openwrt/staging/blogic.git
13 years agobe2net: Change the data type of the 'on die temperature' stat.
Somnath Kotur [Fri, 30 Sep 2011 07:23:35 +0000 (07:23 +0000)]
be2net: Change the data type of the 'on die temperature' stat.

This was showing up as junk value on PPC /Big endian machines since
it was marked as a byte.

Signed-off-by: Somnath Kotur <somnath.kotur@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobe2net: Add 60 second delay to allow FAT dump completion on recovery from EEH
Somnath Kotur [Mon, 3 Oct 2011 08:10:57 +0000 (08:10 +0000)]
be2net: Add 60 second delay to allow FAT dump completion on recovery from EEH

Add 60s delay before timeout on polling Bit 31 so that FAT dump can
complete when reset occurs.

Signed-off-by: Somnath Kotur <somnath.kotur@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoRPS: Ensure that an expired hardware filter can be re-added later
Ben Hutchings [Mon, 3 Oct 2011 04:42:46 +0000 (04:42 +0000)]
RPS: Ensure that an expired hardware filter can be re-added later

Amir Vadai wrote:
> When a stream is paused, and its rule is expired while it is paused,
> no new rule will be configured to the HW when traffic resume.
[...]
> - When stream was resumed, traffic was steered again by RSS, and
> because current-cpu was equal to desired-cpu,  ndo_rx_flow_steer
> wasn't called and no rule was configured to the HW.

Fix this by setting the flow's current CPU only in the table for the
newly selected RX queue.

Reported-and-tested-by: Amir Vadai <amirv@dev.mellanox.co.il>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonetdev/phy/icplus: Use mdiobus_write() and mdiobus_read() for proper locking.
David Daney [Fri, 30 Sep 2011 12:17:48 +0000 (12:17 +0000)]
netdev/phy/icplus: Use mdiobus_write() and mdiobus_read() for proper locking.

Usually you have to take the bus lock.  Why not here too?

I saw this when working on something else.  Not even compile tested.

Signed-off-by: David Daney <david.daney@cavium.com>
Cc: Greg Dietsche <Gregory.Dietsche@cuw.edu>
Cc: "Uwe Kleine-Konig" <u.kleine-koenig@pengutronix.de>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonetdev/phy: Use mdiobus_read() so that proper locks are taken.
David Daney [Fri, 30 Sep 2011 11:51:22 +0000 (11:51 +0000)]
netdev/phy: Use mdiobus_read() so that proper locks are taken.

Accesses to the mdio busses must be done with the mdio_lock to ensure
proper operation.  Conveniently we have the helper function
mdiobus_read() to do that for us.  Lets use it in get_phy_id() instead
of accessing the bus without the lock held.

Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge branch 'master' of git://git.infradead.org/users/linville/wireless-next into...
John W. Linville [Fri, 30 Sep 2011 18:52:16 +0000 (14:52 -0400)]
Merge branch 'master' of git://git.infradead.org/users/linville/wireless-next into for-davem

Conflicts:
drivers/net/wireless/iwlwifi/iwl-pci.c
drivers/net/wireless/wl12xx/main.c

13 years agobna: Driver Version changed to 3.0.2.2
Rasesh Mody [Tue, 27 Sep 2011 10:39:11 +0000 (10:39 +0000)]
bna: Driver Version changed to 3.0.2.2

Signed-off-by: Gurunatha Karaje <gkaraje@brocade.com>
Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobna: Add Callback to Fix RXQ Stop
Rasesh Mody [Tue, 27 Sep 2011 10:39:10 +0000 (10:39 +0000)]
bna: Add Callback to Fix RXQ Stop

Change details:
 - Add a callback in the BNA, which is called before sending FW command to stop
   RxQs. After this callback is called, driver should not post anymore Rx
   buffers to the RxQ. This addresses a small window where driver posts Rx
   buffers while FW is stopping/has stopped the RxQ.
 - Registering callback function, rx_stall_cbfn, during bna_rx_create.
   Invoking callback function, rx_stall_cbfn, before sending rx_cfg_clr
   command to FW
 - Bnad_cb_rx_stall implementation - set a flag in the Rxq to mark buffer
   posting disabled state. While posting buffers check for the above flag.

Signed-off-by: Gurunatha Karaje <gkaraje@brocade.com>
Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobna: PLL Init Fix and Add Stats Attributes
Rasesh Mody [Tue, 27 Sep 2011 10:39:09 +0000 (10:39 +0000)]
bna: PLL Init Fix and Add Stats Attributes

Change details:
 - Fix to release soft reset in PLL init for HW
 - Added stats attributes and new bfi msg class
 - Removed some unused code and typo fixes

Signed-off-by: Gurunatha Karaje <gkaraje@brocade.com>
Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobna: Brocade 1860 HW Enablement
Rasesh Mody [Tue, 27 Sep 2011 10:39:08 +0000 (10:39 +0000)]
bna: Brocade 1860 HW Enablement

This patch enables new HW Brocade 1860. Add BFA_CM_NIC capability mask to
bfa_ioc_attr, Sub-System Device ID Info and support for Brocade 1860 device
ID to bfa_ioc.c and bnad.c.

Signed-off-by: Gurunatha Karaje <gkaraje@brocade.com>
Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobna: Implement FW Download for New HW
Rasesh Mody [Tue, 27 Sep 2011 10:39:07 +0000 (10:39 +0000)]
bna: Implement FW Download for New HW

Add new device ID 0x22 and new asic generation BFI_ASIC_GEN_CT2 for 1860.
Implement FW download from user space for new Brocade HW.

Signed-off-by: Gurunatha Karaje <gkaraje@brocade.com>
Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobna: Capability Map and MFG Block Changes for New HW
Rasesh Mody [Tue, 27 Sep 2011 10:39:06 +0000 (10:39 +0000)]
bna: Capability Map and MFG Block Changes for New HW

Add capability map and generic model name scheme for manufacturing block.
Add card types for new HW.

Remove bfa_mfg_is_card_type_valid and ibfa_mfg_adapter_prop_init_flash_ct
macros.

Signed-off-by: Gurunatha Karaje <gkaraje@brocade.com>
Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobna: Brocade 1860 IOC PLL, Reg Defs and ASIC Mode Changes
Rasesh Mody [Tue, 27 Sep 2011 10:39:05 +0000 (10:39 +0000)]
bna: Brocade 1860 IOC PLL, Reg Defs and ASIC Mode Changes

Add logic to set ASIC specfic interface in IOC, HW interface initialization
APIs, mode based initialization and MSI-X resource allocation for 1860 with
no asic block. Add new h/w specific register definitions and setup registers
used by IOC logic.

Use normal kernel declaration style, c99 initializers and const for mailbox
structures. Remove unneeded parentheses.

