GitHub/moto-9609/android_kernel_motorola_exynos9610.git
8 years agoigb: Only DMA sync frame length
Andrew Lunn [Fri, 3 Jun 2016 21:03:25 +0000 (23:03 +0200)]
igb: Only DMA sync frame length

On some platforms, syncing a buffer for DMA is expensive. Rather than
sync the whole 2K receive buffer, only synchronise the length of the
frame, which will typically be the MTU, or a much smaller TCP ACK.

For an IMX6Q, this gives around 6% increased TCP receive performance,
which is cache operations bound and reduces CPU load for TCP transmit.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoixgbe: fix spoofed packets with macvlans
Emil Tantilov [Thu, 2 Jun 2016 01:59:44 +0000 (18:59 -0700)]
ixgbe: fix spoofed packets with macvlans

When setting spoofing, both VLAN and MAC need to be set together.
This change resolves an issue where MAC-VLANs on the VF fail to pass
traffic due to spoofed packets.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoixgbe: Correct reporting of timestamping for x550
Tony Nguyen [Wed, 1 Jun 2016 16:50:43 +0000 (09:50 -0700)]
ixgbe: Correct reporting of timestamping for x550

Update ixgbe_ethtool_get_ts_info() to show that x550 supports hardware
timestamping of all packets.

Reported-by: Guy Harris <guy@alum.mit.edu>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agofm10k: Remove create_workqueue
Bhaktipriya Shridhar [Wed, 1 Jun 2016 15:40:09 +0000 (21:10 +0530)]
fm10k: Remove create_workqueue

alloc_workqueue replaces deprecated create_workqueue().

A dedicated workqueue has been used since the workitem (viz
fm10k_service_task, which manages and runs other subtasks) is involved in
normal device operation and requires forward progress under memory
pressure.

create_workqueue has been replaced with alloc_workqueue with max_active
as 0 since there is no need for throttling the number of active work
items.

Since network devices may be used in memory reclaim path,
WQ_MEM_RECLAIM has been set to guarantee forward progress.

flush_workqueue is unnecessary since destroy_workqueue() itself calls
drain_workqueue() which flushes repeatedly till the workqueue
becomes empty. Hence the call to flush_workqueue() has been dropped.

Signed-off-by: Bhaktipriya Shridhar <bhaktipriya96@gmail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoigb: call igb_ptp_suspend during suspend/resume cycle
Jacob Keller [Tue, 24 May 2016 20:56:31 +0000 (13:56 -0700)]
igb: call igb_ptp_suspend during suspend/resume cycle

Properly stop the extra workqueue items and ensure that we resume
cleanly. This is better than using igb_ptp_init and igb_ptp_stop since
these functions destroy the PHC device, which will cause other problems
if we do so. Since igb_ptp_reset now re-schedules the work-queue item we
don't need an equivalent igb_ptp_resume in the resume workflow.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoigb: implement igb_ptp_suspend
Jacob Keller [Tue, 24 May 2016 20:56:30 +0000 (13:56 -0700)]
igb: implement igb_ptp_suspend

Make igb_ptp_stop take advantage of this new function to reduce code
duplication.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoigb: re-use igb_ptp_reset in igb_ptp_init
Jacob Keller [Tue, 24 May 2016 20:56:29 +0000 (13:56 -0700)]
igb: re-use igb_ptp_reset in igb_ptp_init

Modify igb_ptp_init to take advantage of igb_ptp_reset, and remove
duplicated work that was occurring in both igb_ptp_reset and
igb_ptp_init.

In total, resetting the TSAUXC register, and resetting the system time
both happen in igb_ptp_reset already. igb_ptp_reset now also takes care
of starting the delayed work item for overflow checks, as well.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoigb: introduce IGB_PTP_OVERFLOW_CHECK flag
Jacob Keller [Tue, 24 May 2016 20:56:28 +0000 (13:56 -0700)]
igb: introduce IGB_PTP_OVERFLOW_CHECK flag

Don't continue to use complex MAC type checks for handling various cases
where we have overflow check code. Make this code more obvious by
introducing a flag which is enabled for hardware that needs these
checks.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoigb: introduce ptp_flags variable and use it to replace IGB_FLAG_PTP
Jacob Keller [Tue, 24 May 2016 20:56:27 +0000 (13:56 -0700)]
igb: introduce ptp_flags variable and use it to replace IGB_FLAG_PTP

Upcoming patches will introduce new PTP specific flags. To avoid
cluttering the normal flags variable, introduce PTP specific "ptp_flags"
variable for this purpose, and move IGB_FLAG_PTP to become
IGB_PTP_ENABLED.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoixgbe: Error handler for duplicate filter locations in hardware for cls_u32 offloads
Amritha Nambiar [Tue, 17 May 2016 01:33:20 +0000 (18:33 -0700)]
ixgbe: Error handler for duplicate filter locations in hardware for cls_u32 offloads

For u32 classifier filters, avoid overwriting existing filter
in a hardware location without removing it first, to clean up
inconsistencies due to duplicate values for filter location.

Verified with the following filters:

Create child hash tables:
handle 1: u32 divisor 1
handle 2: u32 divisor 1

Link to the child hash table from parent hash table:
handle 800:0:11 u32 ht 800: link 1: \
offset at 0 mask 0f00 shift 6 plus 0 eat \
match ip protocol 6 ff match ip dst 15.0.0.1/32

handle 800:0:12 u32 ht 800: link 2: \
offset at 0 mask 0f00 shift 6 plus 0 eat \
match ip protocol 17 ff match ip dst 16.0.0.1/32

Add filter into child hash table:
handle 1:0:3 u32 ht 1: \
match tcp src 22 ffff action drop

Add another filter to the same location:
handle 2:0:3 u32 ht 2: \
match tcp src 33 ffff action drop

Signed-off-by: Amritha Nambiar <amritha.nambiar@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoixgbe: Fix deleting link filters for cls_u32 offloads
Amritha Nambiar [Sat, 7 May 2016 02:09:51 +0000 (19:09 -0700)]
ixgbe: Fix deleting link filters for cls_u32 offloads

On deleting filters which are links to a child hash table, the filters
in the child hash table must be cleared from the hardware if there
is no link between the parent and child hash table.

Verified with the following filters:

Create a child hash table:
handle 1: u32 divisor 1

Link to the child hash table from parent hash table:
handle 800:0:10 u32 ht 800: link 1: \
offset at 0 mask 0f00 shift 6 plus 0 eat \
match ip protocol 6 ff match ip dst 15.0.0.1/32

Add filters into child hash table:
handle 1:0:2 u32 ht 1: \
match tcp src 22 ffff action drop
        handle 1:0:3 u32 ht 1: \
        match tcp src 33 ffff action drop

Delete link filter from parent hash table:
handle 800:0:10 u32

Signed-off-by: Amritha Nambiar <amritha.nambiar@intel.com>
Acked-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoe1000e: prevent division by zero if TIMINCA is zero
Denys Vlasenko [Fri, 6 May 2016 19:41:51 +0000 (21:41 +0200)]
e1000e: prevent division by zero if TIMINCA is zero

Users report that under VMWare, er32(TIMINCA) returns zero.
This causes division by zero at init time as follows:

 ==>       incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
           for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
                   /* latch SYSTIMH on read of SYSTIML */
                   systim_next = (cycle_t)er32(SYSTIML);
                   systim_next |= (cycle_t)er32(SYSTIMH) << 32;

                   time_delta = systim_next - systim;
                   temp = time_delta;
 ====>             rem = do_div(temp, incvalue);

This change makes kernel survive this, and users report that
NIC does work after this change.

