platform/kernel/linux-starfive.git
2 years agosctp: move hlist_node and hashent out of sctp_ep_common
Xin Long [Tue, 21 Dec 2021 21:40:30 +0000 (16:40 -0500)]
sctp: move hlist_node and hashent out of sctp_ep_common

Struct sctp_ep_common is included in both asoc and ep, but hlist_node
and hashent are only needed by ep after asoc_hashtable was dropped by
Commit b5eff7128366 ("sctp: drop the old assoc hashtable of sctp").

So it is better to move hlist_node and hashent from sctp_ep_common to
sctp_endpoint, and it saves some space for each asoc.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'add-tests-for-vxlan-with-ipv6-underlay'
Jakub Kicinski [Thu, 23 Dec 2021 01:14:44 +0000 (17:14 -0800)]
Merge branch 'add-tests-for-vxlan-with-ipv6-underlay'

Amit Cohen says:

====================
Add tests for VxLAN with IPv6 underlay

mlxsw driver lately added support for VxLAN with IPv6 underlay.
This set adds the relevant tests for IPv6, most of them are same to
IPv4 tests with the required changes.

Patch set overview:
Patch #1 relaxes requirements for offloading TC filters that
match on 802.1q fields. The following selftests make use of these
newly-relaxed filters.
Patch #2 adds preparation as part of selftests API, which will be used
later.
Patches #3-#4 add tests for VxLAN with bridge aware and unaware.
Patche #5 cleans unused function.
Patches #6-#7 add tests for VxLAN symmetric and asymmetric.
Patch #8 adds test for Q-in-VNI.
====================

Link: https://lore.kernel.org/r/20211221144949.2527545-1-amcohen@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: forwarding: Add Q-in-VNI test for IPv6
Amit Cohen [Tue, 21 Dec 2021 14:49:49 +0000 (16:49 +0200)]
selftests: forwarding: Add Q-in-VNI test for IPv6

Add test to check Q-in-VNI traffic with IPv6 underlay and overlay.
The test is similar to the existing IPv4 test.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: forwarding: Add a test for VxLAN symmetric routing with IPv6
Amit Cohen [Tue, 21 Dec 2021 14:49:48 +0000 (16:49 +0200)]
selftests: forwarding: Add a test for VxLAN symmetric routing with IPv6

In a similar fashion to the asymmetric test, add a test for symmetric
routing. In symmetric routing both the ingress and egress VTEPs perform
routing in the overlay network into / from the VxLAN tunnel. Packets in
different directions use the same VNI - the L3 VNI.
Different tenants (VRFs) use different L3 VNIs.

Add a test which is similar to the existing IPv4 test to check IPv6.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: forwarding: Add a test for VxLAN asymmetric routing with IPv6
Amit Cohen [Tue, 21 Dec 2021 14:49:47 +0000 (16:49 +0200)]
selftests: forwarding: Add a test for VxLAN asymmetric routing with IPv6

In asymmetric routing the ingress VTEP routes the packet into the
correct VxLAN tunnel, whereas the egress VTEP only bridges the packet to
the correct host. Therefore, packets in different directions use
different VNIs - the target VNI.

Add a test which is similar to the existing IPv4 test to check IPv6.

The test uses a simple topology with two VTEPs and two VNIs and verifies
that ping passes between hosts (local / remote) in the same VLAN (VNI)
and in different VLANs belonging to the same tenant (VRF).

While the test does not check VM mobility, it does configure an anycast
gateway using a macvlan device on both VTEPs.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: forwarding: vxlan_bridge_1q: Remove unused function
Amit Cohen [Tue, 21 Dec 2021 14:49:46 +0000 (16:49 +0200)]
selftests: forwarding: vxlan_bridge_1q: Remove unused function

Remove `vxlan_ping_test()` which is not used and probably was copied
mistakenly from vxlan_bridge_1d.sh.

This was found while adding an equivalent test for IPv6.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: forwarding: Add VxLAN tests with a VLAN-aware bridge for IPv6
Amit Cohen [Tue, 21 Dec 2021 14:49:45 +0000 (16:49 +0200)]
selftests: forwarding: Add VxLAN tests with a VLAN-aware bridge for IPv6

The tests are very similar to their VLAN-unaware counterpart
(vxlan_bridge_1d_ipv6.sh and vxlan_bridge_1d_port_8472_ipv6.sh),
but instead of using multiple VLAN-unaware bridges, a single VLAN-aware
bridge is used with multiple VLANs.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: forwarding: Add VxLAN tests with a VLAN-unaware bridge for IPv6
Amit Cohen [Tue, 21 Dec 2021 14:49:44 +0000 (16:49 +0200)]
selftests: forwarding: Add VxLAN tests with a VLAN-unaware bridge for IPv6

Add tests similar to vxlan_bridge_1d.sh and vxlan_bridge_1d_port_8472.sh.

The tests set up a topology with three VxLAN endpoints: one
"local", possibly offloaded, and two "remote", formed using veth pairs
and likely purely software bridges. The "local" endpoint is connected to
host systems by a VLAN-unaware bridge.

Since VxLAN tunnels must be unique per namespace, each of the "remote"
endpoints is in its own namespace. H3 forms the bridge between the three
domains.

Send IPv4 packets and IPv6 packets with IPv6 underlay.
Use `TC_FLAG`, which is defined in `forwarding.config` file, for TC
checks. `TC_FLAG` allows testing that on HW datapath, the traffic
actually goes through HW.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: lib.sh: Add PING_COUNT to allow sending configurable amount of packets
Amit Cohen [Tue, 21 Dec 2021 14:49:43 +0000 (16:49 +0200)]
selftests: lib.sh: Add PING_COUNT to allow sending configurable amount of packets

Currently `ping_do()` and `ping6_do()` send 10 packets.

There are cases that it is not possible to catch only the interesting
packets using tc rule, so then, it is possible to send many packets and
verify that at least this amount of packets hit the rule.

Add `PING_COUNT` variable, which is set to 10 by default, to allow tests
sending more than 10 packets using the existing ping API.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomlxsw: spectrum_flower: Make vlan_id limitation more specific
Amit Cohen [Tue, 21 Dec 2021 14:49:42 +0000 (16:49 +0200)]
mlxsw: spectrum_flower: Make vlan_id limitation more specific

Spectrum ASICs do not support matching of VLAN ID at egress.
Currently, mlxsw driver forbids matching of all VLAN related fields at
egress, which is too strict check.

For example, the following filter is not supported by the driver:
$ tc filter add dev swpX egress protocol 802.1q pref 1 handle 101 flower
vlan_ethtype ipv4 src_ip .. dst_ip .. skip_sw action pass
Error: mlxsw_spectrum: vlan_id key is not supported on egress.
We have an error talking to the kernel

The filter above does not match on VLAN ID, but is bounced anyway.

Make the check more specific, forbid only matching of 'vlan_id' at egress.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'mlx5-updates-2021-12-21' of git://git.kernel.org/pub/scm/linux/kernel...
Jakub Kicinski [Thu, 23 Dec 2021 01:13:00 +0000 (17:13 -0800)]
Merge tag 'mlx5-updates-2021-12-21' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-updates-2021-12-21

1) From Shay Drory: Devlink user knobs to control device's EQ size

This series provides knobs which will enable users to
minimize memory consumption of mlx5 Functions (PF/VF/SF).
mlx5 exposes two new generic devlink params for EQ size
configuration and uses devlink generic param max_macs.

LINK: https://lore.kernel.org/netdev/20211208141722.13646-1-shayd@nvidia.com/
2) From Tariq and Lama, allocate software channel objects and statistics
  of a mlx5 netdevice private data dynamically upon first demand to save on
  memory.

* tag 'mlx5-updates-2021-12-21' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
  net/mlx5e: Take packet_merge params directly from the RX res struct
  net/mlx5e: Allocate per-channel stats dynamically at first usage
  net/mlx5e: Use dynamic per-channel allocations in stats
  net/mlx5e: Allow profile-specific limitation on max num of channels
  net/mlx5e: Save memory by using dynamic allocation in netdev priv
  net/mlx5e: Add profile indications for PTP and QOS HTB features
  net/mlx5e: Use bitmap field for profile features
  net/mlx5: Remove the repeated declaration
  net/mlx5: Let user configure max_macs generic param
  devlink: Clarifies max_macs generic devlink param
  net/mlx5: Let user configure event_eq_size param
  devlink: Add new "event_eq_size" generic device param
  net/mlx5: Let user configure io_eq_size param
  devlink: Add new "io_eq_size" generic device param
====================

Link: https://lore.kernel.org/r/20211222031604.14540-1-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agocodel: remove unnecessary pkt_sched.h include
Jakub Kicinski [Tue, 21 Dec 2021 19:39:41 +0000 (11:39 -0800)]
codel: remove unnecessary pkt_sched.h include

Commit d068ca2ae2e6 ("codel: split into multiple files") moved all
Qdisc-related code to codel_qdisc.h, move the include of pkt_sched.h
as well.