Signed-off-by: Gurunatha Karaje <gkaraje@brocade.com>
Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge git://github.com/Jkirsher/net-next
David S. Miller [Thu, 29 Sep 2011 19:41:20 +0000 (15:41 -0400)]
Merge git://github.com/Jkirsher/net-next

13 years agousbnet: add timestamping support
Michael Riesch [Thu, 29 Sep 2011 04:06:26 +0000 (04:06 +0000)]
usbnet: add timestamping support

In order to make USB-to-Ethernet-adapters (depending on usbnet) support
timestamping, the "skb_defer_rx_timestamp" and "skb_tx_timestamp" function
calls are added.

Signed-off-by: Michael Riesch <michael@riesch.at>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet/fec: add poll controller function for fec nic
Xiao Jiang [Thu, 29 Sep 2011 02:15:57 +0000 (02:15 +0000)]
net/fec: add poll controller function for fec nic

Add poll controller function for fec nic.

Signed-off-by: Xiao Jiang <jgq516@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet/fec: replace hardcoded irq num with macro
Xiao Jiang [Thu, 29 Sep 2011 02:15:56 +0000 (02:15 +0000)]
net/fec: replace hardcoded irq num with macro

Don't use hardcoded irq num and replace it with
FEC_IRQ_NUM macro.

Signed-off-by: Xiao Jiang <jgq516@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoixgbe: add ECC warning for legacy interrupts
Don Skidmore [Thu, 4 Aug 2011 02:07:48 +0000 (02:07 +0000)]
ixgbe: add ECC warning for legacy interrupts

Noticed that the legacy Interrupt handler didn't have the same
ECC warning as did the MSI.  So this patch adds it.

Signed-off-by: Don Skidmore <donald.c.skidmore>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: cleanup ixgbe_setup_gpie() for X540
Don Skidmore [Wed, 17 Aug 2011 10:15:21 +0000 (10:15 +0000)]
ixgbe: cleanup ixgbe_setup_gpie() for X540

The X540 thermal sensor interrupt isn't a General Purpose Interrupt
so doesn't need to be enabled in ixgbe_setup_gpie().  Likewise X540 doesn't
use the SDP0 for thermal sensor so it doesn't need to be enabled for any
device other than 82599.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe add thermal sensor support for x540 hardware
Jacob Keller [Sat, 20 Aug 2011 04:49:45 +0000 (04:49 +0000)]
ixgbe add thermal sensor support for x540 hardware

Add code to enable thermal sensors for the x540 hardware, as well as a
thermal interrupt check which will exit with a critical message of a
thermal overheat is detected. Intent of code allows other mac types to
be added with different configuration in the future.

Fixed in this version is the addition of setting the temp_sensor
capable flag which was previously only set for a specific mac.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: update {P}FC thresholds to account for X540 and loopback
John Fastabend [Tue, 23 Aug 2011 03:14:22 +0000 (03:14 +0000)]
ixgbe: update {P}FC thresholds to account for X540 and loopback

Revise high and low threshold marks wrt flow control to account
for the X540 devices and latency introduced by the loopback
switch.

Without this it was in theory possible to drop frames on a
supposedly lossless link with X540 or SR-IOV enabled.

Previously we used a magic number in a define to calculate the
threshold values. This made it difficult to sort out exactly
which latencies were or were not being accounted for. Here
I was overly explicit and tried to used #define names that would
be recognizable after reading the IEEE 802.1Qbb specification.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: disable LLI for FCoE
Vasu Dev [Thu, 18 Aug 2011 06:20:07 +0000 (06:20 +0000)]
ixgbe: disable LLI for FCoE

Disable LLI for FCoE since regular interrupt
and their moderation rate works slightly better
for FCoE also.

Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbe: Cleanup q_vector interrupt throttle rate logic
Emil Tantilov [Wed, 31 Aug 2011 00:01:16 +0000 (00:01 +0000)]
ixgbe: Cleanup q_vector interrupt throttle rate logic

This patch is meant to help cleanup the interrupt throttle rate logic by
storing the interrupt throttle rate as a value in microseconds instead of
interrupts per second.  The advantage to this approach is that the value
can now be stored in an 16 bit field and doesn't require as much math to
flip the value back and forth since the hardware already used microseconds
when setting the rate.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoixgbevf: Fix broken trunk vlan
Greg Rose [Sat, 27 Aug 2011 02:06:25 +0000 (02:06 +0000)]
ixgbevf: Fix broken trunk vlan

Changes to clean up the vlan rx path broke trunk vlan.  Trunk vlans in
a VF driver are those set using:

"ip link set <pfdev> vf <n> <vlanid>"

Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
CC: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agoe1000: don't enable dma receives until after dma address has been setup
Dean Nelson [Fri, 16 Sep 2011 16:52:54 +0000 (16:52 +0000)]
e1000: don't enable dma receives until after dma address has been setup

Doing an 'ifconfig ethN down' followed by an 'ifconfig ethN up' on a qemu-kvm
guest system configured with two e1000 NICs can result in an 'unable to handle
kernel paging request at 0000000100000000' or 'bad page map in process ...' or
something similar.

These result from a 4096-byte page being corrupted with the following two-word
pattern (16-bytes) repeated throughout the entire page:

  0x0000000000000000
  0x0000000100000000

There can be other bits set as well. What is a constant is that the 2nd word
has the 32nd bit set. So one could see:

        :
  0x0000000000000000
  0x0000000100000000
  0x0000000000000000
  0x0000000172adc067    <<< bad pte
  0x800000006ec60067
  0x0000000700000040
  0x0000000000000000
  0x0000000100000000
        :

Which came from from a process' page table I dumped out when the marked line
was seen as bad by print_bad_pte().

The repeating pattern represents the e1000's two-word receive descriptor:

struct e1000_rx_desc {
        __le64 buffer_addr;   /* Address of the descriptor's data buffer */
        __le16 length;        /* Length of data DMAed into data buffer */
        __le16 csum;          /* Packet checksum */
        u8 status;            /* Descriptor status */
        u8 errors;            /* Descriptor Errors */
        __le16 special;
};

And the 32nd bit of the 2nd word maps to the 'u8 status' member, and
corresponds to E1000_RXD_STAT_DD which indicates the descriptor is done.

The corruption appears to result from the following...

 . An 'ifconfig ethN down' gets us into e1000_close(), which through a number
   of subfunctions results in:
     1. E1000_RCTL_EN being cleared in RCTL register.  [e1000_down()]
     2. dma_free_coherent() being called.  [e1000_free_rx_resources()]

 . An 'ifconfig ethN up' gets us into e1000_open(), which through a number of
   subfunctions results in:
     1. dma_alloc_coherent() being called.  [e1000_setup_rx_resources()]
     2. E1000_RCTL_EN being set in RCTL register.  [e1000_setup_rctl()]
     3. E1000_RCTL_EN being cleared in RCTL register.  [e1000_configure_rx()]
     4. RDLEN, RDBAH and RDBAL registers being set to reflect the dma page
        allocated in step 1.  [e1000_configure_rx()]
     5. E1000_RCTL_EN being set in RCTL register.  [e1000_configure_rx()]

During the 'ifconfig ethN up' there is a window opened, starting in step 2
where the receives are enabled up until they are disabled in step 3, in which
the address of the receive descriptor dma page known by the NIC is still the
previous one which was freed during the 'ifconfig ethN down'. If this memory
has been reallocated for some other use and the NIC feels so inclined, it will
write to that former dma page with predictably unpleasant results.

