platform/kernel/linux-rpi.git
4 years agor8169: remove mask argument from r8168ep_ocp_read
Heiner Kallweit [Sat, 23 May 2020 11:23:21 +0000 (13:23 +0200)]
r8169: remove mask argument from r8168ep_ocp_read

Remove the mask argument as it's not used by r8168ep_ocp_read().

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agor8169: remove mask argument from r8168dp_ocp_read
Heiner Kallweit [Sat, 23 May 2020 11:22:22 +0000 (13:22 +0200)]
r8169: remove mask argument from r8168dp_ocp_read

All callers read the full 32bit value, therefore the mask argument can
be removed.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agor8169: remove mask argument from rtl_w0w1_eri
Heiner Kallweit [Sat, 23 May 2020 11:21:36 +0000 (13:21 +0200)]
r8169: remove mask argument from rtl_w0w1_eri

rtl_eri_read() returns the full 32bit value, therefore there's no
benefit in writing back parts of it only. handle it like the vendor
driver and write the full 32 bit always. Omitting the mask argument
avoids some overhead and makes the code better readable.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Sat, 23 May 2020 23:51:26 +0000 (16:51 -0700)]
Merge branch '100GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
100GbE Intel Wired LAN Driver Updates 2020-05-22

This series contains updates to virtchnl and the ice driver.

Geert Uytterhoeven fixes a data structure alignment issue in the
virtchnl structures.

Henry adds Flow Director support which allows for the redirection on
ntuple rules over six patches.  Initially Henry adds the initial
infrastructure for Flow Director, and then later adds IPv4 and IPv6
support, as well as being able to display the ntuple rules.

Bret add Accelerated Receive Flow Steering (aRFS) support which is used
to steer receive flows to a specific queue.  Fixes a transmit timeout
when the VF link transitions from up/down/up because the transmit and
receive queue interrupts are not enabled as part of VF's link up.  Fixed
an issue when the default VF LAN address is changed and after reset the
PF will attempt to add the new MAC, which fails because it already
exists. This causes the VF to be disabled completely until it is removed
and enabled via sysfs.

Anirudh (Ani) makes a fix where the ice driver needs to call set_mac_cfg
to enable jumbo frames, so ensure it gets called during initialization
and after reset.  Fix bad register reads during a register dump in
ethtool by removing the bad registers.

Paul fixes an issue where the receive Malicious Driver Detection (MDD)
auto reset message was not being logged because it occurred after the VF
reset.

Victor adds a check for compatibility between the Dynamic Device
Personalization (DDP) package and the NIC firmware to ensure that
everything aligns.

Jesse fixes a administrative queue string call with the appropriate
error reporting variable.  Also fixed the loop variables that are
comparing or assigning signed against unsigned values.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Sat, 23 May 2020 23:47:41 +0000 (16:47 -0700)]
Merge branch '1GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
1GbE Intel Wired LAN Driver Updates 2020-05-22

This series contains updates to e1000e, igc and igb.

Many of the patches in this series are fixes, but many of the igc fixes
are based on the recent filter rule handling Andre has been working,
which will not backport to earlier/stable kernels.  The remaining fixes
for e1000e and igb have CC'd stable where applicable.

Andre continue with his refactoring of the filter rule code to help with
reducing the complexity, in multiple patches.  Fix the inconsistent size
of a struct field.  Fixed an issue where filter rules stay active in the
hardware, even after it was deleted, so make sure to disable the filter
rule before deleting.  Fixed an issue with NFC rules which were dropping
valid multicast MAC address.  Fixed how the NFC rules are restored after
the NIC is reset or brought up, so that they are restored in the same order
they were initially setup in.  Fix a potential memory leak when the
driver is unloaded and the NFC rules are not flushed from memory
properly.  Fixed how NFC rule validation handles when a request to
overwrite an existing rule.  Changed the locking around the NFC rule API
calls from spin_locks to mutex locks to avoid unnecessary busy waiting
on lock contention.

Sasha clean up more unused code in the igc driver.

Kai-Heng Feng from Canonical provides three fixes, first has igb report
the speed and duplex as unknown when in runtime suspend.  Fixed e1000e
to pass up the error when disabling ULP mode.  Fixed e1000e performance
by disabling TSO by default for certain MACs.

Vitaly disables S0ix entry and exit flows for ME systems.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge tag 'mlx5-updates-2020-05-22' of git://git.kernel.org/pub/scm/linux/kernel...
David S. Miller [Sat, 23 May 2020 23:37:00 +0000 (16:37 -0700)]
Merge tag 'mlx5-updates-2020-05-22' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-updates-2020-05-22

This series includes two updates and one cleanup patch

1) Tang Bim, clean-up with IS_ERR() usage

2) Vlad introduces a new mlx5 kconfig flag for TC support

   This is required due to the high volume of current and upcoming
   development in the eswitch and representors areas where some of the
   feature are TC based such as the downstream patches of MPLSoUDP and
   the following representor bonding support for VF live migration and
   uplink representor dynamic loading.
   For this Vlad kept TC specific code in tc.c and rep/tc.c and
   organized non TC code in representors specific files.

3) Eli Cohen adds support for MPLS over UPD encap and decap TC offloads.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: psample: fix build error when CONFIG_INET is not enabled
Randy Dunlap [Fri, 22 May 2020 20:05:26 +0000 (13:05 -0700)]
net: psample: fix build error when CONFIG_INET is not enabled

Fix psample build error when CONFIG_INET is not set/enabled by
bracketing the tunnel code in #ifdef CONFIG_NET / #endif.

../net/psample/psample.c: In function ‘__psample_ip_tun_to_nlattr’:
../net/psample/psample.c:216:25: error: implicit declaration of function ‘ip_tunnel_info_opts’; did you mean ‘ip_tunnel_info_opts_set’? [-Werror=implicit-function-declaration]

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Yotam Gigi <yotam.gi@gmail.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: at803x: fix PHY ID masks
Michael Walle [Fri, 22 May 2020 09:53:31 +0000 (11:53 +0200)]
net: phy: at803x: fix PHY ID masks

Ever since its first commit 0ca7111a38f05 ("phy: add AT803x driver") the
PHY ID mask was set to 0xffffffef. It is unclear to me why this mask was
chosen in the first place. Both the AR8031/AR8033 and the AR8035
datasheets mention it is always the given value:
 - for AR8031/AR8033 its 0x004d/0xd074
 - for AR8035 its 0x004d/0xd072

Unfortunately, I don't have a datasheet for the AR8030. Therefore, we
leave its PHY ID mask untouched. For the PHYs mentioned before use the
handy PHY_ID_MATCH_EXACT() macro.

I've tried to contact the author of the initial commit, but received no
answer so far.

Cc: Matus Ujhelyi <ujhelyi.m@gmail.com>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoice: cleanup unsigned loops
Jesse Brandeburg [Sat, 16 May 2020 00:36:37 +0000 (17:36 -0700)]
ice: cleanup unsigned loops

Fix loop variables that are comparing or assigning signed against
unsigned values, mostly by declaring loop counters as unsigned.

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>
4 years agoice: fix usage of incorrect variable
Jesse Brandeburg [Sat, 16 May 2020 00:36:36 +0000 (17:36 -0700)]
ice: fix usage of incorrect variable

The driver was using rq_last_status where it should have been
using sq_last_status. Fix the string to be using the correct
error reporting variable.

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>
4 years agoice: Fix bad register reads
Anirudh Venkataramanan [Sat, 16 May 2020 00:36:35 +0000 (17:36 -0700)]
ice: Fix bad register reads

The "ethtool -d" handler reads registers in the ice_regs_dump_list array
and returns read values back to the userspace.

The register offsets PFINT0_ITR* are not valid as per the specification
and reading these causes a "unable to handle kernel paging request" bug
in the driver. Remove these registers from ice_regs_dump_list.

Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: check for compatibility between DDP package and firmware
Victor Raj [Sat, 16 May 2020 00:36:34 +0000 (17:36 -0700)]
ice: check for compatibility between DDP package and firmware

Require the Dynamic Device Personalization (DDP) file to have the same
major version number and the same or older minor number than the firmware
version major and minor, respectively.

Check the OS and NVM package versions before downloading the package.
If the OS package version is not compatible with NVM then return an
appropriate error.

