linux-2.6-microblaze.git
2 years agodocs: ctucanfd: CTU CAN FD open-source IP core documentation.
Pavel Pisa [Tue, 22 Mar 2022 07:21:09 +0000 (08:21 +0100)]
docs: ctucanfd: CTU CAN FD open-source IP core documentation.

CTU CAN FD IP core documentation based on Martin Jeřábek's diploma theses
Open-source and Open-hardware CAN FD Protocol Support
https://dspace.cvut.cz/handle/10467/80366
.

Link: https://lore.kernel.org/all/692b965999ff6c272239df0fe1c76b68d02b134d.1647932262.git.pisa@cmp.felk.cvut.cz
Signed-off-by: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Signed-off-by: Martin Jerabek <martin.jerabek01@gmail.com>
Signed-off-by: Ondrej Ille <ondrej.ille@gmail.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agocan: ctucanfd: CTU CAN FD open-source IP core - platform/SoC support.
Pavel Pisa [Mon, 21 Mar 2022 23:32:32 +0000 (00:32 +0100)]
can: ctucanfd: CTU CAN FD open-source IP core - platform/SoC support.

Platform bus adaptation for CTU CAN FD open-source IP core.

The core has been tested together with OpenCores SJA1000
modified to be CAN FD frames tolerant on MicroZed Zynq based
MZ_APO education kits designed by Petr Porazil from PiKRON.com
company. FPGA design

  https://gitlab.fel.cvut.cz/canbus/zynq/zynq-can-sja1000-top.

The kit description at the Computer Architectures course pages

  https://cw.fel.cvut.cz/wiki/courses/b35apo/documentation/mz_apo/start .

Kit carrier board and mechanics design source files

  https://gitlab.com/pikron/projects/mz_apo/microzed_apo

The work is documented in Martin Jeřábek's diploma theses
Open-source and Open-hardware CAN FD Protocol Support

  https://dspace.cvut.cz/handle/10467/80366
.

Link: https://lore.kernel.org/all/4d5c53499bafe7717815f948801bd5aedaa05c12.1647904780.git.pisa@cmp.felk.cvut.cz
Signed-off-by: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Signed-off-by: Martin Jerabek <martin.jerabek01@gmail.com>
Signed-off-by: Ondrej Ille <ondrej.ille@gmail.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agocan: ctucanfd: CTU CAN FD open-source IP core - PCI bus support.
Pavel Pisa [Mon, 21 Mar 2022 23:32:31 +0000 (00:32 +0100)]
can: ctucanfd: CTU CAN FD open-source IP core - PCI bus support.

PCI bus adaptation for CTU CAN FD open-source IP core.

The project providing FPGA design for Intel EP4CGX15 based DB4CGX15
PCIe board with PiKRON.com designed transceiver riser shield is available
at https://gitlab.fel.cvut.cz/canbus/pcie-ctucanfd .

Link: https://lore.kernel.org/all/a81333e206a9bcf9434797f6f54d8664775542e2.1647904780.git.pisa@cmp.felk.cvut.cz
Signed-off-by: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Signed-off-by: Martin Jerabek <martin.jerabek01@gmail.com>
Signed-off-by: Ondrej Ille <ondrej.ille@gmail.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agocan: ctucanfd: add support for CTU CAN FD open-source IP core - bus independent part.
Martin Jerabek [Mon, 21 Mar 2022 23:32:30 +0000 (00:32 +0100)]
can: ctucanfd: add support for CTU CAN FD open-source IP core - bus independent part.

This driver adds support for the CTU CAN FD open-source IP core.
More documentation and core sources at project page
(https://gitlab.fel.cvut.cz/canbus/ctucanfd_ip_core).
The core integration to Xilinx Zynq system as platform driver
is available (https://gitlab.fel.cvut.cz/canbus/zynq/zynq-can-sja1000-top).
Implementation on Intel FPGA based PCI Express board is available
from project (https://gitlab.fel.cvut.cz/canbus/pcie-ctucanfd).

More about CAN bus related projects used and developed at CTU FEE at
https://canbus.pages.fel.cvut.cz/ .

Link: https://lore.kernel.org/all/1906e4941560ae2ce4b8d181131fd4963aa31611.1647904780.git.pisa@cmp.felk.cvut.cz
Signed-off-by: Martin Jerabek <martin.jerabek01@gmail.com>
Signed-off-by: Ondrej Ille <ondrej.ille@gmail.com>
Signed-off-by: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agodt-bindings: net: can: binding for CTU CAN FD open-source IP core.
Pavel Pisa [Mon, 21 Mar 2022 23:32:29 +0000 (00:32 +0100)]
dt-bindings: net: can: binding for CTU CAN FD open-source IP core.

The device-tree bindings for open-source/open-hardware CAN FD IP core
designed at the Czech Technical University in Prague.

CTU CAN FD IP core and other CTU CAN bus related projects
listing and documentation page

   http://canbus.pages.fel.cvut.cz/

Link: https://lore.kernel.org/all/c5a37fc470ae065b21e79caa65863539393c0d7c.1647904780.git.pisa@cmp.felk.cvut.cz
Signed-off-by: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agodt-bindings: vendor-prefix: add prefix for the Czech Technical University in Prague.
Pavel Pisa [Mon, 21 Mar 2022 23:32:28 +0000 (00:32 +0100)]
dt-bindings: vendor-prefix: add prefix for the Czech Technical University in Prague.

The Czech Technical University in Prague (CTU) is one of
the biggest and oldest (founded 1707) technical universities
in Europe. The abbreviation in Czech language is ČVUT according
to official name in Czech language

  České vysoké učení technické v Praze

The English translation

  The Czech Technical University in Prague

The university pages in English

  https://www.cvut.cz/en

Link: https://lore.kernel.org/all/ff3a7216114fcd83530e70b994ef0e4277ddf000.1647904780.git.pisa@cmp.felk.cvut.cz
Signed-off-by: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Acked-by: Rob Herring <robh@kernel.org>
Acked-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agocan: mcp251xfd: add support for mcp251863
Marc Kleine-Budde [Tue, 22 Mar 2022 08:19:13 +0000 (09:19 +0100)]
can: mcp251xfd: add support for mcp251863

The MCP251863 device is a CAN-FD controller (MCP2518FD) with an
integrated transceiver (ATA6563). This patch add support for the new
device.

Link: https://lore.kernel.org/all/20220419072805.2840340-3-mkl@pengutronix.de
Cc: Thomas Kopp <thomas.kopp@microchip.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agodt-binding: can: mcp251xfd: add binding information for mcp251863
Marc Kleine-Budde [Tue, 22 Mar 2022 08:19:13 +0000 (09:19 +0100)]
dt-binding: can: mcp251xfd: add binding information for mcp251863

The MCP251863 device is a CAN-FD controller (MCP2518FD) with an
integrated Transceiver (ATA6563). Add the microchip,mcp251863 as a new
compatible to the binding.