This is similar to the previous commit, although we don't care as
much about incremental builds after pkt_sched.h was touched itself
it is included by net/sch_generic.h which is modified ~20 times
a year.

This decreases the incremental build size after touching pkt_sched.h
from 1592 to 617 objects.

Fix unmasked missing includes in WiFi drivers.

Acked-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/20211221193941.3805147-2-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agocodel: remove unnecessary sock.h include
Jakub Kicinski [Tue, 21 Dec 2021 19:39:40 +0000 (11:39 -0800)]
codel: remove unnecessary sock.h include

Since sock.h is modified relatively often (60 times in the last
12 months) it seems worthwhile to decrease the incremental build
work.

CoDel's header includes net/inet_ecn.h which in turn includes net/sock.h.
codel.h is itself included by mac80211 which is included by much of
the WiFi stack and drivers. Removing the net/inet_ecn.h include from
CoDel breaks the dependecy between WiFi and sock.h.

Commit d068ca2ae2e6 ("codel: split into multiple files") moved all
the code which actually needs ECN helpers out to net/codel_impl.h,
the include can be moved there as well.

This decreases the incremental build size after touching sock.h
from 4999 objects to 4051 objects.

Fix unmasked missing includes in WiFi drivers.

Acked-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/20211221193941.3805147-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: broadcom: bcm4908enet: remove redundant variable bytes
Colin Ian King [Wed, 22 Dec 2021 00:39:37 +0000 (00:39 +0000)]
net: broadcom: bcm4908enet: remove redundant variable bytes

The variable bytes is being used to summate slot lengths,
however the value is never used afterwards. The summation
is redundant so remove variable bytes.

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Link: https://lore.kernel.org/r/20211222003937.727325-1-colin.i.king@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoice: trivial: fix odd indenting
Jesse Brandeburg [Tue, 21 Dec 2021 23:05:38 +0000 (15:05 -0800)]
ice: trivial: fix odd indenting

Fix an odd indent where some code was left indented, and causes smatch
to warn:
ice_log_pkg_init() warn: inconsistent indenting

While here, for consistency, add a break after the default case.

This commit has a Fixes: but we caught this while it was only in net-next.

Fixes: 247dd97d713c ("ice: Refactor status flow for DDP load")
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Link: https://lore.kernel.org/r/20211221230538.2546315-1-jesse.brandeburg@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next...
Jakub Kicinski [Wed, 22 Dec 2021 22:25:41 +0000 (14:25 -0800)]
Merge branch '100GbE' of git://git./linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
100GbE Intel Wired LAN Driver Updates 2021-12-21

This series contains updates to ice driver only.

Karol modifies the reset flow to correct issues with PTP reset.

Jake extends PTP support for E822 based devices. This includes a few
cleanup patches, that fix some minor issues. In addition, there are some
slight refactors to ease the addition of E822 support, followed by adding
the new hardware implementation ice_ptp_hw.c.

There are a few major differences with E822 support compared to E810
support:

*) The E822 device has a Clock Generation Unit which must be initialized in
order to generate proper clock frequencies on the output that drives the PTP
hardware clock registers

*) The E822 PHY is a bit different and requires a more complex
initialization procedure which must be rerun any time the link configuration
changes.

*) The E822 devices support enhanced timestamp calibration by making use of
a process called Vernier offset measurement. This allows the hardware to
measure phase offset related to the PHY clocks for Serdes and FEC, reducing
the inaccuracy of the timestamp relative to the actual packet transmission
and receipt. Making use of this requires data gathered from the first
transmitted and received packets, and waiting for the PHY to complete the
calibration measurements. This is done as part of a new kthread, ov_work.
Note that to avoid delay in enabling timestamps, we start the PHY in
'bypass' mode which allows timestamps to be captured without the Vernier
calibration measurement. Once the first packets have been sent and received,
we then complete the calibration setup and exit bypass mode and begin using
the more precise timestamps. According to the datasheet, timestamps without
calibration data can be incorrect relative to actual receipt or transmission
by up to 1 clock cycle (~1.25 nanoseconds), while calibrated timestamps
should be correct to within 1/8th of a clock cycle (~0.15 nanoseconds).

*) E822 devices support crosstimestamping via PCIe PTM, which we enable when
available on the platform.

There is a fair amount of logic required to perform PHY and CGU
initialization, which is the vast majority of the new code, but it is fairly
self contained within ice_ptp_hw.c, with the exception of monitoring for
offset validity being handled by a kthread.

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
  ice: support crosstimestamping on E822 devices if supported
  ice: exit bypass mode once hardware finishes timestamp calibration
  ice: ensure the hardware Clock Generation Unit is configured
  ice: implement basic E822 PTP support
  ice: convert clk_freq capability into time_ref
  ice: introduce ice_ptp_init_phc function
  ice: use 'int err' instead of 'int status' in ice_ptp_hw.c
  ice: PTP: move setting of tstamp_config
  ice: introduce ice_base_incval function
  ice: Fix E810 PTP reset flow
====================

Link: https://lore.kernel.org/r/20211221174845.3063640-1-anthony.l.nguyen@intel.com
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/mlx5e: Take packet_merge params directly from the RX res struct
Tariq Toukan [Tue, 14 Sep 2021 11:46:05 +0000 (14:46 +0300)]
net/mlx5e: Take packet_merge params directly from the RX res struct

As packet_merge params structure is saved on the RX resources structure, there
is no need to pass it separately.

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Allocate per-channel stats dynamically at first usage
Lama Kayal [Wed, 22 Sep 2021 06:32:33 +0000 (09:32 +0300)]
net/mlx5e: Allocate per-channel stats dynamically at first usage

Make stats allocation per-channel dynamic on demand, at channel open
operation.

Previously the stats array was pre-allocated for the maximum possible
number of channels. Here we defer the per-channel stats instance allocation
upon its first usage, so that it's allocated only if really needed.

Allocating stats on demand helps maintain a more memory-efficient code,
as we're saving memory when the used number of channels is smaller than
the maximum.

The stats memory instances are still freed in mlx5e_priv_arrays_free(),
so that they are persistent to channels' closure.

Memory size allocated for struct mlx5e_channel_stats is 3648 bytes.
If maximum number of channel stands for 64, the total memory space
allocated for stats is 3648x64 = 228K bytes. In scenarios where the
number of channels in use is significantly smaller than maximum number,
the memory saved can be remarkable.

Signed-off-by: Lama Kayal <lkayal@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Use dynamic per-channel allocations in stats
Tariq Toukan [Sun, 5 Dec 2021 14:04:14 +0000 (16:04 +0200)]
net/mlx5e: Use dynamic per-channel allocations in stats

Make stats array an array of pointer. This patch comes in to prepare for
the next patch where allocations of the stats are to be performed
dynamically on first usage.

Signed-off-by: Lama Kayal <lkayal@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Allow profile-specific limitation on max num of channels
Tariq Toukan [Wed, 29 Sep 2021 14:29:24 +0000 (17:29 +0300)]
net/mlx5e: Allow profile-specific limitation on max num of channels

Let SF/VF representor's netdev use profile-specific limitation on
max_nch to reduce its memory and HW resources consumption.

This is particularly important for environments with limited memory
and high number of SFs.

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Vu Pham <vuhuong@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Save memory by using dynamic allocation in netdev priv
Tariq Toukan [Thu, 2 Sep 2021 07:24:06 +0000 (10:24 +0300)]
net/mlx5e: Save memory by using dynamic allocation in netdev priv

Many arrays in priv are statically allocated with a pre-defined maximum
(for num channels, num TCs, etc...), that is in some cases significantly
larger than the actual maximum. Examples:
- The more VFs are supported, the less MSIX vectors each of them could
  have. This limits the max_nch for each.
- Systems with limited number of cores or MSIX (< 64).
- Netdev profiles that do not support: QoS (DCB / HTB), PTP TX port
  timestamping.

Here we save some amount of memory by moving several structures
and arrays to follow the actual maximum instead.
This patch also prepares the code for even more savings to follow.

For example, on a system where the maximum num of channel is 8,
the channels stats structs alone go down from 3648*64 = 228 KB to
3648*8 = 28.5 KB per interface.