I realize that in the guest, we're dealing with an e1000 NIC that is software
emulated by qemu-kvm. The problem doesn't appear to occur on bare-metal. Andy
suspects that this is because in the emulator link-up is essentially instant
and traffic can start flowing immediately. Whereas on bare-metal, link-up
usually seems to take at least a few milliseconds. And this might be enough
to prevent traffic from flowing into the device inside the window where
E1000_RCTL_EN is set.

So perhaps a modification needs to be made to the qemu-kvm e1000 NIC emulator
to delay the link-up. But in defense of the emulator, it seems like a bad idea
to enable dma operations before the address of the memory to be involved has
been made known.

The following patch no longer enables receives in e1000_setup_rctl() but leaves
them however they were. It only enables receives in e1000_configure_rx(), and
only after the dma address has been made known to the hardware.

There are two places where e1000_setup_rctl() gets called. The one in
e1000_configure() is followed immediately by a call to e1000_configure_rx(), so
there's really no change functionally (except for the removal of the problem
window. The other is in __e1000_shutdown() and is not followed by a call to
e1000_configure_rx(), so there is a change functionally. But consider...

 . An 'ifconfig ethN down' (just as described above).

 . A 'suspend' of the system, which (I'm assuming) will find its way into
   e1000_suspend() which calls __e1000_shutdown() resulting in:
     1. E1000_RCTL_EN being set in RCTL register.  [e1000_setup_rctl()]

And again we've re-opened the problem window for some unknown amount of time.

Signed-off-by: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Dean Nelson <dnelson@redhat.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
13 years agocandev: allow SJW user setting for bittiming calculation
Oliver Hartkopp [Wed, 28 Sep 2011 02:50:11 +0000 (02:50 +0000)]
candev: allow SJW user setting for bittiming calculation

This patch adds support for SJW user settings to not set the synchronization
jump width (SJW) to 1 in any case when using the in-kernel bittiming
calculation.

The ip-tool from iproute2 already supports to pass the user defined SJW
value. The given SJW value is sanitized with the controller specific sjw_max
and the calculated tseg2 value. As the SJW can have values up to 4 providing
this value will lead to the maximum possible SJW automatically. A higher SJW
allows higher controller oscillator tolerances.

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: sh_eth: move the asm/sh_eth.h to include/linux/
Yoshihiro Shimoda [Tue, 27 Sep 2011 21:49:12 +0000 (21:49 +0000)]
net: sh_eth: move the asm/sh_eth.h to include/linux/

Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agosh: modify prototype in sh_eth.h
Yoshihiro Shimoda [Tue, 27 Sep 2011 21:49:05 +0000 (21:49 +0000)]
sh: modify prototype in sh_eth.h

Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: sh_eth: use ioremap()
Yoshihiro Shimoda [Tue, 27 Sep 2011 21:48:58 +0000 (21:48 +0000)]
net: sh_eth: use ioremap()

This patch also changes writel/readl to iowrite32/ioread32.

Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agocan/sja1000: add driver for EMS PCMCIA card
Oliver Hartkopp [Fri, 23 Sep 2011 06:59:48 +0000 (06:59 +0000)]
can/sja1000: add driver for EMS PCMCIA card

This patch adds the driver for the SJA1000 based PCMCIA card 'CPC-Card' from
EMS Dr. Thomas Wuensche (http://www.ems-wuensche.de).

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Markus Plessing <plessing@ems-wuensche.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoconnector: add comm change event report to proc connector
Vladimir Zapolskiy [Wed, 21 Sep 2011 09:26:44 +0000 (09:26 +0000)]
connector: add comm change event report to proc connector

Add an event to monitor comm value changes of tasks.  Such an event
becomes vital, if someone desires to control threads of a process in
different manner.

A natural characteristic of threads is its comm value, and helpfully
application developers have an opportunity to change it in runtime.
Reporting about such events via proc connector allows to fine-grain
monitoring and control potentials, for instance a process control daemon
listening to proc connector and following comm value policies can place
specific threads to assigned cgroup partitions.

It might be possible to achieve a pale partial one-shot likeness without
this update, if an application changes comm value of a thread generator
task beforehand, then a new thread is cloned, and after that proc
connector listener gets the fork event and reads new thread's comm value
from procfs stat file, but this change visibly simplifies and extends the
matter.

Signed-off-by: Vladimir Zapolskiy <vzapolskiy@gmail.com>
Acked-by: Evgeniy Polyakov <zbr@ioremap.net>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: rps: fix the support for PPPOE
Changli Gao [Tue, 20 Sep 2011 22:36:07 +0000 (22:36 +0000)]
net: rps: fix the support for PPPOE

The upper protocol numbers of PPPOE are different, and should be treated
specially.

Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoaf_unix: dont send SCM_CREDENTIALS by default
Eric Dumazet [Mon, 19 Sep 2011 05:52:27 +0000 (05:52 +0000)]
af_unix: dont send SCM_CREDENTIALS by default

Since commit 7361c36c5224 (af_unix: Allow credentials to work across
user and pid namespaces) af_unix performance dropped a lot.

This is because we now take a reference on pid and cred in each write(),
and release them in read(), usually done from another process,
eventually from another cpu. This triggers false sharing.

# Events: 154K cycles
#
# Overhead  Command       Shared Object        Symbol
# ........  .......  ..................  .........................
#
    10.40%  hackbench  [kernel.kallsyms]   [k] put_pid
     8.60%  hackbench  [kernel.kallsyms]   [k] unix_stream_recvmsg
     7.87%  hackbench  [kernel.kallsyms]   [k] unix_stream_sendmsg
     6.11%  hackbench  [kernel.kallsyms]   [k] do_raw_spin_lock
     4.95%  hackbench  [kernel.kallsyms]   [k] unix_scm_to_skb
     4.87%  hackbench  [kernel.kallsyms]   [k] pid_nr_ns
     4.34%  hackbench  [kernel.kallsyms]   [k] cred_to_ucred
     2.39%  hackbench  [kernel.kallsyms]   [k] unix_destruct_scm
     2.24%  hackbench  [kernel.kallsyms]   [k] sub_preempt_count
     1.75%  hackbench  [kernel.kallsyms]   [k] fget_light
     1.51%  hackbench  [kernel.kallsyms]   [k]
__mutex_lock_interruptible_slowpath
     1.42%  hackbench  [kernel.kallsyms]   [k] sock_alloc_send_pskb

This patch includes SCM_CREDENTIALS information in a af_unix message/skb
only if requested by the sender, [man 7 unix for details how to include
ancillary data using sendmsg() system call]

Note: This might break buggy applications that expected SCM_CREDENTIAL
from an unaware write() system call, and receiver not using SO_PASSCRED
socket option.