Split the 32-byte segment name into a 28-byte segment name and
a 4-byte Track-ID. Older packages will still work with this change
because no package has a name that will take up more than 28 bytes;
in this case the Track-ID will be 0.

Note that the driver will store the segment name as 32-bytes in the
ice_hw structure, in order to normalize the length of the various
package name strings that it uses.

Also add section ID and structure for the segment metadata section.

Signed-off-by: Victor Raj <victor.raj@intel.com>
Signed-off-by: Dan Nowlin <dan.nowlin@intel.com>
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: Check if unicast MAC exists before setting VF MAC
Brett Creeley [Sat, 16 May 2020 00:36:33 +0000 (17:36 -0700)]
ice: Check if unicast MAC exists before setting VF MAC

Currently if a unicast MAC is set via ndo_set_vf_mac, the PF driver will
set the VF's dflt_lan_addr.addr once some basic checks have passed. The
VF is then reset. During reset the PF driver will attempt to program the
VF's MAC from the dflt_lan_addr.addr field. This fails when the MAC
already exists on the PF's switch.

This is causing the VF to be completely disabled until removing/enabling
any VFs via sysfs.

Fix this by checking if the unicast MAC exists before triggering a VF
reset directly in ndo_set_vf_mac. Also, add a check if the unicast MAC
is set to the same value as before and return 0 if that is the case.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: Fix Tx timeout when link is toggled on a VF's interface
Brett Creeley [Sat, 16 May 2020 00:36:32 +0000 (17:36 -0700)]
ice: Fix Tx timeout when link is toggled on a VF's interface

Currently if the iavf is loaded and a VF link transitions from up to
down to up again a Tx timeout will be triggered. This happens because
Tx/Rx queue interrupts are only enabled when receiving the
VIRTCHNL_OP_CONFIG_MAP_IRQ message, which happens on reset or initial
iavf driver load, but not when bringing link up. This is problematic
because they are disabled on the VIRTCHNL_OP_DISABLE_QUEUES message,
which is part of bringing a VF's link down. However, they are not
enabled on the VIRTCHNL_OP_ENABLE_QUEUES message, which is part of
bringing a VF's link up.

Fix this by re-enabling the VF's Rx and Tx queue interrupts when they
were previously configured. This is done by first checking to make
sure the previous value in QINT_[R|T]QCTL.MSIX_INDX is not 0, which
is used to represent the OICR in the VF's interrupt space. If the
MSIX_INDX is non-zero then enable the interrupt by setting the
QINT_[R|T]CTL.CAUSE_ENA bit to 1.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: print Rx MDD auto reset message before VF reset
Paul Greenwalt [Sat, 16 May 2020 00:36:31 +0000 (17:36 -0700)]
ice: print Rx MDD auto reset message before VF reset

Rx MDD auto reset message was not being logged because logging occurred
after the VF reset and the VF MDD data was reinitialized.

Log the Rx MDD auto reset message before triggering the VF reset.

Signed-off-by: Paul Greenwalt <paul.greenwalt@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: Call ice_aq_set_mac_cfg
Anirudh Venkataramanan [Sat, 16 May 2020 00:36:30 +0000 (17:36 -0700)]
ice: Call ice_aq_set_mac_cfg

As per the specification, the driver needs to call set_mac_cfg
(opcode 0x0603) to be able to exercise jumbo frames. Call the
function during initialization and the post reset rebuild flow.

Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: Implement aRFS
Brett Creeley [Tue, 12 May 2020 01:01:46 +0000 (18:01 -0700)]
ice: Implement aRFS

Enable accelerated Receive Flow Steering (aRFS). It is used to steer Rx
flows to a specific queue. This functionality is triggered by the network
stack through ndo_rx_flow_steer and requires Flow Director (ntuple on) to
function.

The fltr_info is used to add/remove/update flow rules in the HW, the
fltr_state is used to determine what to do with the filter with respect
to HW and/or SW, and the flow_id is used in co-ordination with the
network stack.

The work for aRFS is split into two paths: the ndo_rx_flow_steer
operation and the ice_service_task. The former is where the kernel hands
us an Rx SKB among other items to setup aRFS and the latter is where
the driver adds/updates/removes filter rules from HW and updates filter
state.

In the Rx path the following things can happen:
        1. New aRFS entries are added to the hash table and the state is
           set to ICE_ARFS_INACTIVE so the filter can be updated in HW
           by the ice_service_task path.
        2. aRFS entries have their Rx Queue updated if we receive a
           pre-existing flow_id and the filter state is ICE_ARFS_ACTIVE.
           The state is set to ICE_ARFS_INACTIVE so the filter can be
           updated in HW by the ice_service_task path.
        3. aRFS entries marked as ICE_ARFS_TODEL are deleted

In the ice_service_task path the following things can happen:
        1. New aRFS entries marked as ICE_ARFS_INACTIVE are added or
           updated in HW.
           and their state is updated to ICE_ARFS_ACTIVE.
        2. aRFS entries are deleted from HW and their state is updated
           to ICE_ARFS_TODEL.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Signed-off-by: Madhu Chittim <madhu.chittim@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: Restore filters following reset
Henry Tieman [Tue, 12 May 2020 01:01:45 +0000 (18:01 -0700)]
ice: Restore filters following reset

Following a reset, Flow Director filters are cleared from the hardware.
Rebuild the filters using the software structures containing the filter
rules.

Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: Enable flex-bytes support
Henry Tieman [Tue, 12 May 2020 01:01:44 +0000 (18:01 -0700)]
ice: Enable flex-bytes support

Flex-bytes allows for packet matching based on an offset and value. This
is supported via the ethtool user-def option.  It is specified by providing
an offset followed by a 2 byte match value. Offset is measured from the
start of the MAC address.

The following restrictions apply to flex-bytes. The specified offset must
be an even number and be smaller than 0x1fe.

Example usage:

ethtool -N eth0 flow-type tcp4 src-ip 192.168.0.55 dst-ip 172.16.0.55 \
src-port 12 dst-port 13 user-def 0x10ffff action 32

Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: Support IPv6 Flow Director filters
Henry Tieman [Tue, 12 May 2020 01:01:43 +0000 (18:01 -0700)]
ice: Support IPv6 Flow Director filters

Extend supported filters to allow for IPv6 filters.

Supported fields are: src-ip, dst-ip, src-port, and dst-port
Supported flow-types are: tcp6, udp6, sctp6, ip6

Example usage:

ethtool -N eth0 flow-type tcp6 src-port 12 dst-port 13 \
src-ip fce0::1:34 dst-ip fce0::1:35 action 32

Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: Support IPv4 Flow Director filters
Henry Tieman [Tue, 12 May 2020 01:01:42 +0000 (18:01 -0700)]
ice: Support IPv4 Flow Director filters

Support the addition and deletion of IPv4 filters.

Supported fields are: src-ip, dst-ip, src-port, and dst-port
Supported flow-types are: tcp4, udp4, sctp4, ip4

Example usage:

ethtool -N eth0 flow-type tcp4 src-ip 192.168.0.55 dst-ip 172.16.0.55 \
src-port 16 dst-port 12 action 32

Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: Support displaying ntuple rules
Henry Tieman [Tue, 12 May 2020 01:01:41 +0000 (18:01 -0700)]
ice: Support displaying ntuple rules

Add functionality for ethtool --show-ntuple, allowing for filters to be
displayed when set functionality is added. Add statistics related to
Flow Director matches and status.

Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoice: Initialize Flow Director resources
Henry Tieman [Tue, 12 May 2020 01:01:40 +0000 (18:01 -0700)]
ice: Initialize Flow Director resources

Flow Director allows for redirection based on ntuple rules. Rules are
programmed using the ethtool set-ntuple interface. Supported actions are
redirect to queue and drop.

Setup the initial framework to process Flow Director filters. Create and
allocate resources to manage and program filters to the hardware. Filters
are processed via a sideband interface; a control VSI is created to manage
communication and process requests through the sideband. Upon allocation of
resources, update the hardware tables to accept perfect filters.

