GitHub/moto-9609/android_kernel_motorola_exynos9610.git
8 years agoi40e/i40evf: use logical operators, not bitwise
Mitch Williams [Thu, 14 Jan 2016 00:51:50 +0000 (16:51 -0800)]
i40e/i40evf: use logical operators, not bitwise

Mr. Spock would certainly raise an eyebrow to see us using bitwise
operators, when we should clearly be relying on logic. Fascinating.

Change-ID: Ie338010c016f93e9faa2002c07c90b15134b7477
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e/i40evf: use pages correctly in Rx
Mitch Williams [Thu, 14 Jan 2016 00:51:49 +0000 (16:51 -0800)]
i40e/i40evf: use pages correctly in Rx

Refactor the packet split Rx code to properly use half-pages for
receives. The previous code was doing way more mapping and unmapping
than it needed to, and wasn't properly using half-pages.

Increment the page use count each time we give a half-page to an skb,
knowing that the stack will probably process and release the page before
we need it again. Only free and reallocate pages if the count shows that
both half-pages are in use. Add counters to track reallocations and page
reuse.

Change-ID: I534b299196036b64be82b4861a0a4036310a8f22
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e/i40evf: use __GFP_NOWARN
Jesse Brandeburg [Thu, 14 Jan 2016 00:51:48 +0000 (16:51 -0800)]
i40e/i40evf: use __GFP_NOWARN

The i40e and i40evf drivers now cleanly handle allocation
failures and can avoid kernel log spew from the memory allocator
when allocations fail, so set __GFP_NOWARN on Rx buffer alloc.

Change-ID: Ic9e1b83c495e2a3ef6b069ba7fb6e52ce134cd23
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: dump descriptor indexes in hex
Jesse Brandeburg [Thu, 14 Jan 2016 00:51:47 +0000 (16:51 -0800)]
i40e: dump descriptor indexes in hex

The debugging helpers for showing descriptor rings were
dumping the indexes in decimal and the offsets in hex.

Put everything in hex and at least be consistent.

Also update copyright year in file header.

Change-ID: Ia35a21411a2ddb713772dffb4e8718889fcfc895
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e/i40evf: try again after failure
Jesse Brandeburg [Thu, 14 Jan 2016 00:51:46 +0000 (16:51 -0800)]
i40e/i40evf: try again after failure

This is the "Don't Give Up" patch.  Previously the
driver could fail an allocation, and then possibly stall
a queue forever, by never coming back to continue receiving
or allocating buffers.

With this patch, the driver will keep polling trying to allocate
receive buffers until it succeeds.  This should keep all receive
queues running even in the face of memory pressure.

Also update copyright year in file header.

Change-ID: I2b103d1ce95b9831288a7222c3343ffa1988b81b
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e/i40evf: don't lose interrupts
Jesse Brandeburg [Thu, 14 Jan 2016 00:51:45 +0000 (16:51 -0800)]
i40e/i40evf: don't lose interrupts

While re-enabling interrupts the driver would clear all pending
causes. This meant that if an interrupt was generated while the driver
was cleaning or polling with interrupts disabled, then that interrupt
was lost.  This could cause a queue to become dead, especially for
receive.  Refactored the enable_icr0 function in order to allow
it to be decided by the caller whether the CLEARPBA (clear pending
events) bit will be set while re-enabling the interrupt.

Also update copyright year in file headers.

Change-ID: Ic1db100a05e13c98919057696db147a258ca365a
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40evf: Change vf driver string to reflect all products i40evf supports
Catherine Sullivan [Thu, 14 Jan 2016 00:51:44 +0000 (16:51 -0800)]
i40evf: Change vf driver string to reflect all products i40evf supports

Change the driver string to 40-10 Gigabit instead of XL710/X710 for X722
and all future products.

Also update copyright year in file header.

Change-ID: I57fae656b36dc4eb682b2b7a054f8f48f3589149
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: Refactor force_wb and WB_ON_ITR functionality code
Anjali Singhai Jain [Thu, 14 Jan 2016 00:51:43 +0000 (16:51 -0800)]
i40e: Refactor force_wb and WB_ON_ITR functionality code

Now that the Force-WriteBack functionality in X710/XL710 devices
has been moved out of the clean routine and into the service task,
we need to make sure WriteBack-On-ITR is separated out since it
is still called from clean.

In the X722 devices, Force-WriteBack implies WriteBack-On-ITR but
without the interrupt, which put the driver into a missed
interrupt scenario and a potential tx-timeout report.

With this patch, we break the two functions out, and call the
appropriate ones at the right place. This will avoid creating missed
interrupt like scenarios for X722 devices.

Also update copyright year in file headers.

Change-ID: Iacbde39f95f332f82be8736864675052c3583a40
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: use new add_veb calling with VEB stats control
Shannon Nelson [Thu, 14 Jan 2016 00:51:42 +0000 (16:51 -0800)]
i40e: use new add_veb calling with VEB stats control

The new parameters for add_veb allow us to enable and disable VEB stats,
so let's use them.

Update copyright year.

Change-ID: Ie6e68c68e2d1d459e42168eda661051b56bf0a65
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: add VEB stat control and remove L2 cloud filter
Shannon Nelson [Thu, 14 Jan 2016 00:51:41 +0000 (16:51 -0800)]
i40e: add VEB stat control and remove L2 cloud filter

With the latest firmware, statistics gathering can now be enabled and
disabled in the HW switch, so we need to add a parameter to allow the
driver to set it as desired.  At the same time, the L2 cloud filtering
parameter has been removed as it was never used.

Older drivers working with the newer firmware and newer drivers working
with older firmware will not run into problems with these bits as the
defaults are reasonable and there is no overlap in the bit definitions.
Also, newer drivers will be forced to update because of the change in
function call parameters, a reminder that the functionality exists.

Also update copyright year.

Change-ID: I9acb9160b892ca3146f2f11a88fdcd86be3cadcc
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: set shared bit for multicast filters
Shannon Nelson [Thu, 14 Jan 2016 00:51:40 +0000 (16:51 -0800)]
i40e: set shared bit for multicast filters

Add the use of the new Shared MAC filter bit for multicast and broadcast
filters in order to make better use of the filters available from the
device.  The FW folks have assured me that setting this bit on older FW
will have no affect, so we don't need a version check.

Also fixed a stray indent problem nearby.

Also update copyright year.

