firefly-linux-kernel-4.4.55.git
9 years agosmsc911x: Remove dev==NULL check.
Jeremy Linton [Wed, 19 Aug 2015 16:46:43 +0000 (11:46 -0500)]
smsc911x: Remove dev==NULL check.

The dev==NULL check in smsc911x_probe_config is useless
and isn't providing any additional protection. If a fwnode
doesn't exist then an appropriate error should be returned
by device_get_phy_mode() covering the original case
of a missing of/fwnode.

Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agodevice property: Add ETH_ALEN check, update comments.
Jeremy Linton [Wed, 19 Aug 2015 16:46:42 +0000 (11:46 -0500)]
device property: Add ETH_ALEN check, update comments.

This patch adds MAC address length check back into
the device_get_mac_addr() function before calling
is_valid_ether_addr() similar to the way the OF
routine does it.

Update the comments for the two new functions.

Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge tag 'wireless-drivers-next-for-davem-2015-08-19' of git://git.kernel.org/pub...
David S. Miller [Thu, 20 Aug 2015 21:13:25 +0000 (14:13 -0700)]
Merge tag 'wireless-drivers-next-for-davem-2015-08-19' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
Major changes:

ath10k:

* add support for qca99x0 family of devices
* improve performance of tx_lock
* add support for raw mode (802.11 frame format) and software crypto
  engine enabled via a module parameter

ath9k:

* add fast-xmit support

wil6210:

* implement TSO support
* support bootloader v1 and onwards

iwlwifi:

* Deprecate -10.ucode
* Clean ups towards multiple Rx queues
* Add support for longer CMD IDs. This will be required by new
  firmwares since we are getting close to the u8 limit.
* bugfixes for the D0i3 power state
* Add basic support for FTM
* polish the Miracast operation
* fix a few power consumption issues
* scan cleanup
* fixes for D0i3 system state
* add paging for devices that support it
* add again the new RBD allocation model
* add more options to the firmware debug system
* add support for frag SKBs in Tx
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipv4: Make fib_encap_match static
Ying Xue [Wed, 19 Aug 2015 08:04:51 +0000 (16:04 +0800)]
ipv4: Make fib_encap_match static

Make fib_encap_match() static as it isn't used outside the file.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'ewma'
David S. Miller [Thu, 20 Aug 2015 21:10:23 +0000 (14:10 -0700)]
Merge branch 'ewma'

Johannes Berg says:

====================
average: convert users to inline implementation

Since there's very little benefit of the out-of-line implementation
(a single byte of .text in one driver as far as I've seen), convert
all drivers to the inline implementation, saving memory, and remove
the out-of-line implementation.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoaverage: remove out-of-line implementation
Johannes Berg [Wed, 19 Aug 2015 07:46:22 +0000 (09:46 +0200)]
average: remove out-of-line implementation

Since all users are now converted to the inline implementation,
remove the out-of-line implementation entirely.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agort2x00: use DECLARE_EWMA
Johannes Berg [Wed, 19 Aug 2015 07:46:21 +0000 (09:46 +0200)]
rt2x00: use DECLARE_EWMA

Instead of using the out-of-line EWMA calculation, use DECLARE_EWMA()
to create static inlines. On x86/64 this results in code that's one
byte larger (for me), but reduces struct link_ant and struct link
size by the two unsigned long values that store the parameters each.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoath5k: use DECLARE_EWMA
Johannes Berg [Wed, 19 Aug 2015 07:46:20 +0000 (09:46 +0200)]
ath5k: use DECLARE_EWMA

This reduces code size slightly (at least on x86/64) while also
removing memory consumption by two unsigned long values for each
ath5k device.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovirtio_net: use DECLARE_EWMA
Johannes Berg [Wed, 19 Aug 2015 07:48:40 +0000 (09:48 +0200)]
virtio_net: use DECLARE_EWMA

Instead of using the out-of-line EWMA calculation, use DECLARE_EWMA()
to create static inlines. On x86/64 this results in no change in code
size for me, but reduces the struct receive_queue size by the two
unsigned long values that store the parameters.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobnx2x: Fix vxlan endianity issue
Yuval Mintz [Wed, 19 Aug 2015 07:21:58 +0000 (10:21 +0300)]
bnx2x: Fix vxlan endianity issue

Commit f34fa14cc033 ("bnx2x: Add vxlan RSS support") has introduced an
endianity issue when passing the vxlan UDP port to the HW.

Reported-by: <fengguang.wu@intel.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'vrf-cleanups-part-2'
David S. Miller [Thu, 20 Aug 2015 20:01:57 +0000 (13:01 -0700)]
Merge branch 'vrf-cleanups-part-2'

Nikolay Aleksandrov says:

====================
vrf: cleanups part 2

This is the next part of vrf cleanups, patch 1 drops the SLAB_PANIC
when creating kmem cache since it's handled, patch 02 removes a slave
duplicate check which is already done by the lower/upper code, patch 3
moves the ndo_add_slave code around a bit so we can drop an error
label and patch 4 drops the master device checks which are unnecessary
because the ops are taken from the master device itself so it can't be
different.
====================

Acked-by: David Ahern <dsa@cumulusnetworks.com>
9 years agovrf: ndo_add|del_slave drop unnecessary checks
Nikolay Aleksandrov [Wed, 19 Aug 2015 03:27:10 +0000 (06:27 +0300)]
vrf: ndo_add|del_slave drop unnecessary checks

When ndo_add|del_slave ops are used, they're taken from the respective
master device's netdev ops, so if the master device is a VRF only then
the VRF ops will get called thus no need to check the type of the
master.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovrf: move vrf_insert_slave so we can drop a goto label
Nikolay Aleksandrov [Wed, 19 Aug 2015 03:27:09 +0000 (06:27 +0300)]
vrf: move vrf_insert_slave so we can drop a goto label

We can simplify do_vrf_add_slave by moving vrf_insert_slave in the end
of the enslaving and thus eliminate an error goto label. It always
succeeds and isn't needed before that anyway.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovrf: remove unnecessary duplicate check
Nikolay Aleksandrov [Wed, 19 Aug 2015 03:27:08 +0000 (06:27 +0300)]
vrf: remove unnecessary duplicate check

The upper/lower functions already check for duplicate slaves so no need
to do it again.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovrf: don't panic on cache create failure
Nikolay Aleksandrov [Wed, 19 Aug 2015 03:27:07 +0000 (06:27 +0300)]
vrf: don't panic on cache create failure

It's pointless to panic on cache create failure when that case is handled
and even more so since it's not a kernel-wide fatal problem so don't
panic.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovrf: plug skb leaks
Nikolay Aleksandrov [Wed, 19 Aug 2015 03:12:29 +0000 (06:12 +0300)]
vrf: plug skb leaks

Currently whenever a packet different from ETH_P_IP is sent through the
VRF device it is leaked so plug the leaks and properly drop these
packets.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Acked-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovrf: vrf_master_ifindex_rcu is not always called with rcu read lock
Nikolay Aleksandrov [Tue, 18 Aug 2015 18:40:16 +0000 (21:40 +0300)]
vrf: vrf_master_ifindex_rcu is not always called with rcu read lock

