

Pull ARM updates from Russell King:
 "First ARM push of this merge window, post me coming back from holiday.
  This is what has been in linux-next for the last few weeks.  Not much
  to say which isn't described by the commit summaries."

* 'for-linus' of git://git.linaro.org/people/rmk/linux-arm: (32 commits)
  ARM: 7463/1: topology: Update cpu_power according to DT information
  ARM: 7462/1: topology: factorize the update of sibling masks
  ARM: 7461/1: topology: Add arch_scale_freq_power function
  ARM: 7456/1: ptrace: provide separate functions for tracing syscall {entry,exit}
  ARM: 7455/1: audit: move syscall auditing until after ptrace SIGTRAP handling
  ARM: 7454/1: entry: don't bother with syscall tracing on ret_from_fork path
  ARM: 7453/1: audit: only allow syscall auditing for pure EABI userspace
  ARM: 7452/1: delay: allow timer-based delay implementation to be selected
  ARM: 7451/1: arch timer: implement read_current_timer and get_cycles
  ARM: 7450/1: dcache: select DCACHE_WORD_ACCESS for little-endian ARMv6+ CPUs
  ARM: 7449/1: use generic strnlen_user and strncpy_from_user functions
  ARM: 7448/1: perf: remove arm_perf_pmu_ids global enumeration
  ARM: 7447/1: rwlocks: remove unused branch labels from trylock routines
  ARM: 7446/1: spinlock: use ticket algorithm for ARMv6+ locking implementation
  ARM: 7445/1: mm: update CONTEXTIDR register to contain PID of current process
  ARM: 7444/1: kernel: add arch-timer C3STOP feature
  ARM: 7460/1: remove asm/locks.h
  ARM: 7439/1: head.S: simplify initial page table mapping
  ARM: 7437/1: zImage: Allow DTB command line concatenation with ATAG_CMDLINE
  ARM: 7436/1: Do not map the vectors page as write-through on UP systems
  ...

12 years agoMerge branches 'audit', 'delay', 'fixes', 'misc' and 'sta2x11' into for-linus
Russell King [Fri, 27 Jul 2012 22:06:32 +0000 (23:06 +0100)]
Merge branches 'audit', 'delay', 'fixes', 'misc' and 'sta2x11' into for-linus

12 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Fri, 27 Jul 2012 21:25:40 +0000 (14:25 -0700)]
Merge branch 'for-davem' of git://git./linux/kernel/git/linville/wireless

John W. Linville says:

====================
These fixes are intended for the 3.6 stream.

Hauke Mehrtens provides a pair of bcma fixes, one to fix a build
regression on mips and another to correct a pair of missing iounmap
calls.

Thomas Huehn offers a mac80211_hwsim fix to avoid a possible
use-after-free bug.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotcp: perform DMA to userspace only if there is a task waiting for it
Jiri Kosina [Fri, 27 Jul 2012 10:38:50 +0000 (10:38 +0000)]
tcp: perform DMA to userspace only if there is a task waiting for it

Back in 2006, commit 1a2449a87b ("[I/OAT]: TCP recv offload to I/OAT")
added support for receive offloading to IOAT dma engine if available.

The code in tcp_rcv_established() tries to perform early DMA copy if
applicable. It however does so without checking whether the userspace
task is actually expecting the data in the buffer.

This is not a problem under normal circumstances, but there is a corner
case where this doesn't work -- and that's when MSG_TRUNC flag to
recvmsg() is used.

If the IOAT dma engine is not used, the code properly checks whether
there is a valid ucopy.task and the socket is owned by userspace, but
misses the check in the dmaengine case.

This problem can be observed in real trivially -- for example 'tbench' is a
good reproducer, as it makes a heavy use of MSG_TRUNC. On systems utilizing
IOAT, you will soon find tbench waiting indefinitely in sk_wait_data(), as they
have been already early-copied in tcp_rcv_established() using dma engine.

This patch introduces the same check we are performing in the simple
iovec copy case to the IOAT case as well. It fixes the indefinite
recvmsg(MSG_TRUNC) hangs.

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoRevert "openvswitch: potential NULL deref in sample()"
Jesse Gross [Fri, 27 Jul 2012 10:19:40 +0000 (10:19 +0000)]
Revert "openvswitch: potential NULL deref in sample()"

This reverts commit 5b3e7e6cb5771bedda51cdb6f715d1da8cd9e644.

The problem that the original commit was attempting to fix can
never happen in practice because validation is done one a per-flow
basis rather than a per-packet basis.  Adding additional checks at
runtime is unnecessary and inconsistent with the rest of the code.

CC: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv4: fix TCP early demux
Eric Dumazet [Fri, 27 Jul 2012 06:23:40 +0000 (06:23 +0000)]
ipv4: fix TCP early demux

commit 92101b3b2e317 (ipv4: Prepare for change of rt->rt_iif encoding.)
invalidated TCP early demux, because rx_dst_ifindex is not properly
initialized and checked.