Change-ID: I4c5826a32594382a7937a592a24d228588cee7aa
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: Make the DCB firmware checks for X710/XL710 only
Neerav Parikh [Thu, 14 Jan 2016 00:51:39 +0000 (16:51 -0800)]
i40e: Make the DCB firmware checks for X710/XL710 only

Make the DCB firmware version related checks specific to
X710 and XL710 only. These checks are not required for
X722 family of devices.

Introduced an inline routine to help determine if the
MAC type is X710/XL710 or not.

Moved the firmware version related checks in i40e_sw_init()
and defined flags for different cases

Fix the version check to allow using "Set LLDP MIB" AQ
for beyond FVL4 FW releases.

Change-ID: Ib78288343de983aa0354fc28aa36e99b073662c0
Signed-off-by: Neerav Parikh <neerav.parikh@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: move sync_vsi_filters up in service_task
Jesse Brandeburg [Thu, 14 Jan 2016 00:51:38 +0000 (16:51 -0800)]
i40e: move sync_vsi_filters up in service_task

The sync_vsi_filters function is moved up in the service_task because
it may need to request a reset, and we don't want to wait another round
of service task time.

NOTE: Filters will be replayed by sync_vsi_filters including broadcast
and promiscuous settings.

Also, added some error handling in this space in case any of these
fail the driver will retry correctly.

Also update copyright year.

Change-ID: I23f3d552100baecea69466339f738f27614efd47
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoMerge branch 'inet_lro-remove'
David S. Miller [Wed, 17 Feb 2016 21:15:47 +0000 (16:15 -0500)]
Merge branch 'inet_lro-remove'

Ben Hutchings says:

====================
Remove the inet_lro library

The old inet_lro library has been deprecated ever since GRO was
introduced, but there are still a few drivers using it.  Convert
them to GRO and remove it.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: Remove inet_lro library
Ben Hutchings [Mon, 15 Feb 2016 21:25:57 +0000 (21:25 +0000)]
ipv4: Remove inet_lro library

There are no longer any in-tree drivers that use it.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDMA/nes: Replace LRO with GRO
Ben Hutchings [Mon, 15 Feb 2016 21:25:44 +0000 (21:25 +0000)]
RDMA/nes: Replace LRO with GRO

GRO is simpler to use than the old inet_lro library, and is compatible
with forwarding and bridging configurations.

Compile-tested only.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agopasemi_mac: Replace LRO with GRO
Ben Hutchings [Mon, 15 Feb 2016 21:25:11 +0000 (21:25 +0000)]
pasemi_mac: Replace LRO with GRO

GRO is simpler to use than the old inet_lro library, and is compatible
with forwarding and bridging configurations.

Compile-tested only.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoaf_llc: fix types on llc_ui_wait_for_conn
One Thousand Gnomes [Mon, 15 Feb 2016 19:41:51 +0000 (19:41 +0000)]
af_llc: fix types on llc_ui_wait_for_conn

The timeout is a long, we return it truncated if it is huge. Basically
harmless as the only caller does a boolean check, but tidy it up anyway.

(64bit build tested this time. Thank you 0day)

Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed/qede: use 8.7.3.0 FW.
Yuval Mintz [Mon, 15 Feb 2016 18:22:35 +0000 (13:22 -0500)]
qed/qede: use 8.7.3.0 FW.

This patch moves the qed* driver into utilizing the 8.7.3.0 FW.
This new FW is required for a lot of new SW features, including:
  - Vlan filtering offload
  - Encapsulation offload support
  - HW ingress aggregations
As well as paving the way for the possibility of adding storage protocols
in the future.

V2:
 - Fix kbuild test robot error/warnings.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@qlogic.com>
Signed-off-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'sctp-cleanups'
David S. Miller [Wed, 17 Feb 2016 20:41:55 +0000 (15:41 -0500)]
Merge branch 'sctp-cleanups'

Xin Long says:

====================
sctp: some cleanups for sctp

There are some unused function and redundant rcu locks should be
removed.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosctp: remove the unused sctp_datamsg_free()
Xin Long [Mon, 15 Feb 2016 06:28:05 +0000 (14:28 +0800)]
sctp: remove the unused sctp_datamsg_free()

Since commit 8b570dc9f7b6 ("sctp: only drop the reference on the datamsg
after sending a msg") used sctp_datamsg_put in sctp_sendmsg, instead of
sctp_datamsg_free, this function has no use in sctp.

So we will remove it.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosctp: remove rcu_read_lock in sctp_seq_dump_remote_addrs()
Xin Long [Mon, 15 Feb 2016 06:28:04 +0000 (14:28 +0800)]
sctp: remove rcu_read_lock in sctp_seq_dump_remote_addrs()

sctp_seq_dump_remote_addrs is only called by sctp_assocs_seq_show()
and it has been protected by rcu_read_lock that is from
rhashtable_walk_start().

So we will remove this one.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosctp: move rcu_read_lock from __sctp_lookup_association to sctp_lookup_association
Xin Long [Mon, 15 Feb 2016 06:28:03 +0000 (14:28 +0800)]
sctp: move rcu_read_lock from __sctp_lookup_association to sctp_lookup_association

__sctp_lookup_association() is only invoked by sctp_v4_err() and
sctp_rcv(), both which run on the rx BH, and it has been protected
by rcu_read_lock [see ip_local_deliver_finish() / ipv6_rcv()].

So we can move it to sctp_lookup_association, only let
sctp_lookup_association use rcu_read_lock.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocore: remove unneded headers for net cgroup controllers.
Rosen, Rami [Mon, 15 Feb 2016 00:39:43 +0000 (02:39 +0200)]
core: remove unneded headers for net cgroup controllers.

commit 3ed80a6 (cgroup: drop module support) made including
module.h redundant in the net cgroup controllers,
netclassid_cgroup.c and netprio_cgroup.c. This patch
removes them.

Signed-off-by: Rami Rosen <rami.rosen@intel.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'renesas-cleanups'
David S. Miller [Wed, 17 Feb 2016 20:30:24 +0000 (15:30 -0500)]
Merge branch 'renesas-cleanups'

Sergei Shtylyov says:

====================
Kill useless *switch* defaults in the Renesas Ethernet drivers

   Here's a set of 2 patches against DaveM's 'net-next.git' repo removing
often used *default* cases doing nothing.

[1/2] ravb: kill useless *switch* defaults
[2/2] sh_eth: kill useless *switch* defaults
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosh_eth: kill useless *switch* defaults
Sergei Shtylyov [Sun, 14 Feb 2016 19:56:33 +0000 (22:56 +0300)]
sh_eth: kill useless *switch* defaults

The driver often has the *default* cases doing nothing in the *switch*
statements with  the integer expressions -- remove them.