This is important for environments with high number of VFs/SFs or
limited memory.

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Add profile indications for PTP and QOS HTB features
Tariq Toukan [Tue, 7 Dec 2021 06:38:42 +0000 (08:38 +0200)]
net/mlx5e: Add profile indications for PTP and QOS HTB features

Let the profile indicate support of the PTP and HTB (QOS) features.
This unifies the logic that calculates the number of netdev queues needed
for the features, and allows simplification of mlx5e_create_netdev(),
which no longer requires number of rx/tx queues as parameters.

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Aya Levin <ayal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Use bitmap field for profile features
Tariq Toukan [Sun, 5 Dec 2021 10:31:47 +0000 (12:31 +0200)]
net/mlx5e: Use bitmap field for profile features

Use a features bitmap field in mlx5e_profile to declare profile support
state of the different features.  Let it replace the existing
rx_ptp_support boolean. It will be extended to cover more features in a
downstream patch.

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5: Remove the repeated declaration
Shaokun Zhang [Tue, 7 Dec 2021 12:35:15 +0000 (20:35 +0800)]
net/mlx5: Remove the repeated declaration

Function 'mlx5_esw_vport_match_metadata_supported' and
'mlx5_esw_offloads_vport_metadata_set' are declared twice, so remove
the repeated declaration and blank line.

Cc: Saeed Mahameed <saeedm@nvidia.com>
Cc: Leon Romanovsky <leon@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5: Let user configure max_macs generic param
Shay Drory [Thu, 9 Dec 2021 10:09:29 +0000 (12:09 +0200)]
net/mlx5: Let user configure max_macs generic param

Currently, max_macs is taking 70Kbytes of memory per function. This
size is not needed in all use cases, and is critical with large scale.
Hence, allow user to configure the number of max_macs.

For example, to reduce the number of max_macs to 1, execute::
$ devlink dev param set pci/0000:00:0b.0 name max_macs value 1 \
              cmode driverinit
$ devlink dev reload pci/0000:00:0b.0

Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agodevlink: Clarifies max_macs generic devlink param
Shay Drory [Thu, 9 Dec 2021 10:09:28 +0000 (12:09 +0200)]
devlink: Clarifies max_macs generic devlink param

The generic param max_macs documentation isn't clear.
Replace it with a more descriptive documentation

Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5: Let user configure event_eq_size param
Shay Drory [Thu, 9 Dec 2021 10:09:27 +0000 (12:09 +0200)]
net/mlx5: Let user configure event_eq_size param

Event EQ is an EQ which received the notification of almost all the
events generated by the NIC.
Currently, each event EQ is taking 512KB of memory. This size is not
needed in most use cases, and is critical with large scale. Hence,
allow user to configure the size of the event EQ.

For example to reduce event EQ size to 64, execute::
$ devlink dev param set pci/0000:00:0b.0 name event_eq_size value 64 \
              cmode driverinit
$ devlink dev reload pci/0000:00:0b.0

Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agodevlink: Add new "event_eq_size" generic device param
Shay Drory [Thu, 9 Dec 2021 10:09:26 +0000 (12:09 +0200)]
devlink: Add new "event_eq_size" generic device param

Add new device generic parameter to determine the size of the
asynchronous control events EQ.

For example, to reduce event EQ size to 64, execute:
$ devlink dev param set pci/0000:06:00.0 \
              name event_eq_size value 64 cmode driverinit
$ devlink dev reload pci/0000:06:00.0

Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5: Let user configure io_eq_size param
Shay Drory [Thu, 9 Dec 2021 10:09:25 +0000 (12:09 +0200)]
net/mlx5: Let user configure io_eq_size param

Currently, each I/O EQ is taking 128KB of memory. This size
is not needed in all use cases, and is critical with large scale.
Hence, allow user to configure the size of I/O EQs.

For example, to reduce I/O EQ size to 64, execute:
$ devlink dev param set pci/0000:00:0b.0 name io_eq_size value 64 \
              cmode driverinit
$ devlink dev reload pci/0000:00:0b.0

Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agodevlink: Add new "io_eq_size" generic device param
Shay Drory [Thu, 9 Dec 2021 10:09:24 +0000 (12:09 +0200)]
devlink: Add new "io_eq_size" generic device param

Add new device generic parameter to determine the size of the
I/O completion EQs.

For example, to reduce I/O EQ size to 64, execute:
$ devlink dev param set pci/0000:06:00.0 \
              name io_eq_size value 64 cmode driverinit
$ devlink dev reload pci/0000:06:00.0

Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agoMerge branch '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue
Jakub Kicinski [Wed, 22 Dec 2021 01:20:31 +0000 (17:20 -0800)]
Merge branch '1GbE' of git://git./linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
1GbE Intel Wired LAN Driver Updates 2021-12-21

This series contains updates to igc, igb, igbvf, and fm10k drivers.

Sasha removes unused defines and enum values from igc driver.

Jason Wang removes a variable whose value never changes and, instead,
returns the value directly for igb.

Karen adjusts a reset message from warning to info for igbvf.

Xiang wangx removes a repeated word for fm10k.

* '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
  fm10k: Fix syntax errors in comments
  igbvf: Refactor trace
  igb: remove never changed variable `ret_val'
  igc: Remove obsolete define
  igc: Remove obsolete mask
  igc: Remove obsolete nvm type
  igc: Remove unused phy type
  igc: Remove unused _I_PHY_ID define
====================

Link: https://lore.kernel.org/r/20211221180200.3176851-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: phy: micrel: Adding interrupt support for Link up/Link down in LAN8814 Quad phy
Divya Koppera [Tue, 21 Dec 2021 11:22:17 +0000 (16:52 +0530)]
net: phy: micrel: Adding interrupt support for Link up/Link down in LAN8814 Quad phy

This patch add support for Link up or Link down
interrupt support in LAN8814 Quad phy

Signed-off-by: Divya Koppera <Divya.Koppera@microchip.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20211221112217.9502-1-Divya.Koppera@microchip.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agofm10k: Fix syntax errors in comments
Xiang wangx [Tue, 14 Dec 2021 12:50:04 +0000 (20:50 +0800)]
fm10k: Fix syntax errors in comments

Delete the redundant word 'by'.

Signed-off-by: Xiang wangx <wangxiang@cdjrlc.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoigbvf: Refactor trace
Karen Sornek [Wed, 7 Jul 2021 06:40:25 +0000 (08:40 +0200)]
igbvf: Refactor trace

Refactoring "PF still resetting" message, because previous version looked
like a bug - it informed about changes that worked as designed but might
confuse users. Changes requested to make message more user-friendly.

Signed-off-by: Karen Sornek <karen.sornek@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoigb: remove never changed variable `ret_val'
Jason Wang [Sun, 14 Nov 2021 06:02:22 +0000 (14:02 +0800)]
igb: remove never changed variable `ret_val'

The variable used for return status in `igb_write_xmdio_reg' function
is never changed  and this function is just need return 0. Thus, the
`ret_val' can be removed and return 0 at the end of the
`igb_write_xmdio_reg' function.

Signed-off-by: Jason Wang <wangborong@cdjrlc.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoigc: Remove obsolete define
Sasha Neftin [Tue, 7 Dec 2021 06:19:18 +0000 (08:19 +0200)]
igc: Remove obsolete define

'MII_CR_FULL_DUPLEX' define not in use. This patch comes to tidy up
 obsolete define.

Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoigc: Remove obsolete mask
Sasha Neftin [Tue, 9 Nov 2021 09:42:16 +0000 (11:42 +0200)]
igc: Remove obsolete mask

'IGC_CTRL_EXT_LINK_MODE_MASK' not in use. This patch comes to tidy up
obsolete define.

Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Nechama Kraus <nechamax.kraus@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoigc: Remove obsolete nvm type
Sasha Neftin [Sun, 31 Oct 2021 10:45:56 +0000 (12:45 +0200)]
igc: Remove obsolete nvm type

i225 devices use only spi nvm type. This patch comes to tidy up
obsolete nvm types.

Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Nechama Kraus <nechamax.kraus@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoigc: Remove unused phy type
Sasha Neftin [Mon, 25 Oct 2021 07:16:36 +0000 (10:16 +0300)]
igc: Remove unused phy type

_phy_none type not in use. Clean up the code accordingly,
and get rid of the unused enum line

Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
Tested-by: Nechama Kraus <nechamax.kraus@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoigc: Remove unused _I_PHY_ID define
Sasha Neftin [Wed, 13 Oct 2021 06:19:26 +0000 (09:19 +0300)]
igc: Remove unused _I_PHY_ID define

_I_PHY_ID not in use. Clean up the code accordingly,
and get rid of the unused define

Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Nechama Kraus <nechamax.kraus@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: support crosstimestamping on E822 devices if supported
Jacob Keller [Wed, 13 Oct 2021 16:00:08 +0000 (09:00 -0700)]
ice: support crosstimestamping on E822 devices if supported

E822 devices on supported platforms can generate a cross timestamp
between the platform ART and the device time. This process allows for
very precise measurement of the difference between the PTP hardware
clock and the platform time.