Also remove the use of inet_iif(skb) in favor or skb->skb_iif

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: fix rtnetlink IFF_PROMISC and IFF_ALLMULTI handling
Jiri Benc [Fri, 27 Jul 2012 02:58:22 +0000 (02:58 +0000)]
net: fix rtnetlink IFF_PROMISC and IFF_ALLMULTI handling

When device flags are set using rtnetlink, IFF_PROMISC and IFF_ALLMULTI
flags are handled specially. Function dev_change_flags sets IFF_PROMISC and
IFF_ALLMULTI bits in dev->gflags according to the passed value but
do_setlink passes a result of rtnl_dev_combine_flags which takes those bits
from dev->flags.

This can be easily trigerred by doing:

tcpdump -i eth0 &
ip l s up eth0

ip sets IFF_UP flag in ifi_flags and ifi_change, which is combined with
IFF_PROMISC by rtnl_dev_combine_flags, causing __dev_change_flags to set
IFF_PROMISC in gflags.

Reported-by: Max Matveev <makc@redhat.com>
Signed-off-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoUSB: kaweth.c: use GFP_ATOMIC under spin_lock
Dan Carpenter [Fri, 27 Jul 2012 01:46:51 +0000 (01:46 +0000)]
USB: kaweth.c: use GFP_ATOMIC under spin_lock

The problem is that we call this with a spin lock held.  The call tree
is:
kaweth_start_xmit() holds kaweth->device_lock.
-> kaweth_async_set_rx_mode()
   -> kaweth_control()
      -> kaweth_internal_control_msg()

The kaweth_internal_control_msg() function is only called from
kaweth_control() which used GFP_ATOMIC for its allocations.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotcp: Add TCP_USER_TIMEOUT negative value check
Hangbin Liu [Thu, 26 Jul 2012 22:52:21 +0000 (22:52 +0000)]
tcp: Add TCP_USER_TIMEOUT negative value check

TCP_USER_TIMEOUT is a TCP level socket option that takes an unsigned int. But
patch "tcp: Add TCP_USER_TIMEOUT socket option"(dca43c75) didn't check the negative
values. If a user assign -1 to it, the socket will set successfully and wait
for 4294967295 miliseconds. This patch add a negative value check to avoid
this issue.

Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge tag 'tty-3.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Fri, 27 Jul 2012 19:52:03 +0000 (12:52 -0700)]
Merge tag 'tty-3.6-rc1' of git://git./linux/kernel/git/gregkh/tty

Pull TTY/Serial patches from Greg Kroah-Hartman:
 "Here's the "tiny" set of patches for 3.6-rc1 for the tty layer and
  serial drivers.  They were cherry-picked from the tty-next branch of
  the tty git tree, as they are small and "obvious" fixes.  The larger
  changes, as mentioned before, will be saved for the 3.7-rc1 merge
  window.

  All of these changes have been in the linux-next releases for quite a
  while.

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>"
* tag 'tty-3.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  pch_uart: Fix parity setting issue
  pch_uart: Fix rx error interrupt setting issue
  pch_uart: Fix missing break for 16 byte fifo
  tty ldisc: Close/Reopen race prevention should check the proper flag
  pch_uart: Add eg20t_port lock field, avoid recursive spinlocks
  vt: fix race in vt_waitactive()
  serial/of-serial: Add LPC3220 standard UART compatible string
  serial/8250: Add LPC3220 standard UART type
  serial_core: Update buffer overrun statistics.
  serial: samsung: Fixed wrong comparison for baudclk_rate

12 years agoMerge branch 'kmap_atomic' of git://github.com/congwang/linux
Linus Torvalds [Fri, 27 Jul 2012 18:26:48 +0000 (11:26 -0700)]
Merge branch 'kmap_atomic' of git://github.com/congwang/linux

Pull final kmap_atomic cleanups from Cong Wang:
 "This should be the final round of cleanup, as the definitions of enum
  km_type finally get removed from the whole tree.  The patches have
  been in linux-next for a long time."

* 'kmap_atomic' of git://github.com/congwang/linux:
  pipe: remove KM_USER0 from comments
  vmalloc: remove KM_USER0 from comments
  feature-removal-schedule.txt: remove kmap_atomic(page, km_type)
  tile: remove km_type definitions
  um: remove km_type definitions
  asm-generic: remove km_type definitions
  avr32: remove km_type definitions
  frv: remove km_type definitions
  powerpc: remove km_type definitions
  arm: remove km_type definitions
  highmem: remove the deprecated form of kmap_atomic
  tile: remove usage of enum km_type
  frv: remove the second parameter of kmap_atomic_primary()
  jbd2: remove the second argument of kmap_atomic

12 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Fri, 27 Jul 2012 15:35:26 +0000 (08:35 -0700)]
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

Pull powerpc fixes from Benjamin Herrenschmidt:
 "Here's a handful of powerpc patches, a couple of regression fixes for
  problems introduced in the main batch in this merge window, a couple
  of defconfig updates, and some trivials.