If SOCK_PASSCRED is set on source or destination socket, we still
include credentials for mere write() syscalls.

Performance boost in hackbench : more than 50% gain on a 16 thread
machine (2 quad-core cpus, 2 threads per core)

hackbench 20 thread 2000

4.228 sec instead of 9.102 sec

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge branch 'master' of git://git.infradead.org/users/linville/wireless
John W. Linville [Wed, 28 Sep 2011 14:28:09 +0000 (10:28 -0400)]
Merge branch 'master' of git://git.infradead.org/users/linville/wireless

Conflicts:
drivers/net/wireless/iwlwifi/iwl-scan.c
net/wireless/nl80211.c

13 years agomac80211: treat the WME sta flag as a bit
Arik Nemtsov [Mon, 26 Sep 2011 06:36:42 +0000 (09:36 +0300)]
mac80211: treat the WME sta flag as a bit

Correct flag usage - use it as a bit index instead of a bit value.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agowl3501_cs: min_t() cast truncates high bits
Dan Carpenter [Mon, 26 Sep 2011 06:30:40 +0000 (09:30 +0300)]
wl3501_cs: min_t() cast truncates high bits

wrqu->encoding.length comes from the network administrator.  It's
size u16.  We want to limit "tocopy" to the smallest value of either
"len_keys", "wrqu->encoding.length" or 100.  But because .length
gets cast from u16 to u8 we might use a random, smaller value than
the was desired.  It's probably not very serious, but we may as well
fix it.

Btw, this is from code auditing and not from testing.  I don't know
if this affects anyone in real life.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomac80211: save tx params per sdata
Eliad Peller [Sun, 25 Sep 2011 17:06:54 +0000 (20:06 +0300)]
mac80211: save tx params per sdata

save and configure tx param per sdata, rather than
per hardware.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agocfg80211/mac80211: add netdev param to set_txq_params()
Eliad Peller [Sun, 25 Sep 2011 17:06:53 +0000 (20:06 +0300)]
cfg80211/mac80211: add netdev param to set_txq_params()

tx params are currently configured per hw, although they
should be configured per interface.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomac80211: Send the management frame at requested rate
Rajkumar Manoharan [Sun, 25 Sep 2011 09:23:31 +0000 (14:53 +0530)]
mac80211: Send the management frame at requested rate

Whenever the scan request or tx_mgmt is requesting not to
use CCK rate for managemet frames through
NL80211_ATTR_TX_NO_CCK_RATE attribute, then mac80211 should
select appropriate least non-CCK rate. This could help to
send P2P probes and P2P action frames at non 11b rates
without diabling 11b rates globally.

Cc: Jouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: Rajkumar Manoharan <rmanohar@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agonl80211/cfg80211: Add support to disable CCK rate for management frame
Rajkumar Manoharan [Sun, 25 Sep 2011 09:23:30 +0000 (14:53 +0530)]
nl80211/cfg80211: Add support to disable CCK rate for management frame

Add a new nl80211 attribute to specify whether to send the management
frames in CCK rate or not. As of now the wpa_supplicant is disabling
CCK rate at P2P init itself. So this patch helps to send P2P probe
request/probe response/action frames being sent at non CCK rate in 2GHz
without disabling 11b rates.

This attribute is used with NL80211_CMD_TRIGGER_SCAN and
NL80211_CMD_FRAME commands to disable CCK rate for management frame
transmission.

Cc: Jouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: Rajkumar Manoharan <rmanohar@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoNFC: use after free on error
Dan Carpenter [Fri, 23 Sep 2011 06:14:35 +0000 (09:14 +0300)]
NFC: use after free on error

We returned a freed variable on some error paths when the intent was
to return a NULL.  Part of the reason this was missed was that the
code was confusing because it had too many gotos so I removed them
and simplified the flow a bit.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Lauro Ramos Venancio <lauro.venancio@openbossa.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: fix scan complete processing
Johannes Berg [Thu, 22 Sep 2011 22:15:00 +0000 (15:15 -0700)]
iwlagn: fix scan complete processing

When we cancel a scan, the completion runs
only from the workqueue. This can cause the
remain-on-channel scan to fail when another
one was just canceled, because we're still
aborting it.

To fix this, run the completion inline with
the lock still held before returning from
iwl_scan_cancel_timeout().

Also, to avoid the scan complete work from
completing a new scan prematurely, add a
new STATUS_SCAN_COMPLETE bit.

Reported-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Tested-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: move iwl_process_scan_complete up
Johannes Berg [Thu, 22 Sep 2011 22:14:59 +0000 (15:14 -0700)]
iwlagn: move iwl_process_scan_complete up

To make the next patch easier to read, move
the function up, it'll be needed earlier in
this file in the next patch.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: refactor scan complete
Johannes Berg [Thu, 22 Sep 2011 22:14:58 +0000 (15:14 -0700)]
iwlagn: refactor scan complete

We'll need to be able to run scan complete
inline, not from the workqueue, so refactor
it.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: make iwl_scan_cancel_timeout void
Johannes Berg [Thu, 22 Sep 2011 22:14:57 +0000 (15:14 -0700)]
iwlagn: make iwl_scan_cancel_timeout void

The return value of iwl_scan_cancel_timeout()
isn't used anywhere, so let's just remove it.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: remove Kelvin support
Johannes Berg [Thu, 22 Sep 2011 22:14:56 +0000 (15:14 -0700)]
iwlagn: remove Kelvin support

Only 5150 series devices report their temperature
in Kelvin, and for those we already convert it to
Celsius when storing into priv->temperature, so
there's no way priv->temperature will ever be in
Kelvin. Remove support for this.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: fix slot programming
Johannes Berg [Thu, 22 Sep 2011 22:14:55 +0000 (15:14 -0700)]
iwlagn: fix slot programming

When an AP mode interface is added with a DTIM
period of two, the slot programming is wrong.
Fix it by taking into account the DTIM period.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: fix dangling scan request
Johannes Berg [Thu, 22 Sep 2011 22:14:54 +0000 (15:14 -0700)]
iwlagn: fix dangling scan request

If iwl_scan_initiate() fails for any reason,
priv->scan_request and priv->scan_vif are left
dangling. This can lead to a crash later when
iwl_bg_scan_completed() tries to run a pending
scan request.

