firefly-linux-kernel-4.4.55.git
14 years agoiwlwifi: clear unattended interrupts in tasklet
Shanyu Zhao [Tue, 16 Mar 2010 17:22:26 +0000 (10:22 -0700)]
iwlwifi: clear unattended interrupts in tasklet

Previously in interrupt handling tasklet, iwlwifi driver only clear/ack
those interrupts that are enabled by the driver through inta_mask.
If the hardware generates unattended interrupts, driver will not ack them,
defeating the interrupt coalescing feature. This results in high number
of interrupts per second and high CPU utilization.

This patch addresses this issue by acking those unattended interrupts
in the tasklet. Local test showed an order of magnitude improvement
in terms of the number of interrupts without sacrificing networking
throughput. This is a workaround for hardware issue.

Signed-off-by: Shanyu Zhao <shanyu.zhao@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
14 years agoiwlwifi: counting number of tfds can be free for 4965
Wey-Yi Guy [Thu, 18 Mar 2010 16:05:00 +0000 (09:05 -0700)]
iwlwifi: counting number of tfds can be free for 4965

Forget one hunk in 4965 during "iwlwifi: error checking for number of tfds
in queue" patch.

Reported-by: Shanyu Zhao <shanyu.zhao@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
CC: stable@kernel.org
14 years agoiwlwifi: fix regulatory
Reinette Chatre [Fri, 12 Mar 2010 19:13:26 +0000 (11:13 -0800)]
iwlwifi: fix regulatory

Commit "cfg80211: convert bools into flags" mistakenly modified iwlwifi's
regulatory settings instead of just converting it. Fix this.

This fixes http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2172

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
CC: stable@kernel.org
14 years agomac80211: move netdev queue enabling to correct spot
Johannes Berg [Mon, 22 Mar 2010 20:42:43 +0000 (13:42 -0700)]
mac80211: move netdev queue enabling to correct spot

"mac80211: fix skb buffering issue" still left a race
between enabling the hardware queues and the virtual
interface queues. In hindsight it's totally obvious
that enabling the netdev queues for a hardware queue
when the hardware queue is enabled is wrong, because
it could well possible that we can fill the hw queue
with packets we already have pending. Thus, we must
only enable the netdev queues once all the pending
packets have been processed and sent off to the device.

In testing, I haven't been able to trigger this race
condition, but it's clearly there, possibly only when
aggregation is being enabled.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoAdd USB ID for Thomson SpeedTouch 120g to p54usb id table
Hans de Goede [Wed, 17 Mar 2010 13:37:16 +0000 (14:37 +0100)]
Add USB ID for Thomson SpeedTouch 120g to p54usb id table

Thanks to Chris Chabot for giving his old wireless usb dongle to me
to test it under Linux.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoAdd a pci-id to the mwl8k driver
Benjamin Larsson [Fri, 19 Mar 2010 00:46:10 +0000 (01:46 +0100)]
Add a pci-id to the mwl8k driver

Signed-off-by: Benjamin Larsson <banan@ludd.ltu.se>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoar9170: add support for NEC WL300NU-G USB dongle
Ben Konrath [Thu, 18 Mar 2010 23:06:57 +0000 (19:06 -0400)]
ar9170: add support for NEC WL300NU-G USB dongle

This patch adds support for the NEC WL300NU-G USB wifi dongle.

Signed-off-by: Ben Konrath <ben@bagu.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agomac80211: fix PREQ processing and one small bug
Porsch, Marco [Wed, 24 Feb 2010 08:53:13 +0000 (09:53 +0100)]
mac80211: fix PREQ processing and one small bug

1st) a PREQ should only be processed, if it has the same SN and better
metric (instead of better or equal).
2nd) next_hop[ETH_ALEN] now actually used to buffer
mpath->next_hop->sta.addr for use out of lock.

Signed-off-by: Marco Porsch <marco.porsch@siemens.com>
Acked-by: Javier Cardona <javier@cozybit.com>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agomac80211: correct typos in "unavailable upon resume" warning
John W. Linville [Wed, 17 Mar 2010 15:28:18 +0000 (11:28 -0400)]
mac80211: correct typos in "unavailable upon resume" warning

Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agowireless: convert reg_regdb_search_lock to mutex
John W. Linville [Tue, 16 Mar 2010 19:40:59 +0000 (15:40 -0400)]
wireless: convert reg_regdb_search_lock to mutex

Stanse discovered that kmalloc is being called with GFP_KERNEL while
holding this spinlock.  The spinlock can be a mutex instead, which also
enables the removal of the unlock/lock around the lock/unlock of
cfg80211_mutex and the call to set_regdom.

Reported-by: Jiri Slaby <jirislaby@gmail.com>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoiwlwifi: Silence tfds_in_queue message
Adel Gadllah [Sun, 14 Mar 2010 18:16:25 +0000 (19:16 +0100)]
iwlwifi: Silence tfds_in_queue message

Commit a239a8b47cc0e5e6d7416a89f340beac06d5edaa introduced a
noisy message, that fills up the log very fast.

The error seems not to be fatal (the connection is stable and
performance is ok), so make it IWL_DEBUG_TX rather than IWL_ERR.

Signed-off-by: Adel Gadllah <adel.gadllah@gmail.com>
Cc: stable@kernel.org
Acked-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoath9k: fix BUG_ON triggered by PAE frames
Felix Fietkau [Fri, 12 Mar 2010 03:02:43 +0000 (04:02 +0100)]
ath9k: fix BUG_ON triggered by PAE frames

When I initially stumbled upon sequence number problems with PAE frames
in ath9k, I submitted a patch to remove all special cases for PAE
frames and let them go through the normal transmit path.
Out of concern about crypto incompatibility issues, this change was
merged instead:

commit 6c8afef551fef87a3bf24f8a74c69a7f2f72fc82
Author: Sujith <Sujith.Manoharan@atheros.com>
Date:   Tue Feb 9 10:07:00 2010 +0530

    ath9k: Fix sequence numbers for PAE frames

After a lot of testing, I'm able to reliably trigger a driver crash on
rekeying with current versions with this change in place.
It seems that the driver does not support sending out regular MPDUs with
the same TID while an A-MPDU session is active.
This leads to duplicate entries in the TID Tx buffer, which hits the
following BUG_ON in ath_tx_addto_baw():

    index  = ATH_BA_INDEX(tid->seq_start, bf->bf_seqno);
    cindex = (tid->baw_head + index) & (ATH_TID_MAX_BUFS - 1);

    BUG_ON(tid->tx_buf[cindex] != NULL);

I believe until we actually have a reproducible case of an
incompatibility with another AP using no PAE special cases, we should
simply get rid of this mess.