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoravb: kill useless *switch* defaults
Sergei Shtylyov [Sun, 14 Feb 2016 19:56:03 +0000 (22:56 +0300)]
ravb: kill useless *switch* defaults

The  driver has the *default* case doing nothing in the *switch* statement
with an integer expression -- remove it.

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Wed, 17 Feb 2016 20:23:53 +0000 (15:23 -0500)]
Merge branch '40GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
40GbE Intel Wired LAN Driver Updates 2016-02-17

This series contains updates to i40e/i40evf only.

Jesse cleans up a duplicate declaration in probe.  Then fixes a bug where the
driver was using a offset based off a DMA handle while mapping and unmapping
using sync_single_range_for_[cpu|device], when it should be using DMA handle
(returned from alloc_coherent) and the offset of the memory to be sync'd.
Fixed an issue where sync_vsi_filter() was allocating memory in a way that
could sleep (GFP_KERNEL) which was causing a problem when called by the
team driver under rcu_read_lock().

Shannon adds a check to ensure we do not attempt to do TSO when
skb->ip_summed is not set to CHECKSUM_PARTIAL.

Kiran add functions related to port mirroring features such as add/delete
mirror rule.

Jacob assigns the i40e_pf structure directly instead of using a large
memcpy, to avoid a sparse warning and lets the compiler optimize the copy
since it know the size of the structure in advance.

Anjali enables GENEVE capability for XL710/X710 devices with firmware API
version higher than 1.4.  Added flag for automatic rule eviction feature
for X722, which is disabled by default.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'amd-xgbe-next'
David S. Miller [Wed, 17 Feb 2016 20:22:21 +0000 (15:22 -0500)]
Merge branch 'amd-xgbe-next'

Tom Lendacky says:

====================
amd-xgbe: AMD XGBE driver updates 2016-02-12

The following updates and fixes are included in this driver update series:

- Disable VLAN filtering in promiscuous mode
- Change from using napi_complete to napi_complete_done
- Use __napi_schedule_irqoff when running in interrupt context
- Verify ethtool speed setting is valid for the selected speedset
- Enable PFC based on the pfc_en setting
- Fix the mapping of priorities to traffic classes
- Do traffic class setup when DCB nl callbacks are invoked
- Check Rx queue fifos before stopping Rx queue DMA
- Switch from disable_irq to masking interrupts for auto-negotiation

This patch series is based on net-next.

Changes from v1:
- Removed #ifndef and #define of CRCPOLY_LE as part of the patch to
  disable VLAN filtering in promiscuous mode
- Reworked changes to xgbe_setup_tc to resolve conflicts with commit
  16e5cc647173 (net: rework setup_tc ndo op to consume general tc operand)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoamd-xgbe: Mask auto-negotiation interrupts in ISR
Lendacky, Thomas [Wed, 17 Feb 2016 17:49:28 +0000 (11:49 -0600)]
amd-xgbe: Mask auto-negotiation interrupts in ISR

Currently the auto-negotiation interrupt handling disables the irq
instead of masking off the interrupts. This was done because the phy
library was originally used to read and write the PCS registers, which
could not be performed in interrupt context. Now that the phy library is
no longer used to read and write the PCS registers the interrupts can be
masked off in the interrupt service routine eliminating the need to call
disable_irq/enable_irq. This also requires changing the protection mutex
to a spinlock.

Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoamd-xgbe: Check Rx queue fifos before stopping Rx DMA
Lendacky, Thomas [Wed, 17 Feb 2016 17:49:16 +0000 (11:49 -0600)]
amd-xgbe: Check Rx queue fifos before stopping Rx DMA

Check to be sure that the Rx queue fifos are empty before stopping the
Rx DMA channels.

Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoamd-xgbe: Do traffic class setup when called through dcbnl
Lendacky, Thomas [Wed, 17 Feb 2016 17:49:08 +0000 (11:49 -0600)]
amd-xgbe: Do traffic class setup when called through dcbnl

Currently the netdev traffic class setup is only performed when invoked
through the ndo_setup_tc interface. However, the same setup should be
performed when the dcbnl interface (ieee_setets) is invoked. Rework the
netdev traffic class setup to be invokable through either interface and
also provide the priority to traffic class mapping if available.

Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoamd-xgbe: Fix the mapping of priorities to traffic classes
Lendacky, Thomas [Wed, 17 Feb 2016 17:48:57 +0000 (11:48 -0600)]
amd-xgbe: Fix the mapping of priorities to traffic classes

The driver is checking the pfc_en field of the ieee_pfc structure to
determine whether to associate a priority with a traffic class. This is
incorrect since the pfc_en field is for determining if PFC is enabled
for a traffic class.

The association of priority to traffic class does not depend on whether
the traffic class is enabled for PFC, so remove that check. Also, the
mapping of priorities to traffic classes should be done when configuring
the traffic classes and not the PFC support so move the priority to traffic
class association from xgbe_config_dcb_pfc to xgbe_config_dcb_tc.

Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoamd-xgbe: Enable/disable PFC per traffic class
Lendacky, Thomas [Wed, 17 Feb 2016 17:48:48 +0000 (11:48 -0600)]
amd-xgbe: Enable/disable PFC per traffic class

Currently the PFC flow control is enabled on all traffic classes if
one or more traffic classes request it. The PFC enable setting of the
traffic class should be used to determine whether to enable or disable
flow control for the traffic class.

Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoamd-xgbe: Verify forced speed matches the active speedset
Lendacky, Thomas [Wed, 17 Feb 2016 17:48:39 +0000 (11:48 -0600)]
amd-xgbe: Verify forced speed matches the active speedset

When using ethtool to set the speed for the device, verify that the
specified speed is valid within the active speedset.

Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoamd-xgbe: Use __napi_schedule_irqoff
Lendacky, Thomas [Wed, 17 Feb 2016 17:48:29 +0000 (11:48 -0600)]
amd-xgbe: Use __napi_schedule_irqoff

Change from calling __napi_schedule to __napi_schedule_irqoff when running
in interrupt context or when called by netpoll with interrupts already
disabled. The Tx timer function will continue to use __napi_schedule.

Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoamd-xgbe: Change from napi_complete to napi_complete_done
Lendacky, Thomas [Wed, 17 Feb 2016 17:48:19 +0000 (11:48 -0600)]
amd-xgbe: Change from napi_complete to napi_complete_done

Change from using napi_complete to napi_complete_done to allow for the
use of gro_flush_timeout in tuning network processing.

Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoamd-xgbe: Disable VLAN filtering when in promiscuous mode
Lendacky, Thomas [Wed, 17 Feb 2016 17:48:08 +0000 (11:48 -0600)]
amd-xgbe: Disable VLAN filtering when in promiscuous mode

When the hardware is placed in promiscuous mode it will still perform
VLAN filtering and therefore may not pass all packets to the driver.

Disable all VLAN filtering when entering promiscuous mode and restore
VLAN filtering upon exit from promiscuous mode. In order to avoid adding
forward declarations, move the VLAN related functions earlier in the
file.

Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomacvlan: convert to use IFF_NO_QUEUE
Zhang Shengju [Sun, 14 Feb 2016 14:10:39 +0000 (14:10 +0000)]
macvlan: convert to use IFF_NO_QUEUE

Use IFF_NO_QUEUE to indicate that a device can run without a qdisc.

Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoi40e: use eth_platform_get_mac_address()
Sowmini Varadhan [Wed, 13 Jan 2016 03:32:31 +0000 (19:32 -0800)]
i40e: use eth_platform_get_mac_address()

This commit converts commit b499ffb0a22c ("i40e: Look up MAC address in
Open Firmware or IDPROM") to use eth_platform_get_mac_address()
added by commit c7f5d105495a ("net: Add eth_platform_get_mac_address()
helper.")

Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: add priv flag for automatic rule eviction
Anjali Singhai Jain [Sat, 9 Jan 2016 01:50:21 +0000 (17:50 -0800)]
i40e: add priv flag for automatic rule eviction

The X722 can support automatic rule eviction for automatically added
flow director rules.  Feature is (should be) disabled by default.

Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: Enable Geneve offload for FW API ver > 1.4 for XL710/X710 devices
Anjali Singhai [Wed, 6 Jan 2016 19:49:28 +0000 (11:49 -0800)]
i40e: Enable Geneve offload for FW API ver > 1.4 for XL710/X710 devices

This patch makes sure we check the GENEVE offload capable flag before
we attempt offload.
It also enables the Capability for XL710/X710 devices with FW API
version higher than 1.4

Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e/i40evf: bump version to 1.4.12/1.4.8
Jesse Brandeburg [Mon, 4 Jan 2016 18:33:12 +0000 (10:33 -0800)]
i40e/i40evf: bump version to 1.4.12/1.4.8

Bump driver versions to i40e-1.4.12 and i40evf-1.4.8

Change-ID: I0ad82668c4ded04250391fda396ce191a42ab754
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: avoid large memcpy by assigning struct
Jacob Keller [Mon, 4 Jan 2016 18:33:11 +0000 (10:33 -0800)]
i40e: avoid large memcpy by assigning struct

Assign the i40e_pf structure directly instead of using a large memcpy,
which avoids a sparse warning and lets the compiler optimize the copy
since it knows the size of the structure in advance.

Change-ID: I17604e23be2616521eb760290befcb767b52b3f7
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: count allocation errors
Jesse Brandeburg [Mon, 4 Jan 2016 18:33:10 +0000 (10:33 -0800)]
i40e: count allocation errors

Driver already counted allocation errors, so print
them as part of the ethtool -S output.  Useful for debugging
if your system is having trouble making memory available for
the driver.

Change-ID: I83839fa86e81e6d80f03b917c88dd3ef9a64dde0
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: drop unused function
Jesse Brandeburg [Mon, 4 Jan 2016 18:33:09 +0000 (10:33 -0800)]
i40e: drop unused function

Delete the unused irq_dynamic_disable function.

Change-ID: Ia46071066babd121c7c90f141b6210b00078de3f
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Reviewed-by: Anjali Singhai <anjali.singhai@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: negate PHY int mask bits
Shannon Nelson [Mon, 4 Jan 2016 18:33:08 +0000 (10:33 -0800)]
i40e: negate PHY int mask bits

The PHY interrupt mask bits mask out the events we don't want,
so we need to negate the bitmask of events we want.

Change-ID: I273244da5a8d285b6abc84fd68a90f1e6fa0393e
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: APIs to Add/remove port mirroring rules
Kiran Patil [Mon, 4 Jan 2016 18:33:07 +0000 (10:33 -0800)]
i40e: APIs to Add/remove port mirroring rules

This patch implements necessary functions related to port
mirroring features such as add/delete mirror rule, function
to set promiscuous VLAN mode for VSI if mirror rule_type is
"VLAN Mirroring".

Change-ID: Iaf513fd5f188f99dcb977b48f99e73185dfddc40
Signed-off-by: Kiran Patil <kiran.patil@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: fix: do not sleep in netdev_ops
Jesse Brandeburg [Mon, 4 Jan 2016 18:33:06 +0000 (10:33 -0800)]
i40e: fix: do not sleep in netdev_ops

The driver was being called by VLAN, bonding, teaming operations
that expected to be able to hold locks like rcu_read_lock().

This causes the driver to be held to the requirement to not sleep,
and was found by the kernel debug options for checking sleep
inside critical section, and the locking validator.

Change-ID: Ibc68c835f5ffa8ffe0638ffe910a66fc5649a7f7
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Reviewed-by: Nelson, Shannon <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: allocate memory safer
Jesse Brandeburg [Mon, 4 Jan 2016 18:33:05 +0000 (10:33 -0800)]
i40e: allocate memory safer

The sync_vsi_filter function was allocating memory in such
a way that it could sleep (GFP_KERNEL) which was causing a problem
when called by the team driver under rcu_read_lock(), which cannot
be held while sleeping.  Found with lockdep.

Change-ID: I4e59053cb5eedcf3d0ca151715be3dc42a94bdd5
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: do TSO only if CHECKSUM_PARTIAL is set
Shannon Nelson [Mon, 4 Jan 2016 18:33:04 +0000 (10:33 -0800)]
i40e: do TSO only if CHECKSUM_PARTIAL is set

Don't bother trying to set up a TSO if the skb->ip_summed is not
set to CHECKSUM_PARTIAL.

Change-ID: I6495b3568e404907a2965b48cf3e2effa7c9ab55
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: fix bug in dma sync
Jesse Brandeburg [Mon, 4 Jan 2016 18:33:03 +0000 (10:33 -0800)]
i40e: fix bug in dma sync

Driver was using an offset based off a DMA handle while mapping and
unmapping using sync_single_range_for[cpu|device], where it should
be using DMA handle (returned from alloc_coherent) and the offset of the
memory to be sync'd.

Change-ID: I208256565b1595ff0e9171ab852de06b997917c6
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Reviewed-by: Nelson, Shannon <shannon.nelson@intel.com>
Reviewed-by: Williams, Mitch A <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: trivial: fix missing space
Jesse Brandeburg [Mon, 4 Jan 2016 18:33:02 +0000 (10:33 -0800)]
i40e: trivial: fix missing space

Missing space in comment, fixed.