  The radeonfb one is something that was long standing in SLES which I
  forgot to pickup earlier."

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc/ftrace: Trace function graph entry before updating index
  radeonfb: Add quirk for the graphics adapter in some JSxx
  powerpc: Lack of firmware flash support is not an error
  powerpc: Enable pseries hardware RNG and crypto modules
  powerpc: Update g5_defconfig
  powerpc/kvm/bookehv: Fix build regression
  powerpc: Set stack limit properly in crit_transfer_to_handler

12 years agoMerge tag 'cpumask-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rusty...
Linus Torvalds [Fri, 27 Jul 2012 15:34:16 +0000 (08:34 -0700)]
Merge tag 'cpumask-for-linus' of git://git./linux/kernel/git/rusty/linux-2.6-for-linus

Pull cpumask changes from Rusty Russell:
 "Trivial comment changes to cpumask code.  I guess it's getting boring."

Boring is good.

* tag 'cpumask-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
  cpumask: cpulist_parse() comments correction
  init: add comments to keep initcall-names in sync with initcall levels
  cpumask: add a few comments of cpumask functions

12 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
John W. Linville [Fri, 27 Jul 2012 15:15:03 +0000 (11:15 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless into for-davem

12 years agoMerge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
Linus Torvalds [Fri, 27 Jul 2012 03:29:52 +0000 (20:29 -0700)]
Merge tag 'fixes-for-linus' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Olof Johansson:
 "A mixed bag of fixes, some for merge window fallout (tegra, MXS), and
  a short series of fixes for marvell platforms that didn't make it in
  before 3.5."

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  ARM: mxs: fix compile error caused by prom_update_property change
  ARM: dt: tegra trimslice: enable USB2 port
  ARM: dt: tegra trimslice: add vbus-gpio property
  ARM: vt8500: Add maintainer for VT8500 architecture
  ARM: Kirkwood: Replace mrvl with marvell
  ARM: Orion: fix driver probe error handling with respect to clk
  ARM: Dove: Fixup ge00 initialisation
  ARM: Kirkwood: Fix PHY disable clk problems
  ARM: Kirkwood: Ensure runit clock always ticks.
  ARM: versatile: Don't use platform clock for Integrator & VE
  ARM: tegra: harmony: add regulator supply name and its input supply

12 years agoMerge branch 'for-3.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/cooloney...
Linus Torvalds [Fri, 27 Jul 2012 03:26:27 +0000 (20:26 -0700)]
Merge branch 'for-3.6-rc1' of git://git./linux/kernel/git/cooloney/linux-leds

Pull LED subsystem update from Bryan Wu.

* 'for-3.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/cooloney/linux-leds: (50 commits)
  leds-lp8788: forgotten unlock at lp8788_led_work
  LEDS: propagate error codes in blinkm_detect()
  LEDS: memory leak in blinkm_led_common_set()
  leds: add new lp8788 led driver
  LEDS: add BlinkM RGB LED driver, documentation and update MAINTAINERS
  leds: max8997: Simplify max8997_led_set_mode implementation
  leds/leds-s3c24xx: use devm_gpio_request
  leds: convert Network Space v2 LED driver to devm_kzalloc() and cleanup error exit path
  leds: convert DAC124S085 LED driver to devm_kzalloc()
  leds: convert LM3530 LED driver to devm_kzalloc() and cleanup error exit path
  leds: convert TCA6507 LED driver to devm_kzalloc()
  leds: convert Freescale MC13783 LED driver to devm_kzalloc() and cleanup error exit path
  leds: convert ADP5520 LED driver to devm_kzalloc() and cleanup error exit path
  leds: convert PCA955x LED driver to devm_kzalloc() and cleanup error exit path
  leds: convert Sun Fire LED driver to devm_kzalloc() and cleanup error exit path
  leds: convert PCA9532 LED driver to devm_kzalloc()
  leds: convert LT3593 LED driver to devm_kzalloc()
  leds: convert Renesas TPU LED driver to devm_kzalloc() and cleanup error exit path
  leds: convert LP5523 LED driver to devm_kzalloc() and cleanup error exit path
  leds: convert PCA9633 LED driver to devm_kzalloc()
  ...

12 years agopowerpc/ftrace: Trace function graph entry before updating index
Steven Rostedt [Wed, 18 Jul 2012 12:35:28 +0000 (12:35 +0000)]
powerpc/ftrace: Trace function graph entry before updating index

As Colin Cross ported my x86 change to ARM, he also pointed out that
powerpc is also behind in this fix.