Signed-off-by: Henry Tieman <henry.w.tieman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agovirtchnl: Add missing explicit padding to structures
Geert Uytterhoeven [Thu, 16 Apr 2020 09:50:49 +0000 (11:50 +0200)]
virtchnl: Add missing explicit padding to structures

On e.g. m68k, the alignment of 32-bit values is only 2 bytes, leading
to the following:

    ./include/linux/avf/virtchnl.h:147:36: warning: division by zero [-Wdiv-by-zero]
      { virtchnl_static_assert_##X = (n)/((sizeof(struct X) == (n)) ? 1 : 0) }
^
    ./include/linux/avf/virtchnl.h:577:1: note: in expansion of macro ‘VIRTCHNL_CHECK_STRUCT_LEN’
     VIRTCHNL_CHECK_STRUCT_LEN(272, virtchnl_filter);
     ^~~~~~~~~~~~~~~~~~~~~~~~~
    ./include/linux/avf/virtchnl.h:577:32: error: enumerator value for ‘virtchnl_static_assert_virtchnl_filter’ is not an integer constant
     VIRTCHNL_CHECK_STRUCT_LEN(272, virtchnl_filter);
    ^~~~~~~~~~~~~~~
    ./include/linux/avf/virtchnl.h:147:53: note: in definition of macro ‘VIRTCHNL_CHECK_STRUCT_LEN’
      { virtchnl_static_assert_##X = (n)/((sizeof(struct X) == (n)) ? 1 : 0) }
 ^
    ./include/linux/avf/virtchnl.h:147:36: warning: division by zero [-Wdiv-by-zero]
      { virtchnl_static_assert_##X = (n)/((sizeof(struct X) == (n)) ? 1 : 0) }
^
    ./include/linux/avf/virtchnl.h:619:1: note: in expansion of macro ‘VIRTCHNL_CHECK_STRUCT_LEN’
     VIRTCHNL_CHECK_STRUCT_LEN(16, virtchnl_pf_event);
     ^~~~~~~~~~~~~~~~~~~~~~~~~
    ./include/linux/avf/virtchnl.h:619:31: error: enumerator value for ‘virtchnl_static_assert_virtchnl_pf_event’ is not an integer constant
     VIRTCHNL_CHECK_STRUCT_LEN(16, virtchnl_pf_event);
   ^~~~~~~~~~~~~~~~~
    ./include/linux/avf/virtchnl.h:147:53: note: in definition of macro ‘VIRTCHNL_CHECK_STRUCT_LEN’
      { virtchnl_static_assert_##X = (n)/((sizeof(struct X) == (n)) ? 1 : 0) }
 ^
    ./include/linux/avf/virtchnl.h:147:36: warning: division by zero [-Wdiv-by-zero]
      { virtchnl_static_assert_##X = (n)/((sizeof(struct X) == (n)) ? 1 : 0) }
^
    ./include/linux/avf/virtchnl.h:640:1: note: in expansion of macro ‘VIRTCHNL_CHECK_STRUCT_LEN’
     VIRTCHNL_CHECK_STRUCT_LEN(12, virtchnl_iwarp_qv_info);
     ^~~~~~~~~~~~~~~~~~~~~~~~~
    ./include/linux/avf/virtchnl.h:640:31: error: enumerator value for ‘virtchnl_static_assert_virtchnl_iwarp_qv_info’ is not an integer constant
     VIRTCHNL_CHECK_STRUCT_LEN(12, virtchnl_iwarp_qv_info);
   ^~~~~~~~~~~~~~~~~~~~~~
    ./include/linux/avf/virtchnl.h:147:53: note: in definition of macro ‘VIRTCHNL_CHECK_STRUCT_LEN’
      { virtchnl_static_assert_##X = (n)/((sizeof(struct X) == (n)) ? 1 : 0) }
 ^
    ./include/linux/avf/virtchnl.h:147:36: warning: division by zero [-Wdiv-by-zero]
      { virtchnl_static_assert_##X = (n)/((sizeof(struct X) == (n)) ? 1 : 0) }
^
    ./include/linux/avf/virtchnl.h:647:1: note: in expansion of macro ‘VIRTCHNL_CHECK_STRUCT_LEN’
     VIRTCHNL_CHECK_STRUCT_LEN(16, virtchnl_iwarp_qvlist_info);
     ^~~~~~~~~~~~~~~~~~~~~~~~~
    ./include/linux/avf/virtchnl.h:647:31: error: enumerator value for ‘virtchnl_static_assert_virtchnl_iwarp_qvlist_info’ is not an integer constant
     VIRTCHNL_CHECK_STRUCT_LEN(16, virtchnl_iwarp_qvlist_info);
   ^~~~~~~~~~~~~~~~~~~~~~~~~~
    ./include/linux/avf/virtchnl.h:147:53: note: in definition of macro ‘VIRTCHNL_CHECK_STRUCT_LEN’
      { virtchnl_static_assert_##X = (n)/((sizeof(struct X) == (n)) ? 1 : 0) }
 ^

Fix this by adding explicit padding to structures with holes.

Reported-by: <noreply@ellerman.id.au>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next
David S. Miller [Sat, 23 May 2020 01:30:34 +0000 (18:30 -0700)]
Merge git://git./linux/kernel/git/bpf/bpf-next

Daniel Borkmann says:

====================
pull-request: bpf-next 2020-05-23

The following pull-request contains BPF updates for your *net-next* tree.

We've added 50 non-merge commits during the last 8 day(s) which contain
a total of 109 files changed, 2776 insertions(+), 2887 deletions(-).

The main changes are:

1) Add a new AF_XDP buffer allocation API to the core in order to help
   lowering the bar for drivers adopting AF_XDP support. i40e, ice, ixgbe
   as well as mlx5 have been moved over to the new API and also gained a
   small improvement in performance, from Björn Töpel and Magnus Karlsson.

2) Add getpeername()/getsockname() attach types for BPF sock_addr programs
   in order to allow for e.g. reverse translation of load-balancer backend
   to service address/port tuple from a connected peer, from Daniel Borkmann.

3) Improve the BPF verifier is_branch_taken() logic to evaluate pointers
   being non-NULL, e.g. if after an initial test another non-NULL test on
   that pointer follows in a given path, then it can be pruned right away,
   from John Fastabend.

4) Larger rework of BPF sockmap selftests to make output easier to understand
   and to reduce overall runtime as well as adding new BPF kTLS selftests
   that run in combination with sockmap, also from John Fastabend.

5) Batch of misc updates to BPF selftests including fixing up test_align
   to match verifier output again and moving it under test_progs, allowing
   bpf_iter selftest to compile on machines with older vmlinux.h, and
   updating config options for lirc and v6 segment routing helpers, from
   Stanislav Fomichev, Andrii Nakryiko and Alan Maguire.

6) Conversion of BPF tracing samples outdated internal BPF loader to use
   libbpf API instead, from Daniel T. Lee.

7) Follow-up to BPF kernel test infrastructure in order to fix a flake in
   the XDP selftests, from Jesper Dangaard Brouer.

8) Minor improvements to libbpf's internal hashmap implementation, from
   Ian Rogers.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoe1000e: disable s0ix entry and exit flows for ME systems
Vitaly Lifshits [Thu, 7 May 2020 17:14:06 +0000 (20:14 +0300)]
e1000e: disable s0ix entry and exit flows for ME systems

Since ME systems do not support SLP_S0 in S0ix state, and S0ix entry
and exit flows may cause errors on them it is best to avoid using
e1000e_s0ix_entry_flow and e1000e_s0ix_exit_flow functions.

This was done by creating a struct of all devices that comes with ME
and by checking if the current device has ME.

Signed-off-by: Vitaly Lifshits <vitaly.lifshits@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoe1000e: Disable TSO for buffer overrun workaround
Kai-Heng Feng [Thu, 7 May 2020 14:21:07 +0000 (22:21 +0800)]
e1000e: Disable TSO for buffer overrun workaround

Commit b10effb92e27 ("e1000e: fix buffer overrun while the I219 is
processing DMA transactions") imposes roughly 30% performance penalty.

The commit log states that "Disabling TSO eliminates performance loss
for TCP traffic without a noticeable impact on CPU performance", so
let's disable TSO by default to regain the loss.

CC: stable <stable@vger.kernel.org>
Fixes: b10effb92e27 ("e1000e: fix buffer overrun while the I219 is processing DMA transactions")
BugLink: https://bugs.launchpad.net/bugs/1802691
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoe1000e: Warn if disabling ULP failed
Kai-Heng Feng [Thu, 7 May 2020 06:25:45 +0000 (14:25 +0800)]
e1000e: Warn if disabling ULP failed

The hardware may stop working if driver failed to disable ULP mode.