Change-ID: I8cdf3ce5994b4a97dcc3eeb33422533918546667
Reported-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: trivial: drop duplicate definition
Jesse Brandeburg [Mon, 4 Jan 2016 18:33:01 +0000 (10:33 -0800)]
i40e: trivial: drop duplicate definition

The probe routine already had a u32 val declared, no need
to do it again.  Found by W=2 compile.

Change-ID: Id7b65f6d0ef6bb71067d0557f5be0202b6d8741e
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoMerge branch '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Wed, 17 Feb 2016 14:48:50 +0000 (09:48 -0500)]
Merge branch '40GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
40GbE Intel Wired LAN Driver Updates 2016-02-16

This series contains updates to i40e/i40evf only.

Shannon adds flags to MAC allocation requests to signify that the MAC VLAN
filters should come from the shared resource pool.  Added a new "set switch
config" admin queue command and the new Cisco VXLAN-GPE cloud tunnel type
for the admin queue commands.  Added more detail to the NVM update debug
message in order to see the full ethtool request data.  Also added a few
more bits of netdev data into the debugfs output for dump VSI.

Pandi fixes the width of two datatypes which were being declared a different
size from what they are assigned.

Anjali fixes an issue where we were not doing write-back on interrupt
throttle for legacy case in x722.

Mitch adds a counter for ARQ overflows since sometimes an ever-growing
number indicates that something bad is happening.  Also added 20G speed for
Tx bandwidth calculations.

Jesse refactors the DCB function based on a community suggestion to change
the multi-level if statement into a switch statement.  Cleans up VF device
IDs in the PF, since it does not need to know them.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'tc-hw-offload'
David S. Miller [Wed, 17 Feb 2016 14:47:37 +0000 (09:47 -0500)]
Merge branch 'tc-hw-offload'

John Fastabend says:

====================
tc offload for cls_u32 on ixgbe

This extends the setup_tc framework so it can support more than just
the mqprio offload and push other classifiers and qdiscs into the
hardware. The series here targets the u32 classifier and ixgbe
driver. I worked out the u32 classifier because it is protocol
oblivious and aligns with multiple hardware devices I have access
to. I did an initial implementation on ixgbe because (a) I have one
in my box (b) its a stable driver and (c) it is relatively simple
compared to the other devices I have here but still has enough
flexibility to exercise the features of cls_u32.

I intentionally limited the scope of this series to the basic
feature set. Specifically this uses a 'big hammer' feature bit
to do the offload or not. If the bit is set you get offloaded rules
if it is not then rules will not be offloaded. If we can agree on
this patch series there are some more patches on my queue we can
talk about to make the offload decision per rule using flags similar
to how we do l2 mac updates. Additionally the error strategy can
be improved to be hard aborting, log and continue, etc. I think
these are nice to have improvements but shouldn't block this series.

Also by adding get_parse_graph and set_parse_graph attributes as
in my previous flow_api work we can build programmable devices
and programmatically learn when rules can or can not be loaded
into the hardware. Again future work.

... v3 includes a couple style fixups suggested by Jiri and a
quick fix to ixgbe to check features flag and not dev_features
flag the latter being always set.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ixgbe: abort with cls u32 divisor groups greater than 1
John Fastabend [Wed, 17 Feb 2016 05:19:19 +0000 (21:19 -0800)]
net: ixgbe: abort with cls u32 divisor groups greater than 1

This patch ensures ixgbe will not try to offload hash tables from the
u32 module. The device class does not currently support this so until
it is enabled just abort on these tables.

Interestingly the more flexible your hardware is the less code you
need to implement to guard against these cases.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ixgbe: add support for tc_u32 offload
John Fastabend [Wed, 17 Feb 2016 05:18:53 +0000 (21:18 -0800)]
net: ixgbe: add support for tc_u32 offload

This adds initial support for offloading the u32 tc classifier. This
initial implementation only implements a few base matches and actions
to illustrate the use of the infrastructure patches.

However it is an interesting subset because it handles the u32 next
hdr logic to correctly map tcp packets from ip headers using the ihl
and protocol fields. After this is accepted we can extend the match
and action fields easily by updating the model header file.

Also only the drop action is supported initially.

Here is a short test script,

 #tc qdisc add dev eth4 ingress
 #tc filter add dev eth4 parent ffff: protocol ip \
u32 ht 800: order 1 \
match ip dst 15.0.0.1/32 match ip src 15.0.0.2/32 action drop

<-- hardware has dst/src ip match rule installed -->

 #tc filter del dev eth4 parent ffff: prio 49152
 #tc filter add dev eth4 parent ffff: protocol ip prio 99 \
handle 1: u32 divisor 1
 #tc filter add dev eth4 protocol ip parent ffff: prio 99 \
u32 ht 800: order 1 link 1: \
offset at 0 mask 0f00 shift 6 plus 0 eat match ip protocol 6 ff
 #tc filter add dev eth4 parent ffff: protocol ip \
u32 ht 1: order 3 match tcp src 23 ffff action drop

<-- hardware has tcp src port rule installed -->

 #tc qdisc del dev eth4 parent ffff:

<-- hardware cleaned up -->

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ixgbe: add minimal parser details for ixgbe
John Fastabend [Wed, 17 Feb 2016 05:18:28 +0000 (21:18 -0800)]
net: ixgbe: add minimal parser details for ixgbe

This adds an ixgbe data structure that is used to determine what
headers:fields can be matched and in what order they are supported.

For hardware devices this can be a bit tricky because typically
only pre-programmed (firmware, ucode, rtl) parse graphs will be
supported and we don't yet have an interface to change these from
the OS. So its sort of a you get whatever your friendly vendor
provides affair at the moment.

In the future we can add the get routines and set routines to
update this data structure. One interesting thing to note here
is the data structure here identifies ethernet, ip, and tcp
fields without having to hardcode them as enumerations or use
other identifiers.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: tc: helper functions to query action types
John Fastabend [Wed, 17 Feb 2016 05:18:03 +0000 (21:18 -0800)]
net: tc: helper functions to query action types

This is a helper function drivers can use to learn if the
action type is a drop action.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: add tc offload feature flag
John Fastabend [Wed, 17 Feb 2016 05:17:37 +0000 (21:17 -0800)]
net: add tc offload feature flag

Its useful to turn off the qdisc offload feature at a per device
level. This gives us a big hammer to enable/disable offloading.
More fine grained control (i.e. per rule) may be supported later.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: sched: add cls_u32 offload hooks for netdevs
John Fastabend [Wed, 17 Feb 2016 05:17:09 +0000 (21:17 -0800)]
net: sched: add cls_u32 offload hooks for netdevs

This patch allows netdev drivers to consume cls_u32 offloads via
the ndo_setup_tc ndo op.