Link: https://lore.kernel.org/all/20220419072805.2840340-2-mkl@pengutronix.de
Cc: devicetree@vger.kernel.org
Cc: Thomas Kopp <thomas.kopp@microchip.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agodt-bindings: can: renesas,rcar-canfd: document r8a77961 support
Wolfram Sang [Fri, 1 Apr 2022 15:37:43 +0000 (17:37 +0200)]
dt-bindings: can: renesas,rcar-canfd: document r8a77961 support

This patch adds documentation for the r8a77961 to the
renesas,rcar-canfd binding.

Link: https://lore.kernel.org/all/20220401153743.77871-1-wsa+renesas@sang-engineering.com
Cc: devicetree@vger.kernel.org
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agocan: xilinx_can: mark bit timing constants as const
Marc Kleine-Budde [Thu, 17 Mar 2022 20:29:07 +0000 (21:29 +0100)]
can: xilinx_can: mark bit timing constants as const

This patch marks the bit timing constants as const.

Fixes: c223da689324 ("can: xilinx_can: Add support for CANFD FD frames")
Link: https://lore.kernel.org/all/20220317203119.792552-1-mkl@pengutronix.de
Cc: Appana Durga Kedareswara rao <appana.durga.rao@xilinx.com>
Cc: Naga Sureshkumar Relli <naga.sureshkumar.relli@xilinx.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agoMAINTAINERS: rectify entry for XILINX CAN DRIVER
Lukas Bulwahn [Mon, 21 Mar 2022 12:28:40 +0000 (13:28 +0100)]
MAINTAINERS: rectify entry for XILINX CAN DRIVER

Commit 7843d3c8e5e6 ("dt-bindings: can: xilinx_can: Convert Xilinx CAN
binding to YAML") converts xilinx_can.txt to xilinx,can.yaml, but
missed to adjust its reference in MAINTAINERS.

Hence, ./scripts/get_maintainer.pl --self-test=patterns complains
about a broken reference.

Repair this file reference in XILINX CAN DRIVER.

Fixes: 7843d3c8e5e6 ("dt-bindings: can: xilinx_can: Convert Xilinx CAN binding to YAML")
Link: https://lore.kernel.org/all/20220321122840.17841-1-lukas.bulwahn@gmail.com
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agocan: flexcan: using pm_runtime_resume_and_get instead of pm_runtime_get_sync
Minghao Chi [Tue, 19 Apr 2022 08:14:49 +0000 (08:14 +0000)]
can: flexcan: using pm_runtime_resume_and_get instead of pm_runtime_get_sync

Using pm_runtime_resume_and_get is more appropriate
for simplifing code

Link: https://lore.kernel.org/all/20220419081449.2574026-1-chi.minghao@zte.com.cn
Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agocan: mscan: mpc5xxx_can: Prepare cleanup of powerpc's asm/prom.h
Christophe Leroy [Fri, 1 Apr 2022 17:21:20 +0000 (19:21 +0200)]
can: mscan: mpc5xxx_can: Prepare cleanup of powerpc's asm/prom.h

powerpc's asm/prom.h brings some headers that it doesn't need itself.

In order to clean it up, first add missing headers in users of
asm/prom.h

Link: https://lore.kernel.org/all/878888f9057ad2f66ca0621a0007472bf57f3e3d.1648833432.git.christophe.leroy@csgroup.eu
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agocan: Fix Links to Technologic Systems web resources
Kris Bahnsen [Tue, 29 Mar 2022 20:12:29 +0000 (13:12 -0700)]
can: Fix Links to Technologic Systems web resources

Technologic Systems has rebranded as embeddedTS with the current
domain eventually going offline. Update web/doc URLs to correct
resource locations.

Link: https://lore.kernel.org/all/20220329201229.16279-1-kris@embeddedTS.com
Signed-off-by: Kris Bahnsen <kris@embeddedTS.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agocan: bittiming: can_calc_bittiming(): prefer small bit rate pre-scalers over larger...
Marc Kleine-Budde [Fri, 18 Mar 2022 13:58:44 +0000 (14:58 +0100)]
can: bittiming: can_calc_bittiming(): prefer small bit rate pre-scalers over larger ones

The CiA (CAN in Automation) lists in their Newsletter 1/2018 in the
"Recommendation for the CAN FD bit-timing" [1] article several
recommendations, one of them is:

| Recommendation 3: Choose BRPA and BRPD as low as possible

[1] https://can-newsletter.org/uploads/media/raw/f6a36d1461371a2f86ef0011a513712c.pdf

With the current bit timing algorithm Srinivas Neeli noticed that on
the Xilinx Versal ACAP board the CAN data bit timing parameters are
not calculated optimally. For most bit rates, the bit rate
prescaler (BRP) is != 1, although it's possible to configure the
requested with a bit rate with a prescaler of 1:

| Data Bit timing parameters for xilinx_can_fd2i with 79.999999 MHz ref clock (cmd-line) using algo 'v4.8'
|  nominal                                  real  Bitrt    nom   real  SampP
|  Bitrate TQ[ns] PrS PhS1 PhS2 SJW BRP  Bitrate  Error  SampP  SampP  Error
12000000     12   2    2    2   1   1 11428571   4.8%  75.0%  71.4%   4.8%
10000000     25   1    1    1   1   2  9999999   0.0%  75.0%  75.0%   0.0%
|  8000000     12   3    3    3   1   1  7999999   0.0%  75.0%  70.0%   6.7%
|  5000000     50   1    1    1   1   4  4999999   0.0%  75.0%  75.0%   0.0%
|  4000000     62   1    1    1   1   5  3999999   0.0%  75.0%  75.0%   0.0%
|  2000000    125   1    1    1   1  10  1999999   0.0%  75.0%  75.0%   0.0%
|  1000000    250   1    1    1   1  20   999999   0.0%  75.0%  75.0%   0.0%

The bit timing parameter calculation algorithm iterates effectively
from low to high BRP values. It selects a new best parameter set, if
the sample point error of the current parameter set is equal or less
to old best parameter set.

If the given hardware constraints (clock rate and bit timing parameter
constants) don't allow a sample point error of 0, the algorithm will
first find a valid bit timing parameter set with a low BRP, but then
will accept parameter sets with higher BRPs that have the same sample
point error.