The commit 722b3c746953 "ftrace/graph: Trace function entry before
updating index" fixes an issue with function graph tracing for x86,
where if the called entry function decides not to trace interrupts, it
can fail the check if an interrupt comes in just after the
curr_ret_stack is updated.

The solution is to call the entry function first, then update the
curr_ret_stack if the entry function wants to be traced.

Cc: Colin Cross <ccross@android.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agoradeonfb: Add quirk for the graphics adapter in some JSxx
Tony Breeds [Wed, 18 Jul 2012 06:49:04 +0000 (06:49 +0000)]
radeonfb: Add quirk for the graphics adapter in some JSxx

These devices are set to 640x480 by firmware, switch them to 800x600@60
so that the graphical installer can run on remote console.

Reported by IBM during SLES10 SP2 beta testing:

https://bugzilla.novell.com/show_bug.cgi?id=461002
LTC50817

Signed-off-by: Olaf Hering <olaf@aepfle.de>
Signed-off-by: Tony Breeds <tony@bakeyournoodle.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Lack of firmware flash support is not an error
Anton Blanchard [Sun, 22 Jul 2012 20:42:32 +0000 (20:42 +0000)]
powerpc: Lack of firmware flash support is not an error

Reduce the severity of the warning given when firmware flash is
not supported. Not all platforms have it.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Enable pseries hardware RNG and crypto modules
Anton Blanchard [Sun, 22 Jul 2012 20:26:51 +0000 (20:26 +0000)]
powerpc: Enable pseries hardware RNG and crypto modules

Enable the hardware RNG and crypto modules. I verified they both
autoload via the VIO subsystem, so there is no need to build them in.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Update g5_defconfig
Benjamin Herrenschmidt [Sun, 22 Jul 2012 16:48:32 +0000 (16:48 +0000)]
powerpc: Update g5_defconfig

This updates the g5 defconfig to include nouveau instead of nvidiafb
(which works much better nowadays, in fact the latter crashes on modern
distros), and to set CONFIG_VT_HW_CONSOLE_BINDING without which takeover
from the firmware offb by nouveau doesn't work properly (and leads to
unexplained black screens for some users).

The rest is churn of going through defconfig / savedefconfig

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/kvm/bookehv: Fix build regression
Alexander Graf [Tue, 24 Jul 2012 13:02:34 +0000 (13:02 +0000)]
powerpc/kvm/bookehv: Fix build regression

After merging the register type check patches from Ben's tree, the
hv enabled booke implementation ceased to compile.

This patch fixes things up so everyone's happy again.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Set stack limit properly in crit_transfer_to_handler
Stuart Yoder [Mon, 16 Jul 2012 09:06:48 +0000 (09:06 +0000)]
powerpc: Set stack limit properly in crit_transfer_to_handler

Commit 9778b696a0188ad3b3524b383953ee73b31b7b68 incorrectly
changes the code setting the stack limit on entry to the
kernel to mark the thread_info at the bottom of the stack
out of bounds anymore. This fixes it.

Signed-off-by: Stuart Yoder <stuart.yoder@freescale.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc
Linus Torvalds [Fri, 27 Jul 2012 01:14:11 +0000 (18:14 -0700)]
Merge git://git./linux/kernel/git/davem/sparc

Pull sparc updates from David Miller:
 "Nothing much this merge window for sparc.

  1) Fix FPU state management in sparc32, from Tkhai Kirill.

  2) More sparc32 mm layer code cleanups, largely more side effects of
     the sun4c support removal in the 3.5 From Sam Ravnborg.

  3) Remove unused code in sparc64, from Bjorn Helgaas and Kirill Tkhai.

  4) Some declaration and comment tidies in PROM support code, from
     Geert Uytterhoeven."

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc: (24 commits)
  sparc32, copy_thread: Clear TIF_USEDFPU flag of created task instead of current
  sparc32: delete dead code in show_mem()
  sparc32: move kmap_init() to highmem.c
  sparc32: move probe_memory() to srmmu.c
  sparc32: drop unused BAD_PAGE stuff
  sparc32: centralize all mmu context handling in srmmu.c
  sparc32: drop quicklist
  sparc32: drop sparc model check in paging_init
  sparc32: drop sparc_unmapped_base
  sparc32,leon: drop leon_init()
  sparc32: drop fixmap.h
  sparc32: fixmap.h cleanup
  sparc32: drop unused kmap_atomic_to_page
  sparc32: drop swapper_pg_dir
  sparc32: beautify srmmu_inherit_prom_mappings()
  sparc32: use void * in nocache get/free
  sparc32: fix coding-style in srmmu.c
  sparc32: sort includes in srmmu.c
  sparc32: define a few srmmu functions __init
  sparc64: remove unused function straddles_64bit_va_hole()
  ...