This works aligns with how network drivers have been doing qdisc
offloads for mqprio.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: rework setup_tc ndo op to consume general tc operand
John Fastabend [Wed, 17 Feb 2016 05:16:43 +0000 (21:16 -0800)]
net: rework setup_tc ndo op to consume general tc operand

This patch updates setup_tc so we can pass additional parameters into
the ndo op in a generic way. To do this we provide structured union
and type flag.

This lets each classifier and qdisc provide its own set of attributes
without having to add new ndo ops or grow the signature of the
callback.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: rework ndo tc op to consume additional qdisc handle parameter
John Fastabend [Wed, 17 Feb 2016 05:16:15 +0000 (21:16 -0800)]
net: rework ndo tc op to consume additional qdisc handle parameter

The ndo_setup_tc() op was added to support drivers offloading tx
qdiscs however only support for mqprio was ever added. So we
only ever added support for passing the number of traffic classes
to the driver.

This patch generalizes the ndo_setup_tc op so that a handle can
be provided to indicate if the offload is for ingress or egress
or potentially even child qdiscs.

CC: Murali Karicheri <m-karicheri2@ti.com>
CC: Shradha Shah <sshah@solarflare.com>
CC: Or Gerlitz <ogerlitz@mellanox.com>
CC: Ariel Elior <ariel.elior@qlogic.com>
CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: Bruce Allan <bruce.w.allan@intel.com>
CC: Jesse Brandeburg <jesse.brandeburg@intel.com>
CC: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoi40e/i40evf: Bump i40e to 1.4.11 and i40evf to 1.4.7
Catherine Sullivan [Wed, 23 Dec 2015 20:05:56 +0000 (12:05 -0800)]
i40e/i40evf: Bump i40e to 1.4.11 and i40evf to 1.4.7

Bump.

Change-ID: I21aa520a3c8c5f4f562a98019bf8b76b3706c480
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: trivial: remove unnecessary local var
Jesse Brandeburg [Wed, 23 Dec 2015 20:05:55 +0000 (12:05 -0800)]
i40e: trivial: remove unnecessary local var

Probe routine already has too many locals, just convert one
used for kzalloc into a kcalloc, eliminating the local.

Change-ID: I349049872b71f858cbeb91ad7836e6767fc7b7d1
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Reviewed-by: Anjali Singhai <anjali.singhai@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: remove VF device IDs from PF
Jesse Brandeburg [Wed, 23 Dec 2015 20:05:54 +0000 (12:05 -0800)]
i40e: remove VF device IDs from PF

The PF doesn't need to know about the VF's device IDs, so remove them.

Change-ID: I62cf0e0fffa1ace586e58e00bc271b10ae440f05
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: add netdev info to VSI dump
Shannon Nelson [Wed, 23 Dec 2015 20:05:53 +0000 (12:05 -0800)]
i40e: add netdev info to VSI dump

Add a few more bits of netdev data into the debugfs output for dump VSI.
For now, we'll add the features, hw_features, vlan_features, and flags
bitflags and the state. More could be added later if needed.

Also, tweak a couple nearby output lines for output readability.

Change-ID: I9fb5a9da75c9ad7679498ce9ac3ba24d065ddd2e
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Reviewed-by: Brandeburg, Jesse <jesse.brandeburg@intel.com>
Reviewed-by: Wyborny, Carolyn <carolyn.wyborny@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40evf: enable bus master after reset
Mitch Williams [Wed, 23 Dec 2015 20:05:52 +0000 (12:05 -0800)]
i40evf: enable bus master after reset

If the VF is reset via VFLR, the device will be knocked out of bus
master mode, and the driver will fail to recover from the reset. Fix
this by enabling bus mastering after every reset. In a non-VFLR case,
the bus master bit will not be disabled, and this call will have no effect.

Change-ID: Id515859ac7a691db478222228add6d149e96801a
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: add a little more to an NVM update debug message
Shannon Nelson [Wed, 23 Dec 2015 20:05:51 +0000 (12:05 -0800)]
i40e: add a little more to an NVM update debug message

Add a little more detail to an NVM update debug message in order to
see the full ethtool request data.

Change-ID: Iab10437cb32d6fddc67ee347e7c0b42511e152cd
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Acked-by: Kevin Scott <kevin.c.scott@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: refactor DCB function
Jesse Brandeburg [Wed, 23 Dec 2015 20:05:50 +0000 (12:05 -0800)]
i40e: refactor DCB function

This is a simple refactor suggested by the community
to change a multi-level if statement into a switch.

Change-ID: I831cf3c40426022220aa9b43990022d22dfd50db
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: add 20G speed for Tx bandwidth calculations
Mitch Williams [Wed, 23 Dec 2015 20:05:49 +0000 (12:05 -0800)]
i40e: add 20G speed for Tx bandwidth calculations

When calculating TX bandwidth for VFs, we need to know the link speed to
make sure we don't allocate more bandwidth than is available. Add 20G
link speed to the switch statement so we can support devices that link
at that speed.

Change-ID: I5409f6139d549e5832777db9c22ca0664e0c5f8b
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: add counter for arq overflows
Mitch Williams [Wed, 23 Dec 2015 20:05:48 +0000 (12:05 -0800)]
i40e: add counter for arq overflows

Sometimes, ARQ overflows are a big deal and tell us that the
firmware/hardware/driver/something is having problems. But normally
they're no big deal. To assist in assessing this, add a counter to
our Ethtool stats. A handful of ARQ overflows during VF init is no
problem. A large, ever-growing number indicates that Something Bad is
happening.

Change-ID: Ie5348bfbc8a54a890559cb00279c28d976a55096
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: fix write-back-on-itr to work with legacy itr
Anjali Singhai Jain [Wed, 23 Dec 2015 20:05:47 +0000 (12:05 -0800)]
i40e: fix write-back-on-itr to work with legacy itr

We were not doing write-back on interrupt throttle for Legacy case in X722.
This patch fixes that, so we do WB_ON_ITR for Legacy as well. Plus the issue
that we should still be setting NO_ITR if we are touching the DYN_CTLN register
since we do not want to change ITR setting here.

