cascardo/linux.git
11 years agondisc: Unexport ndisc_{build,send}_skb().
YOSHIFUJI Hideaki [Wed, 12 Dec 2012 17:40:26 +0000 (02:40 +0900)]
ndisc: Unexport ndisc_{build,send}_skb().

These symbols were exported for bonding device by commit 305d552a
("bonding: send IPv6 neighbor advertisement on failover").

It bacame obsolete by commit 7c899432 ("bonding, ipv4, ipv6, vlan: Handle
NETDEV_BONDING_FAILOVER like NETDEV_NOTIFY_PEERS") and removed by
commit 4f5762ec ("bonding: Remove obsolete source file 'bond_ipv6.c'").

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agouapi: add missing netconf.h to export list
stephen hemminger [Wed, 12 Dec 2012 06:58:52 +0000 (06:58 +0000)]
uapi: add missing netconf.h to export list

Add netconf.h for use by iproute2.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agopkt_sched: avoid requeues if possible
Eric Dumazet [Tue, 11 Dec 2012 15:54:33 +0000 (15:54 +0000)]
pkt_sched: avoid requeues if possible

With BQL being deployed, we can more likely have following behavior :

We dequeue a packet from qdisc in dequeue_skb(), then we realize target
tx queue is in XOFF state in sch_direct_xmit(), and we have to hold the
skb into gso_skb for later.

This shows in stats (tc -s qdisc dev eth0) as requeues.

Problem of these requeues is that high priority packets can not be
dequeued as long as this (possibly low prio and big TSO packet) is not
removed from gso_skb.

At 1Gbps speed, a full size TSO packet is 500 us of extra latency.

In some cases, we know that all packets dequeued from a qdisc are
for a particular and known txq :

- If device is non multi queue
- For all MQ/MQPRIO slave qdiscs

This patch introduces a new qdisc flag, TCQ_F_ONETXQUEUE to mark
this capability, so that dequeue_skb() is allowed to dequeue a packet
only if the associated txq is not stopped.

This indeed reduce latencies for high prio packets (or improve fairness
with sfq/fq_codel), and almost remove qdisc 'requeues'.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agosolos-pci: fix double-free of TX skb in DMA mode
David Woodhouse [Tue, 11 Dec 2012 14:57:14 +0000 (14:57 +0000)]
solos-pci: fix double-free of TX skb in DMA mode

We weren't clearing card->tx_skb[port] when processing the TX done interrupt.
If there wasn't another skb ready to transmit immediately, this led to a
double-free because we'd free it *again* next time we did have a packet to
send.

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Cc: stable@kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobnx2: Fix accidental reversions.
Michael Chan [Wed, 12 Dec 2012 02:24:20 +0000 (18:24 -0800)]
bnx2: Fix accidental reversions.

Commit 4ce45e02469c382699f4c5f6df727aea9dd2e1ca
"bnx2: Add BNX2 prefix to CHIP ID and name macros"

accidentally reverted 2 commits to use pci_ioumap() and to make
pci_error_handlers const.  This fixes those mistakes.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobna: Driver Version Updated to 3.1.2.1
Rasesh Mody [Tue, 11 Dec 2012 12:24:56 +0000 (12:24 +0000)]
bna: Driver Version Updated to 3.1.2.1

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobna: Firmware update
Rasesh Mody [Tue, 11 Dec 2012 12:24:55 +0000 (12:24 +0000)]
bna: Firmware update

Change Details:
 -      Added Stats clear counter to the bfi_enet_stats_mac structure and
        ethtool stats
 -      Modified the firmware naming convention to contain the firmware image
        version (3.1.0.0). The new convention is
        <firmware-image>-<firmware-version>.bin The change will enforce loading
        only compatible firmware with this driver and also avoid over-writing
        the old firmware image in-order to load new version driver as the
        firmware names used to be the same.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobna: Add RX State
Rasesh Mody [Tue, 11 Dec 2012 12:24:54 +0000 (12:24 +0000)]
bna: Add RX State

Change Details:
 -      BNA state machine for Rx in start_wait state moves it to stop_wait on
        receipt of RX_E_STOP. In Rx stop_wait state, on receipt of
                RX_E_STARTED event does enet stop
                RX_E_STOPPED event does rx_cleanup_cbfn
        rx_cleanup_cbfn in this case is called without post_cbfn. post_cbfn
        happens only after RX_E_STARTED event is received in start_wait. Without
        doing post_cbfn, NAPI remains disabled and in cleanup we try to disable
        again causing endless wait. ifconfig process and other workers can thus
        get stuck.
 -      Introducing start_stop_wait state for Rx. This state handles the case of
        if post_cbfn is not done simply do stop without the cleanup.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobna: Rx Page Based Allocation
Rasesh Mody [Tue, 11 Dec 2012 12:24:53 +0000 (12:24 +0000)]
bna: Rx Page Based Allocation

Change Details:
        Enhanced support for GRO. Page-base allocation method for Rx buffers is
used in GRO. Skb allocation has been removed in Rx path to use always warm-cache
skbs provided by napi_get_frags.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobna: TX Intr Coalescing Fix
Rasesh Mody [Tue, 11 Dec 2012 12:24:52 +0000 (12:24 +0000)]
bna: TX Intr Coalescing Fix

Change Details:
        For Tx IB, IPM was enabled with inter_pkt_timeo of 0. This caused the
Tx IB not to generate interrupt till inter_pkt_count of packets have been
received. Correct definition for BFI_TX_INTERPKT_TIMEO & BFI_TX_INTERPKT_COUNT

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobna: Tx and Rx Optimizations
Rasesh Mody [Tue, 11 Dec 2012 12:24:51 +0000 (12:24 +0000)]
bna: Tx and Rx Optimizations

Change details:
 -      Have contiguous queue pages for TxQ, RxQ and CQ. Data structure and
        QPT changes related to contiguous queue pages
 -      Optimized Tx and Rx unmap structures. Tx and Rx fast path changes due to
        unmap data structure changes
 -      Re-factored Tx and Rx fastpath routines as per the new queue data structures
 -      Implemented bnad_txq_wi_prepare() to program the opcode, flags, frame_len
        and num_vectors in the work item
 -      Reduced Max TxQ and RxQ depth to 2048 while default value for Tx/Rx queue
        depth is unaltered (512)

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobna: Code Cleanup and Enhancements
Rasesh Mody [Tue, 11 Dec 2012 12:24:50 +0000 (12:24 +0000)]
bna: Code Cleanup and Enhancements

Change details:
 -      Remove unnecessary prefetch
 -      Simplify checking & comparison of CQ flags
 -      Dereference & store unmap_array, unmap_cons & current unmap_array
        element only once
 -      Make structures tx_config & rx_config cache line aligned.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
John W. Linville [Tue, 11 Dec 2012 21:24:55 +0000 (16:24 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next into for-davem

11 years agoMerge branch 'for-john' of git://git.sipsolutions.net/mac80211-next
John W. Linville [Tue, 11 Dec 2012 21:04:03 +0000 (16:04 -0500)]
Merge branch 'for-john' of git://git.sipsolutions.net/mac80211-next

11 years agoMerge branch 'for-john' of git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi...
John W. Linville [Tue, 11 Dec 2012 21:03:03 +0000 (16:03 -0500)]
Merge branch 'for-john' of git://git./linux/kernel/git/iwlwifi/iwlwifi-next

11 years agoath9k: check pdata variable before dereferencing it
Gabor Juhos [Mon, 10 Dec 2012 23:06:41 +0000 (00:06 +0100)]
ath9k: check pdata variable before dereferencing it

Due to my recent commit (ath9k: allow to load EEPROM
content via firmware API) smatch complains about that
the 'pdata' variable in 'ath9k_hw_init' can be NULL
and it is dereferenced before checking that. That is
absolutely correct.