12 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Fri, 27 Jul 2012 01:09:01 +0000 (18:09 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking updates and fixes from David Miller:

1) Reinstate the no-ref optimization for input route lookups in ipv4 to
   fix some routing cache removal perf regressions.

2) Make TCP socket pre-demux work on ipv6 side too, from Eric Dumazet.

3) Get RX hash value from correct place in be2net driver, from
   Sarveshwar Bandi.

4) Validation of FIB cached routes missing critical check, from Eric
   Dumazet.

5) EEH support in mlx4 driver, from Kleber Sacilotto de Souza.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (23 commits)
  ipv6: Early TCP socket demux
  ipv4: Fix input route performance regression.
  pch_gbe: vlan skb len fix
  pch_gbe: add extra clean tx
  pch_gbe: fix transmit watchdog timeout
  ixgbe: fix panic while dumping packets on Tx hang with IOMMU
  be2net: Fix to parse RSS hash from Receive completions correctly.
  net/mlx4_en: Limit the RFS filter IDs to be < RPS_NO_FILTER
  hyperv: Add error handling to rndis_filter_device_add()
  hyperv: Add a check for ring_size value
  ipv4: rt_cache_valid must check expired routes
  net/pch_gpe: Cannot disable ethernet autonegation
  qeth: repair crash in qeth_l3_vlan_rx_kill_vid()
  netiucv: cleanup attribute usage
  net: wiznet add missing HAS_IOMEM dependency
  be2net: Missing byteswap in be_get_fw_log_level causes oops on PowerPC
  mlx4: Add support for EEH error recovery
  cdc-ncm: tag Ericsson WWAN devices (eg F5521gw) with FLAG_WWAN
  wanmain: comparing array with NULL
  caif: fix NULL pointer check
  ...

12 years agoBtrfs: using vmalloc and friends needs vmalloc.h
Stephen Rothwell [Fri, 27 Jul 2012 00:11:13 +0000 (10:11 +1000)]
Btrfs: using vmalloc and friends needs vmalloc.h

On powerpc, we don't get the implicit vmalloc.h include, and as a result
the build fails noisily:

  fs/btrfs/send.c: In function 'fs_path_free':
  fs/btrfs/send.c:185:4: error: implicit declaration of function 'vfree' [-Werror=implicit-function-declaration]
  fs/btrfs/send.c: In function 'fs_path_ensure_buf':
  fs/btrfs/send.c:215:4: error: implicit declaration of function 'vmalloc' [-Werror=implicit-function-declaration]
  fs/btrfs/send.c:215:12: warning: assignment makes pointer from integer without a cast [enabled by default]
  fs/btrfs/send.c:225:12: warning: assignment makes pointer from integer without a cast [enabled by default]
  fs/btrfs/send.c:233:13: warning: assignment makes pointer from integer without a cast [enabled by default]
  fs/btrfs/send.c: In function 'iterate_dir_item':
  fs/btrfs/send.c:900:10: warning: assignment makes pointer from integer without a cast [enabled by default]
  fs/btrfs/send.c:909:11: warning: assignment makes pointer from integer without a cast [enabled by default]
  fs/btrfs/send.c: In function 'btrfs_ioctl_send':
  fs/btrfs/send.c:4463:17: warning: assignment makes pointer from integer without a cast [enabled by default]
  fs/btrfs/send.c:4469:17: warning: assignment makes pointer from integer without a cast [enabled by default]
  fs/btrfs/send.c:4475:2: error: implicit declaration of function 'vzalloc' [-Werror=implicit-function-declaration]
  fs/btrfs/send.c:4475:20: warning: assignment makes pointer from integer without a cast [enabled by default]
  fs/btrfs/send.c:4483:21: warning: assignment makes pointer from integer without a cast [enabled by default]

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoleds-lp8788: forgotten unlock at lp8788_led_work
Devendra Naga [Thu, 26 Jul 2012 18:00:31 +0000 (02:00 +0800)]
leds-lp8788: forgotten unlock at lp8788_led_work

at default case of switch we return with out unlocking the mutex.