This is only supported if we know the TSC frequency relative to ART, so
we do not enable this unless the boot CPU has a known TSC frequency (as
required by convert_art_ns_to_tsc).

Because PCIe PTM support is not available on all platforms, introduce
CONFIG_ICE_HWTS and make it depend on X86 where we know the support
exists.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: exit bypass mode once hardware finishes timestamp calibration
Jacob Keller [Wed, 13 Oct 2021 15:56:58 +0000 (08:56 -0700)]
ice: exit bypass mode once hardware finishes timestamp calibration

Once the E822 device has sent and received one packet, the hardware
computes the internal delay of the PHY using a process known as Vernier
calibration. This calibration calculates a more accurate offset for the
Tx and Rx timestamps. To make use of this offset, we need to exit the
bypass mode. This cannot be done until the PHY has completed offset
calibration, as indicated by the offset valid bits.

To handle this, introduce a kthread work item which will poll the offset
valid bits every few milliseconds seeing if it is safe to exit bypass
mode.

Once we have finished calibrating the offsets, we can program the total
Tx and Rx offset registers and turn off the bypass bit. This allows the
hardware to include the more precise vernier calibration offset, and
improves the timestamp precision.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: ensure the hardware Clock Generation Unit is configured
Jacob Keller [Wed, 13 Oct 2021 15:55:58 +0000 (08:55 -0700)]
ice: ensure the hardware Clock Generation Unit is configured

The E822 device has a Clock Generation Unit (CGU) responsible for
determining the clock frequency that drives the timers.

Ensure this function is initialized when bringing up the PTP support, so
that the clock has a known frequency.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: implement basic E822 PTP support
Jacob Keller [Wed, 13 Oct 2021 15:54:51 +0000 (08:54 -0700)]
ice: implement basic E822 PTP support

Implement support for the basic operations needed to enable the PTP
hardware clock on E822 devices.

This includes implementations for the various PHY access functions, as
well as the ability to start and stop the PHY timers. This is different
from the E810 device because the configuration depends on link speed, so
we cannot just start the PHYs immediately. We must wait until the link
is up to get proper values for the speed based initialization.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: convert clk_freq capability into time_ref
Jacob Keller [Wed, 13 Oct 2021 15:53:41 +0000 (08:53 -0700)]
ice: convert clk_freq capability into time_ref

Convert the clk_freq value into the associated time_ref frequency value
for E822 devices. This simplifies determining the time reference value
for the clock.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: introduce ice_ptp_init_phc function
Jacob Keller [Wed, 13 Oct 2021 15:52:29 +0000 (08:52 -0700)]
ice: introduce ice_ptp_init_phc function

When we enable support for E822 devices, there are some additional
steps required to initialize the PTP hardware clock. To make this easier
to implement as device-specific behavior, refactor the register setups
in ice_ptp_init_owner to a new ice_ptp_init_phc function defined in
ice_ptp_hw.c

This function will have a common section, and an e810 specific
sub-implementation.

This will enable easily extending the functionality to cover the E822
specific setup required to initialize the hardware clock generation
unit. It also makes it clear which steps are E810 specific vs which ones
are necessary for all ice devices.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: use 'int err' instead of 'int status' in ice_ptp_hw.c
Jacob Keller [Wed, 13 Oct 2021 15:51:36 +0000 (08:51 -0700)]
ice: use 'int err' instead of 'int status' in ice_ptp_hw.c

The ice_ptp_hw.c file introduced a bunch of uses of "int status" instead
of the more traditional "int err" or "int ret". These are actually
traditional Linux error codes (as opposed to the recently removed
ice_status enumeration values).

We're about to add a bunch of new functions to ice_ptp_hw.c. It's
normally preferred in the ice driver to use "int ret" or "int err" when
dealing with error code values.

Instead of making the new functions use "int status" lets just fix all
of ice_ptp_hw.c to use "int err". This will match the new functions and
ensures a consistent style across at least the PTP related files.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: PTP: move setting of tstamp_config
Jacob Keller [Wed, 13 Oct 2021 15:48:13 +0000 (08:48 -0700)]
ice: PTP: move setting of tstamp_config

The tstamp_config structure is being set inside of
ice_ptp_cfg_timestamp, which is the function used to set Tx and
Rx timestamping during initialization.

This function is also used in order to set the PHY port timestamping
status. However, it makes sense to always set the tstamp_config directly
whenever the ice_set_tx_tstamp or ice_set_rx_tstamp functions are
called.

Move assignment of tstamp_config into the related functions and out of
ice_ptp_cfg_timestamp.

Now that we assign the timestamp mode in the relevant functions, we no
longer modify the config value in ice_set_timestamp_mode. In turn, we
no longer want to copy that config value into the PF cached structure.
Instead, this is now the source of truth for actual configuration. On
success of ice_set_timestamp_mode, copy the real configured mode back to
report it out to userspace.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: introduce ice_base_incval function
Jacob Keller [Wed, 13 Oct 2021 15:47:20 +0000 (08:47 -0700)]
ice: introduce ice_base_incval function

A future change will add additional possible increment values for the
E822 device support. To handle this, we want to look up the increment
value to use instead of hard coding it to the nominal value for E810
devices. Introduce ice_base_incval as a function to get the best nominal
increment value to use.

For now, it just returns the E810 value, but will be refactored in the
future to look up the value based on the device type and configured
clock frequency.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoice: Fix E810 PTP reset flow
Karol Kolacinski [Mon, 20 Dec 2021 16:58:27 +0000 (08:58 -0800)]
ice: Fix E810 PTP reset flow

The PF reset does not reset PHC and PHY clocks so it's unnecessary to
stop them and reinitialize after the reset.
Configuring timestamping changes the VSI fields so it needs to be
performed after VSIs are initialized, which was not done in case of a
reset.

Suggested-by: Patrick Talbert <ptalbert@redhat.com>
Signed-off-by: Karol Kolacinski <karol.kolacinski@intel.com>
Tested-by: Pasi Vaananen <pvaanane@redhat.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
2 years agoMerge tag 'mac80211-next-for-net-next-2021-12-21' of git://git.kernel.org/pub/scm...
Jakub Kicinski [Tue, 21 Dec 2021 15:41:47 +0000 (07:41 -0800)]
Merge tag 'mac80211-next-for-net-next-2021-12-21' of git://git./linux/kernel/git/jberg/mac80211-next

Johannes Berg says:

====================
This time we have:
 * ndo_fill_forward_path support in mac80211, to let drivers use it
 * association comeback notification for userspace, to be able
   to react more sensibly to long delays
 * support for background radar detection hardware in some chipsets
 * SA Query Procedures offload on the AP side
 * more logging if we find problems with HT/VHT/HE
 * various cleanups and minor fixes

Conflicts:

net/wireless/reg.c:
  e08ebd6d7b90 ("cfg80211: Acquire wiphy mutex on regulatory work")
  701fdfe348f7 ("cfg80211: Enable regulatory enforcement checks for drivers supporting mesh iface")
  https://lore.kernel.org/r/20211221111950.57ecc6a7@canb.auug.org.au

drivers/net/wireless/ath/ath10k/wmi.c:
  7f599aeccbd2 ("cfg80211: Use the HE operation IE to determine a 6GHz BSS channel")
  3bf2537ec2e3 ("ath10k: drop beacon and probe response which leak from other channel")
  https://lore.kernel.org/r/20211221115004.1cd6b262@canb.auug.org.au