While running net-next I hit this:
[  634.073119] ===============================
[  634.073150] [ INFO: suspicious RCU usage. ]
[  634.073182] 4.2.0-rc6+ #45 Not tainted
[  634.073213] -------------------------------
[  634.073244] include/net/vrf.h:38 suspicious rcu_dereference_check()
usage!
[  634.073274]
               other info that might help us debug this:

[  634.073307]
               rcu_scheduler_active = 1, debug_locks = 1
[  634.073338] 2 locks held by swapper/0/0:
[  634.073369]  #0:  (((&n->timer))){+.-...}, at: [<ffffffff8112bc35>]
call_timer_fn+0x5/0x480
[  634.073412]  #1:  (slock-AF_INET){+.-...}, at: [<ffffffff8174f0f5>]
icmp_send+0x155/0x5f0
[  634.073450]
               stack backtrace:
[  634.073483] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 4.2.0-rc6+ #45
[  634.073514] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS
VirtualBox 12/01/2006
[  634.073545]  0000000000000000 0593ba8242d9ace4 ffff88002fc03b48
ffffffff81803f1b
[  634.073612]  0000000000000000 ffffffff81e12500 ffff88002fc03b78
ffffffff811003c5
[  634.073642]  0000000000000000 ffff88002ec4e600 ffffffff81f00f80
ffff88002fc03cf0
[  634.073669] Call Trace:
[  634.073694]  <IRQ>  [<ffffffff81803f1b>] dump_stack+0x4c/0x65
[  634.073728]  [<ffffffff811003c5>] lockdep_rcu_suspicious+0xc5/0x100
[  634.073763]  [<ffffffff8174eb56>] icmp_route_lookup+0x176/0x5c0
[  634.073793]  [<ffffffff8174f2fb>] ? icmp_send+0x35b/0x5f0
[  634.073818]  [<ffffffff8174f274>] ? icmp_send+0x2d4/0x5f0
[  634.073844]  [<ffffffff8174f3ce>] icmp_send+0x42e/0x5f0
[  634.073873]  [<ffffffff8170b662>] ipv4_link_failure+0x22/0xa0
[  634.073899]  [<ffffffff8174bdda>] arp_error_report+0x3a/0x80
[  634.073926]  [<ffffffff816d6100>] ? neigh_lookup+0x2c0/0x2c0
[  634.073952]  [<ffffffff816d396e>] neigh_invalidate+0x8e/0x110
[  634.073984]  [<ffffffff816d62ae>] neigh_timer_handler+0x1ae/0x290
[  634.074013]  [<ffffffff816d6100>] ? neigh_lookup+0x2c0/0x2c0
[  634.074013]  [<ffffffff8112bce3>] call_timer_fn+0xb3/0x480
[  634.074013]  [<ffffffff8112bc35>] ? call_timer_fn+0x5/0x480
[  634.074013]  [<ffffffff816d6100>] ? neigh_lookup+0x2c0/0x2c0
[  634.074013]  [<ffffffff8112c2bc>] run_timer_softirq+0x20c/0x430
[  634.074013]  [<ffffffff810af50e>] __do_softirq+0xde/0x630
[  634.074013]  [<ffffffff810afc97>] irq_exit+0x117/0x120
[  634.074013]  [<ffffffff81810976>] smp_apic_timer_interrupt+0x46/0x60
[  634.074013]  [<ffffffff8180e950>] apic_timer_interrupt+0x70/0x80
[  634.074013]  <EOI>  [<ffffffff8106b9d6>] ? native_safe_halt+0x6/0x10
[  634.074013]  [<ffffffff81101d8d>] ? trace_hardirqs_on+0xd/0x10
[  634.074013]  [<ffffffff81027d43>] default_idle+0x23/0x200
[  634.074013]  [<ffffffff8102852f>] arch_cpu_idle+0xf/0x20
[  634.074013]  [<ffffffff810f89ba>] default_idle_call+0x2a/0x40
[  634.074013]  [<ffffffff810f8dcc>] cpu_startup_entry+0x39c/0x4c0
[  634.074013]  [<ffffffff817f9cad>] rest_init+0x13d/0x150
[  634.074013]  [<ffffffff81f69038>] start_kernel+0x4a8/0x4c9
[  634.074013]  [<ffffffff81f68120>] ?
early_idt_handler_array+0x120/0x120
[  634.074013]  [<ffffffff81f68339>] x86_64_start_reservations+0x2a/0x2c
[  634.074013]  [<ffffffff81f68485>] x86_64_start_kernel+0x14a/0x16d

It would seem vrf_master_ifindex_rcu() can be called without RCU held in
other contexts as well so introduce a new helper which acquires rcu and
returns the ifindex.
Also add curly braces around both the "if" and "else" parts as per the
style guide.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agolwtunnel: Fix the sparse warnings in fib_encap_match
Ying Xue [Wed, 19 Aug 2015 07:46:17 +0000 (15:46 +0800)]
lwtunnel: Fix the sparse warnings in fib_encap_match

When CONFIG_LWTUNNEL config is not enabled, the lwtstate_free() is not
declared in lwtunnel.h at all. However, even in this case, the function
is still referenced in fib_semantics.c so that there appears the
following sparse warnings:

net/ipv4/fib_semantics.c:553:17: error: undefined identifier 'lwtstate_free'
  CC      net/ipv4/fib_semantics.o
  net/ipv4/fib_semantics.c: In function ‘fib_encap_match’:
  net/ipv4/fib_semantics.c:553:3: error: implicit declaration of function ‘lwtstate_free’ [-Werror=implicit-function-declaration]
  cc1: some warnings being treated as errors
  make[1]: *** [net/ipv4/fib_semantics.o] Error 1
  make: *** [net/ipv4/fib_semantics.o] Error 2

To eliminate the error, we define an empty function for lwtstate_free()
in lwtunnel.h when CONFIG_LWTUNNEL is disabled.

Fixes: df383e6240ef ("lwtunnel: fix memory leak")
Cc: Jiri Benc <jbenc@redhat.com>
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Acked-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Wed, 19 Aug 2015 03:21:32 +0000 (20:21 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2015-08-18

This series contains updates to igb, e100, e1000e and ixgbe.

Shota Suzuki provides a fix for a possible overflow in
igb_set_interrupt_capability() which leads to an oops.  When changing the
number of queues by "ethtool -L", set IGB_FLAG_QUEUE_PAIRS in the same
manner as when initializing the igb driver.

Vasily Averin provides a fix for a missing rtnl_unlock() for when we
error out due to not being able to allocate memory for our queues.

Stefan Assman provides a couple of fixes for igb/igbvf.  First changes
the igb driver in probe to simply call igb_enable_sriov() instead of
igb_sriov_reinit() since we are starting from scratch.  Then in igbvf,
fix the driver where it does not clear the buffer_info->dma in all
cases after calling dma_unmap_single(), which was found by changing the
MTU twice.

Richard Cochran implements the periodic output function using the
programmable clock outputs available in i210 when possible, falling
back to the target time for longer periods.

Todd adds support for the Marvell PHY 1512 which is required for i354
devices.  Then updates igb to make sure SR-IOV init uses the correct
number of queues, since recent changes could result in the PF holding
onto all of the queues.