This patch changes the algorithm to only accept a new parameter set,
if the resulting sample point error is lower. This leads to the
following data bit timing parameter for the Versal ACAP board:

| Data Bit timing parameters for xilinx_can_fd2i with 79.999999 MHz ref clock (cmd-line) using algo 'can-next'
|  nominal                                  real  Bitrt    nom   real  SampP
|  Bitrate TQ[ns] PrS PhS1 PhS2 SJW BRP  Bitrate  Error  SampP  SampP  Error
12000000     12   2    2    2   1   1 11428571   4.8%  75.0%  71.4%   4.8%
10000000     12   2    3    2   1   1  9999999   0.0%  75.0%  75.0%   0.0%
|  8000000     12   3    3    3   1   1  7999999   0.0%  75.0%  70.0%   6.7%
|  5000000     12   5    6    4   1   1  4999999   0.0%  75.0%  75.0%   0.0%
|  4000000     12   7    7    5   1   1  3999999   0.0%  75.0%  75.0%   0.0%
|  2000000     12  14   15   10   1   1  1999999   0.0%  75.0%  75.0%   0.0%
|  1000000     25  14   15   10   1   2   999999   0.0%  75.0%  75.0%   0.0%

Note: Due to HW constraints a data bit rate of 1 MBit/s with BRP = 1 is not possible.

Link: https://lore.kernel.org/all/20220318144913.873614-1-mkl@pengutronix.de
Link: https://lore.kernel.org/all/20220113203004.jf2rqj2pirhgx72i@pengutronix.de
Cc: Srinivas Neeli <sneeli@xilinx.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agocan: rx-offload: rename can_rx_offload_queue_sorted() -> can_rx_offload_queue_timestamp()
Marc Kleine-Budde [Mon, 14 Mar 2022 22:09:10 +0000 (23:09 +0100)]
can: rx-offload: rename can_rx_offload_queue_sorted() -> can_rx_offload_queue_timestamp()

This patch renames the function can_rx_offload_queue_sorted() to
can_rx_offload_queue_timestamp(). This better describes what the
function does, it adds a newly RX'ed skb to the sorted queue by its
timestamp.

Link: https://lore.kernel.org/all/20220417194327.2699059-1-mkl@pengutronix.de
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agoMerge branch 'rtnetlink-improve-alt_ifname-config-and-fix-dangerous-group-usage'
Paolo Abeni [Tue, 19 Apr 2022 11:39:01 +0000 (13:39 +0200)]
Merge branch 'rtnetlink-improve-alt_ifname-config-and-fix-dangerous-group-usage'

Florent Fourcot says:

====================
rtnetlink: improve ALT_IFNAME config and fix dangerous GROUP usage

First commit forbids dangerous calls when both IFNAME and GROUP are
given, since it can introduce unexpected behaviour when IFNAME does not
match any interface.

Second patch achieves primary goal of this patchset to fix/improve
IFLA_ALT_IFNAME attribute, since previous code was never working for
newlink/setlink. ip-link command is probably getting interface index
before, and was not using this feature.

Last two patches are improving error code on corner cases.

Changes in v2:
  * Remove ifname argument in rtnl_dev_get/do_setlink
    functions (simplify code)
  * Use a boolean to avoid condition duplication in __rtnl_newlink

Changes in v3:
  * Simplify rtnl_dev_get signature

Changes in v4:
  * Rename link_lookup to link_specified

Changes in v5:
  * Re-order patches
====================

Link: https://lore.kernel.org/r/20220415165330.10497-1-florent.fourcot@wifirst.fr
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agortnetlink: return EINVAL when request cannot succeed
Florent Fourcot [Fri, 15 Apr 2022 16:53:30 +0000 (18:53 +0200)]
rtnetlink: return EINVAL when request cannot succeed

A request without interface name/interface index/interface group cannot
work. We should return EINVAL

Signed-off-by: Florent Fourcot <florent.fourcot@wifirst.fr>
Signed-off-by: Brian Baboch <brian.baboch@wifirst.fr>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agortnetlink: return ENODEV when IFLA_ALT_IFNAME is used in dellink
Florent Fourcot [Fri, 15 Apr 2022 16:53:29 +0000 (18:53 +0200)]
rtnetlink: return ENODEV when IFLA_ALT_IFNAME is used in dellink

If IFLA_ALT_IFNAME is set and given interface is not found,
we should return ENODEV and be consistent with IFLA_IFNAME
behaviour
This commit extends feature of commit 76c9ac0ee878,
"net: rtnetlink: add possibility to use alternative names as message handle"

CC: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Florent Fourcot <florent.fourcot@wifirst.fr>
Signed-off-by: Brian Baboch <brian.baboch@wifirst.fr>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agortnetlink: enable alt_ifname for setlink/newlink
Florent Fourcot [Fri, 15 Apr 2022 16:53:28 +0000 (18:53 +0200)]
rtnetlink: enable alt_ifname for setlink/newlink

buffer called "ifname" given in function rtnl_dev_get
is always valid when called by setlink/newlink,
but contains only empty string when IFLA_IFNAME is not given. So
IFLA_ALT_IFNAME is always ignored

This patch fixes rtnl_dev_get function with a remove of ifname argument,
and move ifname copy in do_setlink when required.

It extends feature of commit 76c9ac0ee878,
"net: rtnetlink: add possibility to use alternative names as message
handle""

CC: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Florent Fourcot <florent.fourcot@wifirst.fr>
Signed-off-by: Brian Baboch <brian.baboch@wifirst.fr>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agortnetlink: return ENODEV when ifname does not exist and group is given
Florent Fourcot [Fri, 15 Apr 2022 16:53:27 +0000 (18:53 +0200)]
rtnetlink: return ENODEV when ifname does not exist and group is given

When the interface does not exist, and a group is given, the given
parameters are being set to all interfaces of the given group. The given
IFNAME/ALT_IF_NAME are being ignored in that case.

That can be dangerous since a typo (or a deleted interface) can produce
weird side effects for caller:

Case 1:

 IFLA_IFNAME=valid_interface
 IFLA_GROUP=1
 MTU=1234

Case 1 will update MTU and group of the given interface "valid_interface".

Case 2:

 IFLA_IFNAME=doesnotexist
 IFLA_GROUP=1
 MTU=1234

Case 2 will update MTU of all interfaces in group 1. IFLA_IFNAME is
ignored in this case

This behaviour is not consistent and dangerous. In order to fix this issue,
we now return ENODEV when the given IFNAME does not exist.

Signed-off-by: Florent Fourcot <florent.fourcot@wifirst.fr>
Signed-off-by: Brian Baboch <brian.baboch@wifirst.fr>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoMerge branch 'net-sched-allow-user-to-select-txqueue'
Paolo Abeni [Tue, 19 Apr 2022 10:20:48 +0000 (12:20 +0200)]
Merge branch 'net-sched-allow-user-to-select-txqueue'

Tonghao Zhang says:

====================
net: sched: allow user to select txqueue

From: Tonghao Zhang <xiangxia.m.yue@gmail.com>

Patch 1 allow user to select txqueue in clsact hook.
Patch 2 support skbhash to select txqueue.
====================

Link: https://lore.kernel.org/r/20220415164046.26636-1-xiangxia.m.yue@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: sched: support hash selecting tx queue
Tonghao Zhang [Fri, 15 Apr 2022 16:40:46 +0000 (00:40 +0800)]
net: sched: support hash selecting tx queue

This patch allows users to pick queue_mapping, range
from A to B. Then we can load balance packets from A
to B tx queue. The range is an unsigned 16bit value
in decimal format.