This patch completely fixes my crash issues in STA mode and makes it
stay connected without throughput drops or connectivity issues even
when the AP is configured to a very short group rekey interval.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agowl1251: fix potential crash
Grazvydas Ignotas [Thu, 11 Mar 2010 15:45:26 +0000 (17:45 +0200)]
wl1251: fix potential crash

In case debugfs does not init for some reason (or is disabled
on older kernels) driver does not allocate stats.fw_stats
structure, but tries to clear it later and trips on a NULL
pointer:

Unable to handle kernel NULL pointer dereference at virtual address
00000000
PC is at __memzero+0x24/0x80
Backtrace:
[<bf0ddb88>] (wl1251_debugfs_reset+0x0/0x30 [wl1251])
[<bf0d6a2c>] (wl1251_op_stop+0x0/0x12c [wl1251])
[<bf0bc228>] (ieee80211_stop_device+0x0/0x74 [mac80211])
[<bf0b0d10>] (ieee80211_stop+0x0/0x4ac [mac80211])
[<c02deeac>] (dev_close+0x0/0xb4)
[<c02deac0>] (dev_change_flags+0x0/0x184)
[<c031f478>] (devinet_ioctl+0x0/0x704)
[<c0320720>] (inet_ioctl+0x0/0x100)

Add a NULL pointer check to fix this.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Acked-by: Kalle Valo <kalle.valo@iki.fi>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoMerge branch 'wireless-2.6' of git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi...
John W. Linville [Wed, 10 Mar 2010 21:34:38 +0000 (16:34 -0500)]
Merge branch 'wireless-2.6' of git://git./linux/kernel/git/iwlwifi/iwlwifi-2.6

14 years agomac80211: Fix memory leak in ieee80211_if_write()
Eric Dumazet [Wed, 10 Mar 2010 16:13:36 +0000 (17:13 +0100)]
mac80211: Fix memory leak in ieee80211_if_write()

Fix memory leak and use kmalloc() instead of kzalloc() as we are going
to overwrite the allocated buffer.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agomac80211: Fix (dynamic) power save entry
Juuso Oikarinen [Tue, 9 Mar 2010 12:25:02 +0000 (14:25 +0200)]
mac80211: Fix (dynamic) power save entry

Currently hardware with !IEEE80211_HW_PS_NULLFUNC_STACK and
IEEE80211_HW_REPORTS_TX_ACK_STATUS will never enter PSM due to the
conditions in the power save entry functions.

Fix those conditions.

Signed-off-by: Juuso Oikarinen <juuso.oikarinen@nokia.com>
Cc: stable@kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoipw2200: use kmalloc for large local variables
Zhu Yi [Tue, 9 Mar 2010 08:05:31 +0000 (16:05 +0800)]
ipw2200: use kmalloc for large local variables

Fixed below compiler warning:

drivers/net/wireless/ipw2x00/ipw2200.c: In function ‘ipw_load_firmware’:
drivers/net/wireless/ipw2x00/ipw2200.c:3260: warning: the frame size of
1168 bytes is larger than 1024 bytes

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoath5k: read eeprom IQ calibration values correctly for G mode
Bruno Randolf [Tue, 9 Mar 2010 07:56:10 +0000 (16:56 +0900)]
ath5k: read eeprom IQ calibration values correctly for G mode

we read the IQ correction values (i_cal and q_cal) for G mode from a wrong
location (the same shifts as for A mode is applied which is incorrect). use
correct locations, matching the docs and HAL sources.

also we should write IQ correction only when we have that information in the
EEPROM, starting from version 4. also write it in the same way as we do in the
periodic recalibration (enable last), just to be sure.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Acked-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoath5k: fix I/Q calibration (for real)
Bruno Randolf [Tue, 9 Mar 2010 07:56:05 +0000 (16:56 +0900)]
ath5k: fix I/Q calibration (for real)

I/Q calibration was completely broken, resulting in a high number of CRC errors
on received packets. before i could see around 10% to 20% CRC errors, with this
patch they are between 0% and 3%.

1.) the removal of the mask in commit "ath5k: Fix I/Q calibration
(f1cf2dbd0f798b71b1590e7aca6647f2caef1649)" resulted in no mask beeing used
when writing the I/Q values into the register. additional errors in the
calculation of the values (see 2.) resulted too high numbers, exceeding the
masks, so wrong values like 0xfffffffe were written. to be safe we should
always use the bitmask when writing parts of a register.

2.) using a (s32) cast for q_coff is a wrong conversion to signed, since we
convert to a signed value later by substracting 128. this resulted in too low
numbers for Q many times, which were limited to -16 by the boundary check later
on.

3.) checked everything against the HAL sources and took over comments and minor
optimizations from there.

4.) we can't use ENABLE_BITS when we want to write a number (the number can
contain zeros). also always write the correction values first and set ENABLE
bit last, like the HAL does.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Cc: stable@kernel.org
Acked-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoath5k: fix TSF reset
Bruno Randolf [Tue, 9 Mar 2010 07:55:33 +0000 (16:55 +0900)]
ath5k: fix TSF reset

to reset the TSF, AR5K_BEACON_RESET_TSF has to be 1, not 0. also we have a
function for that so use it.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Acked-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoath5k: use fixed antenna for tx descriptors
Bruno Randolf [Tue, 9 Mar 2010 07:55:23 +0000 (16:55 +0900)]
ath5k: use fixed antenna for tx descriptors

when using a fixed antenna we should use the antenna number in all tx
descriptors, otherwise the hardware will sometimes send the frame out on the
other antenna. it seems like the hardware does not always respect the default
antenna and diversity settings (esp.  AR5K_STA_ID1_DEFAULT_ANTENNA).

also i would like to note that antenna diversity does not always work correctly
on 5414 (at least) when only one antenna is connected: for example all frames
might be received on antenna A but still the HW tries to send on antenna B some
times, causing packet loss.

this is both verified with the antenna statistics output of the previous patch
and a spectrum analyzer.

Signed-off-by: Bruno Randolf <br1@einfach.org>
Acked-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agolibipw: split ieee->networks into small pieces
Zhu Yi [Mon, 8 Mar 2010 05:18:03 +0000 (13:18 +0800)]
libipw: split ieee->networks into small pieces

The ieee->networks consists of 128 struct libipw_network entries. If
we allocate this chunk of memory altogether, it ends up with an
order 4 page allocation. High order page allocation is likely to fail
on system high load. This patch splits the big chunk memory allocation
into small pieces, each is 344 bytes, allocates them with 128 times.

The patch fixed bug http://bugzilla.kernel.org/show_bug.cgi?id=14989

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agomac80211: Fix sta_mtx unlocking on insert STA failure path
Jouni Malinen [Sat, 6 Mar 2010 16:35:08 +0000 (18:35 +0200)]
mac80211: Fix sta_mtx unlocking on insert STA failure path

Commit 34e895075e21be3e21e71d6317440d1ee7969ad0 introduced sta_mtx
locking into sta_info_insert() (now sta_info_insert_rcu), but forgot
to unlock this mutex on one of the error paths. Fix this by adding
the missing mutex_unlock() call for the case where STA insert fails
due to an entry existing already. This may happen at least in AP mode
when a STA roams between two BSSes (vifs).