Take the return value of e1000_disable_ulp_lpt_lp() into account, and
pass up the error if it fails.

Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigb: Report speed and duplex as unknown when device is runtime suspended
Kai-Heng Feng [Tue, 5 May 2020 04:01:54 +0000 (12:01 +0800)]
igb: Report speed and duplex as unknown when device is runtime suspended

igb device gets runtime suspended when there's no link partner. We can't
get correct speed under that state:
$ cat /sys/class/net/enp3s0/speed
1000

In addition to that, an error can also be spotted in dmesg:
[  385.991957] igb 0000:03:00.0 enp3s0: PCIe link lost

Since device can only be runtime suspended when there's no link partner,
we can skip reading register and let the following logic set speed and
duplex with correct status.

The more generic approach will be wrap get_link_ksettings() with begin()
and complete() callbacks. However, for this particular issue, begin()
calls igb_runtime_resume() , which tries to rtnl_lock() while the lock
is already hold by upper ethtool layer.

So let's take this approach until the igb_runtime_resume() no longer
needs to hold rtnl_lock.

CC: stable <stable@vger.kernel.org>
Suggested-by: Alexander Duyck <alexander.duyck@gmail.com>
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Remove unused descriptor's flags
Sasha Neftin [Mon, 4 May 2020 06:29:25 +0000 (09:29 +0300)]
igc: Remove unused descriptor's flags

Enable Tidv register, Report Packet Sent, Report Status and
Ethernet CRC flags not in use.
This patch comes to clean up these flags.

Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Remove igc_nfc_rule_exit()
Andre Guedes [Fri, 24 Apr 2020 20:16:23 +0000 (13:16 -0700)]
igc: Remove igc_nfc_rule_exit()

During igc_down(), we call igc_nfc_rule_exit() which traverse the NFC
rule list disabling filters one by one. Later on in igc_down() flow
we issue an hardware reset which also clear all filters.  Since we
already reset the hardware, we don't actually need to disable each
filter manually. In order to simplify the code, this patch removes
igc_nfc_rule() altogether.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Change adapter->nfc_rule_lock to mutex
Andre Guedes [Fri, 24 Apr 2020 20:16:22 +0000 (13:16 -0700)]
igc: Change adapter->nfc_rule_lock to mutex

This patch changes adapter->nfc_rule_lock type from spin_lock to mutex
so we avoid unnecessary busy waiting on lock contention.

A closer look at the execution context of NFC rule API users shows that
all of them run in process context. The API users are: ethtool ops,
igc_configure(), called when interface is brought up by user or reset
workequeue thread, igc_down(), called when interface is brought down,
and igc_remove(), called when driver is unloaded.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Change return type from igc_disable_nfc_rule()
Andre Guedes [Fri, 24 Apr 2020 20:16:21 +0000 (13:16 -0700)]
igc: Change return type from igc_disable_nfc_rule()

None of igc_disable_nfc_rule() callers actually check its returning
value. A closer look at why this function would fail shows that the
only situation is when we try to delete an Ethertype or MAC filter that
doesn't exist.

That situation is very unlikely so we can change igc_del_etype_filter()
and igc_del_mac_filter() logic to "if the filter doesn't exist, we are
done", and keep the logic in igc_disable_nfc_rule() callers simple.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Fix NFC rule validation
Andre Guedes [Fri, 24 Apr 2020 20:16:20 +0000 (13:16 -0700)]
igc: Fix NFC rule validation

If we try to overwrite an existing rule with the same filter but
different action, we get EEXIST error as shown below.

$ ethtool -N eth0 flow-type ether dst <MACADDR> action 1 loc 10
$ ethtool -N eth0 flow-type ether dst <MACADDR> action 2 loc 10
rmgr: Cannot insert RX class rule: File exists

The second command is expected to overwrite the previous rule in location
10 and succeed.

This patch fixes igc_ethtool_check_nfc_rule() so it also checks the
rules location. In case they match, the rule under evaluation should not
be considered invalid.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Fix NFC rules leak when driver is unloaded
Andre Guedes [Fri, 24 Apr 2020 20:16:19 +0000 (13:16 -0700)]
igc: Fix NFC rules leak when driver is unloaded

If we have RFC rules in adapter->nfc_rule_list when the IGC driver
is unloaded, all rules are leaked. This patch fixes the issue by
introducing the helper igc_flush_nfc_rules() and calling it in
igc_remove(). It also updates igc_set_features() so is reuses the
new helper instead of re-implementing it.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Refactor igc_ethtool_update_nfc_rule()
Andre Guedes [Fri, 24 Apr 2020 20:16:18 +0000 (13:16 -0700)]
igc: Refactor igc_ethtool_update_nfc_rule()

Current implementation of igc_ethtool_update_nfc_rule() is a bit
convoluted since it handles too many things: rule lookup, deletion
and addition. This patch breaks it into three functions so we simplify
the code and improve code reuse.

Code related to rule lookup is refactored out to a new function called
igc_get_nfc_rule().

Code related to rule addition is refactored out to a new function called
igc_add_nfc_rule(). This function enables the rule in hardware and adds
it to the adapter's list.

Code related to rule deletion is refactored out to a new function called
igc_del_nfc_rule(). This function disables the rule in hardware, removes
it from adapter's list, and deletes it.

As a byproduct of this refactoring, igc_enable_nfc_rule() and
igc_disable_nfc_rule() are moved to igc_main.c since they are not used
in igc_ethtool.c anymore, and igc_restore_nfc_rules() and igc_nfc_rule_
exit() are moved around to avoid forward declaration.

Also, since this patch already touches igc_ethtool_get_nfc_rule(), it
takes the opportunity to remove the 'match_flags' check. Empty flags
are not allowed to be added so no need to check that.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Fix NFC rules restoration
Andre Guedes [Fri, 24 Apr 2020 20:16:17 +0000 (13:16 -0700)]
igc: Fix NFC rules restoration

When network interface is brought up, the driver re-enables the NFC
rules previously configured. However, this is done in reverse order
the rules were added and hardware filters are configured differently.

For example, consider the following rules:

$ ethtool -N eth0 flow-type ether dst 00:00:00:00:00:AA queue 0
$ ethtool -N eth0 flow-type ether dst 00:00:00:00:00:BB queue 1
$ ethtool -N eth0 flow-type ether dst 00:00:00:00:00:CC queue 2
$ ethtool -N eth0 flow-type ether dst 00:00:00:00:00:DD queue 3

RAL/RAH registers are configure so filter index 1 has address ending
with AA, filter index 2 has address ending in BB, and so on.

If we bring the interface down and up again, RAL/RAH registers are
configured so filter index 1 has address ending in DD, filter index 2
has CC, and so on. IOW, in reverse order we had before bringing the
interface down.

This issue can be fixed by traversing adapter->nfc_rule_list in
backwards when restoring the rules. Since hlist doesn't support
backwards traversal, this patch replaces it by list_head and fixes
igc_restore_nfc_rules() accordingly.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Fix NFC rules with multicast addresses
Andre Guedes [Fri, 24 Apr 2020 20:16:16 +0000 (13:16 -0700)]
igc: Fix NFC rules with multicast addresses

Multicast MAC addresses are valid address for NFC rules but
igc_add_mac_filter() is currently rejecting them. In fact, the I225
controller doesn't impose any constraint on the address value so this
patch gets rid of the address validation check in MAC filter APIs.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Fix NFC rule overwrite cases
Andre Guedes [Fri, 24 Apr 2020 20:16:15 +0000 (13:16 -0700)]
igc: Fix NFC rule overwrite cases

When the 'loc' argument is passed in ethtool, the input rule overwrites
any rule present in that location. In this situation we must disable the
old rule otherwise it is left enabled in hardware. This patch fixes
the issue by always calling igc_disable_nfc_rule() when deleting the
old rule, no matter the value of 'input' argument.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Fix locking issue when retrieving NFC rules
Andre Guedes [Fri, 24 Apr 2020 20:16:14 +0000 (13:16 -0700)]
igc: Fix locking issue when retrieving NFC rules

Access to NFC rules stored in adapter->nfc_rule_list is protect by
adapter->nfc_rule_lock. The functions igc_ethtool_get_nfc_rule()
and igc_ethtool_get_nfc_rules() are missing to hold the lock while
accessing rule objects.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Fix 'sw_idx' type in struct igc_nfc_rule
Andre Guedes [Fri, 24 Apr 2020 20:16:13 +0000 (13:16 -0700)]
igc: Fix 'sw_idx' type in struct igc_nfc_rule

The 'sw_idx' field from 'struct igc_nfc_rule' is u16 type but it is
assigned an u32 value in igc_ethtool_init_nfc_rule(). This patch changes
'sw_idx' type to u32 so they match. Also, it makes more sense to call
this field 'location' since it holds the NFC rule location.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agoigc: Refactor igc_ethtool_add_nfc_rule()
Andre Guedes [Fri, 24 Apr 2020 20:16:12 +0000 (13:16 -0700)]
igc: Refactor igc_ethtool_add_nfc_rule()

Current implementation of igc_ethtool_add_nfc_rule() is quite long and a
bit convoluted so this patch does a code refactoring to improve the
code.