$ tc filter ... action skbedit queue_mapping skbhash A B

"skbedit queue_mapping QUEUE_MAPPING" (from "man 8 tc-skbedit")
is enhanced with flags: SKBEDIT_F_TXQ_SKBHASH

  +----+      +----+      +----+
  | P1 |      | P2 |      | Pn |
  +----+      +----+      +----+
    |           |           |
    +-----------+-----------+
                |
                | clsact/skbedit
                |      MQ
                v
    +-----------+-----------+
    | q0        | qn        | qm
    v           v           v
  HTB/FQ       FIFO   ...  FIFO

For example:
If P1 sends out packets to different Pods on other host, and
we want distribute flows from qn - qm. Then we can use skb->hash
as hash.

setup commands:
$ NETDEV=eth0
$ ip netns add n1
$ ip link add ipv1 link $NETDEV type ipvlan mode l2
$ ip link set ipv1 netns n1
$ ip netns exec n1 ifconfig ipv1 2.2.2.100/24 up

$ tc qdisc add dev $NETDEV clsact
$ tc filter add dev $NETDEV egress protocol ip prio 1 \
        flower skip_hw src_ip 2.2.2.100 action skbedit queue_mapping skbhash 2 6
$ tc qdisc add dev $NETDEV handle 1: root mq
$ tc qdisc add dev $NETDEV parent 1:1 handle 2: htb
$ tc class add dev $NETDEV parent 2: classid 2:1 htb rate 100kbit
$ tc class add dev $NETDEV parent 2: classid 2:2 htb rate 200kbit
$ tc qdisc add dev $NETDEV parent 1:2 tbf rate 100mbit burst 100mb latency 1
$ tc qdisc add dev $NETDEV parent 1:3 pfifo
$ tc qdisc add dev $NETDEV parent 1:4 pfifo
$ tc qdisc add dev $NETDEV parent 1:5 pfifo
$ tc qdisc add dev $NETDEV parent 1:6 pfifo
$ tc qdisc add dev $NETDEV parent 1:7 pfifo

$ ip netns exec n1 iperf3 -c 2.2.2.1 -i 1 -t 10 -P 10

pick txqueue from 2 - 6:
$ ethtool -S $NETDEV | grep -i tx_queue_[0-9]_bytes
     tx_queue_0_bytes: 42
     tx_queue_1_bytes: 0
     tx_queue_2_bytes: 11442586444
     tx_queue_3_bytes: 7383615334
     tx_queue_4_bytes: 3981365579
     tx_queue_5_bytes: 3983235051
     tx_queue_6_bytes: 6706236461
     tx_queue_7_bytes: 42
     tx_queue_8_bytes: 0
     tx_queue_9_bytes: 0

txqueues 2 - 6 are mapped to classid 1:3 - 1:7
$ tc -s class show dev $NETDEV
...
class mq 1:3 root leaf 8002:
 Sent 11949133672 bytes 7929798 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
class mq 1:4 root leaf 8003:
 Sent 7710449050 bytes 5117279 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
class mq 1:5 root leaf 8004:
 Sent 4157648675 bytes 2758990 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
class mq 1:6 root leaf 8005:
 Sent 4159632195 bytes 2759990 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
class mq 1:7 root leaf 8006:
 Sent 7003169603 bytes 4646912 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
...

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Jonathan Lemon <jonathan.lemon@gmail.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Alexander Lobakin <alobakin@pm.me>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Talal Ahmad <talalahmad@google.com>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Cc: Antoine Tenart <atenart@kernel.org>
Cc: Wei Wang <weiwan@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
Reviewed-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: sched: use queue_mapping to pick tx queue
Tonghao Zhang [Fri, 15 Apr 2022 16:40:45 +0000 (00:40 +0800)]
net: sched: use queue_mapping to pick tx queue

This patch fixes issue:
* If we install tc filters with act_skbedit in clsact hook.
  It doesn't work, because netdev_core_pick_tx() overwrites
  queue_mapping.

  $ tc filter ... action skbedit queue_mapping 1

And this patch is useful:
* We can use FQ + EDT to implement efficient policies. Tx queues
  are picked by xps, ndo_select_queue of netdev driver, or skb hash
  in netdev_core_pick_tx(). In fact, the netdev driver, and skb
  hash are _not_ under control. xps uses the CPUs map to select Tx
  queues, but we can't figure out which task_struct of pod/containter
  running on this cpu in most case. We can use clsact filters to classify
  one pod/container traffic to one Tx queue. Why ?

  In containter networking environment, there are two kinds of pod/
  containter/net-namespace. One kind (e.g. P1, P2), the high throughput
  is key in these applications. But avoid running out of network resource,
  the outbound traffic of these pods is limited, using or sharing one
  dedicated Tx queues assigned HTB/TBF/FQ Qdisc. Other kind of pods
  (e.g. Pn), the low latency of data access is key. And the traffic is not
  limited. Pods use or share other dedicated Tx queues assigned FIFO Qdisc.
  This choice provides two benefits. First, contention on the HTB/FQ Qdisc
  lock is significantly reduced since fewer CPUs contend for the same queue.
  More importantly, Qdisc contention can be eliminated completely if each
  CPU has its own FIFO Qdisc for the second kind of pods.

  There must be a mechanism in place to support classifying traffic based on
  pods/container to different Tx queues. Note that clsact is outside of Qdisc
  while Qdisc can run a classifier to select a sub-queue under the lock.

  In general recording the decision in the skb seems a little heavy handed.
  This patch introduces a per-CPU variable, suggested by Eric.

  The xmit.skip_txqueue flag is firstly cleared in __dev_queue_xmit().
  - Tx Qdisc may install that skbedit actions, then xmit.skip_txqueue flag
    is set in qdisc->enqueue() though tx queue has been selected in
    netdev_tx_queue_mapping() or netdev_core_pick_tx(). That flag is cleared
    firstly in __dev_queue_xmit(), is useful:
  - Avoid picking Tx queue with netdev_tx_queue_mapping() in next netdev
    in such case: eth0 macvlan - eth0.3 vlan - eth0 ixgbe-phy:
    For example, eth0, macvlan in pod, which root Qdisc install skbedit
    queue_mapping, send packets to eth0.3, vlan in host. In __dev_queue_xmit() of
    eth0.3, clear the flag, does not select tx queue according to skb->queue_mapping
    because there is no filters in clsact or tx Qdisc of this netdev.
    Same action taked in eth0, ixgbe in Host.
  - Avoid picking Tx queue for next packet. If we set xmit.skip_txqueue
    in tx Qdisc (qdisc->enqueue()), the proper way to clear it is clearing it
    in __dev_queue_xmit when processing next packets.