Signed-off-by: Jouni Malinen <j@w1.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agort2x00: remove KSEG1ADDR define from rt2x00soc.h
Helmut Schaa [Fri, 5 Mar 2010 16:44:22 +0000 (17:44 +0100)]
rt2x00: remove KSEG1ADDR define from rt2x00soc.h

Remove the KSEG1ADDR define from rt2x00soc.h as it redefines and covers the
correct one from the arch/mips/include/asm/addrspace.h. Otherwise the driver
oopses on the target platform (Ralink rt3050 board).

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoRevert "iwlwifi: Send broadcast probe request only when asked to"
Reinette Chatre [Thu, 25 Feb 2010 18:02:19 +0000 (10:02 -0800)]
Revert "iwlwifi: Send broadcast probe request only when asked to"

This reverts commit 21b2d8bd2f0d4e0f21ade147fd193c8b9c1fd2b9.

As explained by Johannes:
When we
build a probe request frame in the buffer with the SSID, we could
arrive over the limit of 200 bytes. When we build it in the buffer
without the SSID (wildcard) we don't arrive over 200 bytes, but the
ucode still allows direct probe in addition because it has an internal
buffer that is larger when it inserts the SSID...

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
14 years agoiwl3945: fix memory corruption
Reinette Chatre [Fri, 26 Feb 2010 19:01:36 +0000 (11:01 -0800)]
iwl3945: fix memory corruption

Recent patch "iwlwifi: move 3945 clip groups to 3945 data" exposed a memory
corruption problem. When initializing the clip groups the code was
mistakenly using the iwlagn rate count, not the 3945 rate count. This
resulted in more memory being written than was allocated.

"iwlwifi: move 3945 clip groups to 3945 data" moved the location where the
clip groups are stored and the impact is now severe in that the number of
configured TX queues is modified. Previously the
"temperature" field was overwritten, which did not seem to affect the
operation.

Fix this one instance where wrong rate count was used. I also noticed one
more location where the iwlagn rate count was used to index an iwl3945
array, fix this. I also modified one location that modified the iwlagn rate
count to obtain the iwl3945 rate count ... just use the iwl3945 rate count
directly.

This fixes http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2165 and
http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2168

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
14 years agoMAINTAINERS: add netdev to CAN network layer and drivers entries
Wolfgang Grandegger [Mon, 8 Mar 2010 20:51:41 +0000 (12:51 -0800)]
MAINTAINERS: add netdev to CAN network layer and drivers entries

Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotipc: filter out messages not intended for this host
Neil Horman [Mon, 8 Mar 2010 20:43:56 +0000 (12:43 -0800)]
tipc: filter out messages not intended for this host

Port commit 20deb48d16fdd07ce2fdc8d03ea317362217e085
from git://tipc.cslab.ericsson.net/pub/git/people/allan/tipc.git

Part of the large effort I'm trying to help with getting all the downstreamed
code from windriver forward ported to the upstream tree

Origional commit message
Restore check to filter out inadverdently received messages
This patch reimplements a check that allows TIPC to discard messages
that are not intended for it.  This check was present in TIPC 1.5/1.6,
but was removed by accident during the development of TIPC 1.7; it has
now been updated to account for new features present in TIPC 1.7 and
reinserted into TIPC.  The main benefit of this check is to filter
out messages arriving from orphaned link endpoints, which can arise
when a node exits the network and then re-enters it with a different
TIPC network address (i.e. <Z.C.N> value).

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Origionally-authored-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotipc: fix endianness on tipc subscriber messages
Neil Horman [Mon, 8 Mar 2010 20:20:58 +0000 (12:20 -0800)]
tipc: fix endianness on tipc subscriber messages

Remove htohl implementation from tipc

I was working on forward porting the downstream commits for TIPC and ran accross this one:
http://tipc.cslab.ericsson.net/cgi-bin/gitweb.cgi?p=people/allan/tipc.git;a=commitdiff;h=894279b9437b63cbb02405ad5b8e033b51e4e31e

I was going to just take it, when I looked closer and noted what it was doing.
This is basically a routine to byte swap fields of data in sent/received packets
for tipc, dependent upon the receivers guessed endianness of the peer when a
connection is established.  Asside from just seeming silly to me, it appears to
violate the latest RFC draft for tipc:
http://tipc.sourceforge.net/doc/draft-spec-tipc-02.txt
Which, according to section 4.2 and 4.3.3, requires that all fields of all
commands be sent in network byte order.  So instead of just taking this patch,
instead I'm removing the htohl function and replacing the calls with calls to
ntohl in the rx path and htonl in the send path.

As part of this fix, I'm also changing the subscr_cancel function, which
searches the list of subscribers, using a memcmp of the entire subscriber list,
for the entry to tear down.  unfortunately it memcmps the entire tipc_subscr
structure which has several bits that are private to the local side, so nothing
will ever match.  section 5.2 of the draft spec indicates the <type,upper,lower>
tuple should uniquely identify a subscriber, so convert subscr_cancel to just
match on those fields (properly endian swapped).

I've tested this using the tipc test suite, and its passed without issue.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoethtool: Use noinline_for_stack
Eric Dumazet [Mon, 8 Mar 2010 20:17:04 +0000 (12:17 -0800)]
ethtool: Use noinline_for_stack

Use self documenting noinline_for_stack instead of duplicated comments.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet/sunrpc: Convert (void)snprintf to snprintf
Joe Perches [Mon, 8 Mar 2010 20:15:59 +0000 (12:15 -0800)]
net/sunrpc: Convert (void)snprintf to snprintf

(Applies on top of "Remove uses of NIPQUAD, use %pI4")

Casts to void of snprintf are most uncommon in kernel source.
9 use casts, 1301 do not.

Remove the remaining uses in net/sunrpc/

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet/sunrpc: Remove uses of NIPQUAD, use %pI4
Joe Perches [Mon, 8 Mar 2010 20:15:28 +0000 (12:15 -0800)]
net/sunrpc: Remove uses of NIPQUAD, use %pI4

Originally submitted Jan 1, 2010
http://patchwork.kernel.org/patch/71221/

Convert NIPQUAD to the %pI4 format extension where possible
Convert %02x%02x%02x%02x/NIPQUAD to %08x/ntohl

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocan: fix bfin_can build error after alloc_candev() change
Barry Song [Mon, 8 Mar 2010 20:13:57 +0000 (12:13 -0800)]
can: fix bfin_can build error after alloc_candev() change

Looks like commit a6e4bc530403 didn't include updates to drivers so the
Blackfin CAN driver fails to build now.

Signed-off-by: Barry Song <barry.song@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Acked-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotcp: Fix tcp_make_synack()
Eric Dumazet [Mon, 8 Mar 2010 19:32:01 +0000 (11:32 -0800)]
tcp: Fix tcp_make_synack()

Commit 4957faad (TCPCT part 1g: Responder Cookie => Initiator), part
of TCP_COOKIE_TRANSACTION implementation, forgot to correctly size
synack skb in case user data must be included.

Many thanks to Mika Pentillä for spotting this error.