In practice, this seems to be very rare due to
the STATUS_SCANNING check earlier. That check,
however, is wrong -- it should allow a scan to
be queued when a reset/roc scan is going on.
When a normal scan is already going on, a new
one can't be issued by mac80211, so that code
can be removed completely. I introduced this
bug when adding off-channel support in commit
266af4c745952e9bebf687dd68af58df553cb59d.

Cc: stable@kernel.org [3.0]
Reported-by: Peng Yan <peng.yan@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: use kcalloc when possible for array allocation
Emmanuel Grumbach [Thu, 22 Sep 2011 22:14:53 +0000 (15:14 -0700)]
iwlagn: use kcalloc when possible for array allocation

As everybody knows kcalloc checks the multiplication is safe and
that we don't run into overflow.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: update rate scaling with BA notifications
Emmanuel Grumbach [Thu, 22 Sep 2011 22:14:52 +0000 (15:14 -0700)]
iwlagn: update rate scaling with BA notifications

In the current code, the rate scaling isn't fed with
statistics from the BA notifications.

This is since my patch:

iwlagn: reclaim the packets in transport layer

Fix that.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: add debugging to show probe related info in scan notification
Wey-Yi Guy [Thu, 22 Sep 2011 22:14:51 +0000 (15:14 -0700)]
iwlagn: add debugging to show probe related info in scan notification

Add debugging to show the status of probe in scan notification
to help debug probe related issues

Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: set the sequence control from the transport layer
Emmanuel Grumbach [Thu, 22 Sep 2011 22:14:50 +0000 (15:14 -0700)]
iwlagn: set the sequence control from the transport layer

Since all the queue logic has been moved to the transport layer,
the sequence number is set in the transport layer.
While doing that I forgot that the mac header is copied to the
TB of the TX cmd in the upper layer before the call to the transport
layer. So basically we used the sequence number from mac80211...
This was fine for the first assocation but after the second, mac80211
resets its counters while we don't hence a shift that led to terrible
impact on performance.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoMAINTAINERS: update iwlwifi
Wey-Yi Guy [Thu, 22 Sep 2011 22:14:49 +0000 (15:14 -0700)]
MAINTAINERS: update iwlwifi

Change to iwlwifi.git instead of iwlwifi-2.6.git
iwlwifi-2.6.git still works for backward compatibility

Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: sparse warning priv->temperature is signed
Emmanuel Grumbach [Thu, 22 Sep 2011 14:15:37 +0000 (07:15 -0700)]
iwlagn: sparse warning priv->temperature is signed

Since priv->temperature is signed, we cannot use debugfs_create_u32
to refer to it.
Use a regular debugfs file instead.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: remove warning in iwl_rx_handle
Emmanuel Grumbach [Thu, 22 Sep 2011 14:15:36 +0000 (07:15 -0700)]
iwlagn: remove warning in iwl_rx_handle

Txid was used without being initialized.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoath9k: fix a regression in ath9k_ps_restore
Felix Fietkau [Thu, 22 Sep 2011 14:04:32 +0000 (08:04 -0600)]
ath9k: fix a regression in ath9k_ps_restore

After 'ath9k: optimize ath9k_ps_restore', it would only send the device to
network sleep and not to full sleep anymore, potentially causing more
battery drain.

Reported-by: Vivek Natarajan <nataraja@qca.qualcomm.com>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoNFC: protect nci_data_exchange transactions
Ilan Elias [Thu, 22 Sep 2011 08:36:19 +0000 (11:36 +0300)]
NFC: protect nci_data_exchange transactions

Protect 'cb' and 'cb_context' arguments in nci_data_exchange.
In fact, this implements a queue with max length of 1 data
exchange transactions in parallel.

Signed-off-by: Ilan Elias <ilane@ti.com>
Acked-by: Lauro Ramos Venancio <lauro.venancio@openbossa.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoNFC: implicitly deactivate in nci_start_poll
Ilan Elias [Thu, 22 Sep 2011 08:13:01 +0000 (11:13 +0300)]
NFC: implicitly deactivate in nci_start_poll

When start_poll is called, and a target was implicitly activated,
we need to implicitly deactivate it.
On the other hand, when the target was activated by the user,
we should not deactivate it.

Signed-off-by: Ilan Elias <ilane@ti.com>
Acked-by: Lauro Ramos Venancio <lauro.venancio@openbossa.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoNFC: improve readability of an 'if' in nci core.c
Ilan Elias [Thu, 22 Sep 2011 07:47:52 +0000 (10:47 +0300)]
NFC: improve readability of an 'if' in nci core.c

Signed-off-by: Ilan Elias <ilane@ti.com>
Acked-by: Lauro Ramos Venancio <lauro.venancio@openbossa.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomwifiex: reset skb length before inserting to free queue
Amitkumar Karwar [Thu, 22 Sep 2011 04:43:25 +0000 (21:43 -0700)]
mwifiex: reset skb length before inserting to free queue

After handling command response, cmd skb is inserted into command
free queue(which keeps track of availabile skbs) for reuse purpose.
Skb length is not getting reset to zero here. This patch takes care
of it.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomwifiex: pass correct band parameter to ieee80211_channel_to_frequency()
Amitkumar Karwar [Thu, 22 Sep 2011 04:43:24 +0000 (21:43 -0700)]
mwifiex: pass correct band parameter to ieee80211_channel_to_frequency()

ieee80211_channel_to_frequency() routine expects band parameter in
the form of "enum ieee80211_band band". Currently driver specific
band (BAND_A, BAND_AN etc.) is passed to the routine.

This patch makes sure that correct parameter is passed.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomwifiex: update bss band information
Amitkumar Karwar [Thu, 22 Sep 2011 04:43:23 +0000 (21:43 -0700)]
mwifiex: update bss band information

In recent commit "mwifiex: use cfg80211 dynamic scan..."
(7c6fa2a843..) scan table handling in driver is removed to
make use of cfg80211 dynamic scan table. Now driver sends
beacon buffers found in scanning directly to stack and parse
the buffer for requested BSS only during association.

Beacon buffer doesn't contain bss band information. Driver
gets it from firmware in separate tlv (chan_band_tlv).
Currently since we don't inform stack about bss bandinfo,
there is an issue with 5GHz association.