Since on real hardware incvalue is never zero, this should not affect
real hardware use case.

Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agofm10k: fix incorrect index calculation in fm10k_write_reta
Jacob Keller [Mon, 18 Apr 2016 22:45:00 +0000 (15:45 -0700)]
fm10k: fix incorrect index calculation in fm10k_write_reta

The index calculated when looping through the indir array passed to
fm10k_write_reta was incorrectly calculated as the first part i needs to
be multiplied by 4.

Fixes: 0cfea7a65738 ("fm10k: fix possible null pointer deref after kcalloc", 2016-04-13)
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agofm10k: Align Rx buffers to 512B blocks
Alexander Duyck [Fri, 15 Apr 2016 17:00:46 +0000 (13:00 -0400)]
fm10k: Align Rx buffers to 512B blocks

While reviewing the i40e driver changes to support page based receive I
realized that I had overlooked the fact that the fm10k hardware required a
512 byte alignment for Rx buffers.  This patch is meant to address that by
changing the alignment for Rx buffers to 512 bytes instead of allowing it
to be L1 cache aligned.

Signed-off-by: Alexander Duyck <aduyck@mirantis.com>
Tested-by: Krishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agofm10k: don't use BIT() macro where the value isn't a bitmask
Jacob Keller [Thu, 14 Apr 2016 20:17:27 +0000 (13:17 -0700)]
fm10k: don't use BIT() macro where the value isn't a bitmask

The FM10K_MAX_DATA_PER_TXD is really just using a bitshift as a power of
2 operation in an efficient manner. We shouldn't represent this as a BIT()
because that obscures the intention of the operation.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <Krishneil.k.singh@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, 29 Jun 2016 12:23:08 +0000 (08:23 -0400)]
Merge branch '40GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
40GbE Intel Wired LAN Driver Updates 2016-06-27

This series contains updates to i40e and i40evf only.

Mitch provides several changes, first adds functions to enable and disable
VSI on a VEB, which allows for configuration of limited promiscuous mode
specifically for bridging purposes.  Sets the RSS Hash Enable registers by
default now that VF RSS is configured by the PF driver.  Fixed a issue
where we could overflow the buffer, by checking the address count and bail
out of the loop at the appropriate time.  Removed the need for a reset
when the device enters limited promiscuous mode, since this was causing
heartburn for people who were using VFs and bridging.

Catherine adds a call to set the client interface down when we put the VSI
down.  Fixed an issue where RSS queues was being limited to the number
of CPUs, so if a user wants to use more queues than CPUs, we want to
trust they know what they are doing and let them.

Greg cleans up the driver suspend routine to ensure we are calling
synchronize_irq() before freeing IRQ vectors and explicitly free the other
causes interrupt resources and shut down the MSIX interrupt.

Serey fixes i40e_set_settings() to not fail when a Direct Attach (DA)
cable is used.

Avinash fixes a supported link bug by removing code which was not allowing
100BaseT to show up in the supported link modes for 10GBaseT PHYs.

Shannon adds a bit of information to the error messages to help determine
the source of error by adding VSI info to macaddr messages.

Tushar Dave fixes error received when turning off TSO on some systems,
which was caused by enabling FD_SB without checking availability of
MSIx vectors, so add the check.

Neerav fixes a possible panic when LLDP/DCBX change happens and the
driver tried to notify the client(s) for each of the PF VSIs, which would
panic when it reached a VSI that did not have any netdev associated with
it.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: lpc_eth: use phy_ethtool_{get|set}_link_ksettings
Philippe Reynes [Tue, 28 Jun 2016 21:59:45 +0000 (23:59 +0200)]
net: ethernet: lpc_eth: use phy_ethtool_{get|set}_link_ksettings

There are two generics functions phy_ethtool_{get|set}_link_ksettings,
so we can use them instead of defining the same code in the driver.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: lpc_eth: use phydev from struct net_device
Philippe Reynes [Tue, 28 Jun 2016 21:59:44 +0000 (23:59 +0200)]
net: ethernet: lpc_eth: use phydev from struct net_device

The private structure contain a pointer to phydev, but the structure
net_device already contain such pointer. So we can remove the pointer
phydev in the private structure, and update the driver to use the
one contained in struct net_device.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet_sched: netem: do not call qdisc_drop() with a NULL skb
Eric Dumazet [Tue, 28 Jun 2016 08:30:08 +0000 (10:30 +0200)]
net_sched: netem: do not call qdisc_drop() with a NULL skb

If skb_unshare() fails, we call qdisc_drop() with a NULL skb, which
is no longer supported.

Fixes: 520ac30f4551 ("net_sched: drop packets after root qdisc lock is released")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'qed-static-checker'
David S. Miller [Wed, 29 Jun 2016 11:53:42 +0000 (07:53 -0400)]
Merge branch 'qed-static-checker'

Sudarsana Reddy Kalluru says:

====================
qed*: Fix the static checker warnings.

The patch series addresses the static checker warnings introduced by the
earlier patches related to qed/qede coalesce configuration support.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqede: Fix the static checker warnings.
Sudarsana Reddy Kalluru [Tue, 28 Jun 2016 06:10:59 +0000 (02:10 -0400)]
qede: Fix the static checker warnings.

Static checker warnings:
drivers/net/ethernet/qlogic/qede/qede_ethtool.c:435 qede_get_coalesce()
warn: passing casted pointer '&coal->rx_coalesce_usecs' to
'edev->ops->common->get_coalesce()' 32 vs 16.

The u32 pointer is being typecasted to u16 which may fail for big-endian
platforms.

Fixes: d552fa84cb35 ("qede: Add support for coalescing config read/update.")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Sudarsana Reddy Kalluru <sudarsana.kalluru@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoqed: Fix static checker warnings.
Sudarsana Reddy Kalluru [Tue, 28 Jun 2016 06:10:58 +0000 (02:10 -0400)]
qed: Fix static checker warnings.

Static checker warnings:
drivers/net/ethernet/qlogic/qed/qed_int.c:2450 qed_init_cau_sb_entry()
warn: always true condition '(cdev->rx_coalesce_usecs <= 255) =>
(0-255 <= 255)'
drivers/net/ethernet/qlogic/qed/qed_int.c:2511 qed_int_cau_conf_sb()
warn: always true condition '(p_hwfn->cdev->rx_coalesce_usecs <= 255)
=> (0-255 <= 255)'
..

The data types for rx/tx_coalesce_usecs should be u16.