Reported-by: Penttillä Mika <mika.penttila@ixonos.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: fix route cache rebuilds
Eric Dumazet [Mon, 8 Mar 2010 03:20:00 +0000 (03:20 +0000)]
net: fix route cache rebuilds

We added an automatic route cache rebuilding in commit 1080d709fb9d8cd43
but had to correct few bugs. One of the assumption of original patch,
was that entries where kept sorted in a given way.

This assumption is known to be wrong (commit 1ddbcb005c395518 gave an
explanation of this and corrected a leak) and expensive to respect.

Paweł Staszewski reported to me one of his machine got its routing cache
disabled after few messages like :

[ 2677.850065] Route hash chain too long!
[ 2677.850080] Adjust your secret_interval!
[82839.662993] Route hash chain too long!
[82839.662996] Adjust your secret_interval!
[155843.731650] Route hash chain too long!
[155843.731664] Adjust your secret_interval!
[155843.811881] Route hash chain too long!
[155843.811891] Adjust your secret_interval!
[155843.858209] vlan0811: 5 rebuilds is over limit, route caching
disabled
[155843.858212] Route hash chain too long!
[155843.858213] Adjust your secret_interval!

This is because rt_intern_hash() might be fooled when computing a chain
length, because multiple entries with same keys can differ because of
TOS (or mark/oif) bits.

In the rare case the fast algorithm see a too long chain, and before
taking expensive path, we call a helper function in order to not count
duplicates of same routes, that only differ with tos/mark/oif bits. This
helper works with data already in cpu cache and is not be very
expensive, despite its O(N^2) implementation.

Paweł Staszewski sucessfully tested this patch on his loaded router.

Reported-and-tested-by: Paweł Staszewski <pstaszewski@itcare.pl>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlcnic: remove extra space from board names
Amit Kumar Salecha [Mon, 8 Mar 2010 00:14:50 +0000 (00:14 +0000)]
qlcnic: remove extra space from board names

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlcnic: fix bios version check
Amit Kumar Salecha [Mon, 8 Mar 2010 00:14:49 +0000 (00:14 +0000)]
qlcnic: fix bios version check

Bios sub version from unified fw image is calculated incorrect.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlcnic: validate unified fw image
Sucheta Chakraborty [Mon, 8 Mar 2010 00:14:48 +0000 (00:14 +0000)]
qlcnic: validate unified fw image

Validate all sections of unified fw image, before accessing them,
to avoid seg fault.

Signed-off-by: Sucheta Chakraborty <sucheta@dut6195.unminc.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlcnic: fix multicast handling
Sucheta Chakraborty [Mon, 8 Mar 2010 00:14:47 +0000 (00:14 +0000)]
qlcnic: fix multicast handling

For promiscuous mode, driver send request to device for deleting
multicast addresses and again it send request for adding them back
while exiting from this mode, this is bad for performance.
Just setting device in promiscuous mode is enough, no need to del/add
multicast addresses.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlcnic: additional driver statistics.
Sucheta Chakraborty [Mon, 8 Mar 2010 00:14:46 +0000 (00:14 +0000)]
qlcnic: additional driver statistics.

Statistics added for lro/lso bytes, count for tx stop queue and
wake queue and skb alloc failure count.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoqlcnic: fix tx csum status
Sucheta Chakraborty [Mon, 8 Mar 2010 00:14:45 +0000 (00:14 +0000)]
qlcnic: fix tx csum status

Kernel default tx csum function (ethtool_op_get_tx_csum) doesn't show
correct csum status. It takes various FLAGS (NETIF_F_ALL_CSUM) in account
to show tx csum status, which driver doesn't set while disabling tx csum.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: remove unused code in be_load_fw
Ajit Khaparde [Sun, 7 Mar 2010 14:23:44 +0000 (14:23 +0000)]
be2net: remove unused code in be_load_fw

This patch cleans up some unused code from be_load_fw().

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: remove usage of be_pci_func
Ajit Khaparde [Sun, 7 Mar 2010 14:21:27 +0000 (14:21 +0000)]
be2net: remove usage of be_pci_func

When PCI functions are virtuialized in applications by assigning PCI
functions to VM (PCI passthrough), the be2net driver in the VM sees a

different function number. So, use of PCI function number in any
calculation will break existing code. This patch takes care of it.

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotcp: Add SNMP counters for backlog and min_ttl drops
Eric Dumazet [Sun, 7 Mar 2010 23:21:57 +0000 (23:21 +0000)]
tcp: Add SNMP counters for backlog and min_ttl drops

Commit 6b03a53a (tcp: use limited socket backlog) added the possibility
of dropping frames when backlog queue is full.

Commit d218d111 (tcp: Generalized TTL Security Mechanism) added the
possibility of dropping frames when TTL is under a given limit.

This patch adds new SNMP MIB entries, named TCPBacklogDrop and
TCPMinTTLDrop, published in /proc/net/netstat in TcpExt: line

netstat -s | egrep "TCPBacklogDrop|TCPMinTTLDrop"
    TCPBacklogDrop: 0
    TCPMinTTLDrop: 0

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: add __must_check to sk_add_backlog
Zhu Yi [Sun, 7 Mar 2010 16:21:39 +0000 (16:21 +0000)]
net: add __must_check to sk_add_backlog

Add the "__must_check" tag to sk_add_backlog() so that any failure to
check and drop packets will be warned about.

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobridge: Fix RCU race in br_multicast_stop
Herbert Xu [Fri, 5 Mar 2010 21:03:35 +0000 (21:03 +0000)]
bridge: Fix RCU race in br_multicast_stop

Thanks to Paul McKenny for pointing out that it is incorrect to use
synchronize_rcu_bh to ensure that pending callbacks have completed.
Instead we should use rcu_barrier_bh.

Reported-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobridge: Use RCU list primitive in __br_mdb_ip_get
Herbert Xu [Fri, 5 Mar 2010 21:07:39 +0000 (21:07 +0000)]
bridge: Use RCU list primitive in __br_mdb_ip_get

As Paul McKenney correctly pointed out, __br_mdb_ip_get needs
to use the RCU list walking primitive in order to work correctly
on platforms where data-dependency ordering is not guaranteed.

Reported-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoipv6: Optmize translation between IPV6_PREFER_SRC_xxx and RT6_LOOKUP_F_xxx.
YOSHIFUJI Hideaki / 吉藤英明 [Sun, 7 Mar 2010 00:14:44 +0000 (00:14 +0000)]
ipv6: Optmize translation between IPV6_PREFER_SRC_xxx and RT6_LOOKUP_F_xxx.