Change-ID: I5db8491ee1544118a389db839cecc93e1bbc480e
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: Store lan_vsi_idx and lan_vsi_id in the right size
Pandi Maharajan [Wed, 23 Dec 2015 20:05:46 +0000 (12:05 -0800)]
i40e: Store lan_vsi_idx and lan_vsi_id in the right size

lan_vsi_idx and lan_vsi_id are assigned to u16 data sized variables but
declared in u8. This patch fixes the width of the datatype.

Change-ID: If4bcbcc7d32f2b287c51cb33d17879691258dce2
Signed-off-by: Pandi Kumar Maharajan <pandi.maharajan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: Bump AQ minor version to 1.5 for new FW features
Shannon Nelson [Wed, 23 Dec 2015 20:05:45 +0000 (12:05 -0800)]
i40e: Bump AQ minor version to 1.5 for new FW features

Bump AQ minor version to 1.5 for new FW features.

Change-ID: I5a790f7f519a2a8921aaa1c5663727dd1897ffec
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Acked-by: Kevin Scott <kevin.c.scott@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: AQ thermal sensor control struct
Shannon Nelson [Wed, 23 Dec 2015 20:05:44 +0000 (12:05 -0800)]
i40e: AQ thermal sensor control struct

Add the new AQ command and struct for managing a thermal sensor.

Change-ID: I6f5631839a0f3dca352a6c222f1269a960e2310a
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoMerge branch 'ip-sysctl-namespaceify'
David S. Miller [Wed, 17 Feb 2016 01:42:55 +0000 (20:42 -0500)]
Merge branch 'ip-sysctl-namespaceify'

Nikolay Borisov says:

====================
Namespacify various ip sysctl knobs

This series continues namespacifying more net related knobs.
The focus here is on ip options. Patches 1,3,4,5 namespacify
the respective sysctl knobs. Patch 2 moves some igmp code to the
correct file (and function) and also adds some #ifdef guards to
silence compilation warnings.

Finally, patch 5 exposes the ip fragmentation related sysctls
since all of the knobs are namespaced.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: Export ip fragment sysctl to unprivileged users
Nikolay Borisov [Mon, 15 Feb 2016 10:11:32 +0000 (12:11 +0200)]
net: Export ip fragment sysctl to unprivileged users

Now that all the ip fragmentation related sysctls are namespaceified
there is no reason to hide them anymore from "root" users inside
containers.

Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: namespacify ip fragment max dist sysctl knob
Nikolay Borisov [Mon, 15 Feb 2016 10:11:31 +0000 (12:11 +0200)]
ipv4: namespacify ip fragment max dist sysctl knob

Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: namespacify ip_early_demux sysctl knob
Nikolay Borisov [Mon, 15 Feb 2016 10:11:30 +0000 (12:11 +0200)]
ipv4: namespacify ip_early_demux sysctl knob

Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: Namespacify ip_dynaddr sysctl knob
Nikolay Borisov [Mon, 15 Feb 2016 10:11:29 +0000 (12:11 +0200)]
ipv4: Namespacify ip_dynaddr sysctl knob

Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoigmp: net: Move igmp namespace init to correct file
Nikolay Borisov [Mon, 15 Feb 2016 10:11:28 +0000 (12:11 +0200)]
igmp: net: Move igmp namespace init to correct file

When igmp related sysctl were namespacified their initializatin was
erroneously put into the tcp socket namespace constructor. This
patch moves the relevant code into the igmp namespace constructor to
keep things consistent.

Also sprinkle some #ifdefs to silence warnings

Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: Namespaceify ip_default_ttl sysctl knob
Nikolay Borisov [Mon, 15 Feb 2016 10:11:27 +0000 (12:11 +0200)]
ipv4: Namespaceify ip_default_ttl sysctl knob

Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'wireless-drivers-next-for-davem-2016-02-12' of git://git.kernel.org/pub...
David S. Miller [Wed, 17 Feb 2016 01:38:29 +0000 (20:38 -0500)]
Merge tag 'wireless-drivers-next-for-davem-2016-02-12' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
Major changes:

wl12xx

* add device tree support for SPI

mwifiex

* add debugfs file to read chip information
* add MSIx support for newer pcie chipsets (8997 onwards)
* add schedule scan support
* add WoWLAN net-detect support
* firmware dump support for w8997 chipset

iwlwifi

* continue the work on multiple Rx queues
* add support for beacon storing used in low power states
* use the regular firmware image of WoWLAN
* fix 8000 devices for Big Endian machines
* more firmware debug hooks
* add support for P2P Client snoozing
* make the beacon filtering for AP mode configurable
* fix transmit queues overflow with LSO

libertas

* add support for setting power save via cfg80211
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: phy: spi_ks8995: include linux/gpio/consumer.h
Arnd Bergmann [Fri, 12 Feb 2016 10:42:34 +0000 (11:42 +0100)]
net: phy: spi_ks8995: include linux/gpio/consumer.h

The ks8995 phy driver just started using gpiod_* functions, which are
declared in linux/gpio/consumer.h, not linux/gpio.h, resulting in a
build failure in randconfig builds that do not have CONFIG_GPIOLIB
enabled:

drivers/net/phy/spi_ks8995.c: In function 'ks8995_probe':
drivers/net/phy/spi_ks8995.c:477:3: error: implicit declaration of function 'gpiod_set_value' [-Werror=implicit-function-declaration]

This changes the header inclusion so it builds in all configurations.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: cd6f288cbaab ("net: phy: spi_ks8995: add support for resetting switch using GPIO")
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoi40e: AQ Add VXLAN-GPE tunnel type
Shannon Nelson [Wed, 23 Dec 2015 20:05:43 +0000 (12:05 -0800)]
i40e: AQ Add VXLAN-GPE tunnel type

Add the new Cisco VXLAN-GPE cloud tunnel type for the Add Cloud Filter
and UDP tunnel AQ commands.

Change-ID: I2c093c7d79726c7fca08a36e5c63581a905da3d2
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Acked-by: Kevin Scott <kevin.c.scott@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agotcp: add tcpi_min_rtt and tcpi_notsent_bytes to tcp_info
Eric Dumazet [Fri, 12 Feb 2016 06:02:53 +0000 (22:02 -0800)]
tcp: add tcpi_min_rtt and tcpi_notsent_bytes to tcp_info

tcpi_min_rtt reports the minimal rtt observed by TCP stack for the flow,
in usec unit. Might be ~0U if not yet known.

tcpi_notsent_bytes reports the amount of bytes in the write queue that
were not yet sent.