Check the 'pdata' variable before using it to avoid
a NULL pointer dereference.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath5k: RX timestamp is reported at end of frame
Thomas Pedersen [Mon, 10 Dec 2012 22:48:03 +0000 (14:48 -0800)]
ath5k: RX timestamp is reported at end of frame

This is true for at least AR5213, and shouldn't be different for other
ath5k PHYs. Tested on AR2413 and AR5414.

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
Tested-by: Bob Copeland <me@bobcopeland.com>
Acked-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_htc: RX timestamp is reported at end of frame
Thomas Pedersen [Mon, 10 Dec 2012 22:48:02 +0000 (14:48 -0800)]
ath9k_htc: RX timestamp is reported at end of frame

Accurate RX timestamp reporting is important for proper IBSS merging,
mesh synchronization, and MCCA scheduling. Namely, knowing where the TSF
is recorded is needed to sync with the beacon timestamp field.

Tested with AR9271.

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: RX timestamp is reported at end of frame
Thomas Pedersen [Mon, 10 Dec 2012 22:48:01 +0000 (14:48 -0800)]
ath9k: RX timestamp is reported at end of frame

Accurate RX timestamp reporting is important for proper IBSS merging,
mesh synchronization, and MCCA scheduling. Namely, knowing where the TSF
is recorded is needed to sync with the beacon timestamp field.

Tested with AR9280.

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agort2800usb: reorganize 2001:3c1e in usb id table Wi-Fi adapter
John W. Linville [Tue, 11 Dec 2012 20:55:02 +0000 (15:55 -0500)]
rt2800usb: reorganize 2001:3c1e in usb id table Wi-Fi adapter

Someone who physically disassembled the device confirms that its
chipset is Ralink RT5370n.

(Fixed-up after having already merged original patch. -- JWL)

Signed-off-by: Maia Kozheva <sikon@ubuntu.com>
Reviewed-by: Xose Vazquez Perez <xose.vazquez@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agonet: gro: avoid double copy in skb_gro_receive()
Eric Dumazet [Tue, 11 Dec 2012 08:38:29 +0000 (08:38 +0000)]
net: gro: avoid double copy in skb_gro_receive()

__copy_skb_header(nskb, p) already copied p->cb[], no need to copy
it again.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobridge: fix seq check in br_mdb_dump()
Cong Wang [Mon, 10 Dec 2012 02:15:35 +0000 (02:15 +0000)]
bridge: fix seq check in br_mdb_dump()

In case of rehashing, introduce a global variable 'br_mdb_rehash_seq'
which gets increased every time when rehashing, and assign
net->dev_base_seq + br_mdb_rehash_seq to cb->seq.

In theory cb->seq could be wrapped to zero, but this is not
easy to fix, as net->dev_base_seq is not visible inside
br_mdb_rehash(). In practice, this is rare.

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Stephen Hemminger <shemminger@vyatta.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Thomas Graf <tgraf@suug.ch>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: smc911x: use io{read,write}*_rep accessors
Matthew Leach [Tue, 11 Dec 2012 04:49:49 +0000 (04:49 +0000)]
net: smc911x: use io{read,write}*_rep accessors

The {read,write}s{b,w,l} operations are not defined by all
architectures and are being removed from the asm-generic/io.h
interface.

This patch replaces the usage of these string functions in the smc911x
accessors with io{read,write}{8,16,32}_rep calls instead.

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Ben Herrenschmidt <benh@kernel.crashing.org>
Cc: netdev@vger.kernel.org
Signed-off-by: Matthew Leach <matthew@mattleach.net>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: remove obsolete simple_strto<foo>
Abhijit Pawar [Mon, 10 Dec 2012 19:30:52 +0000 (19:30 +0000)]
net: remove obsolete simple_strto<foo>

This patch removes the redundant occurences of simple_strto<foo>

Signed-off-by: Abhijit Pawar <abhi.c.pawar@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotun: allow setting ethernet addresss while running
stephen hemminger [Mon, 10 Dec 2012 15:16:00 +0000 (15:16 +0000)]
tun: allow setting ethernet addresss while running

This is a pure software device, and ok with live address change.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: gro: dev_gro_receive() cleanup
Eric Dumazet [Mon, 10 Dec 2012 13:28:16 +0000 (13:28 +0000)]
net: gro: dev_gro_receive() cleanup

__napi_gro_receive() is inlined from two call sites for no good reason.

Lets move the prep stuff in a function of its own, called only if/when
needed. This saves 300 bytes on x86 :

# size net/core/dev.o.after net/core/dev.o.before
   text    data     bss     dec     hex filename
  51968    1238    1040   54246    d3e6 net/core/dev.o.before
  51664    1238    1040   53942    d2b6 net/core/dev.o.after

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: fix a race in gro_cell_poll()
Eric Dumazet [Mon, 10 Dec 2012 12:32:03 +0000 (12:32 +0000)]
net: fix a race in gro_cell_poll()

Dmitry Kravkov reported packet drops for GRE packets since GRO support
was added.

There is a race in gro_cell_poll() because we call napi_complete()
without any synchronization with a concurrent gro_cells_receive()

Once bug was triggered, we queued packets but did not schedule NAPI
poll.

We can fix this issue using the spinlock protected the napi_skbs queue,
as we have to hold it to perform skb dequeue anyway.

As we open-code skb_dequeue(), we no longer need to mask IRQS, as both
producer and consumer run under BH context.

Bug added in commit c9e6bc644e (net: add gro_cells infrastructure)

Reported-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobnx2x: use netdev_alloc_frag()
Eric Dumazet [Mon, 10 Dec 2012 12:16:06 +0000 (12:16 +0000)]
bnx2x: use netdev_alloc_frag()

Using netdev_alloc_frag() instead of kmalloc() permits better GRO or
TCP coalescing behavior, as skb_gro_receive() doesn't have to fallback
to frag_list overhead.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Dmitry Kravkov <dmitry@broadcom.com>
Cc: Eilon Greenstein <eilong@broadcom.com>
Acked-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agodoc: Tighten-up and clarify description of tcp_fin_timeout
Rick Jones [Mon, 10 Dec 2012 11:33:00 +0000 (11:33 +0000)]
doc: Tighten-up and clarify description of tcp_fin_timeout

The description for tcp_fin_timeout should be tigher and more clear.