Use "priv" field of struct cfg80211_bss to store bandinfo.
This fixes 5GHz association issue.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomwifiex: fix 5GHz association issue
Amitkumar Karwar [Thu, 22 Sep 2011 04:43:22 +0000 (21:43 -0700)]
mwifiex: fix 5GHz association issue

Sometimes association in 5GHz doesn't work. Dmesg log shows
"Can not find requested SSID xyz" error message. Currently
while preparing scan channel list for firmware Null entries
are created for disabled channels. The routine which retrieves
this list ignores channels after Null entry. Hence sometimes
driver doesn't scan the channel of requested AP and association
fails. The issue is fixed by avoiding those NULL entries.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agob43: LCN-PHY: minor clean ups
Rafał Miłecki [Wed, 21 Sep 2011 19:44:15 +0000 (21:44 +0200)]
b43: LCN-PHY: minor clean ups

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agob43: update dummy transmission
Rafał Miłecki [Wed, 21 Sep 2011 19:44:14 +0000 (21:44 +0200)]
b43: update dummy transmission

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agob43: add missing MMIO defines
Rafał Miłecki [Wed, 21 Sep 2011 19:44:13 +0000 (21:44 +0200)]
b43: add missing MMIO defines

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agolibertas: scan behaviour consistency improvements
Daniel Drake [Wed, 21 Sep 2011 17:43:59 +0000 (18:43 +0100)]
libertas: scan behaviour consistency improvements

When scanning for the broadcast SSID, there is no need to add the
SSID TLV (restoring the behaviour of the driver behaviour in the wext
days, confirmed in Marvell specifications).

If bssid is unspecified, the current scan code will usually fire off an
active scan probing for the specific requested SSID. However, if a scan
is ongoing (or has just finished), those scan results will be used
instead (even if that scan is totally different, e.g. a passive scan on
channel 4 for a different SSID). Fix this inconsistency by always
firing off a scan when associating without a bssid.

Signed-off-by: Daniel Drake <dsd@laptop.org>
Acked-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agocfg80211: Validate cipher suite against supported ciphers
Jouni Malinen [Wed, 21 Sep 2011 15:14:56 +0000 (18:14 +0300)]
cfg80211: Validate cipher suite against supported ciphers

Instead of using a hardcoded list of cipher suites in nl80211.c, use a
shared function in util.c to verify that the driver advertises support
for the specified cipher. This provides more accurate validation of the
values and allows vendor-specific cipher suites to be added in drivers.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agocfg80211: Remove strict validation of AKM suites
Jouni Malinen [Wed, 21 Sep 2011 15:11:33 +0000 (18:11 +0300)]
cfg80211: Remove strict validation of AKM suites

NL80211_ATTR_AKM_SUITES can be used to configure new AKMs, like FT or
the SHA-256 -based AKMs or FT from 802.11r/802.11w. In addition, vendor
specific AKMs could be used. The current validation code for the connect
command prevents cfg80211-based drivers from using these mechanisms even
if the driver would not actually use this AKM value (i.e., it uses
WPA/RSN IE from user space). mac80211-based drivers allow any AKM to be
used since this value is not used there.

Remove the unnecessary validation step in cfg80211 to allow drivers to
decide what AKMs are supported. In theory, we could handle this by
advertising supported AKMs, but that would not be very effective unless
we enforce all drivers (including mac80211) to advertise the set of
supported AKMs. This would require additional changes in many places
whenever a new AKM is introduced even though no actually functionality
changes may be required in most drivers.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agomac80211: add ieee80211_vif param to tsf functions
Eliad Peller [Wed, 21 Sep 2011 11:06:11 +0000 (14:06 +0300)]
mac80211: add ieee80211_vif param to tsf functions

TSF can be kept per vif.
Add ieee80211_vif param to set/get/reset_tsf, and move
the debugfs entries to the per-vif directory.

Update all the drivers that implement these callbacks.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoath9k: add Block ACK bitmap in sample debug
Mohammed Shafi Shajakhan [Wed, 21 Sep 2011 08:52:49 +0000 (14:22 +0530)]
ath9k: add Block ACK bitmap in sample debug

this represents the bitmap of block ACK received after the
successful transmission of an aggregate frame. also made few
changes to beautify the display

Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoMerge branch 'for-linville' of git://github.com/lucacoelho/wl12xx
John W. Linville [Tue, 27 Sep 2011 18:20:23 +0000 (14:20 -0400)]
Merge branch 'for-linville' of git://github.com/lucacoelho/wl12xx

13 years agoMerge branch 'master' of git://github.com/padovan/bluetooth-next
John W. Linville [Tue, 27 Sep 2011 18:15:00 +0000 (14:15 -0400)]
Merge branch 'master' of git://github.com/padovan/bluetooth-next

13 years agoskge: handle irq better on single port card
Stephen Hemminger [Tue, 27 Sep 2011 17:41:37 +0000 (13:41 -0400)]
skge: handle irq better on single port card

Most boards with SysKonnect/Marvell Ethernet have only a single port.
For the single port case, use the standard Ethernet driver convention
of allocating IRQ when device is brought up rather than at probe time.

This patch also adds some additional read after writes to avoid any
PCI posting problems when setting the IRQ mask.

The error handling of dual port cards is also changed.  If second port
can not be brought up, then just fail. No point in continuing, since
the failure is most certainly because of out of memory.

It is worth noting that the dual port skge device has a single irq but two
seperate status rings and therefore has two NAPI objects, one for
each port.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobe2net: Show newly flashed FW ver in ethtool
Sathya Perla [Tue, 27 Sep 2011 17:30:27 +0000 (13:30 -0400)]
be2net: Show newly flashed FW ver in ethtool

This fix provides a newly flashed FW version (appended, in braces)
along with the currently running FW version via ethtool. The newly
flashed version runs only after a system reset.

Signed-off-by: Suresh Reddy <Suresh.Reddy@emulex.com>
Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agobe2net: fix multicast filter programming
Sathya Perla [Tue, 27 Sep 2011 17:29:38 +0000 (13:29 -0400)]
be2net: fix multicast filter programming

 Re-posting with subject fixed!

 Multicast programming has been broken since commit 5b8821b7. Setting the
 MULTICAST flag while sending the cmd to the FW was missing. Fixed this.

Also fixed-up some indentation in the adjacent lines.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agotcp: rename tcp_skb_cb flags
Eric Dumazet [Tue, 27 Sep 2011 17:25:05 +0000 (13:25 -0400)]
tcp: rename tcp_skb_cb flags

Rename struct tcp_skb_cb "flags" to "tcp_flags" to ease code review and
maintenance.

Its content is a combination of FIN/SYN/RST/PSH/ACK/URG/ECE/CWR flags

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agotcp: unalias tcp_skb_cb flags and ip_dsfield
Eric Dumazet [Tue, 27 Sep 2011 06:20:08 +0000 (02:20 -0400)]
tcp: unalias tcp_skb_cb flags and ip_dsfield

struct tcp_skb_cb contains a "flags" field containing either tcp flags
or IP dsfield depending on context (input or output path)

Introduce ip_dsfield to make the difference clear and ease maintenance.
If later we want to save space, we can union flags/ip_dsfield

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoenic: Add support for port profile association on a enic SRIOV VF
Roopa Prabhu [Thu, 22 Sep 2011 03:44:43 +0000 (03:44 +0000)]
enic: Add support for port profile association on a enic SRIOV VF

This patch touchs most of the enic port profile handling code.
Tried to break it into sub patches without success.