Alex Williamson provides a fix in the case where a guest OS does not
support hot-unplug, so disable SR-IOV prior to unregister_netdev() to
avoid the problem.

Jia-Ju Bai provides several patches, first knocks some collecting dust
off an old e100 driver to add a check to avoid a null pointer
dereference.  Then cleans up a possible resource leak by releasing the
skb buffer allocated when the e100_xmit_prepare() runs into an issue
in the DMA mapping.  In igb, add a missing rtnl_unlock() for when we
error out due to igb_sriov_reinit() in the igb_init_interrupt_scheme().
Provides a e1000e fix, based on suggestions from Alex Duyck to move
head/tail register writing to e1000_configure_tx/rx() to avoid a
possible null pointer dereference (similar to igb driver).  Lastly,
fix a possible memory leak in igb_probe(), where the memory shadow_vfta
allocated by kcalloc in igb_sw_init() is not freed.

Mark simplifies port-specific macros for ixgbe by eliminating explicit
comparisons with 0 and enclose formal parameters in parens to eliminate
the risk of an operator precedence issue.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'vrf-next'
David S. Miller [Wed, 19 Aug 2015 03:16:53 +0000 (20:16 -0700)]
Merge branch 'vrf-next'

Nikolay Aleksandrov says:

====================
vrf: a few simplifications and cleanups

These patches remove some unnecessary checks (patches 3, 4), unnecessary
num_slaves member and refcnt manipulations which are already done by the
upper functions.
====================

Acked-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovrf: simplify the netdev notifier function
Nikolay Aleksandrov [Tue, 18 Aug 2015 17:28:04 +0000 (20:28 +0300)]
vrf: simplify the netdev notifier function

We can drop the check because if vrf_ptr is present then we must have
the vrf device as a master and since we're running with rtnl it can't go
away.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovrf: don't check for dstats and rth in uninit path
Nikolay Aleksandrov [Tue, 18 Aug 2015 17:28:03 +0000 (20:28 +0300)]
vrf: don't check for dstats and rth in uninit path

dstats and rth are always present because we fail the device registration
if they can't be allocated in vrf_init() (ndo_init) so drop the unnecessary
checks.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovrf: drop unused num_slaves member
Nikolay Aleksandrov [Tue, 18 Aug 2015 17:28:02 +0000 (20:28 +0300)]
vrf: drop unused num_slaves member

slave_queue has a num_slaves member which is unused, drop it.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovrf: drop unnecessary dev refcnt changes
Nikolay Aleksandrov [Tue, 18 Aug 2015 17:28:01 +0000 (20:28 +0300)]
vrf: drop unnecessary dev refcnt changes

netdev_master_upper_dev_link/unlink already do a dev_hold/put on the
devices being linked, so no need to take another reference.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agohv_netvsc: Fix dereference of nvdev before check
Andrew Schwartzmeyer [Wed, 19 Aug 2015 03:06:32 +0000 (20:06 -0700)]
hv_netvsc: Fix dereference of nvdev before check

Passes static analysis by Smatch.

Signed-off-by: Andrew Schwartzmeyer <andschwa@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agolwtunnel: ip tunnel: fix multiple routes with different encap
Jiri Benc [Tue, 18 Aug 2015 16:42:09 +0000 (18:42 +0200)]
lwtunnel: ip tunnel: fix multiple routes with different encap

Currently, two routes going through the same tunnel interface are considered
the same even when they are routed to a different host after encapsulation.
This causes all routes added after the first one to have incorrect
encapsulation parameters.

This is nicely visible by doing:

  # ip r a 192.168.1.2/32 dev vxlan0 tunnel dst 10.0.0.2
  # ip r a 192.168.1.3/32 dev vxlan0 tunnel dst 10.0.0.3
  # ip r
  [...]
  192.168.1.2/32 tunnel id 0 src 0.0.0.0 dst 10.0.0.2 [...]
  192.168.1.3/32 tunnel id 0 src 0.0.0.0 dst 10.0.0.2 [...]

Implement the missing comparison function.

Fixes: 3093fbe7ff4bc ("route: Per route IP tunnel metadata via lightweight tunnel")
Signed-off-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agolwtunnel: fix memory leak
Jiri Benc [Tue, 18 Aug 2015 16:41:13 +0000 (18:41 +0200)]
lwtunnel: fix memory leak

The built lwtunnel_state struct has to be freed after comparison.

Fixes: 571e722676fe3 ("ipv4: support for fib route lwtunnel encap attributes")
Signed-off-by: Jiri Benc <jbenc@redhat.com>
Acked-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agocxgb4: memory corruption in debugfs
Dan Carpenter [Tue, 18 Aug 2015 09:31:44 +0000 (12:31 +0300)]
cxgb4: memory corruption in debugfs

You can't use kstrtoul() with an int or it causes memory corruption.
Also j should be unsigned or we have underflow bugs.

I considered changing "j" to unsigned long but everything fits in a u32.

Fixes: 8e3d04fd7d70 ('cxgb4: Add MPS tracing support')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoenic: Fix namespace pollution causing build errors.
David S. Miller [Tue, 18 Aug 2015 21:24:18 +0000 (14:24 -0700)]
enic: Fix namespace pollution causing build errors.

drivers/net/built-in.o: In function `.vnic_wq_devcmd2_alloc':
(.text+0x49fe40): multiple definition of `.vnic_wq_devcmd2_alloc'
drivers/scsi/built-in.o:(.text+0xb4318): first defined here
drivers/net/built-in.o:(.opd+0x2af00): multiple definition of `vnic_wq_devcmd2_alloc'
drivers/scsi/built-in.o:(.opd+0xad70): first defined here
drivers/net/built-in.o: In function `.vnic_wq_init_start':
(.text+0x49f9c0): multiple definition of `.vnic_wq_init_start'
drivers/scsi/built-in.o:(.text+0xb3b58): first defined here
drivers/net/built-in.o:(.opd+0x2ae88): multiple definition of `vnic_wq_init_start'
drivers/scsi/built-in.o:(.opd+0xace0): first defined here

Rename these to 'enic_*' to avoid the conflict with the functiosn of
the same name in the snic scsi driver.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobnx2x: Add vxlan RSS support
Rajesh Borundia [Tue, 18 Aug 2015 07:22:59 +0000 (10:22 +0300)]
bnx2x: Add vxlan RSS support

Latest FW submission added some vxlan offload capabilities to our device.
This patch adds the ability to connect to the vxlan NDOs and configure
the UDP port associated with it in the HW.

The device would now be capable of performing RSS according to the
inner headers of the vxlan packets.

Signed-off-by: Rajesh Borundia <Rajesh.Borundia@qlogic.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'dsa-multi-swtich'
David S. Miller [Tue, 18 Aug 2015 21:17:22 +0000 (14:17 -0700)]
Merge branch 'dsa-multi-swtich'

Andrew Lunn says:

====================
D in DSA patches

The D in DSA is distributed, meaning multiple switches can be
connected together. Currently no mainline system does this, and so the
code is broken. This patchset contains two fixes, and a small helper.