In addition to being tighter, we should make the spelling of the
state name consistent with what utilities report, remove the now
dated reference to 2.2 and put the default in the consistent place.

Signed-off-by: Rick Jones <rick.jones2@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agominstrel: update stats after processing status
Johannes Berg [Thu, 15 Nov 2012 17:27:56 +0000 (18:27 +0100)]
minstrel: update stats after processing status

Instead of updating stats before sending a packet,
update them after processing the packet's status.
This makes minstrel in line with minstrel_ht.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoiwlwifi: don't handle masked interrupt
Emmanuel Grumbach [Wed, 28 Nov 2012 08:51:34 +0000 (10:51 +0200)]
iwlwifi: don't handle masked interrupt

This can lead to a panic if the driver isn't ready to
handle them. Since our interrupt line is shared, we can get
an interrupt at any time (and CONFIG_DEBUG_SHIRQ checks
that even when the interrupt is being freed).

If the op_mode has gone away, we musn't call it. To avoid
this the transport disables the interrupts when the hw is
stopped and the op_mode is leaving.
If there is an event that would cause an interrupt the INTA
register is updated regardless of the enablement of the
interrupts: even if the interrupts are disabled, the INTA
will be changed, but the device won't issue an interrupt.
But the ISR can be called at any time, so we ought ignore
the value in the INTA otherwise we can call the op_mode
after it was freed.

I found this bug when the op_mode_start failed, and called
iwl_trans_stop_hw(trans, true). Then I played with the
RFKILL button, and removed the module.
While removing the module, the IRQ is freed, and the ISR is
called (CONFIG_DEBUG_SHIRQ enabled). Panic.

Cc: stable@vger.kernel.org
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Reviewed-by: Gregory Greenman <gregory.greenman@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoiwlwifi: silently ignore fw flaws in Tx path
Emmanuel Grumbach [Sun, 2 Dec 2012 07:56:44 +0000 (09:56 +0200)]
iwlwifi: silently ignore fw flaws in Tx path

We know that we have issues with the fw in the reclaim path.
This is why iwl_reclaim doesn't complain too loud when it
happens since it is recoverable. Somehow, the caller of
iwl_reclaim however WARNed when it happens. This doesn't
make any sense.

When I digged into the history of that code, I discovered
that this bug occurs only when we receive a BA notification.
So move the W/A in the BA notification handling code where
it was before.

This patch addresses:
http://bugzilla.intellinuxwireless.org/show_bug.cgi?id=2387

Cc: stable@vger.kernel.org
Reported-by: Florian Reitmeir <florian@reitmeir.org>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoiwlwifi: iwlagn_request_scan: Fix check for priv->scan_request
Tim Gardner [Fri, 7 Dec 2012 13:28:34 +0000 (06:28 -0700)]
iwlwifi: iwlagn_request_scan: Fix check for priv->scan_request

The WARN_ON_ONCE() check for scan_request will not correctly detect
a NULL pointer for scan_type == IWL_SCAN_NORMAL. Make it explicit
that the check only applies to normal scans.

Convert WARN_ON_ONCE to WARN_ON since priv->scan_request really _can't_
be NULL for normal scans. If it is then we should emit frequent warnings.

This smatch warning led to scrutiny of iwlagn_request_scan():

drivers/net/wireless/iwlwifi/dvm/scan.c:894 iwlagn_request_scan() error: we previously assumed 'priv->scan_request' could be null (see line 792)

Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agob43: fix tx path skb leaks
Felix Fietkau [Mon, 10 Dec 2012 16:40:21 +0000 (17:40 +0100)]
b43: fix tx path skb leaks

ieee80211_free_txskb() needs to be used instead of dev_kfree_skb_any for
tx packets passed to the driver from mac80211

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: stable@vger.kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath5k: fix tx path skb leaks
Felix Fietkau [Mon, 10 Dec 2012 15:40:41 +0000 (16:40 +0100)]
ath5k: fix tx path skb leaks

ieee80211_free_txskb() needs to be used instead of dev_kfree_skb_any for
tx packets passed to the driver from mac80211

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: stable@vger.kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: allow to load EEPROM content via firmware API
Gabor Juhos [Mon, 10 Dec 2012 14:30:28 +0000 (15:30 +0100)]
ath9k: allow to load EEPROM content via firmware API

The calibration data for devices w/o a separate
EEPROM chip can be specified via the 'eeprom_data'
field of 'ath9k_platform_data'. The 'eeprom_data'
is usually filled from board specific setup
functions. It is easy if the EEPROM data is mapped
to the memory, but it can be complicated if it is
stored elsewhere.

The patch adds support for loading of the EEPROM
data via the firmware API to avoid this limitation.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: use 'struct ath_hw *' as the first argument for 'ath9k_hw_nvram_read'
Gabor Juhos [Mon, 10 Dec 2012 14:30:27 +0000 (15:30 +0100)]
ath9k: use 'struct ath_hw *' as the first argument for 'ath9k_hw_nvram_read'

The 'ath9k_hw_nvram_read' function takes a
'struct ath_common *' as its first argument.
Almost each of its caller has a 'struct ath_hw *'
parameter in their argument list, and that is
dereferenced in order to get the 'struct ath_common'
pointer.

Change the first argument of 'ath9k_hw_nvram_read'
to be a 'struct ath_hw *', and remove the dereference
calls from the callers.

Also change the type of the first argument of the
ar9300_eeprom_read_{byte,word} functions.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: add EEPROM offset to debug message
Gabor Juhos [Mon, 10 Dec 2012 14:30:26 +0000 (15:30 +0100)]
ath9k: add EEPROM offset to debug message

Show the EEPROM offset of the failed read operation
in 'ath9k_hw_nvram_read'. The debug message is more
informative this way.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: move duplicated debug message to 'ath9k_hw_nvram_read'
Gabor Juhos [Mon, 10 Dec 2012 14:30:25 +0000 (15:30 +0100)]
ath9k: move duplicated debug message to 'ath9k_hw_nvram_read'

The fill_eeprom functions are printing the same
debug message in case the 'ath9k_hw_nvram_read'
function fails. Remove the duplicated code from
fill_eeprom functions and add the ath_dbg call
directly into 'ath9k_hw_nvram_read'.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: Fix signal strength / channel noise reporting
Felix Fietkau [Mon, 10 Dec 2012 13:03:17 +0000 (14:03 +0100)]
ath9k_hw: Fix signal strength / channel noise reporting

While AR_PHY_CCA_NOM_VAL_* does contain the expected internal noise floor
for a chip measured in clean air, it refers to the lowest expected reading.

Depending on the frequency, this measurement can vary by about 6db, thus
causing a higher reported channel noise and signal strength.