This is done in a single patch to not add a temporary 32bit padding hole
in tcp_info.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'unified-tunnel-dst-caching'
David S. Miller [Wed, 17 Feb 2016 01:21:49 +0000 (20:21 -0500)]
Merge branch 'unified-tunnel-dst-caching'

Paolo Abeni says:

====================
net: unify dst caching for tunnel devices

This patch series try to unify the dst cache implementations currently
present in the kernel, namely in ip_tunnel.c and ip6_tunnel.c, introducing a
new generic implementation, replacing the existing ones, and then using
the new implementation in other tunnel devices which currently lack it.

The new dst implementation is compiled, as built-in, only if any device using
it is enabled.

Caching the dst for the tunnel remote address gives small, but measurable,
performance improvement when tunneling over ipv4 (in the 2%-4% range) and
significant ones when tunneling over ipv6 (roughly 60% when no
fragmentation/segmentation take place and the tunnel local address
is not specified).

v2:
- move the vxlan dst_cache usage inside the device lookup functions
- fix usage after free for lwt tunnel moving the dst cache storage inside
  the dst_metadata,
- sparse codying style cleanup
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/ipv4: add dst cache support for gre lwtunnels
Paolo Abeni [Fri, 12 Feb 2016 14:43:59 +0000 (15:43 +0100)]
net/ipv4: add dst cache support for gre lwtunnels

In case of UDP traffic with datagram length below MTU this
gives about 4% performance increase

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Suggested-and-Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agogeneve: add dst caching support
Paolo Abeni [Fri, 12 Feb 2016 14:43:58 +0000 (15:43 +0100)]
geneve: add dst caching support

use generic dst implementation for both plain geneve devices and
lwtunnels.

In case of UDP traffic with datagram length below MTU this give
about 2% performance increase for plain geneve tunnel over ipv4,
about 65% performance increase for ipv6 tunnel.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Suggested-and-Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: add dst_cache to ovs vxlan lwtunnel
Paolo Abeni [Fri, 12 Feb 2016 14:43:57 +0000 (15:43 +0100)]
net: add dst_cache to ovs vxlan lwtunnel

In case of UDP traffic with datagram length
below MTU this give about 2% performance increase
when tunneling over ipv4 and about 60% when tunneling
over ipv6

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Suggested-and-acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: use dst_cache for vxlan device
Paolo Abeni [Fri, 12 Feb 2016 14:43:56 +0000 (15:43 +0100)]
net: use dst_cache for vxlan device

In case of UDP traffic with datagram length
below MTU this give about 3% performance increase
when tunneling over ipv4 and about 70% when
tunneling over ipv6.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Suggested-and-acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoip_tunnel: replace dst_cache with generic implementation
Paolo Abeni [Fri, 12 Feb 2016 14:43:55 +0000 (15:43 +0100)]
ip_tunnel: replace dst_cache with generic implementation

The current ip_tunnel cache implementation is prone to a race
that will cause the wrong dst to be cached on cuncurrent dst cache
miss and ip tunnel update via netlink.

Replacing with the generic implementation fix the issue.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Suggested-and-acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: replace dst_cache ip6_tunnel implementation with the generic one
Paolo Abeni [Fri, 12 Feb 2016 14:43:54 +0000 (15:43 +0100)]
net: replace dst_cache ip6_tunnel implementation with the generic one

This also fix a potential race into the existing tunnel code, which
could lead to the wrong dst to be permanenty cached:

CPU1: CPU2:
  <xmit on ip6_tunnel>
  <cache lookup fails>
  dst = ip6_route_output(...)
<tunnel params are changed via nl>
dst_cache_reset() // no effect,
// the cache is empty
  dst_cache_set() // the wrong dst
// is permanenty stored
// into the cache

With the new dst implementation the above race is not possible
since the first cache lookup after dst_cache_reset will fail due
to the timestamp check

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Suggested-and-acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: add dst_cache support
Paolo Abeni [Fri, 12 Feb 2016 14:43:53 +0000 (15:43 +0100)]
net: add dst_cache support

This patch add a generic, lockless dst cache implementation.
The need for lock is avoided updating the dst cache fields
only in per cpu scope, and requiring that the cache manipulation
functions are invoked with the local bh disabled.

The refresh_ts and reset_ts fields are used to ensure the cache
consistency in case of cuncurrent cache update (dst_cache_set*) and
reset operation (dst_cache_reset).

Consider the following scenario:

CPU1:                                    CPU2:
  <cache lookup with emtpy cache: it fails>
  <get dst via uncached route lookup>
<related configuration changes>
                                         dst_cache_reset()
  dst_cache_set()

The dst entry set passed to dst_cache_set() should not be used
for later dst cache lookup, because it's obtained using old
configuration values.

Since the refresh_ts is updated only on dst_cache lookup, the
cached value in the above scenario will be discarded on the next
lookup.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Suggested-and-acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'bnx2x-next'
David S. Miller [Wed, 17 Feb 2016 01:12:17 +0000 (20:12 -0500)]
Merge branch 'bnx2x-next'

Yuval Mintz says:

====================
bnx2x: driver updates

This series contains several changes - the biggest change is the
addition of Geneve NDO support [allows device to perform RSS according
to inner-headers of encapsulated packet, similar to what it does for
vxlan]. It also extends dcbx support, as well as introducing some minor
changes.

Dave,

Please consider applying this series to `net-next'.
[Do notice patch #3 fails checkpatch due to consistency with existing
HSI]
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnx2x: Warn about grc timeouts in register dump
Yuval Mintz [Tue, 16 Feb 2016 16:08:01 +0000 (18:08 +0200)]
bnx2x: Warn about grc timeouts in register dump

There are several scenarios where taking a register dump from a device
might log benign GRC timeout attentions to system logs.
Most common of those is when taking the dump from a 2-port device.

Sadly, there's no easy way to mask the problematic attentions during the
flow - Changing this behvaior would require a firmware update.
For now, simply warn users to ignore the warnings.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnx2x: extend DCBx support
Yuval Mintz [Tue, 16 Feb 2016 16:08:00 +0000 (18:08 +0200)]
bnx2x: extend DCBx support

This adds support for default application priority.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: Ariel Elior <Ariel.Elior@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnx2x: Add support for single-port DCBx
Yuval Mintz [Tue, 16 Feb 2016 16:07:59 +0000 (18:07 +0200)]
bnx2x: Add support for single-port DCBx

Driver is currently looking at shared information for determining whether
DCBx can be supported for a given port.
On 4-port devices, up-to-date management firmware can support DCBx on
each port of a given engine independently - but that would cause bnx2x to
misinterpert the support and assume DCBx is supported on both.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: Ariel Elior <Ariel.Elior@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>