IPV6_PREFER_SRC_xxx definitions:
| #define IPV6_PREFER_SRC_TMP             0x0001
| #define IPV6_PREFER_SRC_PUBLIC          0x0002
| #define IPV6_PREFER_SRC_COA             0x0004

RT6_LOOKUP_F_xxx definitions:
| #define RT6_LOOKUP_F_SRCPREF_TMP        0x00000008
| #define RT6_LOOKUP_F_SRCPREF_PUBLIC     0x00000010
| #define RT6_LOOKUP_F_SRCPREF_COA        0x00000020

So, we can translate between these two groups by shift operation
instead of multiple 'if's.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocpmac: bump version to 0.5.2
Florian Fainelli [Sun, 7 Mar 2010 00:55:50 +0000 (00:55 +0000)]
cpmac: bump version to 0.5.2

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocpmac: fallback to switch mode if no PHY chip found
Florian Fainelli [Sun, 7 Mar 2010 00:55:47 +0000 (00:55 +0000)]
cpmac: fallback to switch mode if no PHY chip found

If we were unable to detect a PHY on any of the MDIO bus id we tried instead of
bailing out with -ENODEV, assume the MAC is connected to a switch and use MDIO
bus 0. This unbreaks quite a lot of devices out there whose switch cannot be
detected.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocpmac: fix the receiving of 802.1q frames
Florian Fainelli [Sun, 7 Mar 2010 00:55:26 +0000 (00:55 +0000)]
cpmac: fix the receiving of 802.1q frames

Despite what the comment above CPMAC_SKB_SIZE says, the hardware also
needs to account for the FCS length in a received frame. This patch fix
the receiving of 802.1q frames which have 4 more bytes. While at it
unhardcode the definition and use the one from if_vlan.h.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoMAINTAINER: Correct CAN Maintainer responsibilities and paths
Oliver Hartkopp [Sat, 6 Mar 2010 08:31:50 +0000 (08:31 +0000)]
MAINTAINER: Correct CAN Maintainer responsibilities and paths

Update the CAN Maintainer responsibilities and add source paths.
Additional the SocketCAN core ML is not subscribers-only anymore.

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoanother pegasus usb net device
Petko Manolov [Sun, 7 Mar 2010 06:10:01 +0000 (06:10 +0000)]
another pegasus usb net device

This one removes trailing whitespace in pegasus.h and more importantly
adds new Pegasus compatible device.

Signed-off-by: Julian Brown <julian@codesourcery.com>
Signed-off-by: Petko Manolov <petkan@nucleusys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoirda-usb: add error handling and fix leak
Dan Carpenter [Sun, 7 Mar 2010 02:35:42 +0000 (02:35 +0000)]
irda-usb: add error handling and fix leak

If the call to kcalloc() fails then we should return -ENOMEM.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agosock.c: potential null dereference
Dan Carpenter [Sat, 6 Mar 2010 01:04:45 +0000 (01:04 +0000)]
sock.c: potential null dereference

We test that "prot->rsk_prot" is non-null right before we dereference it
on this line.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoems_usb: cleanup: remove uneeded check
Dan Carpenter [Sat, 6 Mar 2010 01:11:38 +0000 (01:11 +0000)]
ems_usb: cleanup: remove uneeded check

"skb" is alway non-null here, but even if it were null the check isn't
needed because dev_kfree_skb() can handle it.

This eliminates a smatch warning about dereferencing a variable before
checking that it is non-null.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobridge: cleanup: remove unneed check
Dan Carpenter [Sat, 6 Mar 2010 01:14:09 +0000 (01:14 +0000)]
bridge: cleanup: remove unneed check

We dereference "port" on the lines immediately before and immediately
after the test so port should hopefully never be null here.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agofix a race in ks8695_poll
Figo.zhang [Fri, 5 Mar 2010 16:36:02 +0000 (16:36 +0000)]
fix a race in ks8695_poll

fix a race at the end of NAPI processing in ks8695_poll() function.

Signed-off-by:Figo.zhang <figo1802@gmail.com>

Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agos2io: Fixing debug message
Breno Leitao [Thu, 4 Mar 2010 10:40:44 +0000 (10:40 +0000)]
s2io: Fixing debug message

Currently s2io is dumping debug messages using the interface name
before it was allocated, showing a message like the following:

s2io: eth%d: Ring Mem PHY: 0x7ef80000
s2io: s2io_reset: Resetting XFrame card eth%d

This patch just fixes it, printing the pci bus information for
the card instead of the interface name.

Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoe1000e: fix packet corruption and tx hang during NFSv2
Jesse Brandeburg [Fri, 5 Mar 2010 02:21:44 +0000 (02:21 +0000)]
e1000e: fix packet corruption and tx hang during NFSv2

when receiving a particular type of NFS v2 UDP traffic, the hardware could
DMA some bad data and then hang, possibly corrupting memory.

Disable the NFS parsing in this hardware, verified to fix the bug.

Originally reported and reproduced by RedHat's Neil Horman
CC: nhorman@tuxdriver.com
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotyphoon: fix incorrect use of smp_wmb()
David Dillow [Thu, 4 Mar 2010 04:37:16 +0000 (04:37 +0000)]
typhoon: fix incorrect use of smp_wmb()

The typhoon driver was incorrectly using smp_wmb() to order memory
accesses against IO to the NIC in a few instances. Use wmb() instead,
which is required to actually order between memory types.

Signed-off-by: David Dillow <dave@thedillows.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoethtool: Add direct access to ops->get_sset_count
Jeff Garzik [Thu, 4 Mar 2010 08:21:53 +0000 (08:21 +0000)]
ethtool: Add direct access to ops->get_sset_count

On 03/04/2010 09:26 AM, Ben Hutchings wrote:
> On Thu, 2010-03-04 at 00:51 -0800, Jeff Kirsher wrote:
>> From: Jeff Garzik<jgarzik@redhat.com>
>>
>> This patch is an alternative approach for accessing string
>> counts, vs. the drvinfo indirect approach.  This way the drvinfo
>> space doesn't run out, and we don't break ABI later.
> [...]
>> --- a/net/core/ethtool.c
>> +++ b/net/core/ethtool.c
>> @@ -214,6 +214,10 @@ static noinline int ethtool_get_drvinfo(struct net_device *dev, void __user *use
>>    info.cmd = ETHTOOL_GDRVINFO;
>>    ops->get_drvinfo(dev,&info);
>>
>> + /*
>> +  * this method of obtaining string set info is deprecated;
>> +  * consider using ETHTOOL_GSSET_INFO instead
>> +  */
>
> This comment belongs on the interface (ethtool.h) not the
> implementation.

Debatable -- the current comment is located at the callsite of
ops->get_sset_count(), which is where an implementor might think to add
a new call.  Not all the numeric fields in ethtool_drvinfo are obtained
from ->get_sset_count().

Hence the "some" in the attached patch to include/linux/ethtool.h,
addressing your comment.

> [...]
>> +static noinline int ethtool_get_sset_info(struct net_device *dev,
>> +                                          void __user *useraddr)
>> +{
> [...]
>> + /* calculate size of return buffer */
>> + for (i = 0; i<  64; i++)
>> + if (sset_mask&  (1ULL<<  i))
>> + n_bits++;
> [...]
>
> We have a function for this:
>
>  n_bits = hweight64(sset_mask);

Agreed.

I've attached a follow-up patch, which should enable my/Jeff's kernel
patch to be applied, followed by this one.

Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoethtool: Add direct access to ops->get_sset_count
Jeff Garzik [Wed, 3 Mar 2010 22:51:50 +0000 (22:51 +0000)]
ethtool: Add direct access to ops->get_sset_count

This patch is an alternative approach for accessing string
counts, vs. the drvinfo indirect approach.  This way the drvinfo
space doesn't run out, and we don't break ABI later.

Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: smc91x: Support Qualcomm MSM development boards.
David Brown [Fri, 5 Mar 2010 09:12:34 +0000 (09:12 +0000)]
net: smc91x: Support Qualcomm MSM development boards.

Signed-off-by: David Brown <davidb@quicinc.com>
Signed-off-by: Daniel Walker <dwalker@codeaurora.org>
Acked-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: backlog functions rename
Zhu Yi [Thu, 4 Mar 2010 18:01:47 +0000 (18:01 +0000)]
net: backlog functions rename

sk_add_backlog -> __sk_add_backlog
sk_add_backlog_limited -> sk_add_backlog

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agox25: use limited socket backlog
Zhu Yi [Thu, 4 Mar 2010 18:01:46 +0000 (18:01 +0000)]
x25: use limited socket backlog

Make x25 adapt to the limited socket backlog change.

Cc: Andrew Hendry <andrew.hendry@gmail.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotipc: use limited socket backlog
Zhu Yi [Thu, 4 Mar 2010 18:01:45 +0000 (18:01 +0000)]
tipc: use limited socket backlog

Make tipc adapt to the limited socket backlog change.

Cc: Jon Maloy <jon.maloy@ericsson.com>
Cc: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agosctp: use limited socket backlog
Zhu Yi [Thu, 4 Mar 2010 18:01:44 +0000 (18:01 +0000)]
sctp: use limited socket backlog

Make sctp adapt to the limited socket backlog change.

Cc: Vlad Yasevich <vladislav.yasevich@hp.com>
Cc: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agollc: use limited socket backlog
Zhu Yi [Thu, 4 Mar 2010 18:01:43 +0000 (18:01 +0000)]
llc: use limited socket backlog

Make llc adapt to the limited socket backlog change.

Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoudp: use limited socket backlog
Zhu Yi [Thu, 4 Mar 2010 18:01:42 +0000 (18:01 +0000)]
udp: use limited socket backlog

Make udp adapt to the limited socket backlog change.

Cc: "David S. Miller" <davem@davemloft.net>
Cc: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Cc: "Pekka Savola (ipv6)" <pekkas@netcore.fi>
Cc: Patrick McHardy <kaber@trash.net>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotcp: use limited socket backlog
Zhu Yi [Thu, 4 Mar 2010 18:01:41 +0000 (18:01 +0000)]
tcp: use limited socket backlog

Make tcp adapt to the limited socket backlog change.

Cc: "David S. Miller" <davem@davemloft.net>
Cc: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Cc: "Pekka Savola (ipv6)" <pekkas@netcore.fi>
Cc: Patrick McHardy <kaber@trash.net>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: add limit for socket backlog
Zhu Yi [Thu, 4 Mar 2010 18:01:40 +0000 (18:01 +0000)]
net: add limit for socket backlog

We got system OOM while running some UDP netperf testing on the loopback
device. The case is multiple senders sent stream UDP packets to a single
receiver via loopback on local host. Of course, the receiver is not able
to handle all the packets in time. But we surprisingly found that these
packets were not discarded due to the receiver's sk->sk_rcvbuf limit.
Instead, they are kept queuing to sk->sk_backlog and finally ate up all
the memory. We believe this is a secure hole that a none privileged user
can crash the system.

The root cause for this problem is, when the receiver is doing
__release_sock() (i.e. after userspace recv, kernel udp_recvmsg ->
skb_free_datagram_locked -> release_sock), it moves skbs from backlog to
sk_receive_queue with the softirq enabled. In the above case, multiple
busy senders will almost make it an endless loop. The skbs in the
backlog end up eat all the system memory.

The issue is not only for UDP. Any protocols using socket backlog is
potentially affected. The patch adds limit for socket backlog so that
the backlog size cannot be expanded endlessly.

Reported-by: Alex Shi <alex.shi@intel.com>
Cc: David Miller <davem@davemloft.net>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru
Cc: "Pekka Savola (ipv6)" <pekkas@netcore.fi>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Vlad Yasevich <vladislav.yasevich@hp.com>
Cc: Sridhar Samudrala <sri@us.ibm.com>
Cc: Jon Maloy <jon.maloy@ericsson.com>
Cc: Allan Stephens <allan.stephens@windriver.com>
Cc: Andrew Hendry <andrew.hendry@gmail.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agorndis_wlan: correct multicast_list handling V3
Jiri Pirko [Thu, 4 Mar 2010 11:32:16 +0000 (03:32 -0800)]
rndis_wlan: correct multicast_list handling V3

My previous patch (655ffee284dfcf9a24ac0343f3e5ee6db85b85c5) added locking in
a bad way. Because rndis_set_oid can sleep, there is need to prepare multicast
addresses into local buffer under netif_addr_lock first, then call
rndis_set_oid outside. This caused reorganizing of the whole function.

Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Reported-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoMAINTAINERS: Add netdev to bridge entry.
David S. Miller [Thu, 4 Mar 2010 08:42:30 +0000 (00:42 -0800)]
MAINTAINERS: Add netdev to bridge entry.

Noticed by Ingo Molnar.

Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocxgb3: fix hot plug removal crash
Divy Le Ray [Wed, 3 Mar 2010 09:49:47 +0000 (09:49 +0000)]
cxgb3: fix hot plug removal crash

queue restart tasklets need to be stopped after napi handlers are stopped
since the latter can restart them.  So stop them after stopping napi.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agogianfar: Fix TX ring processing on SMP machines
Anton Vorontsov [Wed, 3 Mar 2010 08:18:58 +0000 (08:18 +0000)]
gianfar: Fix TX ring processing on SMP machines

Starting with commit a3bc1f11e9b867a4f49505 ("gianfar: Revive SKB
recycling") gianfar driver sooner or later stops transmitting any
packets on SMP machines.

start_xmit() prepares new skb for transmitting, generally it does
three things:

1. sets up all BDs (marks them ready to send), except the first one.
2. stores skb into tx_queue->tx_skbuff so that clean_tx_ring()
   would cleanup it later.
3. sets up the first BD, i.e. marks it ready.

Here is what clean_tx_ring() does:

1. reads skbs from tx_queue->tx_skbuff
2. checks if the *last* BD is ready. If it's still ready [to send]
   then it it isn't transmitted, so clean_tx_ring() returns.
   Otherwise it actually cleanups BDs. All is OK.

Now, if there is just one BD, code flow:

- start_xmit(): stores skb into tx_skbuff. Note that the first BD
  (which is also the last one) isn't marked as ready, yet.
- clean_tx_ring(): sees that skb is not null, *and* its lstatus
  says that it is NOT ready (like if BD was sent), so it cleans
  it up (bad!)
- start_xmit(): marks BD as ready [to send], but it's too late.