Factor in the 6db offset when converting internal noisefloor to channel noise.

This patch makes the reported values more accurate for all chips without
affecting NF calibration behavior.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: stable@vger.kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobcma: mips: fix clearing device IRQ
Rafał Miłecki [Mon, 10 Dec 2012 06:53:56 +0000 (07:53 +0100)]
bcma: mips: fix clearing device IRQ

We were using wrong IRQ number so clearing wasn't working at all.
Depending on a platform this could result in a one device having two
interrupts assigned. On BCM4706 this resulted in all IRQs being broken.

Cc: Hauke Mehrtens <hauke@hauke-m.de>
Cc: stable@vger.kernel.org
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: Update intivals for AR9340
Sujith Manoharan [Mon, 10 Dec 2012 02:50:27 +0000 (08:20 +0530)]
ath9k_hw: Update intivals for AR9340

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: Calculate the correct training power for PAPRD
Sujith Manoharan [Mon, 10 Dec 2012 01:52:38 +0000 (07:22 +0530)]
ath9k_hw: Calculate the correct training power for PAPRD

Assign the training power for PAPRD based on the chip.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: Add HW cap for PAPRD
Sujith Manoharan [Mon, 10 Dec 2012 01:52:37 +0000 (07:22 +0530)]
ath9k_hw: Add HW cap for PAPRD

Add a HW capability to indicate whether PAPRD is enabled
for the card, since PAPRD could be enabled in the EEPROM, but
disabled in the driver. This makes things clearer.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: Fix PAPRD retraining for AR9485
Sujith Manoharan [Mon, 10 Dec 2012 01:52:36 +0000 (07:22 +0530)]
ath9k_hw: Fix PAPRD retraining for AR9485

Retraining of PAPRD based on agc2_pwr is required for
chips other than AR9485.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: Various trivial fixes for PAPRD
Sujith Manoharan [Mon, 10 Dec 2012 01:52:35 +0000 (07:22 +0530)]
ath9k_hw: Various trivial fixes for PAPRD

* Remove unneeded memset.

  All the values in the PAPRD gain table are filled, so there
  is no need to zero out the arrays.

* Use GFP_KERNEL in ar9003_paprd_create_curve

  This is called from the PAPRD work, so the atomic variant
  is not needed.

* Change return type of ar9003_paprd_setup_gain_table

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: Fix redundant PS wrappers
Sujith Manoharan [Mon, 10 Dec 2012 01:52:34 +0000 (07:22 +0530)]
ath9k: Fix redundant PS wrappers

Move the PowerSave wrappers outside ath_paprd_activate(),
since they are already being used in ath_paprd_calibrate().

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: Add a few debug messages for PAPRD
Sujith Manoharan [Mon, 10 Dec 2012 01:52:33 +0000 (07:22 +0530)]
ath9k: Add a few debug messages for PAPRD

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: Fix PAPRD training
Sujith Manoharan [Mon, 10 Dec 2012 01:52:32 +0000 (07:22 +0530)]
ath9k_hw: Fix PAPRD training

The PAPRD training control registers have to be
programmed with values that depend on the chip. This patch
ensures that the correct values are chosen for the chip
in use.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: Fix PAPRD registers for AR9485
Sujith Manoharan [Mon, 10 Dec 2012 01:52:31 +0000 (07:22 +0530)]
ath9k_hw: Fix PAPRD registers for AR9485

Various PAPRD registers are at addresses that are different
from those for the rest of the chips in the AR9003 family.
Fix them.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: ar9003: fix OTP register offsets for AR9340
Gabor Juhos [Sun, 9 Dec 2012 22:57:09 +0000 (23:57 +0100)]
ath9k: ar9003: fix OTP register offsets for AR9340

Trying to access the OTP memory on the AR9340
causes a data bus error like this:

  Data bus error, epc == 86e84164, ra == 86e84164
  Oops[#1]:
  Cpu 0
  $ 0   : 00000000 00000061 deadc0de 00000000
  $ 4   : b8115f18 00015f18 00000007 00000004
  $ 8   : 00000001 7c7c3c7c 7c7c7c7c 7c7c7c7c
  $12   : 7c7c3c7c 001f0041 00000000 7c7c7c3c
  $16   : 86ee0000 00015f18 00000000 00000007
  $20   : 00000004 00000064 00000004 86d71c44
  $24   : 00000000 86e6ca00
  $28   : 86d70000 86d71b20 86ece0c0 86e84164
  Hi    : 00000000
  Lo    : 00000064
  epc   : 86e84164 ath9k_hw_wait+0x58/0xb0 [ath9k_hw]
      Tainted: G           O
  ra    : 86e84164 ath9k_hw_wait+0x58/0xb0 [ath9k_hw]
  Status: 1100d403    KERNEL EXL IE
  Cause : 4080801c
  PrId  : 0001974c (MIPS 74Kc)
  Modules linked in: ath9k(O+) ath9k_common(O) ath9k_hw(O) ath(O) ar934x_nfc
  mac80211(O) usbcore usb_common scsi_mod nls_base nand nand_ecc nand_ids
  crc_ccitt cfg80211(O) compat(O) arc4 aes_generic crypto_blkcipher cryptomgr
  aead crypto_hash crypto_algapi ledtrig_timer ledtrig_default_on leds_gpio
  Process insmod (pid: 459, threadinfo=86d70000, task=87942140, tls=779ac440)
  Stack : 802fb500 000200da 804db150 804e0000 87816130 86ee0000 00010000 86d71b88
          86d71bc0 00000004 00000003 86e9fcd0 80305300 0002c0d0 86e74c50 800b4c20
          000003e8 00000001 00000000 86ee0000 000003ff 86e9fd64 80305300 80123938
          fffffffc 00000004 000058bc 00000000 86ea0000 86ee0000 000001ff 878d6000
          99999999 86e9fdc0 86ee0fcc 86e9e664 0000c0d0 86ee0000 0000700000007000
          ...
  Call Trace:
  [<86e84164>] ath9k_hw_wait+0x58/0xb0 [ath9k_hw]
  [<86e9fcd0>] ath9k_hw_setup_statusring+0x16b8/0x1c7c [ath9k_hw]

  Code: 0000a812  0040f809  00000000 <005310241054000b  24020001  0c05b5dc  2404000a  26520001

The cause of the error is that the OTP register
offsets are different on the AR9340 than the
actually used values.

Cc: <stable@vger.kernel.org> # 3.0+
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agort2800usb: Add support for 2001:3c1e (D-Link DWA-125 rev B1) USB Wi-Fi adapter
Maia Kozheva [Sun, 9 Dec 2012 09:07:40 +0000 (16:07 +0700)]
rt2800usb: Add support for 2001:3c1e (D-Link DWA-125 rev B1) USB Wi-Fi adapter

D-Link DWA-125/B1 is a relatively new USB Wi-Fi adapter, using a
Ralink chipset supported by the rt2800usb driver. Currently, to work
around the problem (it's missing in all present kernel versions,
up to and including 3.7.x), I had to add this to /etc/rc.local:

echo 2001 3c1e >> /sys/bus/usb/drivers/rt2800usb/new_id

After that, the device works without problems. Been using it for over
a week with no bugs in sight.