Fixes: commit 722003ac40c2 ("qed: Add support for coalescing config read/update.")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Sudarsana Reddy Kalluru <sudarsana.kalluru@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: mvpp2: use phy_ethtool_{get|set}_link_ksettings
Philippe Reynes [Mon, 27 Jun 2016 22:08:12 +0000 (00:08 +0200)]
net: ethernet: mvpp2: use phy_ethtool_{get|set}_link_ksettings

There are two generics functions phy_ethtool_{get|set}_link_ksettings,
so we can use them instead of defining the same code in the driver.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: mvpp2: use phydev from struct net_device
Philippe Reynes [Mon, 27 Jun 2016 22:08:11 +0000 (00:08 +0200)]
net: ethernet: mvpp2: use phydev from struct net_device

The private structure contain a pointer to phydev, but the structure
net_device already contain such pointer. So we can remove the pointer
phydev in the private structure, and update the driver to use the
one contained in struct net_device.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agotipc: rename udp_port in struct udp_media_addr
Richard Alpe [Mon, 27 Jun 2016 11:34:07 +0000 (13:34 +0200)]
tipc: rename udp_port in struct udp_media_addr

Context implies that port in struct "udp_media_addr" is referring
to a UDP port.

Signed-off-by: Richard Alpe <richard.alpe@ericsson.com>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agotipc: honor msg2addr return value
Richard Alpe [Mon, 27 Jun 2016 11:34:06 +0000 (13:34 +0200)]
tipc: honor msg2addr return value

The UDP msg2addr function tipc_udp_msg2addr() can return -EINVAL which
prior to this patch was unhanded in the caller.

Signed-off-by: Richard Alpe <richard.alpe@ericsson.com>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: the space is required before the open parenthesis '('
Wei Tang [Mon, 27 Jun 2016 10:12:46 +0000 (18:12 +0800)]
net: the space is required before the open parenthesis '('

The space is missing before the open parenthesis '(', and this
will introduce much more noise when checking patch around.

Signed-off-by: Wei Tang <tangwei@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodsa: b53: avoid 'maybe-uninitialized' warning
Arnd Bergmann [Mon, 27 Jun 2016 09:19:13 +0000 (11:19 +0200)]
dsa: b53: avoid 'maybe-uninitialized' warning

In some configurations, gcc produces a warning for correct code
in this driver:

drivers/net/dsa/b53/b53_mmap.c: In function 'b53_mmap_read64':
drivers/net/dsa/b53/b53_mmap.c:107:10: error: 'hi' may be used uninitialized in this function [-Werror=maybe-uninitialized]
  *val = ((u64)hi << 32) | lo;
          ^~~~~~~
drivers/net/dsa/b53/b53_mmap.c: In function 'b53_mmap_read48':
drivers/net/dsa/b53/b53_mmap.c:91:11: error: 'hi' may be used uninitialized in this function [-Werror=maybe-uninitialized]
   *val = ((u64)hi << 32) | lo;
           ^~~~~~~
drivers/net/dsa/b53/b53_mmap.c:83:11: error: 'hi' may be used uninitialized in this function [-Werror=maybe-uninitialized]
   *val = ((u64)hi << 16) | lo;

I have seen the warning before and at the time thought I had fixed
it with 55e7f6abe131 ("dsa: b53: fix big-endian register access"),
however it now came back in a different randconfig build that happens
to have different inlining decisions in the compiler.

The mistake that gcc makes here is that it thinks the second call to
readl() might fail because the address 'reg + 4' is not a multiple
of four despite having knowing that 'reg' itself is a multiple of four.

By open-coding the two reads without the redundant alignment check,
we can avoid the warning and produce slightly better object code, but
get slightly longer source code instead.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: ti: cpdma: switch to use genalloc
Grygorii Strashko [Mon, 27 Jun 2016 09:05:11 +0000 (12:05 +0300)]
net: ethernet: ti: cpdma: switch to use genalloc

TI CPDMA currently uses a bitmap for tracking descriptors alloactions
allocations, but The genalloc already handles the same and can be used
as with special memory (SRAM) as with DMA cherent memory chank
(dma_alloc_coherent()). Hence, switch to using genalloc and add
desc_num property for each channel for limitation of max number of
allowed descriptors for each CPDMA channel. This patch do not affect
on net throuput.

Acked-by: Mugunthan V N <mugunthanvnm@ti.com>
Tested-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMaxim/driver: Add driver for maxim ds26522
Zhao Qiang [Mon, 27 Jun 2016 01:30:22 +0000 (09:30 +0800)]
Maxim/driver: Add driver for maxim ds26522

Signed-off-by: Zhao Qiang <qiang.zhao@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: dwc_eth_qos: use phy_ethtool_{get|set}_link_ksettings
Philippe Reynes [Sat, 25 Jun 2016 21:05:16 +0000 (23:05 +0200)]
net: ethernet: dwc_eth_qos: use phy_ethtool_{get|set}_link_ksettings

There are two generics functions phy_ethtool_{get|set}_link_ksettings,
so we can use them instead of defining the same code in the driver.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: dwc_eth_qos: use phydev from struct net_device
Philippe Reynes [Sat, 25 Jun 2016 21:05:15 +0000 (23:05 +0200)]
net: ethernet: dwc_eth_qos: use phydev from struct net_device

The private structure contain a pointer to phydev, but the structure
net_device already contain such pointer. So we can remove the pointer
phydev in the private structure, and update the driver to use the
one contained in struct net_device.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: sxgbe: use phy_ethtool_{get|set}_link_ksettings
Philippe Reynes [Sat, 25 Jun 2016 20:05:27 +0000 (22:05 +0200)]
net: ethernet: sxgbe: use phy_ethtool_{get|set}_link_ksettings

There are two generics functions phy_ethtool_{get|set}_link_ksettings,
so we can use them instead of defining the same code in the driver.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: sxgbe: use phydev from struct net_device
Philippe Reynes [Sat, 25 Jun 2016 20:05:26 +0000 (22:05 +0200)]
net: ethernet: sxgbe: use phydev from struct net_device

The private structure contain a pointer to phydev, but the structure
net_device already contain such pointer. So we can remove the pointer
phydev in the private structure, and update the driver to use the
one contained in struct net_device.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: r6040: use phy_ethtool_{get|set}_link_ksettings
Philippe Reynes [Sat, 25 Jun 2016 19:09:02 +0000 (21:09 +0200)]
net: ethernet: r6040: use phy_ethtool_{get|set}_link_ksettings

There are two generics functions phy_ethtool_{get|set}_link_ksettings,
so we can use them instead of defining the same code in the driver.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: r6040: use phydev from struct net_device
Philippe Reynes [Sat, 25 Jun 2016 19:09:01 +0000 (21:09 +0200)]
net: ethernet: r6040: use phydev from struct net_device

The private structure contain a pointer to phydev, but the structure
net_device already contain such pointer. So we can remove the pointer
phydev in the private structure, and update the driver to use the
one contained in struct net_device.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: hix5hd2: use phy_ethtool_{get|set}_link_ksettings
Philippe Reynes [Sat, 25 Jun 2016 14:55:13 +0000 (16:55 +0200)]
net: ethernet: hix5hd2: use phy_ethtool_{get|set}_link_ksettings

There are two generics functions phy_ethtool_{get|set}_link_ksettings,
so we can use them instead of defining the same code in the driver.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: hix5hd2: use phydev from struct net_device
Philippe Reynes [Sat, 25 Jun 2016 14:55:12 +0000 (16:55 +0200)]
net: ethernet: hix5hd2: use phydev from struct net_device

The private structure contain a pointer to phydev, but the structure
net_device already contain such pointer. So we can remove the pointer
phy in the private structure, and update the driver to use the
one contained in struct net_device.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'cpsw-runtime-pm'
David S. Miller [Tue, 28 Jun 2016 12:57:28 +0000 (08:57 -0400)]
Merge branch 'cpsw-runtime-pm'

Grygorii Strashko says:

====================
drivers: net: cpsw: improve runtime pm

This series intended to improve runtime PM and allow CPSW to be
RPM suspended when all ethX netdevices are down.