  For performance reasons, use the static key. If user does not config the NET_EGRESS,
  the patch will not be compiled.

  +----+      +----+      +----+
  | P1 |      | P2 |      | Pn |
  +----+      +----+      +----+
    |           |           |
    +-----------+-----------+
                |
                | clsact/skbedit
                |      MQ
                v
    +-----------+-----------+
    | q0        | q1        | qn
    v           v           v
  HTB/FQ      HTB/FQ  ...  FIFO

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Jonathan Lemon <jonathan.lemon@gmail.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Alexander Lobakin <alobakin@pm.me>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: Talal Ahmad <talalahmad@google.com>
Cc: Kevin Hao <haokexin@gmail.com>
Cc: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Cc: Antoine Tenart <atenart@kernel.org>
Cc: Wei Wang <weiwan@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agodocs: net: dsa: describe issues with checksum offload
Luiz Angelo Daros de Luca [Sat, 16 Apr 2022 05:27:37 +0000 (02:27 -0300)]
docs: net: dsa: describe issues with checksum offload

DSA tags before IP header (categories 1 and 2) or after the payload (3)
might introduce offload checksum issues.

Signed-off-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'mlxsw-line-card'
David S. Miller [Mon, 18 Apr 2022 10:00:19 +0000 (11:00 +0100)]
Merge branch 'mlxsw-line-card'

Ido Schimmel says:

====================
mlxsw: Introduce line card support for modular switch

Jiri says:

This patchset introduces support for modular switch systems and also
introduces mlxsw support for NVIDIA Mellanox SN4800 modular switch.
It contains 8 slots to accommodate line cards - replaceable PHY modules
which may contain gearboxes.
Currently supported line card:
16X 100GbE (QSFP28)
Other line cards that are going to be supported:
8X 200GbE (QSFP56)
4X 400GbE (QSFP-DD)
There may be other types of line cards added in the future.

To be consistent with the port split configuration (splitter cabels),
the line card entities are treated in the similar way. The nature of
a line card is not "a pluggable device", but "a pluggable PHY module".

A concept of "provisioning" is introduced. The user may "provision"
certain slot with a line card type. Driver then creates all instances
(devlink ports, netdevices, etc) related to this line card type. It does
not matter if the line card is plugged-in at the time. User is able to
configure netdevices, devlink ports, setup port splitters, etc. From the
perspective of the switch ASIC, all is present and can be configured.

The carrier of netdevices stays down if the line card is not plugged-in.
Once the line card is inserted and activated, the carrier of
the related netdevices is then reflecting the physical line state,
same as for an ordinary fixed port.

Once user does not want to use the line card related instances
anymore, he can "unprovision" the slot. Driver then removes the
instances.

Patches 1-4 are extending devlink driver API and UAPI in order to
register, show, dump, provision and activate the line card.
Patches 5-17 are implementing the introduced API in mlxsw.
The last patch adds a selftest for mlxsw line cards.

Example:
$ devlink port # No ports are listed
$ devlink lc
pci/0000:01:00.0:
  lc 1 state unprovisioned
    supported_types:
       16x100G
  lc 2 state unprovisioned
    supported_types:
       16x100G
  lc 3 state unprovisioned
    supported_types:
       16x100G
  lc 4 state unprovisioned
    supported_types:
       16x100G
  lc 5 state unprovisioned
    supported_types:
       16x100G
  lc 6 state unprovisioned
    supported_types:
       16x100G
  lc 7 state unprovisioned
    supported_types:
       16x100G
  lc 8 state unprovisioned
    supported_types:
       16x100G

Note that driver exposes list supported line card types. Currently
there is only one: "16x100G".

To provision the slot #8:

$ devlink lc set pci/0000:01:00.0 lc 8 type 16x100G
$ devlink lc show pci/0000:01:00.0 lc 8
pci/0000:01:00.0:
  lc 8 state active type 16x100G
    supported_types:
       16x100G
$ devlink port
pci/0000:01:00.0/0: type notset flavour cpu port 0 splittable false
pci/0000:01:00.0/53: type eth netdev enp1s0nl8p1 flavour physical lc 8 port 1 splittable true lanes 4
pci/0000:01:00.0/54: type eth netdev enp1s0nl8p2 flavour physical lc 8 port 2 splittable true lanes 4
pci/0000:01:00.0/55: type eth netdev enp1s0nl8p3 flavour physical lc 8 port 3 splittable true lanes 4
pci/0000:01:00.0/56: type eth netdev enp1s0nl8p4 flavour physical lc 8 port 4 splittable true lanes 4
pci/0000:01:00.0/57: type eth netdev enp1s0nl8p5 flavour physical lc 8 port 5 splittable true lanes 4
pci/0000:01:00.0/58: type eth netdev enp1s0nl8p6 flavour physical lc 8 port 6 splittable true lanes 4
pci/0000:01:00.0/59: type eth netdev enp1s0nl8p7 flavour physical lc 8 port 7 splittable true lanes 4
pci/0000:01:00.0/60: type eth netdev enp1s0nl8p8 flavour physical lc 8 port 8 splittable true lanes 4
pci/0000:01:00.0/61: type eth netdev enp1s0nl8p9 flavour physical lc 8 port 9 splittable true lanes 4
pci/0000:01:00.0/62: type eth netdev enp1s0nl8p10 flavour physical lc 8 port 10 splittable true lanes 4
pci/0000:01:00.0/63: type eth netdev enp1s0nl8p11 flavour physical lc 8 port 11 splittable true lanes 4
pci/0000:01:00.0/64: type eth netdev enp1s0nl8p12 flavour physical lc 8 port 12 splittable true lanes 4
pci/0000:01:00.0/125: type eth netdev enp1s0nl8p13 flavour physical lc 8 port 13 splittable true lanes 4
pci/0000:01:00.0/126: type eth netdev enp1s0nl8p14 flavour physical lc 8 port 14 splittable true lanes 4
pci/0000:01:00.0/127: type eth netdev enp1s0nl8p15 flavour physical lc 8 port 15 splittable true lanes 4
pci/0000:01:00.0/128: type eth netdev enp1s0nl8p16 flavour physical lc 8 port 16 splittable true lanes 4

To uprovision the slot #8:

$ devlink lc set pci/0000:01:00.0 lc 8 notype
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: mlxsw: Introduce devlink line card provision/unprovision/activation tests
Jiri Pirko [Mon, 18 Apr 2022 06:42:41 +0000 (09:42 +0300)]
selftests: mlxsw: Introduce devlink line card provision/unprovision/activation tests

Introduce basic line card manipulation which consists of provisioning,
unprovisioning and activation of a line card.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum: Add port to linecard mapping
Jiri Pirko [Mon, 18 Apr 2022 06:42:40 +0000 (09:42 +0300)]
mlxsw: spectrum: Add port to linecard mapping

For each port get slot_index using PMLP register. For ports residing
on a linecard, identify it with the linecard by setting mapping
using devlink_port_linecard_set() helper. Use linecard slot index for
PMTDB register queries.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core: Extend driver ops by remove selected ports op
Jiri Pirko [Mon, 18 Apr 2022 06:42:39 +0000 (09:42 +0300)]
mlxsw: core: Extend driver ops by remove selected ports op

In case of line card implementation, the core has to have a way to
remove relevant ports manually. Extend the Spectrum driver ops by an op
that implements port removal of selected ports upon request.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_linecards: Implement line card activation process
Jiri Pirko [Mon, 18 Apr 2022 06:42:38 +0000 (09:42 +0300)]
mlxsw: core_linecards: Implement line card activation process

Allow to process events generated upon line card getting "ready" and
"active".