The attached patch is trivial and simply adds the new USB ID to the
list of devices handled by rt2800usb.

Signed-off-by: Maia Kozheva <sikon@ubuntu.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agossb: use WARN in main.c
Cong Ding [Sat, 8 Dec 2012 23:11:06 +0000 (23:11 +0000)]
ssb: use WARN in main.c

Use WARN rather than printk followed by WARN_ON(1), for conciseness.

Signed-off-by: Cong Ding <dinggnu@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agocarl9170: fix copy and paste mishap in carl9170_handle_mpdu
Christian Lamparter [Sat, 8 Dec 2012 14:24:56 +0000 (15:24 +0100)]
carl9170: fix copy and paste mishap in carl9170_handle_mpdu

This patch fixes a regression which was introduced by:
"carl9170: split up carl9170_handle_mpdu"

Previously, the ieee80211_rx_status was kept on the
stack of carl9170_handle_mpdu. Now it's passed into
the function as a pointer parameter. Hence, the old
memcpy call needs to be fixed.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmsmac: add support for BCM43224 with PCI id of 14e4:a8d8
Hauke Mehrtens [Fri, 7 Dec 2012 16:10:03 +0000 (17:10 +0100)]
brcmsmac: add support for BCM43224 with PCI id of 14e4:a8d8

This device can be found on some embedded devices connected to a
Broadcom SoC like the BCM4718.
I tested this with my Netgear WNDR3400 v1.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmsmac: do a read after the write of the objmem on broken PCIe controllers
Hauke Mehrtens [Fri, 7 Dec 2012 16:04:13 +0000 (17:04 +0100)]
brcmsmac: do a read after the write of the objmem on broken PCIe controllers

As described in the documentation of bcma_wflush16 in drivers/net
/wireless/brcm80211/brcmsmac/types.h some PCIe controllers of Broadcom
SoCs are broken. The PCIe controller on these SoCs are mostly used to
connect some additional wifi device to the SoC and some of these wifi
devices are supported by brcmsmac.
For my BCM43224 connected to the broken PCIe controller of the BCM4718 I
need an extra read after write in brcms_b_write_objmem() to prevent a
Data bus error. This fixes the problem reading tsf_random later.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: Fix regression in 'xmit' debugfs file
Sujith Manoharan [Fri, 7 Dec 2012 14:55:30 +0000 (20:25 +0530)]
ath9k: Fix regression in 'xmit' debugfs file

Commit "ath9k: Fix the 'xmit' debugfs file" changed the
the array size of ath_stats.txstats to IEEE80211_NUM_ACS,
which is wrong because the HW queue number is used to
update the statistics. Revert back to using ATH9K_NUM_TX_QUEUES.

Reported-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobcma: unify naming schema for clock functions
Rafał Miłecki [Fri, 7 Dec 2012 11:56:56 +0000 (12:56 +0100)]
bcma: unify naming schema for clock functions

Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agob43legacy: Fix firmware loading when driver is built into the kernel
Larry Finger [Fri, 7 Dec 2012 03:55:16 +0000 (21:55 -0600)]
b43legacy: Fix firmware loading when driver is built into the kernel

Recent versions of udev cause synchronous firmware loading from the
probe routine to fail because the request to user space times out.
The original fix for b43legacy (commit a3ea2c7) moved the firmware
load from the probe routine to a work queue, but it still used synchronous
firmware loading. This method is OK when b43legacy is built as a module;
however, it fails when the driver is compiled into the kernel.

This version changes the code to load the initial firmware file
using request_firmware_nowait(). A completion event is used to
hold the work queue until that file is available. The remaining
firmware files are read synchronously.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Stable <stable@vger.kernel.org> (V3.4+)
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmsmac: add support for cores with revision 17
Hauke Mehrtens [Thu, 6 Dec 2012 23:35:53 +0000 (00:35 +0100)]
brcmsmac: add support for cores with revision 17

This adds support for bcma wifi core revision 17 which is found on
BCM4716/4717/4718 SoCs. The firmware version 610.812 for brcmsmac found
in linux-firmware does not support these cores, but a firmware
generated with b43-fwcutter from the proprietary broadcom wireless
driver works with these chips. This wifi core contains a revision 5
N-PHY and a revision 7 radio of type 0x2056.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoRevert "ath9k_hw: Update AR9003 high_power tx gain table"
Felix Fietkau [Thu, 6 Dec 2012 17:40:11 +0000 (18:40 +0100)]
Revert "ath9k_hw: Update AR9003 high_power tx gain table"

This reverts commit f74b9d365ddd33a375802b064f96a5d0e99af7c0.

Turns out reverting commit a240dc7b3c7463bd60cf0a9b2a90f52f78aae0fd
"ath9k_hw: Updated AR9003 tx gain table for 5GHz" was not enough to
bring the tx power back to normal levels on devices like the
Buffalo WZR-HP-G450H, this one needs to be reverted as well.

This revert improves tx power by ~10 db on that device

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: stable@vger.kernel.org
Cc: rmanohar@qca.qualcomm.com
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: remove WL_CONN() macro
Arend van Spriel [Wed, 5 Dec 2012 14:26:04 +0000 (15:26 +0100)]
brcmfmac: remove WL_CONN() macro

Get rid of WL_CONN(...) macro in favor of brcmf_dbg(CONN,...)

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: remove WL_SCAN() macro
Arend van Spriel [Wed, 5 Dec 2012 14:26:03 +0000 (15:26 +0100)]
brcmfmac: remove WL_SCAN() macro

Get rid of WL_SCAN(...) macro in favor of brcmf_dbg(SCAN,...)

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: remove WL_TRACE() macro
Arend van Spriel [Wed, 5 Dec 2012 14:26:02 +0000 (15:26 +0100)]
brcmfmac: remove WL_TRACE() macro

Get rid of WL_TRACE(...) macro in favor of brcmf_dbg(TRACE,...)

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: replace WL_INFO() macro
Arend van Spriel [Wed, 5 Dec 2012 14:26:01 +0000 (15:26 +0100)]
brcmfmac: replace WL_INFO() macro

Get rid of WL_INFO() in favor of brcmf_dbg(INFO,...).

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: replace WL_ERR() with brcmf_err()
Arend van Spriel [Wed, 5 Dec 2012 14:26:00 +0000 (15:26 +0100)]
brcmfmac: replace WL_ERR() with brcmf_err()

Getting rid of wl_cfg80211 specific error log macro.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: consolidate debug macros in wl_cfg80211
Arend van Spriel [Wed, 5 Dec 2012 14:25:59 +0000 (15:25 +0100)]
brcmfmac: consolidate debug macros in wl_cfg80211

The source file wl_cfg80211.c has its own debug macros and levels.
This patch maps the macros to the ones used in the rest of the
brcmfmac driver.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agonet: 8390: use io{read,write}*_rep accessors
Matthew Leach [Mon, 10 Dec 2012 09:12:38 +0000 (09:12 +0000)]
net: 8390: use io{read,write}*_rep accessors

The {read,write}s{b,w,l} operations are not defined by all
architectures and are being removed from the asm-generic/io.h
interface.