Signed-off-by: Devendra Naga <develkernel412222@gmail.com>
Signed-off-by: Bryan Wu <bryan.wu@canonical.com>
12 years agoLEDS: propagate error codes in blinkm_detect()
Dan Carpenter [Thu, 26 Jul 2012 08:00:25 +0000 (16:00 +0800)]
LEDS: propagate error codes in blinkm_detect()

Use the error codes from the lower levels instead of -ENODEV.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Jan-Simon Möller <jansimon.moeller@gmx.de>
Signed-off-by: Bryan Wu <bryan.wu@canonical.com>
12 years agoLEDS: memory leak in blinkm_led_common_set()
Dan Carpenter [Thu, 26 Jul 2012 07:59:51 +0000 (15:59 +0800)]
LEDS: memory leak in blinkm_led_common_set()

This can return without freeing the bl_work struct.  Also there was no
check for allocation failure.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Jan-Simon Möller <jansimon.moeller@gmx.de>
Signed-off-by: Bryan Wu <bryan.wu@canonical.com>
12 years agocpumask: cpulist_parse() comments correction
Alex Shi [Thu, 26 Jul 2012 23:59:42 +0000 (09:29 +0930)]
cpumask: cpulist_parse() comments correction

As introduced in Rusty's commit 29c0177e6a4, the function has no
parameter @len, so need to remove it from comments to avoid kernel-doc
warning:

alexs@debian:~/linux-next$ scripts/kernel-doc -man
include/linux/cpumask.h | split-man.pl /tmp/man
....
Warning(include/linux/cpumask.h:602): Excess function parameter 'len'
description in 'cpulist_parse'

and correct the function name in comments to cpulist_parse.

Signed-off-by: Alex Shi <alex.shi@intel.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
12 years agoinit: add comments to keep initcall-names in sync with initcall levels
Jim Cromie [Thu, 14 Jun 2012 22:00:59 +0000 (16:00 -0600)]
init: add comments to keep initcall-names in sync with initcall levels

main.c has initcall_level_names[] for parse_args to print in debug messages,
add comments to keep them in sync with initcalls defined in init.h.

Also add "loadable" into comment re not using *_initcall macros in
modules, to disambiguate from kernel/params.c and other builtins.

Signed-off-by: Jim Cromie <jim.cromie@gmail.com>
Acked-by: Borislav Petkov <borislav.petkov@amd.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
12 years agocpumask: add a few comments of cpumask functions
Alex Shi [Mon, 28 May 2012 14:23:51 +0000 (22:23 +0800)]
cpumask: add a few comments of cpumask functions

Current few cpumask functions' purposes are not quite clear. Stupid
user like myself needs to dig into details for clear function
purpose and return value.
Add few explanation for them is helpful.

Thanks for Srivatsa's comments and correction!

Signed-off-by: Alex Shi <alex.shi@intel.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
12 years agosparc32, copy_thread: Clear TIF_USEDFPU flag of created task instead of current
Tkhai Kirill [Fri, 29 Jun 2012 02:55:25 +0000 (02:55 +0000)]
sparc32, copy_thread: Clear TIF_USEDFPU flag of created task instead of current

FPU state is saved into task_struct of created task. Current task continues
use of the state, so it needs TIF_USEDFPU flag is not cleared.

Created task receives fresh FPU and cleared TIF_USEFPU flag is required for it.

Signed-off-by: Kirill Tkhai <tkhai@yandex.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: delete dead code in show_mem()
Sam Ravnborg [Thu, 26 Jul 2012 11:02:28 +0000 (11:02 +0000)]
sparc32: delete dead code in show_mem()

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: move kmap_init() to highmem.c
Sam Ravnborg [Thu, 26 Jul 2012 11:02:27 +0000 (11:02 +0000)]
sparc32: move kmap_init() to highmem.c

Try to keep highmem support in a more central place.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: move probe_memory() to srmmu.c
Sam Ravnborg [Thu, 26 Jul 2012 11:02:26 +0000 (11:02 +0000)]
sparc32: move probe_memory() to srmmu.c

Only one user so move it to the file using it.
It had nothing to do in fault_32.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: drop unused BAD_PAGE stuff
Sam Ravnborg [Thu, 26 Jul 2012 11:02:25 +0000 (11:02 +0000)]
sparc32: drop unused BAD_PAGE stuff

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: centralize all mmu context handling in srmmu.c
Sam Ravnborg [Thu, 26 Jul 2012 11:02:24 +0000 (11:02 +0000)]
sparc32: centralize all mmu context handling in srmmu.c

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: drop quicklist
Sam Ravnborg [Thu, 26 Jul 2012 11:02:23 +0000 (11:02 +0000)]
sparc32: drop quicklist

The quicklist stuff is not used anymore - drop it.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: drop sparc model check in paging_init
Sam Ravnborg [Thu, 26 Jul 2012 11:02:22 +0000 (11:02 +0000)]
sparc32: drop sparc model check in paging_init

We already check the model in head_32.S so no need to
repeat the check here

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: drop sparc_unmapped_base
Sam Ravnborg [Thu, 26 Jul 2012 11:02:21 +0000 (11:02 +0000)]
sparc32: drop sparc_unmapped_base

The base is always the same so no need to use a variable for this.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32,leon: drop leon_init()
Sam Ravnborg [Thu, 26 Jul 2012 11:02:20 +0000 (11:02 +0000)]
sparc32,leon: drop leon_init()

This function was only used to set of_pdt_build_more to leon_node_init().
But the leon_node_init() was a nop as prom_amba_init was never assigned.