To achieve above goal it is required to relax runtime PM constraints for
Davinci MDIO which blocks CPSW runtime PM now, because Davinci MDIO is always
powered on during probe and powered off only when it's going to be removed.
- Patches 6-11 implement PM runtime autosuspend for Davinci MDIO, but keep it
disabled by default, because Davinci MDIO is integrated in big set of TI devices
and not all of them verified to work correctly with RPM autosuspend enabled:
 expected to work on SoCs where MDIO is defined as part of CPSW in DT
 (cpsw.c DRA7/am57x, am437x, am335x)
The CPSW need to be fixed before RPM suspended can be allowed:
 - Patches 1-5 ensure that CPSW will not cause L3 errors while it is in RPM
   suspended state.

Davinci MDIO RPM autosuspend can be enabled through sysfs:
 echo 100 > /sys/devices/../48484000.ethernet/48485000.mdio/power/autosuspend_delay_ms

Patches 12 - 15: introduce new compatible string "ti,cpsw-mdio" which is used
then to enable RPM for am335x/am437x/dra7 SoCs.

Tested on am335x, am437x, am572x and k2g (on k2g with RPM disabled for Davinci MDIO)
These changes should not affect on errata i877 implementation on DRA7.

Power measurement on am335x GP EVM:
 Without this series:  547.60 mW total SoC power
 With this series + "ifconfig eth0 down": 477.32 mW Total Soc Power

Changes in v2:
- CPSW ethtool interface updated to use .begin()/.complete() callbacks
- kbuild failure fixed
- davinci_mdio DT updated with proper description of allowed compatible strings
  combinations

Link on v1:
 https://lkml.org/lkml/2016/6/15/362
====================

Reviewed-by: Mugunthan V N <mugunthanvnm@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoARM: dts: am335x/am437x/dra7: use new "ti, cpsw-mdio" compat string
Grygorii Strashko [Fri, 24 Jun 2016 18:23:55 +0000 (21:23 +0300)]
ARM: dts: am335x/am437x/dra7: use new "ti, cpsw-mdio" compat string

Add "ti,cpsw-mdio" for am335x/am437x/dra7 SoCs where MDIO is
implemented as part of TI CPSW and, this way, enable PM runtime auto
suspend for Davinci MDIO driver on these paltforms.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: davinci_mdio: enable pm runtime auto for ti cpsw-mdio
Grygorii Strashko [Fri, 24 Jun 2016 18:23:54 +0000 (21:23 +0300)]
drivers: net: davinci_mdio: enable pm runtime auto for ti cpsw-mdio

Use "ti,cpsw-mdio" to enable PM runtime auto-suspend on supported
platforms, where MDIO is implemented as part of TI CPSW.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: davinci_mdio: introduce "ti,cpsw-mdio" compat string
Grygorii Strashko [Fri, 24 Jun 2016 18:23:53 +0000 (21:23 +0300)]
net: davinci_mdio: introduce "ti,cpsw-mdio" compat string

Introduce "ti,cpsw-mdio" compatible string for Davinci MDIO, because
it's required to distinguish the case when MDIO is part of TI CPSW to
enable features supported by TI CPSW (for example, enable PM
management).

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: davinci_mdio: document missed "ti, am4372-mdio" compat string
Grygorii Strashko [Fri, 24 Jun 2016 18:23:52 +0000 (21:23 +0300)]
net: davinci_mdio: document missed "ti, am4372-mdio" compat string

Document missed "ti,am4372-mdio" compat string used for TI am437x SoC
(am4372.dtsi).

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: davinci_mdio: implement pm runtime auto mode
Grygorii Strashko [Fri, 24 Jun 2016 18:23:51 +0000 (21:23 +0300)]
drivers: net: davinci_mdio: implement pm runtime auto mode

Davinci MDIO is always used as slave device which services
read/write requests from MDIO/PHY core. It doesn't use IRQ also.

As result, It's possible to relax PM runtime constraints for Davinci
MDIO and enable it on demand, instead of powering it during probe
and powering off during removal.

Hence, implement PM runtime autosuspend for Davinci MDIO, but keep it
disabled by default, because Davinci MDIO is integrated in big set of
TI devices and not all of them expected to work corectly with RPM
 autosuspend enabled:
- expected to work on SoCs where MDIO is part of TI CPSW
(cpsw.c DRA7/am57x, am437x, am335x, dm814x)
- not verified on Keystone 2 and other SoCs where MDIO is used with TI EMAC IP
(davinci_emac.c:  dm6467-emac, am3517-emac, dm816-emac).

Davinci MDIO RPM autosuspend can be enabled through sysfs:
 echo 100 > /sys/devices/../48484000.ethernet/48485000.mdio/power/autosuspend_delay_ms

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: davinci_mdio: add pm runtime callbacks
Grygorii Strashko [Fri, 24 Jun 2016 18:23:50 +0000 (21:23 +0300)]
drivers: net: davinci_mdio: add pm runtime callbacks

Add PM runtime .runtime_suspend()/.runtime_resume() callbacks and
perform Davinci MDIO enabling/disabling from these callbacks. This
allows to reuse pm_runtime_force_suspend/resume() APIs during System
suspend and required for further implementation of PM runtime
autosuspend.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: davinci_mdio: split reset function on init_clk and enable
Grygorii Strashko [Fri, 24 Jun 2016 18:23:49 +0000 (21:23 +0300)]
drivers: net: davinci_mdio: split reset function on init_clk and enable

The Davinci MDIO MDIO_CONTROL.CLKDIV can be calculated only once
during probe, hence split __davinci_mdio_reset() on
davinci_mdio_init_clk() and davinci_mdio_enable(). Initialize and
save CLKDIV in .probe(). Then just use saved value.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: davinci_mdio: drop suspended and lock fields from mdio_data
Grygorii Strashko [Fri, 24 Jun 2016 18:23:48 +0000 (21:23 +0300)]
drivers: net: davinci_mdio: drop suspended and lock fields from mdio_data

It's not expected Davinci MDIO to be accessible after its suspend
callbacks have been called:
 - all consumers of Davinci MDIO will stop/disconnect phys at Device
suspend stage;
 - all phys are expected to be suspned already by PHY/MDIO core;
 - MDIO locking is done by MDIO Bus code.