This patch replaces the usage of these string functions in the 8390
accessors with io{read,write}{8,16,32}_rep calls instead.

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Ben Herrenschmidt <benh@kernel.crashing.org>
Cc: David Miller <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Signed-off-by: Matthew Leach <matthew@mattleach.net>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: dm9000: use io{read,write}*_rep accessors
Matthew Leach [Mon, 10 Dec 2012 09:12:37 +0000 (09:12 +0000)]
net: dm9000: use io{read,write}*_rep accessors

The {read,write}s{b,w,l} operations are not defined by all
architectures and are being removed from the asm-generic/io.h
interface.

This patch replaces the usage of these string functions in the default
DM9000 accessors with io{read,write}{8,16,32}_rep calls instead. This
is required as the dm9000 driver is in use by the blackfin
architecture which uses the asm-generic io accessors.

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Ben Herrenschmidt <benh@kernel.crashing.org>
Cc: David Miller <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Signed-off-by: Matthew Leach <matthew@mattleach.net>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: smc91x: use io{read, write}*_rep accessors instead of string functions
Will Deacon [Mon, 10 Dec 2012 09:12:35 +0000 (09:12 +0000)]
net: smc91x: use io{read, write}*_rep accessors instead of string functions

The {read,write}s{b,w,l} operations are not defined by all architectures
and are being removed from the asm-generic/io.h interface.

This patch replaces the usage of these string functions in the default
SMC accessors with io{read,write}{8,16,32}_rep calls instead, which are
defined for all architectures.

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Ben Herrenschmidt <benh@kernel.crashing.org>
Cc: Nicolas Pitre <nico@fluxnic.net>
Cc: netdev@vger.kernel.org
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobrcmfmac: error messages should not be suppressed
Arend van Spriel [Fri, 7 Dec 2012 09:49:57 +0000 (10:49 +0100)]
brcmfmac: error messages should not be suppressed

The call to brcmf_dbg(ERROR, ...) only resulted in a log message
when compiled with -DDEBUG. Error messages are valuable for resolving
issues so this patch replaces it with brcmf_err(...) so they always
end up in the log.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: use one list of event defintions
Arend van Spriel [Wed, 5 Dec 2012 14:25:56 +0000 (15:25 +0100)]
brcmfmac: use one list of event defintions

Currently, adding a new event requires modification in two source
files. Use macro definition to have one place and have better
maintainability.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmsmac: fix uninitialized variable warning on arm architecture
Arend van Spriel [Wed, 5 Dec 2012 14:25:55 +0000 (15:25 +0100)]
brcmsmac: fix uninitialized variable warning on arm architecture

Using gcc v4.7.2 gave following warning:

  CC [M]  drivers/net/wireless/brcm80211/brcmsmac/aiutils.o
brcmsmac/aiutils.c: In function 'ai_deviceremoved':
brcmsmac/aiutils.c:733:9: error: 'w' may be used uninitialized
in this function [-Werror=uninitialized]

Inspection of the pci_read_config_dword() function showed it can
return without modifying the output variable 'w' so this patch
initializes it to 0.

Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Piotr Haber <phaber@broadcom.com>
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobrcmfmac: rework bus interface
Arend van Spriel [Wed, 5 Dec 2012 14:25:54 +0000 (15:25 +0100)]
brcmfmac: rework bus interface

Rework the bus interface between common driver part and bus-specific
driver part. It prepares for adding tracing in bus-specific callback
functions.

Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agomac80211: a few whitespace fixes
Johannes Berg [Mon, 10 Dec 2012 12:44:19 +0000 (13:44 +0100)]
mac80211: a few whitespace fixes

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agortnetlink: add missing message types to selinux perm table
Cong Wang [Fri, 7 Dec 2012 18:59:48 +0000 (18:59 +0000)]
rtnetlink: add missing message types to selinux perm table

Rebased on the latest net-next tree.

RTM_NEWNETCONF and RTM_GETNETCONF are missing in this table.

Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: Allow DCBnl to use other namespaces besides init_net
John Fastabend [Sun, 9 Dec 2012 20:48:13 +0000 (20:48 +0000)]
net: Allow DCBnl to use other namespaces besides init_net

Allow DCB and net namespace to work together. This is useful if you
have containers that are bound to 'phys' interfaces that want to
also manage their DCB attributes.

The net namespace is taken from sock_net(skb->sk) of the netlink skb.

CC: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agosmsc95xx: fix async register writes on big endian platforms
Steve Glendinning [Mon, 10 Dec 2012 01:03:08 +0000 (01:03 +0000)]
smsc95xx: fix async register writes on big endian platforms

This patch fixes a missing endian conversion which results in the
interface failing to come up on BE platforms.

It also removes an unnecessary pointer dereference from this
function.

Signed-off-by: Steve Glendinning <steve.glendinning@shawell.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agosmsc95xx: fix register dump of last register
Steve Glendinning [Mon, 10 Dec 2012 01:03:07 +0000 (01:03 +0000)]
smsc95xx: fix register dump of last register

This patch fixes the ethtool register dump for smsc95xx to dump
all 4 bytes of the final register (COE_CR) instead of just the
first byte.

Signed-off-by: Steve Glendinning <steve.glendinning@shawell.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agosmsc75xx: only set mac address once on bind
Steve Glendinning [Mon, 10 Dec 2012 01:01:19 +0000 (01:01 +0000)]
smsc75xx: only set mac address once on bind

This patch changes when we decide what the device's MAC address
is from per ifconfig up to once when the device is connected.

Without this patch, a manually forced device MAC is overwritten
on ifconfig down/up.  Also devices that have no EEPROM are
assigned a new random address on ifconfig down/up instead of
persisting the same one.

Signed-off-by: Steve Glendinning <steve.glendinning@shawell.net>
Reported-by: Robert Cunningham <rcunningham@nsmsurveillance.com>
Cc: Bjorn Mork <bjorn@mork.no>
Cc: Dan Williams <dcbw@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: remove obsolete simple_strto<foo>
Abhijit Pawar [Sun, 9 Dec 2012 23:12:28 +0000 (23:12 +0000)]
net: remove obsolete simple_strto<foo>

This patch replace the obsolete simple_strto<foo> with kstrto<foo>

Signed-off-by: Abhijit Pawar <abhi.c.pawar@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agovirtio_net: fix a typo in virtnet_alloc_queues()
Amerigo Wang [Mon, 10 Dec 2012 02:24:08 +0000 (02:24 +0000)]
virtio_net: fix a typo in virtnet_alloc_queues()

Obviously it should check !vi->rq.

Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobridge: make buffer larger in br_setlink()
Dan Carpenter [Fri, 7 Dec 2012 01:10:46 +0000 (01:10 +0000)]
bridge: make buffer larger in br_setlink()

We pass IFLA_BRPORT_MAX to nla_parse_nested() so we need
IFLA_BRPORT_MAX + 1 elements.  Also Smatch complains that we read past
the end of the array when in br_set_port_flag() when it's called with
IFLA_BRPORT_FAST_LEAVE.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agocaif_usb: Make the driver name check more efficient
Ben Hutchings [Fri, 7 Dec 2012 06:20:27 +0000 (06:20 +0000)]
caif_usb: Make the driver name check more efficient

Use the device model to get just the name, rather than using the
ethtool API to get all driver information.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agocaif_usb: Check driver name before reading driver state in netdev notifier
Ben Hutchings [Fri, 7 Dec 2012 06:17:26 +0000 (06:17 +0000)]
caif_usb: Check driver name before reading driver state in netdev notifier

In cfusbl_device_notify(), the usbnet and usbdev variables are
initialised before the driver name has been checked.  In case the
device's driver is not cdc_ncm, this may result in reading beyond the
end of the netdev private area.  Move the initialisation below the
driver name check.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agovirtio-net: support changing the number of queue pairs through ethtool
Jason Wang [Fri, 7 Dec 2012 07:04:57 +0000 (07:04 +0000)]
virtio-net: support changing the number of queue pairs through ethtool

This patch implements the ethtool_{set|get}_channels method of virtio-net to
allow user to change the number of queues when the device is running on demand.

Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agovirtio_net: multiqueue support
Jason Wang [Fri, 7 Dec 2012 07:04:56 +0000 (07:04 +0000)]
virtio_net: multiqueue support

This patch adds the multiqueue (VIRTIO_NET_F_MQ) support to virtio_net
driver. VIRTIO_NET_F_MQ capable device could allow the driver to do packet
transmission and reception through multiple queue pairs and does the packet
steering to get better performance. By default, one one queue pair is used, user
could change the number of queue pairs by ethtool in the next patch.

When multiple queue pairs is used and the number of queue pairs is equal to the
number of vcpus. Driver does the following optimizations to implement per-cpu
virt queue pairs:

- select the txq based on the smp processor id.
- smp affinity hint to the cpu that owns the queue pairs.

This could be used with the flow steering support of the device to guarantee the
packets of a single flow is handled by the same cpu.

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agovirtio-net: separate fields of sending/receiving queue from virtnet_info
Jason Wang [Fri, 7 Dec 2012 07:04:55 +0000 (07:04 +0000)]
virtio-net: separate fields of sending/receiving queue from virtnet_info

To support multiqueue transmitq/receiveq, the first step is to separate queue
related structure from virtnet_info. This patch introduce send_queue and
receive_queue structure and use the pointer to them as the parameter in
functions handling sending/receiving.

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agovxlan: Add capability of Rx checksum offload for inner packet
Joseph Gasparakis [Fri, 7 Dec 2012 14:14:18 +0000 (14:14 +0000)]
vxlan: Add capability of Rx checksum offload for inner packet

This patch adds capability in vxlan to identify received
checksummed inner packets and signal them to the upper layers of
the stack. The driver needs to set the skb->encapsulation bit
and also set the skb->ip_summed to CHECKSUM_UNNECESSARY.

Signed-off-by: Joseph Gasparakis <joseph.gasparakis@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agovxlan: capture inner headers during encapsulation
Joseph Gasparakis [Fri, 7 Dec 2012 14:14:16 +0000 (14:14 +0000)]
vxlan: capture inner headers during encapsulation

Allow VXLAN to make use of Tx checksum offloading and Tx scatter-gather.
The advantage to these two changes is that it also allows the VXLAN to
make use of GSO.

Signed-off-by: Joseph Gasparakis <joseph.gasparakis@intel.com>
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: Handle encapsulated offloads before fragmentation or handing to lower dev
Alexander Duyck [Fri, 7 Dec 2012 14:14:15 +0000 (14:14 +0000)]
net: Handle encapsulated offloads before fragmentation or handing to lower dev

This change allows the VXLAN to enable Tx checksum offloading even on
devices that do not support encapsulated checksum offloads. The
advantage to this is that it allows for the lower device to change due
to routing table changes without impacting features on the VXLAN itself.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: Add support for hardware-offloaded encapsulation
Joseph Gasparakis [Fri, 7 Dec 2012 14:14:14 +0000 (14:14 +0000)]
net: Add support for hardware-offloaded encapsulation

This patch adds support in the kernel for offloading in the NIC Tx and Rx
checksumming for encapsulated packets (such as VXLAN and IP GRE).

For Tx encapsulation offload, the driver will need to set the right bits
in netdev->hw_enc_features. The protocol driver will have to set the
skb->encapsulation bit and populate the inner headers, so the NIC driver will
use those inner headers to calculate the csum in hardware.

For Rx encapsulation offload, the driver will need to set again the
skb->encapsulation flag and the skb->ip_csum to CHECKSUM_UNNECESSARY.
In that case the protocol driver should push the decapsulated packet up
to the stack, again with CHECKSUM_UNNECESSARY. In ether case, the protocol
driver should set the skb->encapsulation flag back to zero. Finally the
protocol driver should have NETIF_F_RXCSUM flag set in its features.

Signed-off-by: Joseph Gasparakis <joseph.gasparakis@intel.com>
Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agor8169: workaround for missing extended GigaMAC registers
françois romieu [Fri, 7 Dec 2012 11:20:21 +0000 (11:20 +0000)]
r8169: workaround for missing extended GigaMAC registers

GigaMAC registers have been reported left unitialized in several
situations:
- after cold boot from power-off state
- after S3 resume

Tweaking rtl_hw_phy_config takes care of both.

This patch removes an excess entry (",") at the end of the exgmac_reg
array as well.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Wang YanQing <udknight@gmail.com>
Cc: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'tipc_net-next_v2' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Sun, 9 Dec 2012 01:25:45 +0000 (20:25 -0500)]
Merge branch 'tipc_net-next_v2' of git://git./linux/kernel/git/paulg/linux

Paul Gortmaker says:

====================
Changes since v1:
-get rid of essentially unused variable spotted by
 Neil Horman (patch #2)

-drop patch #3; defer it for 3.9 content, so Neil,
 Jon and Ying can discuss its specifics at their
 leisure while net-next is closed.  (It had no
 direct dependencies to the rest of the series, and
 was just an optimization)

-fix indentation of accept() code directly in place
 vs. forking it out to a separate function (was patch
 #10, now patch #9).

Rebuilt and re-ran tests just to ensure nothing odd happened.

Original v1 text follows, updated pull information follows that.

           ---------

Here is another batch of TIPC changes.  The most interesting
thing is probably the non-blocking socket connect - I'm told
there were several users looking forward to seeing this.