We can fix this simply by reordering lstatus/tx_skbuff writes.

Reported-by: Martyn Welch <martyn.welch@ge.com>
Bisected-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Tested-by: Martyn Welch <martyn.welch@ge.com>
Cc: Sandeep Gopalpet <Sandeep.Kumar@freescale.com>
Cc: Stable <stable@vger.kernel.org> [2.6.33]
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agor8169: use correct barrier between cacheable and non-cacheable memory
David Dillow [Wed, 3 Mar 2010 16:33:10 +0000 (16:33 +0000)]
r8169: use correct barrier between cacheable and non-cacheable memory

r8169 needs certain writes to be visible to other CPUs or the NIC before
touching the hardware, but was using smp_wmb() which is only required to
order cacheable memory access. Switch to wmb() which is required to
order both cacheable and non-cacheable memory.

Noticed by Catalin Marinas and Paul Mackerras.

Signed-off-by: David Dillow <dave@thedillows.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agotipc: Fix oops on send prior to entering networked mode (v3)
Neil Horman [Wed, 3 Mar 2010 08:31:23 +0000 (08:31 +0000)]
tipc: Fix oops on send prior to entering networked mode (v3)

Fix TIPC to disallow sending to remote addresses prior to entering NET_MODE

user programs can oops the kernel by sending datagrams via AF_TIPC prior to
entering networked mode.  The following backtrace has been observed:

ID: 13459  TASK: ffff810014640040  CPU: 0   COMMAND: "tipc-client"
[exception RIP: tipc_node_select_next_hop+90]
RIP: ffffffff8869d3c3  RSP: ffff81002d9a5ab8  RFLAGS: 00010202
RAX: 0000000000000001  RBX: 0000000000000001  RCX: 0000000000000001
RDX: 0000000000000000  RSI: 0000000000000001  RDI: 0000000001001001
RBP: 0000000001001001   R8: 0074736575716552   R9: 0000000000000000
R10: ffff81003fbd0680  R11: 00000000000000c8  R12: 0000000000000008
R13: 0000000000000001  R14: 0000000000000001  R15: ffff810015c6ca00
ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
RIP: 0000003cbd8d49a3  RSP: 00007fffc84e0be8  RFLAGS: 00010206
RAX: 000000000000002c  RBX: ffffffff8005d116  RCX: 0000000000000000
RDX: 0000000000000008  RSI: 00007fffc84e0c00  RDI: 0000000000000003
RBP: 0000000000000000   R8: 00007fffc84e0c10   R9: 0000000000000010
R10: 0000000000000000  R11: 0000000000000246  R12: 0000000000000000
R13: 00007fffc84e0d10  R14: 0000000000000000  R15: 00007fffc84e0c30
ORIG_RAX: 000000000000002c  CS: 0033  SS: 002b

What happens is that, when the tipc module in inserted it enters a standalone
node mode in which communication to its own address is allowed <0.0.0> but not
to other addresses, since the appropriate data structures have not been
allocated yet (specifically the tipc_net pointer).  There is nothing stopping a
client from trying to send such a message however, and if that happens, we
attempt to dereference tipc_net.zones while the pointer is still NULL, and
explode.  The fix is pretty straightforward.  Since these oopses all arise from
the dereference of global pointers prior to their assignment to allocated
values, and since these allocations are small (about 2k total), lets convert
these pointers to static arrays of the appropriate size.  All the accesses to
these bits consider 0/NULL to be a non match when searching, so all the lookups
still work properly, and there is no longer a chance of a bad dererence
anywhere.  As a bonus, this lets us eliminate the setup/teardown routines for
those pointers, and elimnates the need to preform any locking around them to
prevent access while their being allocated/freed.

I've updated the tipc_net structure to behave this way to fix the exact reported
problem, and also fixed up the tipc_bearers and media_list arrays to fix an
obvious simmilar problem that arises from issuing tipc-config commands to
manipulate bearers/links prior to entering networked mode

I've tested this for a few hours by running the sanity tests and stress test
with the tipcutils suite, and nothing has fallen over.  There have been a few
lockdep warnings, but those were there before, and can be addressed later, as
they didn't actually result in any deadlock.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
CC: Allan Stephens <allan.stephens@windriver.com>
CC: David S. Miller <davem@davemloft.net>
CC: tipc-discussion@lists.sourceforge.net
 bearer.c |   37 ++++++-------------------------------
 bearer.h |    2 +-
 net.c    |   25 ++++---------------------
 3 files changed, 11 insertions(+), 53 deletions(-)
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agogre: fix hard header destination address checking
Timo Teräs [Wed, 3 Mar 2010 04:01:13 +0000 (04:01 +0000)]
gre: fix hard header destination address checking

ipgre_header() can be called with zero daddr when the gre device is
configured as multipoint tunnel and still has the NOARP flag set (which is
typically cleared by the userspace arp daemon).  If the NOARP packets are
not dropped, ipgre_tunnel_xmit() will take rt->rt_gateway (= NBMA IP) and
use that for route look up (and may lead to bogus xfrm acquires).

The multicast address check is removed as sending to multicast group should
be ok.  In fact, if gre device has a multicast address as destination
ipgre_header is always called with multicast address.

Signed-off-by: Timo Teras <timo.teras@iki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agonet: add scheduler sync hint to tcp_prequeue().
Mike Galbraith [Wed, 3 Mar 2010 04:46:50 +0000 (04:46 +0000)]
net: add scheduler sync hint to tcp_prequeue().

Decreases the odds wakee will suffer from frequent cache misses.

Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoIPv6: fix race between cleanup and add/delete address
stephen hemminger [Wed, 3 Mar 2010 08:19:59 +0000 (08:19 +0000)]
IPv6: fix race between cleanup and add/delete address

This solves a potential race problem during the cleanup process.
The issue is that addrconf_ifdown() needs to traverse address list,
but then drop lock to call the notifier. The version in -next
could get confused if add/delete happened during this window.
Original code (2.6.32 and earlier) was okay because all addresses
were always deleted.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoIPv6: addrconf notify when address is unavailable
stephen hemminger [Tue, 2 Mar 2010 13:32:46 +0000 (13:32 +0000)]
IPv6: addrconf notify when address is unavailable

My recent change in net-next to retain permanent addresses caused regression.
Device refcount would not go to zero when device was unregistered because
left over anycast reference would hold ipv6 dev reference which would hold
device references...

The correct procedure is to call notify chain when address is no longer
available for use.  When interface comes back DAD timer will notify
back that address is available.