Cc: Daniel Hellstrom <daniel@gaisler.com>
Cc: Konrad Eisele <konrad@gaisler.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: drop fixmap.h
Sam Ravnborg [Thu, 26 Jul 2012 11:02:19 +0000 (11:02 +0000)]
sparc32: drop fixmap.h

sparc32 does not support fixmaps - so do not pretend so by
having the fixmap.h file.
Move relevant parts to vaddrs.h.

I looked at simplifying this even more but failed to understand
the reasoning behind the extra guard page involved and due to
missing testing possibilities only the trivial conversion was done.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: fixmap.h cleanup
Sam Ravnborg [Thu, 26 Jul 2012 11:02:18 +0000 (11:02 +0000)]
sparc32: fixmap.h cleanup

Remove all unused stuff from fixmap.h
It is only used for highmem - sparc32 has no fixmap support.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: drop unused kmap_atomic_to_page
Sam Ravnborg [Thu, 26 Jul 2012 11:02:17 +0000 (11:02 +0000)]
sparc32: drop unused kmap_atomic_to_page

No users left of this function - drop it.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: drop swapper_pg_dir
Sam Ravnborg [Thu, 26 Jul 2012 11:02:16 +0000 (11:02 +0000)]
sparc32: drop swapper_pg_dir

We save one page of RAM dropping swapper_pg_dir.
It was only used for an assignment in init-mm.c and we
redid this later in srmmu.c anyway.

This is likely a left-over from the sun4c removal.

To avoid a dummy variable we use a simple #define swapper_pg_dir NULL

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: beautify srmmu_inherit_prom_mappings()
Sam Ravnborg [Thu, 26 Jul 2012 11:02:15 +0000 (11:02 +0000)]
sparc32: beautify srmmu_inherit_prom_mappings()

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: use void * in nocache get/free
Sam Ravnborg [Thu, 26 Jul 2012 11:02:14 +0000 (11:02 +0000)]
sparc32: use void * in nocache get/free

This allowed to us to kill a lot of casts,
with no loss of readability in any places

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: fix coding-style in srmmu.c
Sam Ravnborg [Thu, 26 Jul 2012 11:02:13 +0000 (11:02 +0000)]
sparc32: fix coding-style in srmmu.c

Fix the most annoying issues that distracts me:
- whitespace
- missing space after "if" and "while"
- spaces around operators
and similar simple things.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: sort includes in srmmu.c
Sam Ravnborg [Thu, 26 Jul 2012 11:02:12 +0000 (11:02 +0000)]
sparc32: sort includes in srmmu.c

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosparc32: define a few srmmu functions __init
Sam Ravnborg [Thu, 26 Jul 2012 11:02:11 +0000 (11:02 +0000)]
sparc32: define a few srmmu functions __init

They are only used during early init so lets get rid of them
after init to save some RAM.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge tag 'remoteproc-for-3.6' of git://git.kernel.org/pub/scm/linux/kernel/git/ohad...
Linus Torvalds [Thu, 26 Jul 2012 23:19:08 +0000 (16:19 -0700)]
Merge tag 'remoteproc-for-3.6' of git://git./linux/kernel/git/ohad/remoteproc

Pull remoteproc update from Ohad Ben-Cohen:
 - custom binary format support from Sjur Brændeland
 - groundwork for recovery and runtime pm support
 - some cleanups and API simplifications

Fix up conflicts in drivers/remoteproc/remoteproc_core.c due to clashes
with earlier cleanups by Sjur Brændeland (with part of the cleanups
moved into the new remoteproc_elf_loader.c file).

* tag 'remoteproc-for-3.6' of git://git.kernel.org/pub/scm/linux/kernel/git/ohad/remoteproc:
  MAINTAINERS: add remoteproc's git
  remoteproc: Support custom firmware handlers
  remoteproc: Move Elf related functions to separate file
  remoteproc: Add function rproc_get_boot_addr
  remoteproc: Pass struct fw to load_segments and find_rsc_table.
  remoteproc: adopt the driver core's alloc/add/del/put naming
  remoteproc: remove the get_by_name/put API
  remoteproc: support non-iommu carveout assignment
  remoteproc: simplify unregister/free interfaces
  remoteproc: remove the now-redundant kref
  remoteproc: maintain a generic child device for each rproc
  remoteproc: allocate vrings on demand, free when not needed

12 years agoipv6: Early TCP socket demux
Eric Dumazet [Thu, 26 Jul 2012 12:18:11 +0000 (12:18 +0000)]
ipv6: Early TCP socket demux

This is the IPv6 missing bits for infrastructure added in commit
41063e9dd1195 (ipv4: Early TCP socket demux.)