* tag 'mac80211-next-for-net-next-2021-12-21' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211-next: (32 commits)
  cfg80211: Enable regulatory enforcement checks for drivers supporting mesh iface
  rfkill: allow to get the software rfkill state
  cfg80211: refactor cfg80211_get_ies_channel_number()
  nl82011: clarify interface combinations wrt. channels
  nl80211: Add support to offload SA Query procedures for AP SME device
  nl80211: Add support to set AP settings flags with single attribute
  mac80211: add more HT/VHT/HE state logging
  cfg80211: Use the HE operation IE to determine a 6GHz BSS channel
  cfg80211: rename offchannel_chain structs to background_chain to avoid confusion with ETSI standard
  mac80211: Notify cfg80211 about association comeback
  cfg80211: Add support for notifying association comeback
  mac80211: introduce channel switch disconnect function
  cfg80211: Fix order of enum nl80211_band_iftype_attr documentation
  cfg80211: simplify cfg80211_chandef_valid()
  mac80211: Remove a couple of obsolete TODO
  mac80211: fix FEC flag in radio tap header
  mac80211: use coarse boottime for airtime fairness code
  ieee80211: change HE nominal packet padding value defines
  cfg80211: use ieee80211_bss_get_elem() instead of _get_ie()
  mac80211: Use memset_after() to clear tx status
  ...
====================

Link: https://lore.kernel.org/r/20211221112532.28708-1-johannes@sipsolutions.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/sched: use min() macro instead of doing it manually
Yang Li [Tue, 21 Dec 2021 01:14:55 +0000 (09:14 +0800)]
net/sched: use min() macro instead of doing it manually

Fix following coccicheck warnings:
./net/sched/cls_api.c:3333:17-18: WARNING opportunity for min()
./net/sched/cls_api.c:3389:17-18: WARNING opportunity for min()
./net/sched/cls_api.c:3427:17-18: WARNING opportunity for min()

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'net-amd-xgbe-add-support-for-yellow-carp-ethernet-device'
Jakub Kicinski [Tue, 21 Dec 2021 02:42:43 +0000 (18:42 -0800)]
Merge branch 'net-amd-xgbe-add-support-for-yellow-carp-ethernet-device'

Raju Rangoju says:

====================
net: amd-xgbe: Add support for Yellow Carp Ethernet device

Add support for newer version of Hardware, the Yellow Carp Ethernet device
====================

Link: https://lore.kernel.org/r/20211220135428.1123575-1-rrangoju@amd.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: amd-xgbe: Disable the CDR workaround path for Yellow Carp Devices
Raju Rangoju [Mon, 20 Dec 2021 13:54:28 +0000 (19:24 +0530)]
net: amd-xgbe: Disable the CDR workaround path for Yellow Carp Devices

Yellow Carp Ethernet devices do not require
Autonegotiation CDR workaround, hence disable the same.

Co-developed-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: amd-xgbe: Alter the port speed bit range
Raju Rangoju [Mon, 20 Dec 2021 13:54:27 +0000 (19:24 +0530)]
net: amd-xgbe: Alter the port speed bit range

Newer generation Hardware uses the slightly different
port speed bit widths, so alter the existing port speed
bit range to extend support to the newer generation hardware
while maintaining the backward compatibility with older
generation hardware.

The previously reserved bits are now being used which
then requires the adjustment to the BIT values, e.g.:

Before:
   PORT_PROPERTY_0[22:21] - Reserved
   PORT_PROPERTY_0[26:23] - Supported Speeds

After:
   PORT_PROPERTY_0[21] - Reserved
   PORT_PROPERTY_0[26:22] - Supported Speeds

To make this backwards compatible, the existing BIT
definitions for the port speeds are incremented by one
to maintain the original position.

Co-developed-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: amd-xgbe: Add Support for Yellow Carp Ethernet device
Raju Rangoju [Mon, 20 Dec 2021 13:54:26 +0000 (19:24 +0530)]
net: amd-xgbe: Add Support for Yellow Carp Ethernet device

Yellow Carp Ethernet devices use the existing PCI ID but
the window settings for the indirect PCS access have been
altered. Add the check for Yellow Carp Ethernet devices to
use the new register values.

Co-developed-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomctp: emit RTM_NEWADDR and RTM_DELADDR
Matt Johnston [Mon, 20 Dec 2021 02:31:04 +0000 (10:31 +0800)]
mctp: emit RTM_NEWADDR and RTM_DELADDR

Userspace can receive notification of MCTP address changes via
RTNLGRP_MCTP_IFADDR rtnetlink multicast group.

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
Link: https://lore.kernel.org/r/20211220023104.1965509-1-matt@codeconstruct.com.au
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'lan966x-switchdev-and-vlan'
David S. Miller [Mon, 20 Dec 2021 11:44:06 +0000 (11:44 +0000)]
Merge branch 'lan966x-switchdev-and-vlan'

Horatiu Vultur says:

====================
net: lan966x: Add switchdev and vlan support

This patch series extends lan966x with switchdev and vlan support.
The first patches just adds new registers and extend the MAC table to
handle the interrupts when a new address is learn/forget.

v7->v8:
- remove extra mac learn when the port leaves the bridge
- replace memcpy with ether_addr_copy
- change the order of operations in lan966x_switch_driver_init/exit
- refactor lan966x_port_bridge_flags

v6->v7:
- fix build issues when compiling as a module

v5->v6:
- fix issues with the singletones, they were not really singletons
- simplify the case where lan966x ports are added to bridges with foreign
  ports
- drop the cases NETDEV_PRE_UP and NETDEV_DOWN
- fix the change of MAC address
- drop the callbacks .ndo_set_features, .ndo_vlan_rx_add_vid,
  .ndo_vlan_rx_kill_vid
- remove duplicate code when port was added in a vlan, the MAC entries
  will be added by the fdb

v4->v5:
- make the notifier_block from lan966x to be singletones
- use switchdev_handle_port_obj_add and switchdev_handle_fdb_event_to_device
  when getting callbacks in the lan966x
- merge the two vlan patches in a single one

v3->v4:
- split the last patch in multiple patches
- replace spin_lock_irqsave/restore with spin_lock/spin_unlock
- remove lan966x_port_change_rx_flags because it was copying all the frames to
  the CPU instead of removing all RX filters.
- implement SWITCHDEV_ATTR_ID_PORT_PRE_BRIDGE_FLAGS
- remove calls to __dev_mc_unsync/sync as they are not needed
- replace 0/1 with false/true
- make sure that the lan966x ports are not added to bridges that have other
  interfaces except lan966x
- and allow the lan966x ports to be part of only the same bridge.

v2->v3:
- separate the PVID used when the port is in host mode or vlan unaware
- fix issue when the port was leaving the bridge

v1->v2:
- when allocating entries for the mac table use kzalloc instead of
  devm_kzalloc
- also use GFP_KERNEL instead of GFP_ATOMIC, because is never called
  in atomic context
- when deleting an mac table entry, the order of operations was wrong
- if ana irq is enabled make sure it gets disabled when the driver is
  removed
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: lan966x: Extend switchdev with fdb support
Horatiu Vultur [Sat, 18 Dec 2021 21:49:46 +0000 (22:49 +0100)]
net: lan966x: Extend switchdev with fdb support

Extend lan966x driver with fdb support by implementing the switchdev
calls SWITCHDEV_FDB_ADD_TO_DEVICE and SWITCHDEV_FDB_DEL_TO_DEVICE.

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: lan966x: Extend switchdev bridge flags
Horatiu Vultur [Sat, 18 Dec 2021 21:49:45 +0000 (22:49 +0100)]
net: lan966x: Extend switchdev bridge flags

Currently allow a port to be part or not of the multicast flooding mask.
By implementing the switchdev calls SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS
and SWITCHDEV_ATTR_ID_PORT_PRE_BRIDGE_FLAGS.

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: lan966x: Add vlan support.
Horatiu Vultur [Sat, 18 Dec 2021 21:49:44 +0000 (22:49 +0100)]
net: lan966x: Add vlan support.

Extend the driver to support vlan filtering  by implementing the
switchdev calls SWITCHDEV_OBJ_ID_PORT_VLAN,
SWITCHDEV_ATTR_ID_BRIDGE_VLAN_FILTERING

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: lan966x: Add support to offload the forwarding.
Horatiu Vultur [Sat, 18 Dec 2021 21:49:43 +0000 (22:49 +0100)]
net: lan966x: Add support to offload the forwarding.

This patch adds basic support to offload in the HW the forwarding of the
frames. The driver registers to the switchdev callbacks and implements
the callbacks for attributes SWITCHDEV_ATTR_ID_PORT_STP_STATE and
SWITCHDEV_ATTR_ID_BRIDGE_AGEING_TIME.
It is not allowed to add a lan966x port to a bridge that contains a
different interface than lan966x.

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: lan966x: Remove .ndo_change_rx_flags
Horatiu Vultur [Sat, 18 Dec 2021 21:49:42 +0000 (22:49 +0100)]
net: lan966x: Remove .ndo_change_rx_flags

The function lan966x_port_change_rx_flags() was used only when
IFF_PROMISC flag was set. In that case it was setting to copy all the
frames to the CPU instead of removing any RX filters. Therefore remove
it.