Hence, it's safe to drop "suspended" and "lock" fields from mdio_data.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: davinci_mdio: remove pm runtime calls from suspend callbacks
Grygorii Strashko [Fri, 24 Jun 2016 18:23:47 +0000 (21:23 +0300)]
drivers: net: davinci_mdio: remove pm runtime calls from suspend callbacks

PM runtime is disabled when Davinci MDIO .suspend_late() and
.resume_early() callbacks are called. As result, any PM runtime calls here will
be just a nop and can be removed.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: davinci_mdio: do pm runtime initialization later in probe
Grygorii Strashko [Fri, 24 Jun 2016 18:23:46 +0000 (21:23 +0300)]
drivers: net: davinci_mdio: do pm runtime initialization later in probe

Do PM runtime initialization later in probe - this allows to simplify
error handling a bit.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: cpsw: ndev: fix accessing to suspended device
Grygorii Strashko [Fri, 24 Jun 2016 18:23:45 +0000 (21:23 +0300)]
drivers: net: cpsw: ndev: fix accessing to suspended device

The CPSW might be suspended by RPM if all ethX interfaces are down,
but it still could be accesible through net_device_ops interfce. In
this case net_device_ops operations requiring registers access will
cause L3 errors and CPSW crash.

Hence, fix it by adding RPM get/put calls in net_device_ops callbacks
which need to access CPSW registers: .ndo_set_mac_address(),
.ndo_vlan_rx_add_vid(), .ndo_vlan_rx_kill_vid().

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: cpsw: ethtool: fix accessing to suspended device
Grygorii Strashko [Fri, 24 Jun 2016 18:23:44 +0000 (21:23 +0300)]
drivers: net: cpsw: ethtool: fix accessing to suspended device

The CPSW might be suspended by RPM if all ethX interfaces are down,
but it still could be accesible through ethtool interfce. In this case
ethtool operations, requiring registers access, will cause L3 errors and
CPSW crash.

ethtool callbcaks which need to access CPSW registers now:
.set_coalesce(), .get_ethtool_stats(), .set_pauseparam(), .get_regs()

Hence, fix it by adding .begin()/.complete() ethtool callbacks, which
will be called before/after each ethtool operation runs, and do CPSW
RPM handling in these callbacks. That way CPSW will be active while
handling ethtool requests.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: cpsw: remove pm runtime calls from suspend callbacks
Grygorii Strashko [Fri, 24 Jun 2016 18:23:43 +0000 (21:23 +0300)]
drivers: net: cpsw: remove pm runtime calls from suspend callbacks

PM runtime is properly handled in cpsw_ndo_open/stop(), as result it
isn't required to duplicate these calls in .suspend()/.resume()
callbacks. Moreover, it might cause unnecessary RPM resume of CPSW
during System suspend in the case it's already suspended because
all ethX interfaces are down already, before System suspend started.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: cpsw: check return code from pm runtime calls
Grygorii Strashko [Fri, 24 Jun 2016 18:23:42 +0000 (21:23 +0300)]
drivers: net: cpsw: check return code from pm runtime calls

Add missed check of return code from PM runtime get() calls.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: cpsw: fix suspend when all ethX devices are down
Grygorii Strashko [Fri, 24 Jun 2016 18:23:41 +0000 (21:23 +0300)]
drivers: net: cpsw: fix suspend when all ethX devices are down

The cpsw_suspend() could trigger L3 error and CPSW will stop
functioning if System enters suspend when all ethX net-devices are
down - in this case CPSW could be already suspended by PM runtime, but
cpsw_suspend() will try to call soft_reset_slave() unconditionally
and access CPSW registers.

Hence, fix it by moving soft_reset_slave() from cpsw_suspend() to
cpsw_slave_stop(). This way slave ports will be reset when CPSW is
active and will be in proper state during Suspend.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'stmmac-next'
David S. Miller [Tue, 28 Jun 2016 12:54:28 +0000 (08:54 -0400)]
Merge branch 'stmmac-next'

Giuseppe Cavallaro says:

====================
stmmac: rework and enhance the PCS support

The 3.xx and 4.xx synopsys gmacs have a very similar
PCS embedded module and they share almost the same registers;
for example:
  AN_Control, AN_Status, AN_Advertisement, AN_Link_Partner_Ability,
  AN_Expansion, TBI_Extended_Status.

Just the RGMII/SMII Control/Status register differs.

So these patches aim to reorganize and enhance the PCS support;
to do that, some small inline functions have been provided and
also some rework to the PCS ISR part has been done.

In the end, the SGMII for MAC2MAC connection has been introduced.

All patches have been built on top of net-next git and, as for
the previous version, not fully tested.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: stmmac: add port selection programming
Giuseppe CAVALLARO [Fri, 24 Jun 2016 13:16:26 +0000 (15:16 +0200)]
drivers: net: stmmac: add port selection programming

In case of SGMII more, for example when a MAC2MAC connection
is needed, the port selection bits (inside the MAC configuration
registers) have to be programmed according to the link selected.
So the patch adds a new DT parameter to pass the port selection
and to programmed related PCS and CORE to use it.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: stmmac: rework core ISR to better manage PCS and PMT
Giuseppe CAVALLARO [Fri, 24 Jun 2016 13:16:25 +0000 (15:16 +0200)]
drivers: net: stmmac: rework core ISR to better manage PCS and PMT

By default, all gmac cores disable the PCS block and always
enable the PMT.

Note that this is done in a different way by 3.x and 4.x cores.

With this rework, PCS and PMT interrupt masks can be driven by
parameters now moved inside the mac_device_info structure
and the settings follow what the HW capability register reports.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: stmmac: reworking the PCS code.
Giuseppe CAVALLARO [Fri, 24 Jun 2016 13:16:24 +0000 (15:16 +0200)]
drivers: net: stmmac: reworking the PCS code.

The 3.xx and 4.xx synopsys gmacs have a very similar
PCS embedded module and they share almost the same registers:
for example:
  AN_Control, AN_Status, AN_Advertisement, AN_Link_Partner_Ability,
  AN_Expansion, TBI_Extended_Status.

Just the RGMII/SMII Control/Status register differs.

So This patch aims to reorganize and enhance the PCS support.
It removes the existent support from the dwmac1000/dwmac4_core.c
moving basic PCS functions inside a new file called: stmmac_pcs.h.

The patch also reviews the available APIs to be better shared among
different hardware and easily enhanced to support new features.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoof_mdio: select fixed phy support unconditionally
Arnd Bergmann [Fri, 24 Jun 2016 09:24:08 +0000 (11:24 +0200)]
of_mdio: select fixed phy support unconditionally

Calling the fixed-phy functions when CONFIG_FIXED_PHY=m as a previous
change tried cannot work if the caller is in built-in code:

drivers/of/built-in.o: In function `of_phy_register_fixed_link':
of_reserved_mem.c:(.text+0x85e0): undefined reference to `fixed_phy_register'

Making of_mdio depend on 'FIXED_PHY || !FIXED_PHY' would solve this
dependency by enforcing that OF_MDIO itself becomes a loadable module
when FIXED_PHY=y, but that creates a different dependency as it
breaks any built-in ethernet driver that uses of_mdio.

Making FIXED_PHY a bool option also cannot work, since it depends on
PHYLIB, which again is tristate.