The patch mainly does the following:
- Port profile operations for a SRIOV VF are modified to work
  only via its PF
- Changes the port profile static struct in struct enic to a pointer.
  This is because a SRIOV PF has to now hold the port profile information
  for all its VF's
- Moved address registration for VF's during port profile ASSOCIATE time
- Most changes in port profile handling code are changes related to indexing
  into the port profile struct array of a PF for the VF port profile
  information

Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: Sujith Sankar <ssujith@cisco.com>
Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoenic: Helper code for SRIOV proxy commands
Roopa Prabhu [Thu, 22 Sep 2011 03:44:38 +0000 (03:44 +0000)]
enic: Helper code for SRIOV proxy commands

This patch adds helper functions to use PF as proxy for SRIOV VF firmware
commands.

Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: Sujith Sankar <ssujith@cisco.com>
Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoenic: Add SRIOV support
Roopa Prabhu [Thu, 22 Sep 2011 03:44:33 +0000 (03:44 +0000)]
enic: Add SRIOV support

This patch adds support to enable SRIOV on enic devices. Enic SRIOV VF's are dynamic vnics and will use the same driver code as dynamic vnics.

Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: Sujith Sankar <ssujith@cisco.com>
Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agotcp: ECN blackhole should not force quickack mode
Eric Dumazet [Thu, 22 Sep 2011 20:02:19 +0000 (20:02 +0000)]
tcp: ECN blackhole should not force quickack mode

While playing with a new ADSL box at home, I discovered that ECN
blackhole can trigger suboptimal quickack mode on linux : We send one
ACK for each incoming data frame, without any delay and eventual
piggyback.

This is because TCP_ECN_check_ce() considers that if no ECT is seen on a
segment, this is because this segment was a retransmit.

Refine this heuristic and apply it only if we seen ECT in a previous
segment, to detect ECN blackhole at IP level.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Jamal Hadi Salim <jhs@mojatatu.com>
CC: Jerry Chu <hkchu@google.com>
CC: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
CC: Jim Gettys <jg@freedesktop.org>
CC: Dave Taht <dave.taht@gmail.com>
Acked-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agosky2: manage irq better on single port card
stephen hemminger [Fri, 23 Sep 2011 11:13:40 +0000 (11:13 +0000)]
sky2: manage irq better on single port card

Most sky2 hardware only has a single port, although some variations of the
chip support two interfaces.  For the single port case, use the standard
Ethernet driver convention of allocating IRQ when device is brought up
rather than at probe time.

Also, change the error handling of dual port cards so that if second
port can not be brought up, then just fail. No point in continuing, since
the failure is most certainly because of out of memory.

The dual port sky2 device has a single irq and a single status ring,
therefore it has a single NAPI object shared by both ports.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agopch_gbe: remove unused variable
Jon Mason [Mon, 26 Sep 2011 09:37:38 +0000 (09:37 +0000)]
pch_gbe: remove unused variable

netdev is unused in pch_gbe_setup_rctl.  Remove this declaration to
avoid a compiler warning.

Signed-off-by: Jon Mason <jdmason@kudzu.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoxen/pciback: Add flag indicating device has been assigned by Xen
Konrad Rzeszutek Wilk [Mon, 26 Sep 2011 16:22:01 +0000 (12:22 -0400)]
xen/pciback: Add flag indicating device has been assigned by Xen

Device drivers that create and destroy SR-IOV virtual functions via
calls to pci_enable_sriov() and pci_disable_sriov can cause catastrophic
failures if they attempt to destroy VFs while they are assigned to
guest virtual machines.  By adding a flag for use by the Xen PCI back
to indicate that a device is assigned a device driver can check that
flag and avoid destroying VFs while they are assigned and avoid system
failures.

Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoehea: Remove sleep at .ndo_get_stats
brenohl@br.ibm.com [Mon, 26 Sep 2011 10:11:03 +0000 (10:11 +0000)]
ehea: Remove sleep at .ndo_get_stats

Currently ehea ndo_get_stats can sleep in two places, in a hcall
and in a GFP_KERNEL alloc, which is not correct.
This patch creates a delayed workqueue that grabs the information each 1
sec from the hardware, and place it into the device structure, so that,
.ndo_get_stats quickly returns the device structure statistics block.

Signed-off-by: Breno Leitao <brenohl@br.ibm.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agodp83640: add time stamp insertion for sync messages
Richard Cochran [Tue, 20 Sep 2011 01:43:16 +0000 (01:43 +0000)]
dp83640: add time stamp insertion for sync messages

This commit adds one step support to the phyter. When enabled, the
hardware does not provide time stamps for transmitted sync messages but
instead inserts the stamp into the outgoing packet.

Signed-off-by: Richard Cochran <richard.cochran@omicron.at>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet: introduce ptp one step time stamp mode for sync packets
Richard Cochran [Tue, 20 Sep 2011 01:43:15 +0000 (01:43 +0000)]
net: introduce ptp one step time stamp mode for sync packets

The IEEE 1588 standard (PTP) has a provision for a "one step" mode, where
time stamps on outgoing event packets are inserted into the packet by the
hardware on the fly. This patch adds a new flag for the SIOCSHWTSTAMP
ioctl that lets user space programs request this mode.

Signed-off-by: Richard Cochran <richard.cochran@omicron.at>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agodp83640: enable six external events and one periodic output
Richard Cochran [Tue, 20 Sep 2011 01:43:14 +0000 (01:43 +0000)]
dp83640: enable six external events and one periodic output

This patch enables six external event channels and one periodic output.
One GPIO is reserved for synchronizing multiple PHYs. The assignment
of GPIO functions can be changed via a module parameter.

The code supports multiple simultaneous events by inducing a PTP clock
event for every channel marked in the PHY's extended status word.

Signed-off-by: Richard Cochran <richard.cochran@omicron.at>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoath9k: Fix a dma warning/memory leak
Mohammed Shafi Shajakhan [Fri, 23 Sep 2011 09:03:14 +0000 (14:33 +0530)]
ath9k: Fix a dma warning/memory leak

proper dma_unmapping and freeing of skb's has to be done in the rx
cleanup for EDMA chipsets when the device is unloaded and this also
seems to address the following warning which shows up occasionally when
the device is unloaded