Fixes: d28d6d2e37d10d ("net: lan966x: add port module support")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: lan966x: More MAC table functionality
Horatiu Vultur [Sat, 18 Dec 2021 21:49:41 +0000 (22:49 +0100)]
net: lan966x: More MAC table functionality

This patch adds support for adding/removing mac entries in the SW list
of entries and in the HW table. This is used by the bridge
functionality.

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: lan966x: add support for interrupts from analyzer
Horatiu Vultur [Sat, 18 Dec 2021 21:49:40 +0000 (22:49 +0100)]
net: lan966x: add support for interrupts from analyzer

This patch adds support for handling the interrupts generated by the
analyzer. Currently, only the MAC table generates these interrupts.
The MAC table will generate an interrupt whenever it learns or forgets
an entry in the table. It is the SW responsibility figure out which
entries were added/removed.

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodt-bindings: net: lan966x: Extend with the analyzer interrupt
Horatiu Vultur [Sat, 18 Dec 2021 21:49:39 +0000 (22:49 +0100)]
dt-bindings: net: lan966x: Extend with the analyzer interrupt

Extend dt-bindings for lan966x with analyzer interrupt.
This interrupt can be generated for example when the HW learn/forgets
an entry in the MAC table.

Acked-by: Rob Herring <robh@kernel.org>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: lan966x: Add registers that are used for switch and vlan functionality
Horatiu Vultur [Sat, 18 Dec 2021 21:49:38 +0000 (22:49 +0100)]
net: lan966x: Add registers that are used for switch and vlan functionality

This patch adds the registers that will be used to enable switchdev and
vlan functionality in the HW.

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'mlxsw-devlink=health-reporter-extensions'
David S. Miller [Mon, 20 Dec 2021 11:32:22 +0000 (11:32 +0000)]
Merge branch 'mlxsw-devlink=health-reporter-extensions'

Ido Schimmel says:

====================
mlxsw: devlink health reporter extensions

This patchset extends the devlink health reporter registered by mlxsw to
report new health events and their related parameters. These are meant
to aid in debugging hardware and firmware issues.

Patches #1-#2 are preparations.

Patch #3 adds the definitions of the new events and parameters.

Patch #4 extends the health reporter to report the new events and
parameters.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core: Extend devlink health reporter with new events and parameters
Danielle Ratson [Mon, 20 Dec 2021 10:56:14 +0000 (12:56 +0200)]
mlxsw: core: Extend devlink health reporter with new events and parameters

Extend the devlink health reporter registered by mlxsw to report new
health events and their related parameters. These are meant to aid in
debugging of hardware / firmware issues.

Beside the test event ('MLXSW_REG_MFDE_EVENT_ID_TEST') that is triggered
following the devlink health 'test' sub-command, the new events are used
to report the triggering of asserts in firmware code
('MLXSW_REG_MFDE_EVENT_ID_FW_ASSERT') and hardware issues
('MLXSW_REG_MFDE_EVENT_ID_FATAL_CAUSE').

Each event is accompanied with a severity parameter and per-event
parameters that are meant to help root cause the detected issue.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: reg: Extend MFDE register with new events and parameters
Danielle Ratson [Mon, 20 Dec 2021 10:56:13 +0000 (12:56 +0200)]
mlxsw: reg: Extend MFDE register with new events and parameters

Extend the Monitoring Firmware Debug (MFDE) register with new events and
their related parameters. These events will be utilized by
devlink-health in the next patch.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core: Convert a series of if statements to switch case
Danielle Ratson [Mon, 20 Dec 2021 10:56:12 +0000 (12:56 +0200)]
mlxsw: core: Convert a series of if statements to switch case

Convert a series of if statements that handle different events to a
switch case statement. Encapsulate the per-event code in different
functions to simplify the code.

This is a preparation for subsequent patches that will add more events
that need to be handled.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: Fix naming convention of MFDE fields
Danielle Ratson [Mon, 20 Dec 2021 10:56:11 +0000 (12:56 +0200)]
mlxsw: Fix naming convention of MFDE fields

Currently, the MFDE register field names are using the convention:
reg_mfde_<NAME_OF_FIELD>, and do not consider the name of the MFDE
event.

Fix the field names so they fit the more accurate convention:
reg_mfde_<NAME_OF_EVENT>_<NAME_OF_FIELD>.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobnx2x: Invalidate fastpath HSI version for VFs
Manish Chopra [Fri, 17 Dec 2021 16:55:52 +0000 (08:55 -0800)]
bnx2x: Invalidate fastpath HSI version for VFs

Commit 0a6890b9b4df ("bnx2x: Utilize FW 7.13.15.0.")
added validation for fastpath HSI versions for different
client init which was not meant for SR-IOV VF clients, which
resulted in firmware asserts when running VF clients with
different fastpath HSI version.

This patch along with the new firmware support in patch #1
fixes this behavior in order to not validate fastpath HSI
version for the VFs.

Fixes: 0a6890b9b4df ("bnx2x: Utilize FW 7.13.15.0.")
Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: Prabhakar Kushwaha <pkushwaha@marvell.com>
Signed-off-by: Alok Prasad <palok@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobnx2x: Utilize firmware 7.13.21.0
Manish Chopra [Fri, 17 Dec 2021 16:55:51 +0000 (08:55 -0800)]
bnx2x: Utilize firmware 7.13.21.0

This new firmware addresses few important issues and enhancements
as mentioned below -

- Support direct invalidation of FP HSI Ver per function ID, required for
  invalidating FP HSI Ver prior to each VF start, as there is no VF start
- BRB hardware block parity error detection support for the driver
- Fix the FCOE underrun flow
- Fix PSOD during FCoE BFS over the NIC ports after preboot driver
- Maintains backward compatibility

This patch incorporates this new firmware 7.13.21.0 in bnx2x driver.

Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: Prabhakar Kushwaha <pkushwaha@marvell.com>
Signed-off-by: Alok Prasad <palok@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agocfg80211: Enable regulatory enforcement checks for drivers supporting mesh iface
Sriram R [Thu, 2 Dec 2021 01:38:40 +0000 (07:08 +0530)]
cfg80211: Enable regulatory enforcement checks for drivers supporting mesh iface

Currently cfg80211 checks for invalid channels whenever there is a
regulatory update and stops the active interfaces if it is operating on
an unsupported channel in the new regulatory domain.

This is done based on a regulatory flag REGULATORY_IGNORE_STALE_KICKOFF
set during wiphy registration which disables this enforcement when
unsupported interface modes are supported by driver.

Add support to enable this enforcement when Mesh Point interface type
is advertised by drivers.

Signed-off-by: Sriram R <quic_srirrama@quicinc.com>
Link: https://lore.kernel.org/r/1638409120-28997-1-git-send-email-quic_srirrama@quicinc.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agorfkill: allow to get the software rfkill state
Emmanuel Grumbach [Sun, 19 Dec 2021 19:51:24 +0000 (21:51 +0200)]
rfkill: allow to get the software rfkill state

iwlwifi needs to be able to differentiate between the
software rfkill state and the hardware rfkill state.

The reason for this is that iwlwifi needs to notify any
change in the software rfkill state even when it doesn't
own the device (which means even when the hardware rfkill
is asserted).

In order to be able to know the software rfkill when the
host does not own the device, iwlwifi needs to be able to
ask the state of the software rfkill ignoring the state
of the hardware rfkill.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Link: https://lore.kernel.org/r/20211219195124.125689-1-emmanuel.grumbach@intel.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agocfg80211: refactor cfg80211_get_ies_channel_number()
Johannes Berg [Thu, 2 Dec 2021 12:09:14 +0000 (13:09 +0100)]
cfg80211: refactor cfg80211_get_ies_channel_number()

Now that this is no longer part of the bigger function,
we can get rid of the channel_num variable. Also change
the function to use the struct element helpers, instead
of open-coding the element handling.

Link: https://lore.kernel.org/r/20211202130913.a0adf67a9319.I6db0340a34fff18d78e9cd512f4abf855da4e43a@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agonl82011: clarify interface combinations wrt. channels
Johannes Berg [Wed, 1 Dec 2021 10:09:25 +0000 (11:09 +0100)]
nl82011: clarify interface combinations wrt. channels

Thinking about MLD (Draft 802.11be) now, it's clear that
we'll have new limitations where different stations (or
links) must be on _different_ channels (or in different
frequency ranges even.) Clarify that the current limit of
multiple channels is a maximum and the same one is OK.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20211201110924.2816d91b6862.I2d997abd525574529f88e941d90aeb640dbb1abf@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agonl80211: Add support to offload SA Query procedures for AP SME device
Veerendranath Jakkam [Fri, 26 Nov 2021 07:25:19 +0000 (12:55 +0530)]
nl80211: Add support to offload SA Query procedures for AP SME device

Add a flag attribute to use in ap settings to indicate userspace
supports offloading of SA Query procedures to driver. Also add AP SME
device feature flag to advertise that the SA Query procedures offloaded
to driver when userspace indicates support for offloading of SA Query
procedures.