When DSDSC event with "ready" bit set is delivered, that means the
line card is powered up. Use MDDC register to push the line card to
active state. Once FW is done with that, the DSDSC event with "active"
bit set is delivered.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_linecards: Add line card objects and implement provisioning
Jiri Pirko [Mon, 18 Apr 2022 06:42:37 +0000 (09:42 +0300)]
mlxsw: core_linecards: Add line card objects and implement provisioning

Introduce objects for line cards and an infrastructure around that.
Use devlink_linecard_create/destroy() to register the line card with
devlink core. Implement provisioning ops with a list of supported
line cards.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: reg: Add Management Binary Code Transfer Register
Jiri Pirko [Mon, 18 Apr 2022 06:42:36 +0000 (09:42 +0300)]
mlxsw: reg: Add Management Binary Code Transfer Register

The MBCT register allows to transfer binary INI codes from the host to
the management FW by transferring it by chunks of maximum 1KB.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: reg: Add Management DownStream Device Control Register
Jiri Pirko [Mon, 18 Apr 2022 06:42:35 +0000 (09:42 +0300)]
mlxsw: reg: Add Management DownStream Device Control Register

The MDDC register allows to control downstream devices and line cards.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: reg: Add Management DownStream Device Query Register
Jiri Pirko [Mon, 18 Apr 2022 06:42:34 +0000 (09:42 +0300)]
mlxsw: reg: Add Management DownStream Device Query Register

The MDDQ register allows to query the DownStream device properties.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum: Introduce port mapping change event processing
Jiri Pirko [Mon, 18 Apr 2022 06:42:33 +0000 (09:42 +0300)]
mlxsw: spectrum: Introduce port mapping change event processing

Register PMLPE trap and process the port mapping changes delivered
by it by creating related ports. Note that this happens after
provisioning. The INI of the linecard is processed and merged by FW.
PMLPE is generated for each port. Process this mapping change.

Layout of PMLPE is the same as layout of PMLP.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: Narrow the critical section of devl_lock during ports creation/removal
Jiri Pirko [Mon, 18 Apr 2022 06:42:32 +0000 (09:42 +0300)]
mlxsw: Narrow the critical section of devl_lock during ports creation/removal

No need to hold the lock for alloc and freecpu. So narrow the critical
section. Follow-up patch is going to benefit from this by adding more
code to the functions which will be out of the critical as well.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: reg: Add Ports Mapping Event Configuration Register
Jiri Pirko [Mon, 18 Apr 2022 06:42:31 +0000 (09:42 +0300)]
mlxsw: reg: Add Ports Mapping Event Configuration Register

The PMECR register is used to enable/disable event triggering
in case of local port mapping change.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum: Allocate port mapping array of structs instead of pointers
Jiri Pirko [Mon, 18 Apr 2022 06:42:30 +0000 (09:42 +0300)]
mlxsw: spectrum: Allocate port mapping array of structs instead of pointers

Instead of array of pointers to port mapping structures, allocate the
array of structures directly.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum: Allow lane to start from non-zero index
Jiri Pirko [Mon, 18 Apr 2022 06:42:29 +0000 (09:42 +0300)]
mlxsw: spectrum: Allow lane to start from non-zero index

So far, the lane index always started from zero. That is not true for
modular systems with gearbox-equipped linecards. Loose the check so the
lanes can start from non-zero index.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodevlink: add port to line card relationship set
Jiri Pirko [Mon, 18 Apr 2022 06:42:28 +0000 (09:42 +0300)]
devlink: add port to line card relationship set

In order to properly inform user about relationship between port and
line card, introduce a driver API to set line card for a port. Use this
information to extend port devlink netlink message by line card index
and also include the line card index into phys_port_name and by that
into a netdevice name.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodevlink: implement line card active state
Jiri Pirko [Mon, 18 Apr 2022 06:42:27 +0000 (09:42 +0300)]
devlink: implement line card active state

Allow driver to mark a line card as active. Expose this state to the
userspace over devlink netlink interface with proper notifications.
'active' state means that line card was plugged in after
being provisioned.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodevlink: implement line card provisioning
Jiri Pirko [Mon, 18 Apr 2022 06:42:26 +0000 (09:42 +0300)]
devlink: implement line card provisioning

In order to be able to configure all needed stuff on a port/netdevice
of a line card without the line card being present, introduce line card
provisioning. Basically by setting a type, provisioning process will
start and driver is supposed to create a placeholder for instances
(ports/netdevices) for a line card type.

Allow the user to query the supported line card types over line card
get command. Then implement two netlink command SET to allow user to
set/unset the card type.

On the driver API side, add provision/unprovision ops and supported
types array to be advertised. Upon provision op call, the driver should
take care of creating the instances for the particular line card type.
Introduce provision_set/clear() functions to be called by the driver
once the provisioning/unprovisioning is done on its side. These helpers
are not to be called directly due to the async nature of provisioning.

Example:
$ devlink port # No ports are listed
$ devlink lc
pci/0000:01:00.0:
  lc 1 state unprovisioned
    supported_types:
       16x100G
  lc 2 state unprovisioned
    supported_types:
       16x100G
  lc 3 state unprovisioned
    supported_types:
       16x100G
  lc 4 state unprovisioned
    supported_types:
       16x100G
  lc 5 state unprovisioned
    supported_types:
       16x100G
  lc 6 state unprovisioned
    supported_types:
       16x100G
  lc 7 state unprovisioned
    supported_types:
       16x100G
  lc 8 state unprovisioned
    supported_types:
       16x100G