Code related to NFC rule object initialization is refactored out to the
local helper function igc_ethtool_init_nfc_rule(). Likewise, code
related to NFC rule validation is refactored out to another local
helper, igc_ethtool_is_nfc_rule_valid().

RX_CLS_FLOW_DISC check is removed since it is redundant. The macro is
defined as the max value fsp->ring_cookie can have, so checking if
fsp->ring_cookie >= adapter->num_rx_queues is already sufficient.

Finally, some log messages are improved or added, and obvious comments
are removed.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
4 years agonet/mlx5e: Support pedit on mpls over UDP decap
Eli Cohen [Wed, 8 Apr 2020 06:01:33 +0000 (09:01 +0300)]
net/mlx5e: Support pedit on mpls over UDP decap

Allow to modify ethernet headers while decapsulating mpls over UDP
packets. This is implemented using the same reformat object used for
decapsulation.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: Add support for hw decapsulation of MPLS over UDP
Eli Cohen [Mon, 3 Feb 2020 11:44:14 +0000 (13:44 +0200)]
net/mlx5e: Add support for hw decapsulation of MPLS over UDP

MPLS over UDP is supported in hardware by using a packet reformat object
with reformat type equal L3_TUNNEL_TO_L2 which both decapsulates the
outer L3, L4 and MPLS headers, and allows for setting the L2 headers of
the resulting decapsulated packet. For the hardware to operate
correctly, the configuration of the firmware must have
FLEX_PARSER_PROFILE_ENABLE = 1.

Example tc rule:
  tc filter add dev bareudp0 protocol all prio 1 root flower enc_dst_port \
      6635 enc_src_ip 8.8.8.23 action mpls pop protocol ip pipe \
      action pedit ex munge eth dst set 00:11:22:33:44:21 pipe action \
      mirred egress redirect dev enp59s0f0_0

We use pedit to set the correct destination MAC.

For MPLS over UDP decapsulation to take place, the driver logic requires
the following:

1. flower filter added on bareudp device.
2. action mpls pop
3. zero or more pedit munge actions
4. one redirect action

Current implementation supports only IPv4 and no VLAN.

tc filter show output looks like this:
   filter protocol all pref 1 flower chain 0
   filter protocol all pref 1 flower chain 0 handle 0x1
     enc_src_ip 8.8.8.24
     enc_dst_port 6635
     in_hw in_hw_count 1
            action order 1: mpls  pop protocol ip pipe
             index 2 ref 1 bind 1

            action order 2:  pedit action pipe keys 2
             index 1 ref 1 bind 1
             key #0  at eth+0: val 00112233 mask 00000000
             key #1  at eth+4: val 44210000 mask 0000ffff

            action order 3: mirred (Egress Redirect to device enp59s0f0_0) stolen
            index 2 ref 1 bind 1

Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Eli Britstein <elibr@mellanox.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: Allow to match on mpls parameters
Eli Cohen [Wed, 29 Jan 2020 14:21:16 +0000 (16:21 +0200)]
net/mlx5e: Allow to match on mpls parameters

Support matching on MPLS over UDP parameters using misc2 section of
match parameters.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: Add support for hw encapsulation of MPLS over UDP
Eli Cohen [Sun, 17 Nov 2019 13:32:24 +0000 (15:32 +0200)]
net/mlx5e: Add support for hw encapsulation of MPLS over UDP

MPLS over UDP is supported by adding a rule on a representor net device
which does tunnel_key set, push mpls and forward to a baredup device. At
the hardware level we use a packet_reformat_context object to do the
encapsulation of the packet.

The resulting packet looks as follows (left side transmitted first):
outer L2 | outer IP | UDP | MPLS | inner L3 and data |

Example usage:
  tc filter add dev $rep0 protocol ip prio 1 root flower skip_sw  \
     action tunnel_key set src_ip 8.8.8.21 dst_ip 8.8.8.24 id 555 \
     dst_port 6635 tos 4 ttl 6 csum action mpls push protocol 0x8847 \
     label 555 tc 3 action mirred egress redirect dev bareudp0

This is how the filter is shown with tc filter show:
tc filter show dev enp59s0f0_0 ingress
filter protocol ip pref 1 flower chain 0
filter protocol ip pref 1 flower chain 0 handle 0x1
  eth_type ipv4
  skip_sw
  in_hw in_hw_count 1
        action order 1: tunnel_key  set
        src_ip 8.8.8.21
        dst_ip 8.8.8.24
        key_id 555
        dst_port 6635
        csum
        tos 0x4
        ttl 6 pipe
         index 1 ref 1 bind 1

        action order 2: mpls  push protocol mpls_uc label 555 tc 3 ttl 255 pipe
         index 1 ref 1 bind 1

        action order 3: mirred (Egress Redirect to device bareudp0) stolen
        index 1 ref 1 bind 1

Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Eli Britstein <elibr@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet: Add netif_is_bareudp() API to identify bareudp devices
Eli Cohen [Sun, 17 Nov 2019 12:35:42 +0000 (14:35 +0200)]
net: Add netif_is_bareudp() API to identify bareudp devices

Add netif_is_bareudp() so the device can be identified as a bareudp one.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Eli Britstein <elibr@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: Introduce kconfig var for TC support
Vlad Buslov [Tue, 12 May 2020 15:24:11 +0000 (18:24 +0300)]
net/mlx5e: Introduce kconfig var for TC support

In order to improve code maintainability and readability, introduce new
CONFIG_MLX5_CLS_ACT kconfig variable to control compilation of TC hardware
offloads implementation. This allows distinguishing between features that
require TC support (MPLSoUDP, etc.) and features that just rely on
representor functionality (rep_bond for live migration, etc.).

Modify rep_tc.h, rep_neigh.h, en_tc.h and chains.h files to provide stubs
for functions that are called from generic code.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: Move TC-specific code from en_main.c to en_tc.c
Vlad Buslov [Tue, 12 May 2020 15:08:41 +0000 (18:08 +0300)]
net/mlx5e: Move TC-specific code from en_main.c to en_tc.c

As a preparation for introducing new kconfig option that controls
compilation of all TC offloads code in mlx5, extract TC-specific code from
en_main.c to en_tc.c. This allows easily compiling out the code by
only including new source in make file when corresponding kconfig is
enabled instead of adding multiple ifdef blocks to en_main.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: Extract neigh-specific code from en_rep.c to rep/neigh.c
Vlad Buslov [Tue, 12 May 2020 14:29:22 +0000 (17:29 +0300)]
net/mlx5e: Extract neigh-specific code from en_rep.c to rep/neigh.c

As a preparation for introducing new kconfig option that controls
compilation of all TC offloads code in mlx5, extract neigh-specific code
from en_rep.c to standalone file. This allows easily compiling out the code
by only including new source in make file when corresponding kconfig is
enabled instead of adding multiple ifdef blocks to en_rep.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: Extract TC-specific code from en_rep.c to rep/tc.c
Vlad Buslov [Tue, 12 May 2020 13:41:41 +0000 (16:41 +0300)]
net/mlx5e: Extract TC-specific code from en_rep.c to rep/tc.c

As a preparation for introducing new kconfig option that controls
compilation of all TC offloads code in mlx5, extract TC-specific code from
en_rep.c to standalone file. This allows easily compiling out the code by
only including new source in make file when corresponding kconfig is
enabled instead of adding multiple ifdef blocks to en_rep.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: Use IS_ERR() to check and simplify code
Tang Bin [Fri, 15 May 2020 23:06:33 +0000 (07:06 +0800)]
net/mlx5e: Use IS_ERR() to check and simplify code

Use IS_ERR() and PTR_ERR() instead of PTR_ERR_OR_ZERO() to
simplify code, avoid redundant judgements.

Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: Tang Bin <tangbin@cmss.chinamobile.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agoMerge branch 'bridge-mrp-Add-br_mrp_unique_ifindex-function'
David S. Miller [Fri, 22 May 2020 23:17:15 +0000 (16:17 -0700)]
Merge branch 'bridge-mrp-Add-br_mrp_unique_ifindex-function'

Horatiu Vultur says:

====================
bridge: mrp: Add br_mrp_unique_ifindex function

This patch series adds small fixes to MRP implementation.
The following are fixed in this patch series:
- now is not allow to add the same port to multiple MRP rings
- remove unused variable
- restore the port state according to the bridge state when the MRP instance
  is deleted

v2:
 - use rtnl_dereference instead of rcu_dereference in the first patch
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobridge: mrp: Restore port state when deleting MRP instance
Horatiu Vultur [Thu, 21 May 2020 23:19:07 +0000 (23:19 +0000)]
bridge: mrp: Restore port state when deleting MRP instance

When a MRP instance is deleted, then restore the port according to the
bridge state. If the bridge is up then the ports will be in forwarding
state otherwise will be in disabled state.

Fixes: 9a9f26e8f7ea ("bridge: mrp: Connect MRP API with the switchdev API")
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoswitchdev: mrp: Remove the variable mrp_ring_state
Horatiu Vultur [Thu, 21 May 2020 23:19:06 +0000 (23:19 +0000)]
switchdev: mrp: Remove the variable mrp_ring_state

Remove the variable mrp_ring_state from switchdev_attr because is not
used anywhere.
The ring state is set using SWITCHDEV_OBJ_ID_RING_STATE_MRP.

Fixes: c284b5459008 ("switchdev: mrp: Extend switchdev API to offload MRP")
Acked-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobridge: mrp: Add br_mrp_unique_ifindex function
Horatiu Vultur [Thu, 21 May 2020 23:19:05 +0000 (23:19 +0000)]
bridge: mrp: Add br_mrp_unique_ifindex function

It is not allow to have the same net bridge port part of multiple MRP
rings. Therefore add a check if the port is used already in a different
MRP. In that case return failure.

Fixes: 9a9f26e8f7ea ("bridge: mrp: Connect MRP API with the switchdev API")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'DP83869-Enhancements'
David S. Miller [Fri, 22 May 2020 23:13:10 +0000 (16:13 -0700)]
Merge branch 'DP83869-Enhancements'

Dan Murphy says:

====================
DP83869 Enhancements

These are improvements to the DP83869 Ethernet PHY driver.  OP-mode and port
mirroring may be strapped on the device but the software only retrives these
settings from the device tree.  Reading the straps and initializing the
associated stored variables so when setting the PHY up and down the PHY's
configuration values will be retained.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: dp83869: Set opmode from straps
Dan Murphy [Thu, 21 May 2020 17:47:38 +0000 (12:47 -0500)]
net: phy: dp83869: Set opmode from straps

If the op-mode for the device is not set in the device tree then set
the strapped op-mode and store it for later configuration.

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Dan Murphy <dmurphy@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: dp83869: Update port-mirroring to read straps
Dan Murphy [Thu, 21 May 2020 17:47:37 +0000 (12:47 -0500)]
net: phy: dp83869: Update port-mirroring to read straps

The device tree may not have the property set for port mirroring
because the hardware may have it strapped. If the property is not in the
DT then check the straps and set the port mirroring bit appropriately.

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Dan Murphy <dmurphy@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests/bpf: CONFIG_LIRC required for test_lirc_mode2.sh
Alan Maguire [Fri, 22 May 2020 11:36:29 +0000 (12:36 +0100)]
selftests/bpf: CONFIG_LIRC required for test_lirc_mode2.sh

test_lirc_mode2.sh assumes presence of /sys/class/rc/rc0/lirc*/uevent
which will not be present unless CONFIG_LIRC=y

Fixes: 6bdd533cee9a ("bpf: add selftest for lirc_mode2 type program")
Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/1590147389-26482-3-git-send-email-alan.maguire@oracle.com
4 years agoselftests/bpf: CONFIG_IPV6_SEG6_BPF required for test_seg6_loop.o
Alan Maguire [Fri, 22 May 2020 11:36:28 +0000 (12:36 +0100)]
selftests/bpf: CONFIG_IPV6_SEG6_BPF required for test_seg6_loop.o

test_seg6_loop.o uses the helper bpf_lwt_seg6_adjust_srh();
it will not be present if CONFIG_IPV6_SEG6_BPF is not specified.

Fixes: b061017f8b4d ("selftests/bpf: add realistic loop tests")
Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/1590147389-26482-2-git-send-email-alan.maguire@oracle.com
4 years agoselftests/bpf: Add general instructions for test execution
Alan Maguire [Fri, 22 May 2020 11:24:34 +0000 (12:24 +0100)]
selftests/bpf: Add general instructions for test execution

Getting a clean BPF selftests run involves ensuring latest trunk LLVM/clang
are used, pahole is recent (>=1.16) and config matches the specified
config file as closely as possible.  Add to bpf_devel_QA.rst and point
tools/testing/selftests/bpf/README.rst to it.

Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Link: https://lore.kernel.org/bpf/1590146674-25485-1-git-send-email-alan.maguire@oracle.com
4 years agoRevert "net: mvneta: speed down the PHY, if WoL used, to save energy"
David S. Miller [Fri, 22 May 2020 23:09:42 +0000 (16:09 -0700)]
Revert "net: mvneta: speed down the PHY, if WoL used, to save energy"

This reverts commit 5e3768a436bb70c9c3e27aaba6b73f8ef8f5dcf3.

On request from Russell King, this is a layering violation.

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agocxgb4: add adapter hotplug support for ULDs
Potnuri Bharat Teja [Thu, 21 May 2020 10:34:29 +0000 (16:04 +0530)]
cxgb4: add adapter hotplug support for ULDs

Upon adapter hotplug, cxgb4 registers ULD devices for all the ULDs that
are already loaded, ensuring that ULD's can enumerate the hotplugged
adapter without reloading the ULD.

Signed-off-by: Potnuri Bharat Teja <bharat@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: flow_offload: simplify hw stats check handling
Edward Cree [Wed, 20 May 2020 18:18:10 +0000 (19:18 +0100)]
net: flow_offload: simplify hw stats check handling

Make FLOW_ACTION_HW_STATS_DONT_CARE be all bits, rather than none, so that
 drivers and __flow_action_hw_stats_check can use simple bitwise checks.

Pre-fill all actions with DONT_CARE in flow_rule_alloc(), rather than
 relying on implicit semantics of zero from kzalloc, so that callers which
 don't configure action stats themselves (i.e. netfilter) get the correct
 behaviour by default.

Only the kernel's internal API semantics change; the TC uAPI is unaffected.

v4: move DONT_CARE setting to flow_rule_alloc() for robustness and simplicity.

v3: set DONT_CARE in nft and ct offload.

v2: rebased on net-next, removed RFC tags.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'ip6_tunnel-add-MPLS-support'
David S. Miller [Fri, 22 May 2020 22:49:31 +0000 (15:49 -0700)]
Merge branch 'ip6_tunnel-add-MPLS-support'

Vadim Fedorenko says:

====================
ip6_tunnel: add MPLS support

The support for MPLS-in-IPv4 was added earlier. This patchset adds
support for MPLS-in-IPv6.

Changes in v2:
- Eliminate ifdefs IS_ENABLE(CONFIG_MPLS)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agompls: Add support for IPv6 tunnels
Vadim Fedorenko [Wed, 20 May 2020 15:21:39 +0000 (18:21 +0300)]
mpls: Add support for IPv6 tunnels

Add support for IPv6 tunnel devices in AF_MPLS.

Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoip6_tunnel: add generic MPLS receive support
Vadim Fedorenko [Wed, 20 May 2020 15:21:38 +0000 (18:21 +0300)]
ip6_tunnel: add generic MPLS receive support

Add support for MPLS in receive side.

Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agotunnel6: support for IPPROTO_MPLS
Vadim Fedorenko [Wed, 20 May 2020 15:21:37 +0000 (18:21 +0300)]
tunnel6: support for IPPROTO_MPLS

This patch is just preparation for MPLS support in ip6_tunnel

Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoip6_tunnel: add MPLS transmit support
Vadim Fedorenko [Wed, 20 May 2020 15:21:36 +0000 (18:21 +0300)]
ip6_tunnel: add MPLS transmit support

Add ETH_P_MPLS_UC as supported protocol.

Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoip6_tunnel: simplify transmit path
Vadim Fedorenko [Wed, 20 May 2020 15:21:35 +0000 (18:21 +0300)]
ip6_tunnel: simplify transmit path

Merge ip{4,6}ip6_tnl_xmit functions into one universal
ipxip6_tnl_xmit in preparation for adding MPLS support.

Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'mediatek-add-support-for-MediaTek-Ethernet-MAC'
David S. Miller [Fri, 22 May 2020 21:20:12 +0000 (14:20 -0700)]
Merge branch 'mediatek-add-support-for-MediaTek-Ethernet-MAC'

Bartosz Golaszewski says:

====================
mediatek: add support for MediaTek Ethernet MAC

This series adds support for the STAR Ethernet Controller present on MediaTeK
SoCs from the MT8* family.

First we convert the existing DT bindings for the PERICFG controller to YAML
and add a new compatible string for mt8516 variant of it. Then we add the DT
bindings for the MAC.

Next we do some cleanup of the mediatek ethernet drivers directory.

The largest patch in the series adds the actual new driver.

The rest of the patches add DT fixups for the boards already supported
upstream.

v1 -> v2:
- add a generic helper for retrieving the net_device associated with given
  private data
- fix several typos in commit messages
- remove MTK_MAC_VERSION and don't set the driver version
- use NET_IP_ALIGN instead of a magic number (2) but redefine it as it defaults
  to 0 on arm64
- don't manually turn the carrier off in mtk_mac_enable()
- process TX cleanup in napi poll callback
- configure pause in the adjust_link callback
- use regmap_read_poll_timeout() instead of handcoding the polling
- use devres_find() to verify that struct net_device is managed by devres in
  devm_register_netdev()
- add a patch moving all networking devres helpers into net/devres.c
- tweak the dma barriers: remove where unnecessary and add comments to the
  remaining barriers
- don't reset internal counters when enabling the NIC
- set the net_device's mtu size instead of checking the framesize in
  ndo_start_xmit() callback
- fix a race condition in waking up the netif queue
- don't emit log messages on OOM errors
- use dma_set_mask_and_coherent()
- use eth_hw_addr_random()
- rework the receive callback so that we reuse the previous skb if unmapping
  fails, like we already do if skb allocation fails
- rework hash table operations: add proper timeout handling and clear bits when
  appropriate

v2 -> v3:
- drop the patch adding priv_to_netdev() and store the netdev pointer in the
  driver private data
- add an additional dma_wmb() after reseting the descriptor in
  mtk_mac_ring_pop_tail()
- check the return value of dma_set_mask_and_coherent()
- improve the DT bindings for mtk-eth-mac: make the reg property in the example
  use single-cell address and size, extend the description of the PERICFG
  phandle and document the mdio sub-node
- add a patch converting the old .txt bindings for PERICFG to yaml
- limit reading the DMA memory by storing the mapped addresses in the driver
  private structure
- add a patch documenting the existing networking devres helpers

v3 -> v4:
- drop the devres patches: they will be sent separately
- call netdev_sent_queue() & netdev_completed_queue() where appropriate
- don't redefine NET_IP_ALIGN: define a private constant in the driver
- fix a couple typos
- only disabe/enable the MAC in suspend/resume if netif is running
- drop the count field from the ring structure and instead calculate the number
  of used descriptors from the tail and head indicies
- rework the locking used to protect the ring structures from concurrent
  access: use cheaper spin_lock_bh() and completely disable the internal
  spinlock used by regmap
- rework the interrupt handling to make it more fine-grained: onle re-enable
  TX and RX interrupts while they're needed, process the stats updates in a
  workqueue, not in napi context
- shrink the code responsible for unmapping and freeing skb memory
- rework the barriers as advised by Arnd

v4 -> v5:
- rename the driver to make it less confusing with the existing mtk_eth_soc
  ethernet driver
- unregister the mdiobus at device's detachment
- open-code spin lock calls to avoid calling the _bh variants where unnecessary
- limit read-modify-write operations where possible when accessing descriptor
  memory
- use READ_ONCE/WRITE_ONCE when modifying the status and data_ptr descriptor
  fields
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoARM64: dts: mediatek: enable ethernet on pumpkin boards
Bartosz Golaszewski [Fri, 22 May 2020 12:07:00 +0000 (14:07 +0200)]
ARM64: dts: mediatek: enable ethernet on pumpkin boards

Add remaining properties to the ethernet node and enable it.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoARM64: dts: mediatek: add ethernet pins for pumpkin boards
Bartosz Golaszewski [Fri, 22 May 2020 12:06:59 +0000 (14:06 +0200)]
ARM64: dts: mediatek: add ethernet pins for pumpkin boards

Setup the pin control for the Ethernet MAC.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoARM64: dts: mediatek: add an alias for ethernet0 for pumpkin boards
Bartosz Golaszewski [Fri, 22 May 2020 12:06:58 +0000 (14:06 +0200)]
ARM64: dts: mediatek: add an alias for ethernet0 for pumpkin boards

Add the ethernet0 alias for ethernet so that u-boot can find this node
and fill in the MAC address.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoARM64: dts: mediatek: add the ethernet node to mt8516.dtsi
Bartosz Golaszewski [Fri, 22 May 2020 12:06:57 +0000 (14:06 +0200)]
ARM64: dts: mediatek: add the ethernet node to mt8516.dtsi

Add the Ethernet MAC node to mt8516.dtsi. This defines parameters common
to all the boards based on this SoC.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoARM64: dts: mediatek: add pericfg syscon to mt8516.dtsi
Bartosz Golaszewski [Fri, 22 May 2020 12:06:56 +0000 (14:06 +0200)]
ARM64: dts: mediatek: add pericfg syscon to mt8516.dtsi

This adds support for the PERICFG register range as a syscon. This will
soon be used by the MediaTek Ethernet MAC driver for NIC configuration.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ethernet: mtk-star-emac: new driver
Bartosz Golaszewski [Fri, 22 May 2020 12:06:55 +0000 (14:06 +0200)]
net: ethernet: mtk-star-emac: new driver

This adds the driver for the MediaTek STAR Ethernet MAC currently used
on the MT8* SoC family. For now we only support full-duplex.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ethernet: mediatek: remove unnecessary spaces from Makefile
Bartosz Golaszewski [Fri, 22 May 2020 12:06:54 +0000 (14:06 +0200)]
net: ethernet: mediatek: remove unnecessary spaces from Makefile

The Makefile formatting in the kernel tree usually doesn't use tabs,
so remove them before we add a second driver.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ethernet: mediatek: rename Kconfig prompt
Bartosz Golaszewski [Fri, 22 May 2020 12:06:53 +0000 (14:06 +0200)]
net: ethernet: mediatek: rename Kconfig prompt

We'll soon by adding a second MediaTek Ethernet driver so modify the
Kconfig prompt.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodt-bindings: net: add a binding document for MediaTek STAR Ethernet MAC
Bartosz Golaszewski [Fri, 22 May 2020 12:06:52 +0000 (14:06 +0200)]
dt-bindings: net: add a binding document for MediaTek STAR Ethernet MAC

This adds yaml DT bindings for the MediaTek STAR Ethernet MAC present
on the mt8* family of SoCs.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodt-bindings: add new compatible to mediatek,pericfg
Bartosz Golaszewski [Fri, 22 May 2020 12:06:51 +0000 (14:06 +0200)]
dt-bindings: add new compatible to mediatek,pericfg

The PERICFG controller is present on the MT8516 SoC. Add an appropriate
compatible variant.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodt-bindings: convert the binding document for mediatek PERICFG to yaml
Bartosz Golaszewski [Fri, 22 May 2020 12:06:50 +0000 (14:06 +0200)]
dt-bindings: convert the binding document for mediatek PERICFG to yaml

Convert the DT binding .txt file for MediaTek's peripheral configuration
controller to YAML. There's one special case where the compatible has
three positions. Otherwise, it's a pretty normal syscon.

Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'ENA-features-and-cosmetic-changes'
David S. Miller [Fri, 22 May 2020 21:12:49 +0000 (14:12 -0700)]
Merge branch 'ENA-features-and-cosmetic-changes'

Arthur Kiyanovski says:

====================
ENA features and cosmetic changes

Diff from V1 of this patchset:
Removed error prints patch

This patchset includes:
1. new rx offset feature
2. reduction of the driver load time
3. multiple cosmetic changes to the code
====================

Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: reduce driver load time
Arthur Kiyanovski [Fri, 22 May 2020 09:09:05 +0000 (12:09 +0300)]
net: ena: reduce driver load time

This commit reduces the driver load time by using usec resolution
instead of msec when polling for hardware state change.

Also add back-off mechanism to handle cases where minimal sleep
time is not enough.

Signed-off-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: cosmetic: minor code changes
Arthur Kiyanovski [Fri, 22 May 2020 09:09:04 +0000 (12:09 +0300)]
net: ena: cosmetic: minor code changes

1. Use BIT macro instead of shift operator for code clarity
2. Replace multiple flag assignments to a single assignment of multiple
   flags in ena_com_add_single_rx_desc()
3. Move ENA_HASH_KEY_SIZE from ena_netdev.h to ena_com.h

Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: cosmetic: fix spacing issues
Arthur Kiyanovski [Fri, 22 May 2020 09:09:03 +0000 (12:09 +0300)]
net: ena: cosmetic: fix spacing issues

1. Add leading and trailing spaces to several comments for better
   readability
2. Make tabs and spaces uniform in enum defines in ena_admin_defs.h

Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: cosmetic: code reorderings
Arthur Kiyanovski [Fri, 22 May 2020 09:09:02 +0000 (12:09 +0300)]
net: ena: cosmetic: code reorderings

1. Reorder sanity checks in get_comp_ctxt() to make more sense
2. Reorder variables in ena_com_fill_hash_function() and
   ena_calc_io_queue_size() in reverse christmas tree.
3. Move around member initializations.

Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: cosmetic: remove unnecessary code
Arthur Kiyanovski [Fri, 22 May 2020 09:09:01 +0000 (12:09 +0300)]
net: ena: cosmetic: remove unnecessary code

1. Remove unused definition of DRV_MODULE_VERSION
2. Remove {} from single line-of-code ifs
3. Remove unnecessary comments from ena_get/set_coalesce()
4. Remove unnecessary extra spaces and newlines

Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: cosmetic: fix line break issues
Arthur Kiyanovski [Fri, 22 May 2020 09:09:00 +0000 (12:09 +0300)]
net: ena: cosmetic: fix line break issues

1. Join unnecessarily broken short lines in ena_com.c ena_netdev.c
2. Fix Indentations of broken lines

Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: cosmetic: fix spelling and grammar mistakes in comments
Arthur Kiyanovski [Fri, 22 May 2020 09:08:59 +0000 (12:08 +0300)]
net: ena: cosmetic: fix spelling and grammar mistakes in comments

fix spelling and grammar mistakes in comments in ena_com.h,
ena_com.c and ena_netdev.c

Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: cosmetic: set queue sizes to u32 for consistency
Arthur Kiyanovski [Fri, 22 May 2020 09:08:58 +0000 (12:08 +0300)]
net: ena: cosmetic: set queue sizes to u32 for consistency

Make all types of variables that convey the number and sizeof queues to
be u32, for consistency with the API between the driver and device via
ena_admin_defs.h:ena_admin_get_feat_resp.max_queue_ext fields. Current
code sometimes uses int and there are multiple assignments between these
variables with different types.

Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: cosmetic: rename ena_update_tx/rx_rings_intr_moderation()
Arthur Kiyanovski [Fri, 22 May 2020 09:08:57 +0000 (12:08 +0300)]
net: ena: cosmetic: rename ena_update_tx/rx_rings_intr_moderation()

Rename ena_update_tx/rx_rings_intr_moderation() to
ena_update_tx/rx_rings_nonadaptive_intr_moderation()
to distinguish between adaptive and non adaptive interrupt moderaion.

Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: simplify ena_com_update_intr_delay_resolution()
Arthur Kiyanovski [Fri, 22 May 2020 09:08:56 +0000 (12:08 +0300)]
net: ena: simplify ena_com_update_intr_delay_resolution()

Initialize prev_intr_delay_resolution with ena_dev->intr_delay_resolution
unconditionally, since it is initialized with
ENA_DEFAULT_INTR_DELAY_RESOLUTION in ena_probe(). This approach makes much
more sense than handling errors of not initializing it.

Also added unlikely to if condition.

Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: fix ena_com_comp_status_to_errno() return value
Arthur Kiyanovski [Fri, 22 May 2020 09:08:55 +0000 (12:08 +0300)]
net: ena: fix ena_com_comp_status_to_errno() return value

Default return value should be -EINVAL since the input
in this case was unexpected.
Also remove the now redundant check in the beginning
of the function.

Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: use explicit variable size for clarity
Arthur Kiyanovski [Fri, 22 May 2020 09:08:54 +0000 (12:08 +0300)]
net: ena: use explicit variable size for clarity

Use u64 instead of unsigned long long for clarity

Signed-off-by: Shai Brandes <shaibran@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: rename ena_com_free_desc to make API more uniform
Arthur Kiyanovski [Fri, 22 May 2020 09:08:53 +0000 (12:08 +0300)]
net: ena: rename ena_com_free_desc to make API more uniform

Rename ena_com_free_desc to ena_com_free_q_entries to match
the LLQ mode.

In non-LLQ mode, an entry in an IO ring corresponds to a
a descriptor. In LLQ mode an entry may correspond to several
descriptors (per LLQ definition).

Signed-off-by: Igor Chauskin <igorch@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ena: add support for the rx offset feature
Arthur Kiyanovski [Fri, 22 May 2020 09:08:52 +0000 (12:08 +0300)]
net: ena: add support for the rx offset feature

Newer ENA devices can write data to rx buffers with an offset
from the beginning of the buffer.

This commit adds support for this feature in the driver.

Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'net-atlantic-QoS-implementation'
David S. Miller [Fri, 22 May 2020 21:08:29 +0000 (14:08 -0700)]
Merge branch 'net-atlantic-QoS-implementation'

Igor Russkikh says:

====================
net: atlantic: QoS implementation

This patch series adds support for mqprio rate limiters and multi-TC:
 * max_rate is supported on both A1 and A2;
 * min_rate is supported on A2 only;

This is a joint work of Mark and Dmitry.

To implement this feature, a couple of rearrangements and code
improvements were done, in areas of TC/ring management, allocation
control, etc.

One of the problems we faced is conflicting ptp functionality, which
consumes a whole traffic class due to hardware limitations.
Patches below have a more detailed description on how PTP and multi-TC
co-exist right now.

v2:
 * accommodated review comments (-Wmissing-prototypes and
   -Wunused-but-set-variable findings);
 * added user notification in case of conflicting multi-TC<->PTP
   configuration;
 * added automatic PTP disabling, if a conflicting configuration is
   detected;
 * removed module param, which was used for PTP disabling in v1;

v1: https://patchwork.ozlabs.org/cover/1294380/
====================

Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: atlantic: proper rss_ctrl1 (54c0) initialization
Mark Starovoytov [Fri, 22 May 2020 08:19:48 +0000 (11:19 +0300)]
net: atlantic: proper rss_ctrl1 (54c0) initialization

This patch fixes an inconsistency between code and spec, which
was found while working on the QoS implementation.

When 8TCs are used, 2 is the maximum supported number of index bits.
In a 4TC mode, we do support 3, but we shouldn't really use the bytes,
which are intended for the 8TC mode.

Signed-off-by: Mark Starovoytov <mstarovoitov@marvell.com>
Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>