Also there were some resource limitation changes that had
the right intent back in 2005, but were now apparently causing
needless limitations to people's real use cases; those have
been relaxed/removed.

There is a lockdep splat fix, but no need for a stable backport,
since it is virtually impossible to trigger in mainline; you
have to essentially modify code to force the probabilities
in your favour to see it.

The rest can largely be categorized as general cleanup of things
seen in the process of getting the above changes done.

Tested between 64 and 32 bit nodes with the test suite.  I've
also compile tested all the individual commits on the chain.

I'd originally figured on this queue not being ready for 3.8, but
the extended stabilization window of 3.7 has changed that.  On
the other hand, this can still be 3.9 material, if that simply
works better for folks - no problem for me to defer it to 2013.
If anyone spots any problems then I'll definitely defer it,
rather than rush a last minute respin.
===================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotipc: refactor accept() code for improved readability
Paul Gortmaker [Tue, 4 Dec 2012 16:01:55 +0000 (11:01 -0500)]
tipc: refactor accept() code for improved readability

In TIPC's accept() routine, there is a large block of code relating
to initialization of a new socket, all within an if condition checking
if the allocation succeeded.

Here, we simply flip the check of the if, so that the main execution
path stays at the same indentation level, which improves readability.
If the allocation fails, we jump to an already existing exit label.

Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
11 years agotipc: add lock nesting notation to quiet lockdep warning
Ying Xue [Mon, 3 Dec 2012 08:12:07 +0000 (16:12 +0800)]
tipc: add lock nesting notation to quiet lockdep warning

TIPC accept() call grabs the socket lock on a newly allocated
socket while holding the socket lock on an old socket. But lockdep
worries that this might be a recursive lock attempt:

  [ INFO: possible recursive locking detected ]
  ---------------------------------------------
  kworker/u:0/6 is trying to acquire lock:
  (sk_lock-AF_TIPC){+.+.+.}, at: [<c8c1226c>] accept+0x15c/0x310 [tipc]

  but task is already holding lock:
  (sk_lock-AF_TIPC){+.+.+.}, at: [<c8c12138>] accept+0x28/0x310 [tipc]

  other info that might help us debug this:
  Possible unsafe locking scenario:

          CPU0
          ----
          lock(sk_lock-AF_TIPC);
          lock(sk_lock-AF_TIPC);

          *** DEADLOCK ***

  May be due to missing lock nesting notation
  [...]

Tell lockdep that this locking is safe by using lock_sock_nested().
This is similar to what was done in commit 5131a184a3458d9 for
SCTP code ("SCTP: lock_sock_nested in sctp_sock_migrate").

Also note that this is isn't something that is seen normally,
as it was uncovered with some experimental work-in-progress
code not yet ready for mainline.  So no need for stable
backports or similar of this commit.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
11 years agotipc: eliminate connection setup for implied connect in recv_msg()
Ying Xue [Thu, 29 Nov 2012 08:28:30 +0000 (16:28 +0800)]
tipc: eliminate connection setup for implied connect in recv_msg()

As connection setup is now completed asynchronously in BH context,
in the function filter_connect(), the corresponding code in recv_msg()
becomes redundant.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
11 years agotipc: introduce non-blocking socket connect
Ying Xue [Thu, 29 Nov 2012 23:51:19 +0000 (18:51 -0500)]
tipc: introduce non-blocking socket connect

TIPC has so far only supported blocking connect(), meaning that a call
to connect() doesn't return until either the connection is fully
established, or an error occurs. This has proved insufficient for many
users, so we now introduce non-blocking connect(), analogous to how
this is done in TCP and other protocols.

With this feature, if a connection cannot be established instantly,
connect() will return the error code "-EINPROGRESS".
If the user later calls connect() again, he will either have the
return code "-EALREADY" or "-EISCONN", depending on whether the
connection has been established or not.

The user must have explicitly set the socket to be non-blocking
(SOCK_NONBLOCK or O_NONBLOCK, depending on method used), so unless
for some reason they had set this already (the socket would anyway
remain blocking in current TIPC) this change should be completely
backwards compatible.

It is also now possible to call select() or poll() to wait for the
completion of a connection.

An effect of the above is that the actual completion of a connection
may now be performed asynchronously, independent of the calls from
user space. Therefore, we now execute this code in BH context, in
the function filter_rcv(), which is executed upon reception of
messages in the socket.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
[PG: minor refactoring for improved connect/disconnect function names]
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
11 years agotipc: consolidate connection-oriented message reception in one function
Ying Xue [Thu, 29 Nov 2012 23:39:14 +0000 (18:39 -0500)]
tipc: consolidate connection-oriented message reception in one function

Handling of connection-related message reception is currently scattered
around at different places in the code. This makes it harder to verify
that things are handled correctly in all possible scenarios.
So we consolidate the existing processing of connection-oriented
message reception in a single routine.  In the process, we convert the
chain of if/else into a switch/case for improved readability.

A cast on the socket_state in the switch is needed to avoid compile
warnings on 32 bit, like "net/tipc/socket.c:1252:2: warning: case value
4294967295’ not in enumerated type".  This happens because existing
tipc code pseudo extends the default linux socket state values with:

#define SS_LISTENING    -1      /* socket is listening */
#define SS_READY        -2      /* socket is connectionless */

It may make sense to add these as _positive_ values to the existing
socket state enum list someday, vs. these already existing defines.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
[PG: add cast to fix warning; remove returns from middle of switch]
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
11 years agotipc: standardize across connect/disconnect function naming
Paul Gortmaker [Thu, 29 Nov 2012 18:48:40 +0000 (13:48 -0500)]
tipc: standardize across connect/disconnect function naming

Currently we have tipc_disconnect and tipc_disconnect_port.  It is
not clear from the names alone, what they do or how they differ.
It turns out that tipc_disconnect just deals with the port locking
and then calls tipc_disconnect_port which does all the work.

If we rename as follows: tipc_disconnect_port --> __tipc_disconnect
then we will be following typical linux convention, where:

   __tipc_disconnect: "raw" function that does all the work.

   tipc_disconnect: wrapper that deals with locking and then calls
    the real core __tipc_disconnect function

With this, the difference is immediately evident, and locking
violations are more apt to be spotted by chance while working on,
or even just while reading the code.

On the connect side of things, we currently only have the single
"tipc_connect2port" function.  It does both the locking at enter/exit,
and the core of the work.  Pending changes will make it desireable to
have the connect be a two part locking wrapper + worker function,
just like the disconnect is already.

Here, we make the connect look just like the updated disconnect case,
for the above reason, and for consistency.  In the process, we also
get rid of the "2port" suffix that was on the original name, since
it adds no descriptive value.

On close examination, one might notice that the above connect
changes implicitly move the call to tipc_link_get_max_pkt() to be
within the scope of tipc_port_lock() protected region; when it was
not previously.  We don't see any issues with this, and it is in
keeping with __tipc_connect doing the work and tipc_connect just
handling the locking.

Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>