Also, link local addresses should be purged when interface is brought
down. The address might be changed.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoIPv6: addrconf timer race
stephen hemminger [Tue, 2 Mar 2010 13:32:45 +0000 (13:32 +0000)]
IPv6: addrconf timer race

The Router Solicitation timer races with device state changes
because it doesn't lock the device. Use local variable to avoid
one repeated dereference.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoIPv6: addrconf dad timer unnecessary bh_disable
stephen hemminger [Tue, 2 Mar 2010 13:32:44 +0000 (13:32 +0000)]
IPv6: addrconf dad timer unnecessary bh_disable

Timer code runs in bottom half, so there is no need for
using _bh form of locking.  Also check if device is not ready
to avoid race with address that is no longer active.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
David S. Miller [Thu, 4 Mar 2010 06:42:54 +0000 (22:42 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next-2.6

14 years agort2x00: Export rt2x00soc_probe from rt2x00soc
Helmut Schaa [Wed, 3 Mar 2010 16:42:55 +0000 (17:42 +0100)]
rt2x00: Export rt2x00soc_probe from rt2x00soc

Export rt2x00soc_probe from rt2x00soc as it is used in rt2800pci.

Otherwise loading rt2800pci gives "rt2800pci: Unknown symbol
rt2x00soc_probe".

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agomac80211: Fix HT rate control configuration
Sujith [Mon, 1 Mar 2010 09:12:57 +0000 (14:42 +0530)]
mac80211: Fix HT rate control configuration

Handling HT configuration changes involved setting the channel
with the new HT parameters and then issuing a rate_update()
notification to the driver.

This behavior changed after the off-channel changes. Now, the channel
is not updated with the new HT params in enable_ht() - instead, it
is now done when the scan work terminates. This results in the driver
depending on stale information, defaulting to non-HT mode always.

Fix this by passing the new channel type to the driver.

Cc: stable@kernel.org
Signed-off-by: Sujith <Sujith.Manoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agobridge: depends on INET
Randy Dunlap [Wed, 3 Mar 2010 09:23:22 +0000 (01:23 -0800)]
bridge: depends on INET

br_multicast calls ip_send_check(), so it should depend on INET.

built-in:
br_multicast.c:(.text+0x88cf4): undefined reference to `ip_send_check'

or modular:
ERROR: "ip_send_check" [net/bridge/bridge.ko] undefined!

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobnx2x: Removed FW 5.2.7
Vladislav Zolotarov [Wed, 3 Mar 2010 09:09:21 +0000 (01:09 -0800)]
bnx2x: Removed FW 5.2.7

Removed FW 5.2.7.

Signed-off-by: Vladislav Zolotarov <vladz@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agodavinci_emac: off by one
Dan Carpenter [Tue, 2 Mar 2010 21:07:24 +0000 (21:07 +0000)]
davinci_emac: off by one

This off by one error was found by smatch.

drivers/net/davinci_emac.c +2390 emac_dev_open(13) error: buffer overflow 'priv->mac_addr' 6 <= 6

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocassini: fix off by one
Dan Carpenter [Tue, 2 Mar 2010 22:22:41 +0000 (22:22 +0000)]
cassini: fix off by one

There are only 6 link_modes.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoBluetooth: Use single_open() for inquiry cache within debugfs
Marcel Holtmann [Tue, 2 Mar 2010 15:48:23 +0000 (15:48 +0000)]
Bluetooth: Use single_open() for inquiry cache within debugfs

The inquiry cache information in debugfs should be using seq_file support
and not allocating memory on the stack for the string. Since the usage of
these information is really seldom, using single_open() for it is good
enough.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoaf_packet: move strict addr_len check right before dev_[mc/unicast]_[add/del]
Jiri Pirko [Tue, 2 Mar 2010 20:40:01 +0000 (20:40 +0000)]
af_packet: move strict addr_len check right before dev_[mc/unicast]_[add/del]

My previous patch 914c8ad2d18b62ad1420f518c0cab0b0b90ab308 incorrectly changed
the length check in packet_mc_add to be more strict. The problem is that
userspace is not filling this field (and it stays zeroed) in case of setting
PACKET_MR_PROMISC or PACKET_MR_ALLMULTI. So move the strict check to the point
in path where the addr_len must be set correctly.

Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Reported-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agobe2net: download NCSI section during firmware update
Sarveshwar Bandi [Tue, 2 Mar 2010 22:37:28 +0000 (22:37 +0000)]
be2net: download NCSI section during firmware update

Adding code to update NCSI section while updating firmware on the controller.

Signed-off-by: Sarveshwar Bandi <sarveshwarb@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agocpmac: use after free
Dan Carpenter [Tue, 2 Mar 2010 22:46:10 +0000 (22:46 +0000)]
cpmac: use after free

The original code dereferenced "cpmac_mii" after calling
"mdiobus_free(cpmac_mii);"

Signed-off-by: Dan Carpenter <error27@gmail.com>
Reviewed-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoipsec: Fix bogus bundle flowi
Herbert Xu [Tue, 2 Mar 2010 02:51:56 +0000 (02:51 +0000)]
ipsec: Fix bogus bundle flowi

When I merged the bundle creation code, I introduced a bogus
flowi value in the bundle.  Instead of getting from the caller,
it was instead set to the flow in the route object, which is
totally different.

The end result is that the bundles we created never match, and
we instead end up with an ever growing bundle list.

Thanks to Jamal for find this problem.

Reported-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Steffen Klassert <steffen.klassert@secunet.com>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
14 years agoiwlwifi: load firmware asynchronously before mac80211 registration
Johannes Berg [Fri, 29 Jan 2010 19:54:56 +0000 (11:54 -0800)]
iwlwifi: load firmware asynchronously before mac80211 registration

At the wireless summit in Portland we discussed a way of
loading firmware asynchronously from ->probe() before
registration to mac80211, in order to register with the
wireless subsystems with complete information in cases
where firmware is required to know parameters.

This is not yet the case in iwlwifi, but for some new
features we're working on it will be the case since
those will only be supported by new firmware images.

Hence, to start with, convert iwlwifi to load firmware
asynchronously from probe, unbinding the device when
firmware loading fails, and only registering with the
wireless subsystems after firmware has been loaded
successfully.

Future patches will hook into this to register the
new firmware capabilities, depending on the firmware
API version.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agoar9170: load firmware asynchronously
Johannes Berg [Wed, 23 Dec 2009 12:15:30 +0000 (13:15 +0100)]
ar9170: load firmware asynchronously

This converts ar9170 to load firmware asynchronously
out of ->probe() and only register with mac80211 when
all firmware has been loaded successfully. If, on the
other hand, any firmware fails to load, it will now
unbind from the device.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agort2x00: fix rt2800pci compilation with SoC
Helmut Schaa [Tue, 2 Mar 2010 15:34:49 +0000 (16:34 +0100)]
rt2x00: fix rt2800pci compilation with SoC

Compiling rt2800pci with CONFIG_RT2800PCI_SOC fails with "... rt2880pci.c:
error: incompatible type for argument 2 of 'rt2x00soc_probe'".

Fix this by using &rt2800pci_ops instead of rt2800pci_ops.

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
14 years agort2x00: Add USB ID for CEIVA adapter to rt73usb
Bryan Polk [Mon, 1 Mar 2010 17:23:28 +0000 (12:23 -0500)]
rt2x00: Add USB ID for CEIVA adapter to rt73usb

This adds support for CEIVA USB wireless adapters to the rt73usb driver.

Signed-off-by: Bryan Polk <sainth@eidolons.org>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>