Call Trace:
[<c0148cd2>] warn_slowpath_common+0x72/0xa0
[<c03b669c>] ? dma_debug_device_change+0x19c/0x200
[<c03b669c>] ? dma_debug_device_change+0x19c/0x200
[<c0148da3>] warn_slowpath_fmt+0x33/0x40
[<c03b669c>] dma_debug_device_change+0x19c/0x200
[<c0657f12>] notifier_call_chain+0x82/0xb0
[<c0171370>] __blocking_notifier_call_chain+0x60/0x90
[<c01713bf>] blocking_notifier_call_chain+0x1f/0x30
[<c044f594>] __device_release_driver+0xa4/0xc0
[<c044f647>] driver_detach+0x97/0xa0
[<c044e65c>] bus_remove_driver+0x6c/0xe0
[<c029af0b>] ? sysfs_addrm_finish+0x4b/0x60
[<c0450109>] driver_unregister+0x49/0x80
[<c0299f54>] ? sysfs_remove_file+0x14/0x20
[<c03c3ab2>] pci_unregister_driver+0x32/0x80
[<f92c2162>] ath_pci_exit+0x12/0x20 [ath9k]
[<f92c8467>] ath9k_exit+0x17/0x36 [ath9k]
[<c06523cd>] ? mutex_unlock+0xd/0x10
[<c018e27f>] sys_delete_module+0x13f/0x200
[<c02139bb>] ? sys_munmap+0x4b/0x60
[<c06547c5>] ? restore_all+0xf/0xf
[<c0657a20>] ? spurious_fault+0xe0/0xe0
[<c01832f4>] ? trace_hardirqs_on_caller+0xf4/0x180
[<c065b863>] sysenter_do_call+0x12/0x38
 ---[ end trace 16e1c1521c06bcf9 ]---
Mapped at:
[<c03b7938>] debug_dma_map_page+0x48/0x120
[<f92ba3e8>] ath_rx_init+0x3f8/0x4b0 [ath9k]
[<f92b5ae4>] ath9k_init_device+0x4c4/0x7b0 [ath9k]
[<f92c2813>] ath_pci_probe+0x263/0x330 [ath9k]

Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agortlwifi: rtl8192cu: Fix unitialized struct
Larry Finger [Fri, 23 Sep 2011 03:59:02 +0000 (22:59 -0500)]
rtlwifi: rtl8192cu: Fix unitialized struct

Driver rtl8192cu assigns a new struct rtl_tcb_desc object, but fails to
clear it.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Stable <stable@kernel.org> [2.6.39+]
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoiwlagn: fix dangling scan request
Johannes Berg [Thu, 22 Sep 2011 21:59:04 +0000 (14:59 -0700)]
iwlagn: fix dangling scan request

If iwl_scan_initiate() fails for any reason,
priv->scan_request and priv->scan_vif are left
dangling. This can lead to a crash later when
iwl_bg_scan_completed() tries to run a pending
scan request.

In practice, this seems to be very rare due to
the STATUS_SCANNING check earlier. That check,
however, is wrong -- it should allow a scan to
be queued when a reset/roc scan is going on.
When a normal scan is already going on, a new
one can't be issued by mac80211, so that code
can be removed completely. I introduced this
bug when adding off-channel support in commit
266af4c745952e9bebf687dd68af58df553cb59d.

Cc: stable@kernel.org [3.0]
Reported-by: Peng Yan <peng.yan@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
13 years agoqlcnic: Change CDRP function
Anirban Chakraborty [Tue, 13 Sep 2011 08:06:18 +0000 (08:06 +0000)]
qlcnic: Change CDRP function

Argument list to CDRP function has become unmanageably long. Fix it by properly
declaring a struct that encompasses all the input and output parameters.

Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoqlcnic: Added error logging for firmware abort
Ameen Rahman [Tue, 13 Sep 2011 08:06:17 +0000 (08:06 +0000)]
qlcnic: Added error logging for firmware abort

Signed-off-by: Ameen Rahman <ameen.rahman@qlogic.com>
Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge git://github.com/Jkirsher/net-next
David S. Miller [Fri, 23 Sep 2011 17:56:44 +0000 (13:56 -0400)]
Merge git://github.com/Jkirsher/net-next

13 years agonet/fec: add imx6q enet support
Shawn Guo [Fri, 23 Sep 2011 02:12:48 +0000 (02:12 +0000)]
net/fec: add imx6q enet support

The imx6q enet is a derivative of imx28 enet controller.  It fixed
the frame endian issue found on imx28, and added 1 Gbps support.

It also fixes a typo on vendor name in Kconfig.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet/fec: fix fec1 check in fec_enet_mii_init()
Shawn Guo [Fri, 23 Sep 2011 02:12:47 +0000 (02:12 +0000)]
net/fec: fix fec1 check in fec_enet_mii_init()

In function fec_enet_mii_init(), it uses non-zero pdev->id as part
of the condition to check the second fec instance (fec1).  This works
before the driver supports device tree probe.  But in case of device
tree probe, pdev->id is -1 which is also non-zero, so the logic becomes
broken when device tree probe gets supported.

The patch change the logic to check "pdev->id > 0" as the part of the
condition for identifying fec1.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agonet/fec: fec_reset_phy() does not need to always succeed
Shawn Guo [Fri, 23 Sep 2011 02:12:46 +0000 (02:12 +0000)]
net/fec: fec_reset_phy() does not need to always succeed

FEC can work without a phy reset on some platforms, which means not
very platform necessarily have a phy-reset gpio encoded in device tree.
Even on the platforms that have the gpio, FEC can work without
resetting phy for some cases, e.g. boot loader has done that.

So it makes more sense to have the phy-reset-gpio request failure as
a debug message rather than a warning, and get fec_reset_phy() return
void since the caller does not check the return anyway.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoixgb: finish conversion to ndo_fix_features
Michał Mirosław [Fri, 23 Sep 2011 02:11:30 +0000 (02:11 +0000)]
ixgb: finish conversion to ndo_fix_features

Finish conversion to unified ethtool ops: convert get_flags.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoixgb: eliminate checkstack warnings
Jesse Brandeburg [Fri, 23 Sep 2011 02:11:29 +0000 (02:11 +0000)]
ixgb: eliminate checkstack warnings

Really trivial fix, use kmalloc/kfree instead of stack space.
use static const instead of const to further reduce stack usage.

V2: reflect changes suggested by Joe Perches

before:
[jbrandeb@jbrandeb-mobl2 linux-2.6]$ make checkstack|grep '\[ixgb\]'
0x00000fc1 ixgb_set_multi [ixgb]:                       768
0x00001031 ixgb_set_multi [ixgb]:                       768
0x000010f2 ixgb_set_multi [ixgb]:                       768
0x061c ixgb_check_options [ixgb]:                       448
0x09c3 ixgb_check_options [ixgb]:                       448
0x0000649e ixgb_set_ringparam [ixgb]:                   192
0x0000130d ixgb_xmit_frame [ixgb]:                      184
0x000019e0 ixgb_xmit_frame [ixgb]:                      184
0x00002267 ixgb_clean [ixgb]:                           152
0x00002673 ixgb_clean [ixgb]:                           152

after:
0x000064ee ixgb_set_ringparam [ixgb]:                   192
0x0000135d ixgb_xmit_frame [ixgb]:                      184
0x00001a30 ixgb_xmit_frame [ixgb]:                      184
0x000022b7 ixgb_clean [ixgb]:                           152
0x000026c3 ixgb_clean [ixgb]:                           152

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>