$ devlink lc set pci/0000:01:00.0 lc 8 type 16x100G
$ devlink lc show pci/0000:01:00.0 lc 8
pci/0000:01:00.0:
  lc 8 state active type 16x100G
    supported_types:
       16x100G
$ devlink port
pci/0000:01:00.0/0: type notset flavour cpu port 0 splittable false
pci/0000:01:00.0/53: type eth netdev enp1s0nl8p1 flavour physical lc 8 port 1 splittable true lanes 4
pci/0000:01:00.0/54: type eth netdev enp1s0nl8p2 flavour physical lc 8 port 2 splittable true lanes 4
pci/0000:01:00.0/55: type eth netdev enp1s0nl8p3 flavour physical lc 8 port 3 splittable true lanes 4
pci/0000:01:00.0/56: type eth netdev enp1s0nl8p4 flavour physical lc 8 port 4 splittable true lanes 4
pci/0000:01:00.0/57: type eth netdev enp1s0nl8p5 flavour physical lc 8 port 5 splittable true lanes 4
pci/0000:01:00.0/58: type eth netdev enp1s0nl8p6 flavour physical lc 8 port 6 splittable true lanes 4
pci/0000:01:00.0/59: type eth netdev enp1s0nl8p7 flavour physical lc 8 port 7 splittable true lanes 4
pci/0000:01:00.0/60: type eth netdev enp1s0nl8p8 flavour physical lc 8 port 8 splittable true lanes 4
pci/0000:01:00.0/61: type eth netdev enp1s0nl8p9 flavour physical lc 8 port 9 splittable true lanes 4
pci/0000:01:00.0/62: type eth netdev enp1s0nl8p10 flavour physical lc 8 port 10 splittable true lanes 4
pci/0000:01:00.0/63: type eth netdev enp1s0nl8p11 flavour physical lc 8 port 11 splittable true lanes 4
pci/0000:01:00.0/64: type eth netdev enp1s0nl8p12 flavour physical lc 8 port 12 splittable true lanes 4
pci/0000:01:00.0/125: type eth netdev enp1s0nl8p13 flavour physical lc 8 port 13 splittable true lanes 4
pci/0000:01:00.0/126: type eth netdev enp1s0nl8p14 flavour physical lc 8 port 14 splittable true lanes 4
pci/0000:01:00.0/127: type eth netdev enp1s0nl8p15 flavour physical lc 8 port 15 splittable true lanes 4
pci/0000:01:00.0/128: type eth netdev enp1s0nl8p16 flavour physical lc 8 port 16 splittable true lanes 4

$ devlink lc set pci/0000:01:00.0 lc 8 notype

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodevlink: add support to create line card and expose to user
Jiri Pirko [Mon, 18 Apr 2022 06:42:25 +0000 (09:42 +0300)]
devlink: add support to create line card and expose to user

Extend the devlink API so the driver is going to be able to create and
destroy linecard instances. There can be multiple line cards per devlink
device. Expose this new type of object over devlink netlink API to the
userspace, with notifications.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agotcp: fix signed/unsigned comparison
Eric Dumazet [Sun, 17 Apr 2022 18:34:32 +0000 (11:34 -0700)]
tcp: fix signed/unsigned comparison

Kernel test robot reported:

smatch warnings:
net/ipv4/tcp_input.c:5966 tcp_rcv_established() warn: unsigned 'reason' is never less than zero.

I actually had one packetdrill failing because of this bug,
and was about to send the fix :)

v2: Andreas Schwab also pointed out that @reason needs to be negated
    before we reach tcp_drop_reason()

Fixes: 4b506af9c5b8 ("tcp: add two drop reasons for tcp_ack()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Andreas Schwab <schwab@linux-m68k.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'tcp-drop-reason-additions'
David S. Miller [Sun, 17 Apr 2022 12:31:32 +0000 (13:31 +0100)]
Merge branch 'tcp-drop-reason-additions'

Eric Dumazet says:

====================
tcp: drop reason additions

Currently, TCP is either missing drop reasons,
or pretending that some useful packets are dropped.

This patch series makes "perf record -a -e skb:kfree_skb"
much more usable.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agotcp: add drop reason support to tcp_ofo_queue()
Eric Dumazet [Sat, 16 Apr 2022 00:10:48 +0000 (17:10 -0700)]
tcp: add drop reason support to tcp_ofo_queue()

packets in OFO queue might be redundant, and dropped.

tcp_drop() is no longer needed.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agotcp: add drop reasons to tcp_rcv_synsent_state_process()
Eric Dumazet [Sat, 16 Apr 2022 00:10:47 +0000 (17:10 -0700)]
tcp: add drop reasons to tcp_rcv_synsent_state_process()

Re-use existing reasons.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agotcp: make tcp_rcv_synsent_state_process() drop monitor friend
Eric Dumazet [Sat, 16 Apr 2022 00:10:46 +0000 (17:10 -0700)]
tcp: make tcp_rcv_synsent_state_process() drop monitor friend

1) A valid RST packet should be consumed, to not confuse drop monitor.

2) Same remark for packet validating cross syn setup,
   even if we might ignore part of it.

3) When third packet of 3WHS is delayed, do not pretend
   the SYNACK was dropped.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agotcp: add drop reason support to tcp_prune_ofo_queue()
Eric Dumazet [Sat, 16 Apr 2022 00:10:45 +0000 (17:10 -0700)]
tcp: add drop reason support to tcp_prune_ofo_queue()

Add one reason for packets dropped from OFO queue because
of memory pressure.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agotcp: add two drop reasons for tcp_ack()
Eric Dumazet [Sat, 16 Apr 2022 00:10:44 +0000 (17:10 -0700)]
tcp: add two drop reasons for tcp_ack()

Add TCP_TOO_OLD_ACK and TCP_ACK_UNSENT_DATA drop
reasons so that tcp_rcv_established() can report
them.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agotcp: add drop reasons to tcp_rcv_state_process()
Eric Dumazet [Sat, 16 Apr 2022 00:10:43 +0000 (17:10 -0700)]
tcp: add drop reasons to tcp_rcv_state_process()

Add basic support for drop reasons in tcp_rcv_state_process()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agotcp: make tcp_rcv_state_process() drop monitor friendly
Eric Dumazet [Sat, 16 Apr 2022 00:10:42 +0000 (17:10 -0700)]
tcp: make tcp_rcv_state_process() drop monitor friendly

tcp_rcv_state_process() incorrectly drops packets
instead of consuming it, making drop monitor very noisy,
if not unusable.

Calling tcp_time_wait() or tcp_done() is part
of standard behavior, packets triggering these actions
were not dropped.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agotcp: add drop reason support to tcp_validate_incoming()
Eric Dumazet [Sat, 16 Apr 2022 00:10:41 +0000 (17:10 -0700)]
tcp: add drop reason support to tcp_validate_incoming()

Creates four new drop reasons for the following cases:

1) packet being rejected by RFC 7323 PAWS check
2) packet being rejected by SEQUENCE check
3) Invalid RST packet
4) Invalid SYN packet

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agotcp: get rid of rst_seq_match
Eric Dumazet [Sat, 16 Apr 2022 00:10:40 +0000 (17:10 -0700)]
tcp: get rid of rst_seq_match

Small cleanup in tcp_validate_incoming(), no need for rst_seq_match
setting and testing.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agotcp: consume incoming skb leading to a reset
Eric Dumazet [Sat, 16 Apr 2022 00:10:39 +0000 (17:10 -0700)]
tcp: consume incoming skb leading to a reset

Whenever tcp_validate_incoming() handles a valid RST packet,
we should not pretend the packet was dropped.