This version now uses 'select FIXED_PHY' to ensure that the fixed-phy
portion of of_mdio is not optional. The main downside of this is
a small increase in code size for cases that do not need fixed phy
support, but it should avoid all of the link-time problems.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: d1bd330a229f ("of_mdio: Enable fixed PHY support if driver is a module")
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocaif: Remove unneeded header file
Amitoj Kaur Chawla [Fri, 24 Jun 2016 06:23:54 +0000 (11:53 +0530)]
caif: Remove unneeded header file

Drop redundant include of moduleparam.h

The Coccinelle semantic patch used to make this change is as follows:
@ includesmodule @
@@

#include <linux/module.h>

@ depends on includesmodule @
@@

- #include <linux/moduleparam.h>

Signed-off-by: Amitoj Kaur Chawla <amitoj1606@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: diag: Add support to filter on device index
David Ahern [Fri, 24 Jun 2016 01:42:51 +0000 (18:42 -0700)]
net: diag: Add support to filter on device index

Add support to inet_diag facility to filter sockets based on device
index. If an interface index is in the filter only sockets bound
to that index (sk_bound_dev_if) are returned.

Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: dnet: use phy_ethtool_{get|set}_link_ksettings
Philippe Reynes [Thu, 23 Jun 2016 21:48:59 +0000 (23:48 +0200)]
net: ethernet: dnet: use phy_ethtool_{get|set}_link_ksettings

There are two generics functions phy_ethtool_{get|set}_link_ksettings,
so we can use them instead of defining the same code in the driver.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: dnet: use phydev from struct net_device
Philippe Reynes [Thu, 23 Jun 2016 21:48:58 +0000 (23:48 +0200)]
net: ethernet: dnet: use phydev from struct net_device

The private structure contain a pointer to phydev, but the structure
net_device already contain such pointer. So we can remove the pointer
phydev in the private structure, and update the driver to use the
one contained in struct net_device.

Signed-off-by: Philippe Reynes <tremyfr@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoi40e: Don't notify client(s) for DCB changes on all VSIs
Neerav Parikh [Tue, 7 Jun 2016 16:14:55 +0000 (09:14 -0700)]
i40e: Don't notify client(s) for DCB changes on all VSIs

When LLDP/DCBX change happens the i40e driver code flow tried to
notify the client(s) for each of the PF VSIs. This resulted into
kernel panic on the first VSI that didn't have any netdev
associated to it.

The DCB change notification to the client(s) should be done only
once for the PF/LAN VSI where the client(s) instances have been
added to. Also, move the notification call after the PF driver has
made changes related to the updated DCB configuration.

Signed-off-by: Neerav Parikh <neerav.parikh@intel.com>
Signed-off-by: Usha Ketineni <usha.k.ketineni@intel.com>
Tested-by: Ronald J Bynoe <ronald.j.bynoe@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: Fix errors resulted while turning off TSO
Tushar Dave [Mon, 16 May 2016 19:40:53 +0000 (12:40 -0700)]
i40e: Fix errors resulted while turning off TSO

On systems with 128 CPUs, turning off TSO results in errors,

i40e 0000:03:00.0: failed to get tracking for 1 vectors for VSI 400, err=-12
i40e 0000:03:00.0: Couldn't create FDir VSI
i40e 0000:03:00.0: i40e_ptp_init: PTP not supported on eth0
i40e 0000:03:00.0: couldn't add VEB, err I40E_ERR_ADMIN_QUEUE_ERROR aq_err I40E_AQ_RC_ENOENT
i40e 0000:03:00.0: rebuild of switch failed: -1, will try to set up simple PF connection
i40e 0000:03:00.0 eth0: adding 00:10:e0:8a:24:b6 vid=0

Enabling FD_SB without checking availability of MSI-X vector is the
root cause. This change adds necessary check.

Signed-off-by: Tushar Dave <tushar.n.dave@oracle.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 from 1.5.16 to 1.6.4
Bimmy Pujari [Mon, 16 May 2016 17:26:45 +0000 (10:26 -0700)]
i40e/i40evf: Bump version from 1.5.16 to 1.6.4

Signed-off-by: Bimmy Pujari <bimmy.pujari@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: add VSI info to macaddr messages
Shannon Nelson [Mon, 16 May 2016 17:26:44 +0000 (10:26 -0700)]
i40e: add VSI info to macaddr messages

Since the macaddr add and delete happens asynchronously, error
messages don't easily get associated to the actual request. Here
we add a bit of information to the error messages to help
determine the source of the error.

Change-ID: Id2d6df5287141c3579677d72d8bd21122823d79f
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 default VSI without a reset
Mitch Williams [Mon, 16 May 2016 17:26:43 +0000 (10:26 -0700)]
i40e: set default VSI without a reset

Remove the need for a reset when the device enters limited promiscuous
mode. This was causing heartburn for people who were using VFs and
bridging, since this would require all of the VFs to undergo a reset
each time the PF changed its promiscuity.

Change-ID: I0a83495c5e4d68112bbc7a7a076d20fa8dd3b61c
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 agoi40evf: always activate correct MAC address filter
Mitch Williams [Mon, 16 May 2016 17:26:42 +0000 (10:26 -0700)]
i40evf: always activate correct MAC address filter

Always add MAC address at the tail of the MAC filter list. Since the
device's "real" MAC address is added first, it will always be at the
beginning of the list. This prevents an issue where the "real" MAC
filter might not get added if too many other filters are added before
bringing the interface up.

Change-ID: I34a8aeebeb0cb87a44b24118adc4176c7b943c1c
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 RSS to not be limited by the number of CPUs
Catherine Sullivan [Mon, 16 May 2016 17:26:41 +0000 (10:26 -0700)]
i40e: Fix RSS to not be limited by the number of CPUs

Limiting qcount to pf->num_lan_msix, effectively limits the RSS queues
to only use the number of CPUs, and ignore all other queues. We don't
want to do this. If the user has changed the RSS settings to use more
queues then CPUS, we want to trust they know what they are doing and
let them. More importantly, if we tell them that is what we did, we want
to actually do it and allow traffic into all of the queues we have
allocated. This does not change the default setting to initially
allocate only the number of CPUS of queue pairs.

Change-ID: Ie941a96e806e4bcd016addb4e17affb46770ada5
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: Removing unnecessary code which caused supported link mode bug
Avinash Dayanand [Mon, 16 May 2016 17:26:40 +0000 (10:26 -0700)]
i40e: Removing unnecessary code which caused supported link mode bug

Removing this code which wasn't allowing 100BaseT to show up in the supported
link modes for 10GBaseT PHYs.

Change-ID: Iada2eafa7ef6b4bac9a2a1380ff533ae5de51e1d
Signed-off-by: Avinash Dayanand <avinash.dayanand@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: fix missing DA cable check
Serey Kong [Mon, 16 May 2016 17:26:39 +0000 (10:26 -0700)]
i40e: fix missing DA cable check

When a Direct Attach (DA) cable is used, if the i40e_set_settings
function is called it would return an error. Add the DA type so
the function won't fail.