With three of more switches, the current device tree binding is not
sufficient to express the routing between the switches. The first
patch extends the binding, in a backwards compatible way, to allow a
link between a switch to describe all the switches accessible over the
link, not just the direct neighbor.

The third patch fixes the port configuration on newer devices for
links connecting switches.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agodsa: mv88e6xxx: Set DSA mode based on chip abilities
Andrew Lunn [Mon, 17 Aug 2015 21:52:52 +0000 (23:52 +0200)]
dsa: mv88e6xxx: Set DSA mode based on chip abilities

Older devices only support a single DSA frame format, where as newer
devices have two. Take this into account when configuring a DSA port.
The port needs to be in plain old DSA mode, since this is a DSA link,
where as the newer format can be used for the CPU port.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: dsa: Add dsa_is_dsa_port() helper
Andrew Lunn [Mon, 17 Aug 2015 21:52:51 +0000 (23:52 +0200)]
net: dsa: Add dsa_is_dsa_port() helper

Add an inline helper for determining is a port is a DSA port.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: dsa: Allow multi hop routes to be expressed
Andrew Lunn [Mon, 17 Aug 2015 21:52:50 +0000 (23:52 +0200)]
net: dsa: Allow multi hop routes to be expressed

With more than two switches in a hierarchy, it becomes necessary to
describe multi-hop routes between switches. The current binding does
not allow this, although the older platform_data did. Extend the link
property to be a list rather than a single phandle to a remote switch.
It is then possible to express that a port should be used to reach
more than one switch and the switch maybe more than one hop away.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoixgbe: TRIVIAL fix up double 'the' and comment style
Jacob Keller [Wed, 10 Jun 2015 18:44:45 +0000 (11:44 -0700)]
ixgbe: TRIVIAL fix up double 'the' and comment style

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoixgbe: Simplify port-specific macros
Mark Rustad [Sat, 6 Jun 2015 17:41:03 +0000 (10:41 -0700)]
ixgbe: Simplify port-specific macros

Simplify port-specific macros by eliminating explicit comparison
with 0. More importantly, enclose formal parameter in parens to
eliminate the risk of an operator precedence surprise.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoigb: make sure SR-IOV init uses the right number of queues
Todd Fujinaka [Sat, 8 Aug 2015 00:27:39 +0000 (17:27 -0700)]
igb: make sure SR-IOV init uses the right number of queues

Recent changes to igb_probe_vfs() could lead to the PF holding onto all
of the queues. Reorder igb_probe_vfs() to be before
gb_init_queue_configuration() and add some more error checking.

Signed-off-by: Todd Fujinaka <todd.fujinaka@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoigbvf: clear buffer_info->dma after dma_unmap_single()
Stefan Assmann [Thu, 6 Aug 2015 07:32:17 +0000 (09:32 +0200)]
igbvf: clear buffer_info->dma after dma_unmap_single()

The driver doesn't clear buffer_info->dma after calling
dma_unmap_single() in all cases. This has been discovered by changing
the mtu twice, which caused the following backtrace.

[   68.569280] WARNING: CPU: 2 PID: 1860 at drivers/iommu/intel-iommu.c:3517 intel_unmap+0x20c/0x220()
[   68.579392] Driver unmaps unmatched page at PFN fffc2a40
[   68.585322] Modules linked in: igbvf ipt_MASQUERADE nf_nat_masquerade_ipv4 iptable_nat nf_nat_ipv4 nf_nat kvm_intel kvm igb megs
[   68.599163] CPU: 2 PID: 1860 Comm: ifconfig Not tainted 4.2.0-rc4+ #147
[   68.606543] Hardware name: IBM  -[546025Z]-/00Y7630, BIOS -[VVE134TUS-1.51]- 10/17/2013
[   68.615473]  0000000000000dbd ffff88046441bb08 ffffffff81a5ad0b ffffffff81e2f9ea
[   68.623775]  ffff88046441bb58 ffff88046441bb48 ffffffff81056b55 ffff88047fc583c0
[   68.632075]  0000000000000000 ffff880469a8e600 00000000fffc2a40 ffff880465b32098
[   68.640375] Call Trace:
[   68.643109]  [<ffffffff81a5ad0b>] dump_stack+0x48/0x5d
[   68.648844]  [<ffffffff81056b55>] warn_slowpath_common+0x95/0xe0
[   68.655549]  [<ffffffff81056c56>] warn_slowpath_fmt+0x46/0x70
[   68.661960]  [<ffffffff8158a614>] ? find_iova+0x54/0x90
[   68.667791]  [<ffffffff815988dc>] intel_unmap+0x20c/0x220
[   68.673815]  [<ffffffff8159891e>] intel_unmap_page+0xe/0x10
[   68.680038]  [<ffffffffa0067536>] igbvf_clean_rx_ring+0x96/0x370 [igbvf]
[   68.687516]  [<ffffffffa0067915>] igbvf_down+0x105/0x110 [igbvf]
[   68.694219]  [<ffffffffa0067beb>] igbvf_change_mtu+0x16b/0x180 [igbvf]
[...]

Signed-off-by: Stefan Assmann <sassmann@kpanic.de>
Acked-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>
9 years agoigb: Fix a memory leak in igb_probe
Jia-Ju Bai [Wed, 5 Aug 2015 14:05:16 +0000 (22:05 +0800)]
igb: Fix a memory leak in igb_probe

In error handling code of igb_probe, the memory adapter->shadow_vfta
allocated by kcalloc in igb_sw_init is not freed. So when register_netdev
or igb_init_i2c is failed, a memory leak will occur.
This patch adds kfree to fix it.

Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoe1000e: Modify Tx/Rx configurations to avoid null pointer dereferences in e1000_open
Jia-Ju Bai [Wed, 5 Aug 2015 10:16:10 +0000 (18:16 +0800)]
e1000e: Modify Tx/Rx configurations to avoid null pointer dereferences in e1000_open

When e1000e_setup_rx_resources is failed in e1000_open,
e1000e_free_tx_resources in "err_setup_rx" segment is executed.
"writel(0, tx_ring->head)" statement in e1000_clean_tx_ring
in e1000e_free_tx_resources will cause a null poonter dereference(crash),
because "tx_ring->head" is only assigned in e1000_configure_tx
in e1000_configure, but it is after e1000e_setup_rx_resources.

This patch moves head/tail register writing to e1000_configure_tx/rx,
which can fix this problem. It is inspired by igb_configure_tx_ring
in the igb driver.

Specially, thank Alexander Duyck for his valuable suggestion.

Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoigb: Fix a deadlock in igb_sriov_reinit
Jia-Ju Bai [Mon, 3 Aug 2015 03:36:26 +0000 (11:36 +0800)]
igb: Fix a deadlock in igb_sriov_reinit

When igb_init_interrupt_scheme in igb_sriov_reinit is failed, the lock
acquired by rtnl_lock() is not released, which causes a deadlock.
This patch adds rtnl_unlock() in error handling to fix it.

Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoe100: Release skb when DMA mapping is failed in e100_xmit_prepare
Jia-Ju Bai [Mon, 3 Aug 2015 02:40:48 +0000 (10:40 +0800)]
e100: Release skb when DMA mapping is failed in e100_xmit_prepare

When pci_dma_mapping_error in e100_xmit_prepare is failed, the skb buffer
allocated by netdev_alloc_skb_ip_align in e100_rx_alloc_skb is not
released, which causes a possible resource leak.
This patch adds error handling code to fix it.

Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoe100: Add a check after pci_pool_create to avoid null pointer dereference
Jia-Ju Bai [Mon, 3 Aug 2015 02:17:08 +0000 (10:17 +0800)]
e100: Add a check after pci_pool_create to avoid null pointer dereference

The driver lacks the check of nic->cbs_pool after pci_pool_create
in e100_probe. When this function is failed, a null pointer dereference
occurs when pci_pool_alloc uses nic->cbs_pool in e100_alloc_cbs.
This patch adds a check and related error handling code to fix it.

Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoigb: Teardown SR-IOV before unregister_netdev()
Alex Williamson [Wed, 29 Jul 2015 20:38:15 +0000 (14:38 -0600)]
igb: Teardown SR-IOV before unregister_netdev()

When the .remove() callback for a PF is called, SR-IOV support for the
device is disabled, which requires unbinding and removing the VFs.
The VFs may be in-use either by the host kernel or userspace, such as
assigned to a VM through vfio-pci.  In this latter case, the VFs may
be removed either by shutting down the VM or hot-unplugging the
devices from the VM.  Unfortunately in the case of a Windows 2012 R2
guest, hot-unplug is broken due to the ordering of the PF driver
teardown.  Disabling SR-IOV prior to unregister_netdev() avoids this
issue.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Acked-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoigb: add support for 1512 PHY
Todd Fujinaka [Wed, 29 Jul 2015 14:32:06 +0000 (07:32 -0700)]
igb: add support for 1512 PHY

This patch adds support for Marvell PHY 1512 (required for I354).

Submitted by: Maciej Szwed <maciej.szwed@intel.com>
Signed-off-by: Todd Fujinaka <todd.fujinaka@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoigb: implement high frequency periodic output signals
Richard Cochran [Thu, 23 Jul 2015 21:59:30 +0000 (14:59 -0700)]
igb: implement high frequency periodic output signals

In addition to interrupt driven target time output events, the i210
also has two programmable clock outputs.  These clocks support periods
between 16 nanoseconds and 140 milliseconds.  This patch implements
the periodic output function using the clock outputs when possible,
falling back to the target time for longer periods.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoigb: do not re-init SR-IOV during probe
Stefan Assmann [Fri, 10 Jul 2015 13:01:12 +0000 (15:01 +0200)]
igb: do not re-init SR-IOV during probe

During driver probing the following code path is triggered.
igb_probe
->igb_sw_init
  ->igb_probe_vfs
    ->igb_pci_enable_sriov
      ->igb_sriov_reinit

Doing the SR-IOV re-init is not necessary during probing since we're
starting from scratch. Here we can call igb_enable_sriov() right away.

Running igb_sriov_reinit() during igb_probe() also seems to cause
occasional packet loss on some onboard 82576 NICs. Reproduced on
Dell and HP servers with onboard 82576 NICs.
Example:
Intel Corporation 82576 Gigabit Network Connection [8086:10c9] (rev 01)
Subsystem: Dell Device [1028:0481]

Signed-off-by: Stefan Assmann <sassmann@kpanic.de>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoigb: missing rtnl_unlock in igb_sriov_reinit()
Vasily Averin [Tue, 7 Jul 2015 15:53:45 +0000 (18:53 +0300)]
igb: missing rtnl_unlock in igb_sriov_reinit()

Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoigb: Fix oops caused by missing queue pairing
Shota Suzuki [Wed, 1 Jul 2015 00:25:52 +0000 (09:25 +0900)]
igb: Fix oops caused by missing queue pairing

When initializing igb driver (e.g. 82576, I350), IGB_FLAG_QUEUE_PAIRS is
set if adapter->rss_queues exceeds half of max_rss_queues in
igb_init_queue_configuration().
On the other hand, IGB_FLAG_QUEUE_PAIRS is not set even if the number of
queues exceeds half of max_combined in igb_set_channels() when changing
the number of queues by "ethtool -L".
In this case, if numvecs is larger than MAX_MSIX_ENTRIES (10), the size
of adapter->msix_entries[], an overflow can occur in
igb_set_interrupt_capability(), which in turn leads to an oops.

Fix this problem as follows:
 - When changing the number of queues by "ethtool -L", set
   IGB_FLAG_QUEUE_PAIRS in the same way as initializing igb driver.
 - When increasing the size of q_vector, reallocate it appropriately.
   (With IGB_FLAG_QUEUE_PAIRS set, the size of q_vector gets larger.)

Another possible way to fix this problem is to cap the queues at its
initial number, which is the number of the initial online cpus. But this
is not the optimal way because we cannot increase queues when another
cpu becomes online.

Note that before commit cd14ef54d25b ("igb: Change to use statically
allocated array for MSIx entries"), this problem did not cause oops
but just made the number of queues become 1 because of entering msi_only
mode in igb_set_interrupt_capability().

Fixes: 907b7835799f ("igb: Add ethtool support to configure number of channels")
CC: stable <stable@vger.kernel.org>
Signed-off-by: Shota Suzuki <suzuki_shota_t3@lab.ntt.co.jp>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoMerge branch 'drivers_iff_no_queue'
David S. Miller [Tue, 18 Aug 2015 18:55:08 +0000 (11:55 -0700)]
Merge branch 'drivers_iff_no_queue'

Phil Sutter says:

====================
net: Convert drivers to IFF_NO_QUEUE and cleanup afterwards

This series converts in-tree users away from the old and deprecated
'tx_queue_len = 0' idiom, adds a warning to notify out-of-tree driver
maintainers that there is need for action on their behalf and finally drops any
workarounds in scheduling algorithm implementations.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: sched: drop all special handling of tx_queue_len == 0
Phil Sutter [Tue, 18 Aug 2015 08:30:49 +0000 (10:30 +0200)]
net: sched: drop all special handling of tx_queue_len == 0

Those were all workarounds for the formerly double meaning of
tx_queue_len, which broke scheduling algorithms if untreated.

Now that all in-tree drivers have been converted away from setting
tx_queue_len = 0, it should be safe to drop these workarounds for
categorically broken setups.

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: warn if drivers set tx_queue_len = 0
Phil Sutter [Tue, 18 Aug 2015 08:30:48 +0000 (10:30 +0200)]
net: warn if drivers set tx_queue_len = 0

Due to the introduction of IFF_NO_QUEUE, there is a better way for
drivers to indicate that no qdisc should be attached by default. Though,
the old convention can't be dropped since ignoring that setting would
break drivers still using it. Instead, add a warning so out-of-tree
driver maintainers get a chance to adjust their code before we finally
get rid of any special handling of tx_queue_len == 0.

Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agostaging: wilc1000: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:47 +0000 (10:30 +0200)]
staging: wilc1000: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Johnny Kim <johnny.kim@atmel.com>
Cc: Rachel Kim <rachel.kim@atmel.com>
Cc: Dean Lee <dean.lee@atmel.com>
Cc: Chris Park <chris.park@atmel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: caif: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:46 +0000 (10:30 +0200)]
net: caif: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Dmitry Tarnyagin <dmitry.tarnyagin@lockless.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: hsr: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:45 +0000 (10:30 +0200)]
net: hsr: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Arvid Brodin <arvid.brodin@alten.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: batman-adv: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:44 +0000 (10:30 +0200)]
net: batman-adv: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Marek Lindner <mareklindner@neomailbox.ch>
Cc: Simon Wunderlich <sw@simonwunderlich.de>
Cc: Antonio Quartulli <antonio@meshcoding.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: mac80211_hwsim: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:43 +0000 (10:30 +0200)]
net: mac80211_hwsim: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: hostap: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:42 +0000 (10:30 +0200)]
net: hostap: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Jouni Malinen <j@w1.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: dsa: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:41 +0000 (10:30 +0200)]
net: dsa: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: ipvlan: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:40 +0000 (10:30 +0200)]
net: ipvlan: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: bonding: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:39 +0000 (10:30 +0200)]
net: bonding: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Jay Vosburgh <j.vosburgh@gmail.com>
Cc: Veaceslav Falico <vfalico@gmail.com>
Cc: Andy Gospodarek <gospo@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: 6lowpan: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:38 +0000 (10:30 +0200)]
net: 6lowpan: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Alexander Aring <alex.aring@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: bridge: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:37 +0000 (10:30 +0200)]
net: bridge: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: 8021q: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:36 +0000 (10:30 +0200)]
net: 8021q: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: vxlan: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:35 +0000 (10:30 +0200)]
net: vxlan: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: team: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:34 +0000 (10:30 +0200)]
net: team: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: nlmon: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:33 +0000 (10:30 +0200)]
net: nlmon: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: loopback: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:32 +0000 (10:30 +0200)]
net: loopback: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: geneve: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:31 +0000 (10:30 +0200)]
net: geneve: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: John W. Linville <linville@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: dummy: convert to using IFF_NO_QUEUE
Phil Sutter [Tue, 18 Aug 2015 08:30:30 +0000 (10:30 +0200)]
net: dummy: convert to using IFF_NO_QUEUE

Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: veth: enable noqueue operation by default
Phil Sutter [Tue, 18 Aug 2015 08:30:29 +0000 (10:30 +0200)]
net: veth: enable noqueue operation by default

Signed-off-by: Phil Sutter <phil@nwl.cc>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge tag 'iwlwifi-next-for-kalle-2015-08-18' of https://git.kernel.org/pub/scm/linux...
Kalle Valo [Tue, 18 Aug 2015 14:20:11 +0000 (17:20 +0300)]
Merge tag 'iwlwifi-next-for-kalle-2015-08-18' of https://git./linux/kernel/git/iwlwifi/iwlwifi-next

* polish the Miracast operation
* fix a few power consumption issues
* scan cleanup
* fixes for D0i3 system state
* add paging for devices that support it
* add again the new RBD allocation model
* add more options to the firmware debug system
* add support for frag SKBs in Tx

9 years agoiwlwifi: mvm: fix a race in D0i3 vs. Tx path
Emmanuel Grumbach [Mon, 17 Aug 2015 12:54:41 +0000 (15:54 +0300)]
iwlwifi: mvm: fix a race in D0i3 vs. Tx path

When we enter D0i3, we must stop TXing otherwise the
sequence number we use might conflict with the firmware's
internal TX. In order to do so, we have
IWL_MVM_STATUS_IN_D0I3 which should prevent any Tx while we
enter D0i3. There is a bug in this code since we may Tx even
if IWL_MVM_STATUS_IN_D0I3 is set. This can happen as long as
mvm->d0i3_ap_sta_id is not set.

To make sure that we don't have any packet in the Tx path
while we set mvm->d0i3_ap_sta_id, call synchronize_net only
after we already set mvm->d0i3_ap_sta_id.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: don't disconnect on beacon loss in D0I3
David Spinadel [Thu, 6 Aug 2015 07:26:50 +0000 (10:26 +0300)]
iwlwifi: mvm: don't disconnect on beacon loss in D0I3

Currently if we wake up during D0I3 due to beacon loss we disconnect
immediately. This behaviour causes redundant disconnection, which could
be prevented by polling as it is usually done in mac80211.
Instead, we prefer reporting beacon loss and let mac80211 try polling
before disconnection.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: out-of-bounds access in iwl_init_sband_channels
Adrien Schildknecht [Fri, 14 Aug 2015 00:35:32 +0000 (02:35 +0200)]
iwlwifi: out-of-bounds access in iwl_init_sband_channels

KASan error report:
==================================================================
BUG: KASan: out of bounds access in iwl_init_sband_channels+0x207/0x260 [iwlwifi] at addr ffff8800c2d0aac8
Read of size 4 by task modprobe/329
==================================================================

Both loops of this function compare data from the 'chan' array and then
check if the index is valid.

The 2 conditions should be inverted to avoid an out-of-bounds access.

Signed-off-by: Adrien Schildknecht <adrien+dev@schischi.me>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: bump mvm firmware API to 16
Emmanuel Grumbach [Sun, 16 Aug 2015 07:39:49 +0000 (10:39 +0300)]
iwlwifi: bump mvm firmware API to 16

The driver is now able to handle -16.ucode.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: rs: disable MIMO only if allowed in configuration
Alexander Bondar [Tue, 28 Jul 2015 13:50:51 +0000 (16:50 +0300)]
iwlwifi: rs: disable MIMO only if allowed in configuration

Fix bug where MIMO is disabled for low latency TX on P2P VIF
regardless of configuration. Make it dependent on
IWL_MVM_RS_DISABLE_P2P_MIMO compilation option. Change configuration
so that MIMO will be disabled only in SDIO platforms.

Signed-off-by: Alexander Bondar <alexander.bondar@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agobcma: switch GPIO portions to use GPIOLIB_IRQCHIP
Linus Walleij [Thu, 13 Aug 2015 22:21:45 +0000 (00:21 +0200)]
bcma: switch GPIO portions to use GPIOLIB_IRQCHIP

This switches the BCMA GPIO driver to use GPIOLIB_IRQCHIP to
handle its interrupts instead of rolling its own copy of the
irqdomain handling etc.

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoath9k: add correct MAC/BB name for ar9561
Miaoqing Pan [Wed, 12 Aug 2015 06:20:46 +0000 (14:20 +0800)]
ath9k: add correct MAC/BB name for ar9561

MAC/BB name is"????" if the MAC/BB is unknown.

Signed-off-by: Miaoqing Pan <miaoqing@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agowl18xx: add diversity statistics
Guy Mishol [Tue, 11 Aug 2015 07:02:57 +0000 (10:02 +0300)]
wl18xx: add diversity statistics

Add diversity statistics and sync the driver
statistics acx and debugfs representation
with the current fw api.