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv4: Fix input route performance regression.
David S. Miller [Thu, 26 Jul 2012 11:14:38 +0000 (11:14 +0000)]
ipv4: Fix input route performance regression.

With the routing cache removal we lost the "noref" code paths on
input, and this can kill some routing workloads.

Reinstate the noref path when we hit a cached route in the FIB
nexthops.

With help from Eric Dumazet.

Reported-by: Alexander Duyck <alexander.duyck@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
Linus Torvalds [Thu, 26 Jul 2012 21:48:55 +0000 (14:48 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

Pull large btrfs update from Chris Mason:
 "This pull request is very large, and the two main features in here
  have been under testing/devel for quite a while.

  We have subvolume quotas from the strato developers.  This enables
  full tracking of how many blocks are allocated to each subvolume (and
  all snapshots) and you can set limits on a per-subvolume basis.  You
  can also create quota groups and toss multiple subvolumes into a big
  group.  It's everything you need to be a web hosting company and give
  each user their own subvolume.

  The userland side of the quotas is being refreshed, they'll send out
  details on where to grab it soon.

  Next is the kernel side of btrfs send/receive from Alexander Block.
  This leverages the same infrastructure as the quota code to figure out
  relationships between blocks and their owners.  It can then compute
  the difference between two snapshots and sends the diffs in a neutral
  format into userland.

  The basic model:

        create a snapshot
        send that snapshot as the initial backup
        make changes
        create a second snapshot
        send the incremental as a backup
        delete the first snapshot
        (use the second snapshot for the next incremental)

  The receive portion is all in userland, and in the 'next' branch of my
  btrfs-progs repo.

  There's still some work to do in terms of optimizing the send side
  from kernel to userland.  The really important part is figuring out
  how two snapshots are different, and this is where we are
  concentrating right now.  The initial send of a dataset is a little
  slower than tar, but the incremental sends are dramatically faster
  than what rsync can do.

  On top of all of that, we have a nice queue of fixes, cleanups and
  optimizations."

Fix up trivial modify/del conflict in fs/btrfs/ioctl.c

Also fix up semantic conflict in fs/btrfs/send.c: the interface to
dentry_open() changed in commit 765927b2d508 ("switch dentry_open() to
struct path, make it grab references itself"), and since it now grabs
whatever references it needs, we should no longer do the mntget() on the
mnt (and we need to dput() the dentry reference we took).

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (65 commits)
  Btrfs: uninit variable fixes in send/receive
  Btrfs: introduce BTRFS_IOC_SEND for btrfs send/receive
  Btrfs: add btrfs_compare_trees function
  Btrfs: introduce subvol uuids and times
  Btrfs: make iref_to_path non static
  Btrfs: add a barrier before a waitqueue_active check
  Btrfs: call the ordered free operation without any locks held
  Btrfs: Check INCOMPAT flags on remount and add helper function
  Btrfs: add helper for tree enumeration
  btrfs: allow cross-subvolume file clone
  Btrfs: improve multi-thread buffer read
  Btrfs: make btrfs's allocation smoothly with preallocation
  Btrfs: lock the transition from dirty to writeback for an eb
  Btrfs: fix potential race in extent buffer freeing
  Btrfs: don't return true in releasepage unless we actually freed the eb
  Btrfs: suppress printk() if all device I/O stats are zero
  Btrfs: remove unwanted printk() for btrfs device I/O stats
  Btrfs: rewrite BTRFS_SETGET_FUNCS
  Btrfs: zero unused bytes in inode item
  Btrfs: kill free_space pointer from inode structure
  ...

Conflicts:
fs/btrfs/ioctl.c

12 years agopch_gbe: vlan skb len fix
Andy Cress [Thu, 26 Jul 2012 06:01:17 +0000 (06:01 +0000)]
pch_gbe: vlan skb len fix

pch_gbe_xmit_frame skb->len verification was incorrect in vlan case
causing bogus transfer length errors.  One correction could be:
    offset = skb->protocol == htons(ETH_P_8021Q) ? 0 : 4;
    if (unlikely(skb->len > (adapter->hw.mac.max_frame_size - offset)))
However, this verification is not necessary, so remove it.

Signed-off-by: Andy Cress <andy.cress@us.kontron.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agopch_gbe: add extra clean tx
Andy Cress [Thu, 26 Jul 2012 06:00:11 +0000 (06:00 +0000)]
pch_gbe: add extra clean tx

This adds extra cleaning to the pch_gbe_clean_tx routine to avoid
transmit timeouts on some BCM PHYs that have different timing.
Also update the DRV_VERSION to 1.01, and show it.

Signed-off-by: Andy Cress <andy.cress@us.kontron.com>
Signed-off-by: David S. Miller <davem@davemloft.net>