Driver handles SA Query procedures in driver's SME it self and skip
sending SA Query request or response frames to userspace when userspace
indicates support for SA Query procedures offload. But if userspace
doesn't advertise support for SA Query procedures offload driver shall
not offload SA Query procedures handling.

Also userspace with SA Query procedures offload capability shall skip SA
Query specific validations when driver indicates support for handling SA
Query procedures.

Signed-off-by: Veerendranath Jakkam <vjakkam@codeaurora.org>
Link: https://lore.kernel.org/r/1637911519-21306-2-git-send-email-vjakkam@codeaurora.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agonl80211: Add support to set AP settings flags with single attribute
Veerendranath Jakkam [Fri, 26 Nov 2021 07:25:18 +0000 (12:55 +0530)]
nl80211: Add support to set AP settings flags with single attribute

In previous method each AP settings flag is represented by a top-level
flag attribute and conversion to enum cfg80211_ap_settings_flags had to
be done before sending them to driver. This commit is to make it easier
to define new AP settings flags and sending them to driver.

This commit also deprecate sending of
%NL80211_ATTR_EXTERNAL_AUTH_SUPPORT in %NL80211_CMD_START_AP. But to
maintain backwards compatibility checks for
%NL80211_ATTR_EXTERNAL_AUTH_SUPPORT in %NL80211_CMD_START_AP when
%NL80211_ATTR_AP_SETTINGS_FLAGS not present in %NL80211_CMD_START_AP.

Signed-off-by: Veerendranath Jakkam <vjakkam@codeaurora.org>
Link: https://lore.kernel.org/r/1637911519-21306-1-git-send-email-vjakkam@codeaurora.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agomac80211: add more HT/VHT/HE state logging
Johannes Berg [Tue, 30 Nov 2021 11:16:59 +0000 (13:16 +0200)]
mac80211: add more HT/VHT/HE state logging

Add more logging in places that affect HT/VHT/HE state, so
things get easier to debug.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20211130131608.ac51d574458c.If197b45c5b31d2fbd254fa12c2d7c736f304d4ae@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agocfg80211: Use the HE operation IE to determine a 6GHz BSS channel
Ayala Beker [Thu, 2 Dec 2021 12:36:10 +0000 (14:36 +0200)]
cfg80211: Use the HE operation IE to determine a 6GHz BSS channel

A non-collocated AP whose primary channel is not a PSC channel
may transmit a duplicated beacon on the corresponding PSC channel
in which it would indicate its true primary channel.
Use this inforamtion contained in the HE operation IE to determine
the primary channel of the AP.
In case of invalid infomration ignore it and use the channel
the frame was received on.

Signed-off-by: Ayala Beker <ayala.beker@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20211202143322.71eb2176e54e.I130f678e4aa390973ab39d838bbfe7b2d54bff8e@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agocfg80211: rename offchannel_chain structs to background_chain to avoid confusion...
Lorenzo Bianconi [Mon, 29 Nov 2021 13:11:24 +0000 (14:11 +0100)]
cfg80211: rename offchannel_chain structs to background_chain to avoid confusion with ETSI standard

ETSI standard defines "Offchannel CAC" as:
"Off-Channel CAC is performed by a number of non-continuous checks
spread over a period in time. This period, which is required to
determine the presence of radar signals, is defined as the Off-Channel
CAC Time..
Minimum Off-Channel CAC Time 6 minutes and Maximum Off-Channel CAC Time
4 hours..".
mac80211 implementation refers to a dedicated hw chain used for continuous
radar monitoring. Rename offchannel_* references to background_* in
order to avoid confusion with ETSI standard.

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Link: https://lore.kernel.org/r/4204cc1d648d76b44557981713231e030a3bd991.1638190762.git.lorenzo@kernel.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agomac80211: Notify cfg80211 about association comeback
Ilan Peer [Mon, 29 Nov 2021 13:32:35 +0000 (15:32 +0200)]
mac80211: Notify cfg80211 about association comeback

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20211129152938.d76eac9e51ee.I986cffab95d51adfee6d84964711644392005113@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agocfg80211: Add support for notifying association comeback
Ilan Peer [Mon, 29 Nov 2021 13:32:34 +0000 (15:32 +0200)]
cfg80211: Add support for notifying association comeback

Thought the underline driver MLME can handle association temporal
rejection with comeback, it is still useful to notify this to
user space, as user space might want to handle the temporal
rejection differently. For example, in case the comeback time
is too long, user space can deauthenticate immediately and try
to associate with a different AP.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20211129152938.2467809e8cb3.I45574185b582666bc78eef0c29a4c36b478e5382@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agomac80211: introduce channel switch disconnect function
Nathan Errera [Mon, 29 Nov 2021 13:32:38 +0000 (15:32 +0200)]
mac80211: introduce channel switch disconnect function

Introduce a disconnect function that can be used when a
channel switch error occurs. The channel switch can request to
block the tx, and so, we need to make sure we do not send a deauth
frame in this case.

Signed-off-by: Nathan Errera <nathan.errera@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20211129152938.cd2a615a0702.I9edb14785586344af17644b610ab5be109dcef00@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agocfg80211: Fix order of enum nl80211_band_iftype_attr documentation
Ilan Peer [Mon, 29 Nov 2021 13:32:41 +0000 (15:32 +0200)]
cfg80211: Fix order of enum nl80211_band_iftype_attr documentation

And fix the comment.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20211129152938.4ef43aff0c5d.I96dcb743bcd4f387ba4cfaa61987aeb642ad762b@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agocfg80211: simplify cfg80211_chandef_valid()
Johannes Berg [Mon, 29 Nov 2021 13:32:43 +0000 (15:32 +0200)]
cfg80211: simplify cfg80211_chandef_valid()

There are a lot of duplicate checks in this function to
check the delta between the control channel and CF1.
With the addition of 320 MHz, this will become even more.
Simplify the code so that the common checks are done
only once for multiple bandwidths.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20211129152938.2d0240b07f11.I759e8e990f5386ba2b56ffb2488a8d4e16e22c1b@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agomac80211: Remove a couple of obsolete TODO
Ilan Peer [Mon, 29 Nov 2021 13:32:44 +0000 (15:32 +0200)]
mac80211: Remove a couple of obsolete TODO

The HE capability IE is an extension IE so remove
an irrelevant comments.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/iwlwifi.20211129152938.550b95b5fca7.Ia31395e880172aefcc0a8c70ed060f84b94bdb83@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agomac80211: fix FEC flag in radio tap header
P Praneesh [Tue, 30 Nov 2021 17:50:47 +0000 (23:20 +0530)]
mac80211: fix FEC flag in radio tap header

In mac80211, while building radiotap header
IEEE80211_RADIOTAP_MCS_HAVE_FEC flag is missing when LDPC enabled
from driver, hence LDPC is not updated properly in radiotap header.
Fix that by adding HAVE_FEC flag while building radiotap header.

Signed-off-by: P Praneesh <quic_ppranees@quicinc.com>
Link: https://lore.kernel.org/r/1638294648-844-2-git-send-email-quic_ppranees@quicinc.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agomac80211: use coarse boottime for airtime fairness code
Felix Fietkau [Fri, 17 Dec 2021 11:42:58 +0000 (12:42 +0100)]
mac80211: use coarse boottime for airtime fairness code

The time values used by the airtime fairness code only need to be accurate
enough to cover station activity detection.
Using ktime_get_coarse_boottime_ns instead of ktime_get_boottime_ns will
drop the accuracy down to jiffies intervals, but at the same time saves
a lot of CPU cycles in a hot path

Signed-off-by: Felix Fietkau <nbd@nbd.name>
Acked-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/r/20211217114258.14619-1-nbd@nbd.name
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
2 years agoMerge branch 'tc-action-offload'
David S. Miller [Sun, 19 Dec 2021 14:08:48 +0000 (14:08 +0000)]
Merge branch 'tc-action-offload'

Simon Horman says:

====================
allow user to offload tc action to net device

Baowen Zheng says:

Allow use of flow_indr_dev_register/flow_indr_dev_setup_offload to offload
tc actions independent of flows.