Signed-off-by: Guy Mishol <guym@ti.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agowl18xx: update statistics acx and debugfs files
Eliad Peller [Tue, 11 Aug 2015 07:02:56 +0000 (10:02 +0300)]
wl18xx: update statistics acx and debugfs files

Sync the driver statistics acx and debugfs representation
with the current fw api.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agort2x00: adjust EEPROM_SIZE for rt2500usb
Adrien Schildknecht [Mon, 10 Aug 2015 22:25:53 +0000 (00:25 +0200)]
rt2x00: adjust EEPROM_SIZE for rt2500usb

rt2500usb_validate_eeprom() read data up to 0x6e (EEPROM_CALIBRATE_OFFSET)
but only 0x6a bytes has been allocated and read from the eeprom.

This lead to out-of-bound accesses and invalid values for
EEPROM_BBPTUNE_R17 and EEPROM_CALIBRATE_OFFSET.

Change the EEPROM_SIZE to 0x6e in order to retrieve all the fields.

Tested with a rt2570 device.

Signed-off-by: Adrien Schildknecht <adrien+dev@schischi.me>
Acked-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agomwl8k: refactor some conditionals for clarity
John Linville [Mon, 10 Aug 2015 16:05:39 +0000 (12:05 -0400)]
mwl8k: refactor some conditionals for clarity

CC [M]  drivers/net/wireless/mwl8k.o
drivers/net/wireless/mwl8k.c: In function ‘mwl8k_bss_info_changed’:
drivers/net/wireless/mwl8k.c:3290:2: warning: ‘ap_mcs_rates’ may be used uninitialized in this function [-Wmaybe-uninitialized]
  memcpy(cmd->mcs_set, mcs_rates, 16);
  ^
drivers/net/wireless/mwl8k.c:4987:5: note: ‘ap_mcs_rates’ was declared here
  u8 ap_mcs_rates[16];
     ^

The warning was bogus.  But the conditionals were rather complicated,
with multiple redundant checks.  This consolidates the checking and
makes it more readable IMHO.

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agortlwifi: rtl8192cu: Remove and replace routine in hw.c and mac.c
Taehee Yoo [Mon, 6 Apr 2015 18:13:00 +0000 (03:13 +0900)]
rtlwifi: rtl8192cu: Remove and replace routine in hw.c and mac.c

I remove duplicated routines which related rtl92cu_set_hw_reg().

1. rtl92c_set_qos() and HW_VAR_AC_PARAM routine are similar code.
so i replace code with rtlpriv->cfg->ops->set_hw_reg().

2. rtl92c_set_mac_addr() and 'HW_VAR_ETHER_ADDR' case at
rtl92cu_set_hw_reg() routine are similar code.
so i removed rtl92c_set_mac_addr() function.
also it was not used anywhere.

3. remove HW_VAR_ACM_CTRL routine in rtl92cu_set_hw_reg().
if rtl_usb->acm_method is not EACMWAY2_SW, HW_VAR_ACM_CTRL is called
from HW_VAR_AC_PARAM. but it never called. because acm_method is always
EACMWAY2_SW. so i remove acm_method check routine
and HW_VAR_ACM_CTRL routine.

both usb and pci interface is not used HW_VAR_ACM_CTRL.
but i can't test pci interface module, so i didn't modify pci code.

Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agortlwifi: rtl8192cu: remove duplicated routine in hw.c and mac.c
Taehee Yoo [Mon, 30 Mar 2015 15:55:32 +0000 (00:55 +0900)]
rtlwifi: rtl8192cu: remove duplicated routine in hw.c and mac.c

rtl92c_set_xxx_filter is same routine with rtl92cu_set_hw_reg.
so i remove those functions that are rtl92c_set_xxx_filter.
(rtl92c_get_xxx_filter is also same reason.)
also i add code updating struct rtl_mac member variable in the
rtl92cu_set_hw_reg.
after that, no more _update_mac_setting is not useful. thus i remove that.

Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoMerge tag 'mac80211-next-for-davem-2015-08-14' mac80211-next.git
Kalle Valo [Tue, 18 Aug 2015 05:44:22 +0000 (08:44 +0300)]
Merge tag 'mac80211-next-for-davem-2015-08-14' mac80211-next.git

iwlwifi needs new mac80211 patches so merge mac80211-next.git to
wireless-drivers-next.git.

9 years agoMerge branch 'Identifier-Locator-Addressing'
David S. Miller [Tue, 18 Aug 2015 04:33:06 +0000 (21:33 -0700)]
Merge branch 'Identifier-Locator-Addressing'

Tom Herbert says:

====================
net: Identifier Locator Addressing - Part I

This patch set provides rudimentary support for Identifier Locator
Addressing or ILA. The basic concept of ILA is that we split an IPv6
address into a 64 bit locator and 64 bit identifier. The identifier is
the identity of an entity in communication ("who"), and the locator
expresses the location of the entity ("where"). Applications
use externally visible address that contains the identifier.
When a packet is actually sent, a translation is done that
overwrites the first 64 bits of the address with a locator.
The packet can then be forwarded over the network to the host where
the addressed entity is located. At the receiver, the reverse
translation is done so the that the application sees the original,
untranslated address. Presumably an external control plane will
provide identifier->locator mappings.

v2:
  - Fix compilation erros when LWT not configured
  - Consolidate ILA into a single ila.c

v3:
  - Change pseudohdr argument od inet_proto_csum_replace functions to
    be a bool

v4:
  - In ila_build_state check locator being in netlink params before
    allocating tunnel state

The data path for ILA is a simple NAT translation that only operates
on the upper 64 bits of a destination address in IPv6 packets. The
basic process is:

   1) Lookup 64 bit identifier (lower 64 bits of destination)
   2) If a match is found
      a) Overwrite locator (upper 64 bits of destination) with
         the new locator
      b) Adjust any checksum that has destination address included in
         pseudo header
   3) Send or receive packet

ILA is a means to implement tunnels or network virtualization without
encapsulation. Since there is no encapsulation involved, we assume that
stateless support in the network for IPv6 (e.g. RSS, ECMP, TSO, etc.)
just works. Also, since we're minimally changing the packet many of
the worries about encapsulation (MTU, checksum, fragmentation) are
not relevant. The downside is that, ILA is not extensible like other
encapsulations (GUE for instance) so it might not be appropriate for
all use cases. Also, this only makes sense to do in IPv6!

A key aspect of ILA is performance. The intent is that ILA would be
used in data centers in virtualizing tasks or jobs. In the fullest
incarnation all intra data center communications might be targeted to
virtual ILA addresses. This is basically adding a new virtualization
capability to the existing services in a datacenter, so there is a
strong expectation is that this does not degrade performance for
existing applications.

Performance seems to be dependent on how ILA is hooked into kernel.
ILA can be implemented under some different models:

  - Mechanically it is a form a stateless DNAT
  - It can be thought of as a type of (source) routing
  - As a functional replacement of encapsulation

In this patch set we hook into the data path using Light Weight
Tunnels (LWT) infrastructure. As part of that, we add support in LWT
to redirect dst input. iproute will be modified to take a new ila encap
type. ILA can be configured like:

ip route add 3333:0:0:1:5555:0:2:0/128 \
   encap ila 2001:0:0:2 via 2401:db00:20:911a:face:0:27:0

ip -6 addr add 3333:0:0:1:5555:0:1:0/128 dev eth0

ip route add table local local 2001:0:0:1:5555:0:1:0/128
   encap ila 3333:0:0:1 dev lo

So sending to destination 3333:0:0:1:5555:0:2:0 will have destination
of 2001:0:0:2:5555:0:2:0 on the wire.