Change-ID: I2b802f27a5d91cfefa72fd1f852acb4d74647a8e
Signed-off-by: Serey Kong <serey.kong@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: Save PCI state before suspend
Greg Rose [Mon, 16 May 2016 17:26:38 +0000 (10:26 -0700)]
i40e: Save PCI state before suspend

The i40e_suspend() function was failing to save PCI state
and this would result in a kernel stack trace from a WARN_ONCE in the
pci_legacy_suspend() function.

Add a call to pci_save_state() to fix that problem.

Change-ID: I4736e62bb660966bd208cc8af617a14cb07fc4bd
Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40e: Clean up MSIX IRQs before suspend
Greg Rose [Mon, 16 May 2016 17:26:37 +0000 (10:26 -0700)]
i40e: Clean up MSIX IRQs before suspend

The i40e_suspend() function calls another function that preps the device
for the power save and resume by freeing all the Tx/Rx resources and
interrupts but that function does not free the "other" causes interrupt
vector and IRQ. It also fails to call synchronize_irq() before freeing
the IRQ vectors.  This sometimes may result in some AER errors on those
systems with that PCIe error reporting feature enabled.

Call synchronize_irq() before freeing IRQ vectors and explicitly free
the other causes interrupt resources and shut down that MSIX interrupt.

Change-ID: Ib88e4536756518a352446da0232189716618ad81
Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
8 years agoi40evf: don't overflow buffer
Mitch Williams [Mon, 16 May 2016 17:26:36 +0000 (10:26 -0700)]
i40evf: don't overflow buffer

If the user adds an obscene amount of MAC addresses, the driver will run
into the situation where it has too many address requests to fit into a
single PF message. The driver checks for this case, and calculates the
maximum number of messages that it can send. Then it completely ignores
this count and overflows the buffer.

Fix this by checking the address count and bailing out of the loop at
the appropriate time.

Change-ID: If8dcbb04602c75941dc0cd8309065e1de9ca791c
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 call to set the client interface down
Catherine Sullivan [Mon, 16 May 2016 17:26:34 +0000 (10:26 -0700)]
i40e: Add a call to set the client interface down

We were failing to set the client interface down when we put the VSI
down. Add this call so that the client doesn't get an open called with
no close.

Also remove an un-needed delay. The VF should not be affected at all by
i40e_down.

Change-ID: I1135dffef534bf84e6fed57cf51bcf590e6cfaf7
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: write HENA for VFs
Mitch Williams [Mon, 16 May 2016 17:26:33 +0000 (10:26 -0700)]
i40e: write HENA for VFs

Now that VF RSS is configured by the PF driver, it needs to set the RSS
Hash Enable registers by default. Without this, no packets will be
hashed and they'll all end up on queue 0.

Change-ID: I38e425f40ddb81e3b19a951cfbb939fa5b1123f1
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 hw struct local variable
Mitch Williams [Mon, 16 May 2016 17:26:32 +0000 (10:26 -0700)]
i40e: add hw struct local variable

This function uses the i40e_hw struct all over the place, so why doesn't
it keep a pointer to the struct? Add this pointer as a local variable
and use it consistently throughout the function.

Change-ID: I10eb688fe40909433fcb8ac7ac891cef67445d72
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 functions to control default VSI
Mitch Williams [Mon, 16 May 2016 17:26:31 +0000 (10:26 -0700)]
i40e: add functions to control default VSI

Add functions to enable and disable default VSI on a VEB. This allows
for configuration of limited promiscuous mode specifically for bridging
purposes.

Change-ID: I0cc5bd68b31c500fdff4d47e1f15d50d2739faf4
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 agoMerge branch 'sfp-infra'
David S. Miller [Mon, 27 Jun 2016 14:41:02 +0000 (10:41 -0400)]
Merge branch 'sfp-infra'

Russell King says:

====================
Initial SFP support patches

Please review and merge this initial patch set, which is part of a
larger set previously posted adding SFP support to phy and mvneta.

This initial set are focused on cleaning up and reorganising the
fixed-phy code to allow the core software-phy code to be re-used.

These are based on net-next.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agophy: improve safety of fixed-phy MII register reading
Russell King [Thu, 23 Jun 2016 13:50:25 +0000 (14:50 +0100)]
phy: improve safety of fixed-phy MII register reading

There is no prevention of a concurrent call to both fixed_mdio_read()
and fixed_phy_update_state(), which can result in the state being
modified while it's being inspected.  Fix this by using a seqcount
to detect modifications, and memcpy()ing the state.

We remain slightly naughty here, calling link_update() and updating
the link status within the read-side loop - which would need rework
of the design to change.

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agophy: generate swphy registers on the fly
Russell King [Thu, 23 Jun 2016 13:50:20 +0000 (14:50 +0100)]
phy: generate swphy registers on the fly

Generate software phy registers as and when requested, rather than
duplicating the state in fixed_phy.  This allows us to eliminate
the duplicate storage of of the same data, which is only different
in format.

As fixed_phy_update_regs() no longer updates register state, rename
it to fixed_phy_update().

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agophy: separate swphy state validation from register generation
Russell King [Thu, 23 Jun 2016 13:50:15 +0000 (14:50 +0100)]
phy: separate swphy state validation from register generation

Separate out the generation of MII registers from the state validation.
This allows us to simplify the error handing in fixed_phy() by allowing
earlier error detection.

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agophy: convert swphy register generation to tabular form
Russell King [Thu, 23 Jun 2016 13:50:10 +0000 (14:50 +0100)]
phy: convert swphy register generation to tabular form

Convert the swphy register generation to tabular form which allows us
to eliminate multiple switch() statements.  This results in a smaller
object code size, more efficient, and easier to add support for faster
speeds.

Before:

Idx Name          Size      VMA       LMA       File off  Algn
  0 .text         00000164  00000000  00000000  00000034  2**2

   text    data     bss     dec     hex filename
    388       0       0     388     184 swphy.o

After:

Idx Name          Size      VMA       LMA       File off  Algn
  0 .text         000000fc  00000000  00000000  00000034  2**2
  5 .rodata       00000028  00000000  00000000  00000138  2**2

   text    data     bss     dec     hex filename
    324       0       0     324     144 swphy.o

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agophy: move fixed_phy MII register generation to a library
Russell King [Thu, 23 Jun 2016 13:50:05 +0000 (14:50 +0100)]
phy: move fixed_phy MII register generation to a library

Move the fixed_phy MII register generation to a library to allow other
software phy implementations to use this code.

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'linux-can-next-for-4.8-20160623' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Mon, 27 Jun 2016 14:33:42 +0000 (10:33 -0400)]
Merge tag 'linux-can-next-for-4.8-20160623' of git://git./linux/kernel/git/mkl/linux-can-next

Marc Kleine-Budde says:

====================
pull-request: can-next 2016-06-17

this is a pull request of 4 patches for net-next/master.

Arnd Bergmann's patch fixes a regresseion in af_can introduced in
linux-can-next-for-4.8-20160617. There are two patches by Ramesh
Shanmugasundaram, which add CAN-2.0 support to the rcar_canfd driver.
And a patch by Ed Spiridonov that adds better error diagnoses messages
to the Ed Spiridonov driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agotipc: Use kmemdup instead of kmalloc and memcpy
Amitoj Kaur Chawla [Thu, 23 Jun 2016 04:49:37 +0000 (10:19 +0530)]
tipc: Use kmemdup instead of kmalloc and memcpy

Replace calls to kmalloc followed by a memcpy with a direct call to
kmemdup.