The motivation for this work is to prepare for using TC police action
instances to provide hardware offload of OVS metering feature - which calls
for policers that may be used by multiple flows and whose lifecycle is
independent of any flows that use them.

This patch includes basic changes to offload drivers to return EOPNOTSUPP
if this feature is used - it is not yet supported by any driver.

Tc cli command to offload and quote an action:

 # tc qdisc del dev $DEV ingress && sleep 1 || true
 # tc actions delete action police index 200 || true

 # tc qdisc add dev $DEV ingress
 # tc qdisc show dev $DEV ingress

 # tc actions add action police rate 100mbit burst 10000k index 200 skip_sw
 # tc -s -d actions list action police
 total acts 1

         action order 0:  police 0xc8 rate 100Mbit burst 10000Kb mtu 2Kb action reclassify
         overhead 0b linklayer ethernet
         ref 1 bind 0  installed 142 sec used 0 sec
         Action statistics:
         Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
         backlog 0b 0p requeues 0
         skip_sw in_hw in_hw_count 1
         used_hw_stats delayed

 # tc filter add dev $DEV protocol ip parent ffff: \
         flower skip_sw ip_proto tcp action police index 200
 # tc -s -d filter show dev $DEV protocol ip parent ffff:
 filter pref 49152 flower chain 0
 filter pref 49152 flower chain 0 handle 0x1
   eth_type ipv4
   ip_proto tcp
   skip_sw
   in_hw in_hw_count 1
         action order 1:  police 0xc8 rate 100Mbit burst 10000Kb mtu 2Kb action
         reclassify overhead 0b linklayer ethernet
         ref 2 bind 1  installed 300 sec used 0 sec
         Action statistics:
         Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
         backlog 0b 0p requeues 0
         skip_sw in_hw in_hw_count 1
         used_hw_stats delayed

 # tc filter add dev $DEV protocol ipv6 parent ffff: \
         flower skip_sw ip_proto tcp action police index 200
 # tc -s -d filter show dev $DEV protocol ipv6 parent ffff:
   filter pref 49151 flower chain 0
   filter pref 49151 flower chain 0 handle 0x1
   eth_type ipv6
   ip_proto tcp
   skip_sw
   in_hw in_hw_count 1
         action order 1:  police 0xc8 rate 100Mbit burst 10000Kb mtu 2Kb action
         reclassify overhead 0b linklayer ethernet
         ref 3 bind 2  installed 761 sec used 0 sec
         Action statistics:
         Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
         backlog 0b 0p requeues 0
         skip_sw in_hw in_hw_count 1
         used_hw_stats delayed

 # tc -s -d actions list action police
 total acts 1

          action order 0:  police 0xc8 rate 100Mbit burst 10000Kb mtu 2Kb action reclassify overhead 0b linklayer ethernet
          ref 3 bind 2  installed 917 sec used 0 sec
          Action statistics:
          Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
          backlog 0b 0p requeues 0
          skip_sw in_hw in_hw_count 1
         used_hw_stats delayed

Changes between v7 and v8:
* Rename enum offload_act_command as enum flow_offload_act_command
* Add a single patch to rename exts stats update function for readability.
* Fix the kernel test robot issue reported by Oliver Sang.

Changes between v6 and v7:
* Add a single patch to rename offload functions with offload for readability.
* Post 166b6a46b78b ("flow_offload: return EOPNOTSUPP for the unsupported mpls action type") as a bug fix to netdev.
* Rename enum flow_act_command as enum flow_offload_act_command
* Rename the new added action offload setup ops with offload instead of flow
* Rename the new added action offload function with offload instead of flow.
* Add more selftest cases for validate filter and actions.

Changes between v5 and v6:
* Fix issue reported by Dan Carpenter found using Smatch.

Changes beteeen v4 and v5:
* Made changes of code style according to the public review comments.
* Add a fix for unsupported mpls action type in flow action setup stage.
* Add ops to tc_action_ops for flow action setup to facilitate
  adding a standalone action module.
* Add notification process when deleting action in reoffload process.

Changes between v3 and v4:
* Made changes according to the public review comments.
* Validate flags inside tcf_action_init() instead of creating new
  tcf_exts_validate_actions() function.
* Exactly match when validating flags of actions and filters.
* Add index to flow_action_entry for driver to identify actions.

Changes between v2 and v3:
* Made changes according to the review comments.
* Delete in_hw and not_in_hw flag and user can judge if the action is
  offloaded to any hardware by in_hw_count.
* Split the main patch of the action offload to three single patch to
  facilitate code review.

Changes between v1 and v2:
* Add the skip_hw/skip_sw for user to specify if the action should be in
  hardware or software.
* Fix issue of sleeping function called from invalid context.
* Change the action offload/delete from batch to one by one.
* Add some parameters to the netlink message for user space to look up
  the offload status of the actions.
* Add reoffload process to update action hw_count when driver is inserted
  or removed.

Changes between v1 and RFC:
* Fix robot test failure.
* Change actions offload process in action add function rather than action
  init.
* Change actions offload delete process after tcf_del_notify to keep
  undeleted actions.
* Add process to update actions stats from hardware.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: tc-testing: add action offload selftest for action and filter
Baowen Zheng [Fri, 17 Dec 2021 18:16:29 +0000 (19:16 +0100)]
selftests: tc-testing: add action offload selftest for action and filter

Add selftest cases in action police with skip_hw.
Add selftest case to validate flags of filter and action.
These tests depend on corresponding iproute2 command support.

Signed-off-by: Baowen Zheng <baowen.zheng@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoflow_offload: validate flags of filter and actions
Baowen Zheng [Fri, 17 Dec 2021 18:16:28 +0000 (19:16 +0100)]
flow_offload: validate flags of filter and actions

Add process to validate flags of filter and actions when adding
a tc filter.

We need to prevent adding filter with flags conflicts with its actions.

Signed-off-by: Baowen Zheng <baowen.zheng@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoflow_offload: add reoffload process to update hw_count
Baowen Zheng [Fri, 17 Dec 2021 18:16:27 +0000 (19:16 +0100)]
flow_offload: add reoffload process to update hw_count

Add reoffload process to update hw_count when driver
is inserted or removed.

We will delete the action if it is with skip_sw flag and
not offloaded to any hardware in reoffload process.

When reoffloading actions, we still offload the actions
that are added independent of filters.

Signed-off-by: Baowen Zheng <baowen.zheng@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: sched: save full flags for tc action
Baowen Zheng [Fri, 17 Dec 2021 18:16:26 +0000 (19:16 +0100)]
net: sched: save full flags for tc action

Save full action flags and return user flags when return flags to
user space.

Save full action flags to distinguish if the action is created
independent from classifier.

We made this change mainly for further patch to reoffload tc actions.

Signed-off-by: Baowen Zheng <baowen.zheng@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoflow_offload: add process to update action stats from hardware
Baowen Zheng [Fri, 17 Dec 2021 18:16:25 +0000 (19:16 +0100)]
flow_offload: add process to update action stats from hardware

When collecting stats for actions update them using both
hardware and software counters.

Stats update process should not run in context of preempt_disable.

Signed-off-by: Baowen Zheng <baowen.zheng@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoflow_offload: rename exts stats update functions with hw
Baowen Zheng [Fri, 17 Dec 2021 18:16:24 +0000 (19:16 +0100)]
flow_offload: rename exts stats update functions with hw

Rename exts stats update functions with hw for readability.

We make this change also to update stats from hw for an action
when it is offloaded to hw as a single action.

Signed-off-by: Baowen Zheng <baowen.zheng@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoflow_offload: add skip_hw and skip_sw to control if offload the action
Baowen Zheng [Fri, 17 Dec 2021 18:16:23 +0000 (19:16 +0100)]
flow_offload: add skip_hw and skip_sw to control if offload the action

We add skip_hw and skip_sw for user to control if offload the action
to hardware.

We also add in_hw_count for user to indicate if the action is offloaded
to any hardware.

Signed-off-by: Baowen Zheng <baowen.zheng@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoflow_offload: allow user to offload tc action to net device
Baowen Zheng [Fri, 17 Dec 2021 18:16:22 +0000 (19:16 +0100)]
flow_offload: allow user to offload tc action to net device

Use flow_indr_dev_register/flow_indr_dev_setup_offload to
offload tc action.

We need to call tc_cleanup_flow_action to clean up tc action entry since
in tc_setup_action, some actions may hold dev refcnt, especially the mirror
action.

Signed-off-by: Baowen Zheng <baowen.zheng@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>