Performance results are below. With ILA we see about a 10% drop in
pps compared to non-ILA. Much of this drop can be attributed to the
loss of early demux on input (translation occurs after it is attempted).
We will address this in the next patch set. Also, IPvlan input path
does not work with ILA since the routing is bypassed-- this will
be addressed in a future patch.

Performance testing:

Performing netperf TCP_RR with 200 clients:

Non-ILA baseline
  84.92% CPU utilization
  1861922.9 tps
  93/163/330 50/90/99% latencies

ILA single destination
  83.16% CPU utilization
  1679683.4 tps
  105/180/332 50/90/99% latencies

References:

Slides from netconf:
http://vger.kernel.org/netconf2015Herbert-ILA.pdf

Slides from presentation at IETF:
https://www.ietf.org/proceedings/92/slides/slides-92-nvo3-1.pdf

I-D:
https://tools.ietf.org/html/draft-herbert-nvo3-ila-00
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: Identifier Locator Addressing module
Tom Herbert [Mon, 17 Aug 2015 20:42:27 +0000 (13:42 -0700)]
net: Identifier Locator Addressing module

Adding new module name ila. This implements ILA translation. Light
weight tunnel redirection is used to perform the translation in
the data path. This is configured by the "ip -6 route" command
using the "encap ila <locator>" option, where <locator> is the
value to set in destination locator of the packet. e.g.

ip -6 route add 3333:0:0:1:5555:0:1:0/128 \
      encap ila 2001:0:0:1 via 2401:db00:20:911a:face:0:25:0

Sets a route where 3333:0:0:1 will be overwritten by
2001:0:0:1 on output.

Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: Add inet_proto_csum_replace_by_diff utility function
Tom Herbert [Mon, 17 Aug 2015 20:42:26 +0000 (13:42 -0700)]
net: Add inet_proto_csum_replace_by_diff utility function

This function updates a checksum field value and skb->csum based on
a value which is the difference between the old and new checksum.

Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: Change pseudohdr argument of inet_proto_csum_replace* to be a bool
Tom Herbert [Mon, 17 Aug 2015 20:42:25 +0000 (13:42 -0700)]
net: Change pseudohdr argument of inet_proto_csum_replace* to be a bool

inet_proto_csum_replace4,2,16 take a pseudohdr argument which indicates
the checksum field carries a pseudo header. This argument should be a
boolean instead of an int.

Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agolwt: Add support to redirect dst.input
Tom Herbert [Mon, 17 Aug 2015 20:42:24 +0000 (13:42 -0700)]
lwt: Add support to redirect dst.input

This patch adds the capability to redirect dst input in the same way
that dst output is redirected by LWT.

Also, save the original dst.input and and dst.out when setting up
lwtunnel redirection. These can be called by the client as a pass-
through.

Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoenic: Fix sparse warning in vnic_devcmd_init().
David S. Miller [Tue, 18 Aug 2015 04:24:59 +0000 (21:24 -0700)]
enic: Fix sparse warning in vnic_devcmd_init().

>> drivers/net/ethernet/cisco/enic/vnic_dev.c:1095:13: sparse: incorrect type in assignment (different address spaces)
   drivers/net/ethernet/cisco/enic/vnic_dev.c:1095:13:    expected void *res
   drivers/net/ethernet/cisco/enic/vnic_dev.c:1095:13:    got void [noderef] <asn:2>*

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agomlx5e: Fix sparse warnings in mlx5e_handle_csum().
David S. Miller [Tue, 18 Aug 2015 04:22:26 +0000 (21:22 -0700)]
mlx5e: Fix sparse warnings in mlx5e_handle_csum().

>> drivers/net/ethernet/mellanox/mlx5/core/en_rx.c:173:44: sparse: incorrect type in argument 1 (different base types)
   drivers/net/ethernet/mellanox/mlx5/core/en_rx.c:173:44:    expected restricted __sum16 [usertype] n
   drivers/net/ethernet/mellanox/mlx5/core/en_rx.c:173:44:    got restricted __be16 [usertype] check_sum

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoinet: Move VRF table lookup to inlined function
David Ahern [Sun, 16 Aug 2015 23:13:27 +0000 (17:13 -0600)]
inet: Move VRF table lookup to inlined function

Table lookup compiles out when VRF is not enabled.

Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: Fix docbook warning for IFF_VRF_MASTER enum
David Ahern [Sun, 16 Aug 2015 16:26:49 +0000 (10:26 -0600)]
net: Fix docbook warning for IFF_VRF_MASTER enum

kbuild test robot reported:
tree:   git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next.git master
head:   d52736e24fe2e927c26817256f8d1a3c8b5d51a0
commit: 4e3c89920cd3a6cfce22c6f537690747c26128dd [751/762] net: Introduce VRF related flags and helpers
reproduce: make htmldocs

>> Warning(include/linux/netdevice.h:1293): Enum value 'IFF_VRF_MASTER' not described in enum 'netdev_priv_flags'

Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: Updates to netif_index_is_vrf
David Ahern [Sun, 16 Aug 2015 13:49:20 +0000 (07:49 -0600)]
net: Updates to netif_index_is_vrf

As Eric noted netif_index_is_vrf is not called with rcu_read_lock held,
so wrap the dev_get_by_index_rcu in rcu_read_lock and unlock.

If VRF is not enabled or oif is 0 skip the device lookup. In both cases
index cannot be the VRF master.

Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'mlx5e-next'
David S. Miller [Mon, 17 Aug 2015 22:51:36 +0000 (15:51 -0700)]
Merge branch 'mlx5e-next'

Achiad Shochat says:

====================
Driver updates 16-Aug-2015

This patchset contains bug fixes, new RSS and pause parameters ethtool
options, and support for RX CHECKSUM_COMPLETE.

Patchset was applied and tested over commit adc6310 ("Merge branch
'mv88e6xxx-switchdev-fdb'").
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/mlx5e: Support RX CHECKSUM_COMPLETE
Achiad Shochat [Sun, 16 Aug 2015 13:04:52 +0000 (16:04 +0300)]
net/mlx5e: Support RX CHECKSUM_COMPLETE

Only for packets with first ethertype set to IPv4/6 for now.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/mlx5e: Support ethtool get/set_pauseparam
Achiad Shochat [Sun, 16 Aug 2015 13:04:51 +0000 (16:04 +0300)]
net/mlx5e: Support ethtool get/set_pauseparam

Only rx/tx pause settings.
Autoneg setting is currently not supported.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/mlx5e: Ethtool link speed setting fixes
Achiad Shochat [Sun, 16 Aug 2015 13:04:50 +0000 (16:04 +0300)]
net/mlx5e: Ethtool link speed setting fixes

- Port speed settings are applied by the device only upon
  port admin status transition from DOWN to UP.
  So we enforce this transition regardless of the port's
  current operation state (which may be occasionally DOWN if
  for example the network cable is disconnected).
- Fix the PORT_UP/DOWN device interface enum
- Set the local_port bit in the device PAOS register
- EXPORT the PAOS (Port Administrative and Operational Status)
  register set/query access functions.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>