The Coccinelle semantic patch used to make this change is as follows:
@@
expression from,to,size,flag;
statement S;
@@

-  to = \(kmalloc\|kzalloc\)(size,flag);
+  to = kmemdup(from,size,flag);
   if (to==NULL || ...) S
-  memcpy(to, from, size);

Signed-off-by: Amitoj Kaur Chawla <amitoj1606@gmail.com>
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoethernet: xircom: fix spelling mistakes on "excessive collisions"
Colin Ian King [Wed, 22 Jun 2016 17:35:05 +0000 (18:35 +0100)]
ethernet: xircom: fix spelling mistakes on "excessive collisions"

trivial fixes to spelling mistakes of the words "excessive collisions"

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: tc35815: fix spelling mistake on "descriptors"
Colin Ian King [Wed, 22 Jun 2016 16:42:21 +0000 (17:42 +0100)]
net: tc35815: fix spelling mistake on "descriptors"

trivial fixes to spelling mistakes of the word "descriptors"

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'mlx5e-100G-extensions'
David S. Miller [Mon, 27 Jun 2016 08:10:47 +0000 (04:10 -0400)]
Merge branch 'mlx5e-100G-extensions'

Saeed Mahameed says:

====================
Mellanox 100G mlx5e Ethernet extensions

This series includes multiple features extensions for mlx5 Ethernet netdevice driver.
Namely, TX Rate limiting, RX interrupt moderation, ethtool settings.

TX Rate limiting:
- ConnectX-4 rate limiting infrastructure
- Set max rate NDO support

RX interrupt moderation:
- CQE based coalescing option (controlled via priv flags)
- Adaptive RX coalescing

ethtool settings:
- priv flags callbacks
- Support new ksettings API
- Add 50G missing link mode
- Support auto negotiation on/off

Applied on top: 0e9390ebf1fe ("Merge branch 'mlxsw-next'")
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Report correct auto negotiation and allow toggling
Gal Pressman [Thu, 23 Jun 2016 14:02:46 +0000 (17:02 +0300)]
net/mlx5e: Report correct auto negotiation and allow toggling

Previous to this patch auto negotiation was reported off although it was
on by default in hardware. This patch reports the correct information to
ethtool and allows the user to toggle it on/off.

Added another parameter to set port proto function in order to pass
the auto negotiation field to the hardware.

Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Use new ethtool get/set link ksettings API
Gal Pressman [Thu, 23 Jun 2016 14:02:45 +0000 (17:02 +0300)]
net/mlx5e: Use new ethtool get/set link ksettings API

Use new get/set link ksettings and remove get/set settings legacy
callbacks.
This allows us to use bitmasks longer than 32 bit for supported and
advertised link modes and use modes that were previously not supported.

Signed-off-by: Gal Pressman <galp@mellanox.com>
CC: Ben Hutchings <bwh@kernel.org>
CC: David Decotigny <decot@googlers.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Add missing 50G baseSR2 link mode
Gal Pressman [Thu, 23 Jun 2016 14:02:44 +0000 (17:02 +0300)]
net/mlx5e: Add missing 50G baseSR2 link mode

Add MLX5E_50GBASE_SR2 as ETHTOOL_LINK_MODE_50000baseSR2_Full_BIT.

Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Cc: Ben Hutchings <bwh@kernel.org>
Cc: David Decotigny <decot@googlers.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoethtool: Add 50G baseSR2 link mode
Gal Pressman [Thu, 23 Jun 2016 14:02:43 +0000 (17:02 +0300)]
ethtool: Add 50G baseSR2 link mode

Add ETHTOOL_LINK_MODE_50000baseSR2_Full_BIT bit.

Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Cc: Ben Hutchings <bwh@kernel.org>
Cc: David Decotigny <decot@googlers.com>
Acked-By: David Decotigny <decot@googlers.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Toggle link only after modifying port parameters
Gal Pressman [Thu, 23 Jun 2016 14:02:42 +0000 (17:02 +0300)]
net/mlx5e: Toggle link only after modifying port parameters

Add a dedicated function to toggle port link. It should be called only
after setting a port register.
Toggle will set port link to down and bring it back up in case that it's
admin status was up.

Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Support adaptive RX coalescing
Gil Rockah [Thu, 23 Jun 2016 14:02:41 +0000 (17:02 +0300)]
net/mlx5e: Support adaptive RX coalescing

Striving for high message rate and low interrupt rate.

Usage:
        ethtool -C <interface> adaptive-rx on/off

Signed-off-by: Gil Rockah <gilr@mellanox.com>
Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
CC: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: CQE based moderation
Tariq Toukan [Thu, 23 Jun 2016 14:02:40 +0000 (17:02 +0300)]
net/mlx5e: CQE based moderation

In this mode the moderation timer will restart upon
new completion (CQE) generation rather than upon interrupt
generation.

The outcome is that for bursty traffic the period timer will never
expire and thus only the moderation frames counter will dictate
interrupt generation, thus the interrupt rate will be relative
to the incoming packets size.
If the burst seizes for "moderation period" time then an interrupt
will be issued immediately.

CQE based moderation is off by default and can be controlled
via ethtool set_priv_flags.

Performance tested on ConnectX4-Lx 50G.

Less packet loss in netperf UDP and TCP tests, with no bw degradation,
for both single and multi streams, with message sizes of
64, 1024, 1472 and 32768 byte.

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Gil Rockah <gilr@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Introduce net device priv flags infrastructure
Gal Pressman [Thu, 23 Jun 2016 14:02:39 +0000 (17:02 +0300)]
net/mlx5e: Introduce net device priv flags infrastructure

Introduce an infrastructure for getting/setting private net device
flags.

Currently a 'nop' priv flag is added, following patches will override
the flag will actual feature specific flags.

Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Add TXQ set max rate support
Yevgeny Petrilin [Thu, 23 Jun 2016 14:02:38 +0000 (17:02 +0300)]
net/mlx5e: Add TXQ set max rate support

Implement set_maxrate ndo.
Use the rate index from the hardware table to attach to channel SQ/TXQ.
In case of failure to configure new rate, the queue remains with
unlimited rate.

We save the configuration on priv structure and apply it each time
Send Queues are being reinitialized (after open/close) operations.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5: Rate limit tables support
Yevgeny Petrilin [Thu, 23 Jun 2016 14:02:37 +0000 (17:02 +0300)]
net/mlx5: Rate limit tables support

Configuring and managing HW rate limit tables.
The HW holds a table of rate limits, each rate is
associated with an index in that table.
Later a Send Queue uses this index to set the rate limit.
Multiple Send Queues can have the same rate limit, which is
represented by a single entry in this table.
Even though a rate can be shared, each queue is being rate
limited independently of others.

The SW shadow of this table holds the rate itself,
the index in the HW table and the refcount (number of queues)
working with this rate.

The exported functions are mlx5_rl_add_rate and mlx5_rl_remove_rate.
Number of different rates and their values are derived
from HW capabilities.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>