Create a special section at the end of tcp_validate_incoming()
to handle this case.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'qca8k_preiv-shrink'
David S. Miller [Sun, 17 Apr 2022 12:28:29 +0000 (13:28 +0100)]
Merge branch 'qca8k_preiv-shrink'

Ansuel Smith says:

====================
net: Reduce qca8k_priv space usage

These 6 patch is a first attempt at reducting qca8k_priv space.
The code changed a lot during times and we have many old logic
that can be replaced with new implementation

The first patch drop the tracking of MTU. We mimic what was done
for mtk and we change MTU only when CPU port is changed.

The second patch finally drop a piece of story of this driver.
The ar8xxx_port_status struct was used by the first implementation
of this driver to put all sort of status data for the port...
With the evolution of DSA all that stuff got dropped till only
the enabled state was the only part of the that struct.
Since it's overkill to keep an array of int, we convert the variable
to a simple u8 where we store the status of each port. This is needed
to don't reanable ports on system resume.

The third patch is a preparation for patch 4. As Vladimir explained
in another patch, we waste a tons of space by keeping a duplicate of
the switch dsa ops in qca8k_priv. The only reason for this is to
dynamically set the correct mdiobus configuration (a legacy dsa one,
or a custom dedicated one)
To solve this problem, we just drop the phy_read/phy_write and we
declare a custom mdiobus in any case.
This way we can use a static dsa switch ops struct and we can drop it
from qca8k_priv

Patch 4 drop the duplicated dsa_switch_ops.

Patch 5 is a fixup for mdio read error.

Patch 6 is an effort to standardize how bus name are done.

This series is just a start of more cleanup.

The idea is to move this driver to the qca dir and split common code
from specific code. Also the mgmt eth code still requires some love
and can totally be optimized by recycling the same skb over time.

Also while working on the MTU it was notice some problem with
the stmmac driver and with the reloading phase that cause all
sort of problems with qca8k.

I'm sending this here just to try to keep small series instead of
proposing monster series hard to review.

v2:
- Rework MTU patch
v3:
- Drop unrealated changes from patch 3
- Add fixup patch for mdio read
- Unify bus name for legacy and OF mdio bus
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: qca8k: unify bus id naming with legacy and OF mdio bus
Ansuel Smith [Fri, 15 Apr 2022 23:30:17 +0000 (01:30 +0200)]
net: dsa: qca8k: unify bus id naming with legacy and OF mdio bus

Add support for multiple switch with OF mdio bus declaration.
Unify the bus id naming and use the same logic for both legacy and OF
mdio bus.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: qca8k: correctly handle mdio read error
Ansuel Smith [Fri, 15 Apr 2022 23:30:16 +0000 (01:30 +0200)]
net: dsa: qca8k: correctly handle mdio read error

Restore original way to handle mdio read error by returning 0xffff.
This was wrongly changed when the internal_mdio_read was introduced,
now that both legacy and internal use the same function, make sure that
they behave the same way.

Fixes: ce062a0adbfe ("net: dsa: qca8k: fix kernel panic with legacy mdio mapping")
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: qca8k: drop dsa_switch_ops from qca8k_priv
Ansuel Smith [Fri, 15 Apr 2022 23:30:15 +0000 (01:30 +0200)]
net: dsa: qca8k: drop dsa_switch_ops from qca8k_priv

Now that dsa_switch_ops is not switch specific anymore, we can drop it
from qca8k_priv and use the static ops directly for the dsa_switch
pointer.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: qca8k: rework and simplify mdiobus logic
Ansuel Smith [Fri, 15 Apr 2022 23:30:14 +0000 (01:30 +0200)]
net: dsa: qca8k: rework and simplify mdiobus logic

In an attempt to reduce qca8k_priv space, rework and simplify mdiobus
logic.
We now declare a mdiobus instead of relying on DSA phy_read/write even
if a mdio node is not present. This is all to make the qca8k ops static
and not switch specific. With a legacy implementation where port doesn't
have a phy map declared in the dts with a mdio node, we declare a
'qca8k-legacy' mdiobus. The conversion logic is used as legacy read and
write ops are used instead of the internal one.
Also drop the legacy_phy_port_mapping as we now declare mdiobus with ops
that already address the workaround.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: qca8k: drop port_sts from qca8k_priv
Ansuel Smith [Fri, 15 Apr 2022 23:30:13 +0000 (01:30 +0200)]
net: dsa: qca8k: drop port_sts from qca8k_priv

Port_sts is a thing of the past for this driver. It was something
present on the initial implementation of this driver and parts of the
original struct were dropped over time. Using an array of int to store if
a port is enabled or not to handle PM operation seems overkill. Switch
and use a simple u8 to store the port status where each bit correspond
to a port. (bit is set port is enabled, bit is not set, port is disabled)
Also add some comments to better describe why we need to track port
status.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: qca8k: drop MTU tracking from qca8k_priv
Ansuel Smith [Fri, 15 Apr 2022 23:30:12 +0000 (01:30 +0200)]
net: dsa: qca8k: drop MTU tracking from qca8k_priv

DSA set the CPU port based on the largest MTU of all the slave ports.
Based on this we can drop the MTU array from qca8k_priv and set the
port_change_mtu logic on DSA changing MTU of the CPU port as the switch
have a global MTU settingfor each port.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/ipv6: Introduce accept_unsolicited_na knob to implement router-side changes for...
Arun Ajith S [Fri, 15 Apr 2022 08:34:02 +0000 (08:34 +0000)]
net/ipv6: Introduce accept_unsolicited_na knob to implement router-side changes for RFC9131

Add a new neighbour cache entry in STALE state for routers on receiving
an unsolicited (gratuitous) neighbour advertisement with
target link-layer-address option specified.
This is similar to the arp_accept configuration for IPv4.
A new sysctl endpoint is created to turn on this behaviour:
/proc/sys/net/ipv6/conf/interface/accept_unsolicited_na.

Signed-off-by: Arun Ajith S <aajith@arista.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoipv6: fix NULL deref in ip6_rcv_core()
Eric Dumazet [Wed, 13 Apr 2022 20:56:53 +0000 (13:56 -0700)]
ipv6: fix NULL deref in ip6_rcv_core()

idev can be NULL, as the surrounding code suggests.

Fixes: 4daf841a2ef3 ("net: ipv6: add skb drop reasons to ip6_rcv_core()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Menglong Dong <imagedong@tencent.com>
Cc: Jiang Biao <benbjiang@tencent.com>
Cc: Hao Peng <flyingpeng@tencent.com>
Link: https://lore.kernel.org/r/20220413205653.1178458-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet_sched: make qdisc_reset() smaller
Eric Dumazet [Thu, 14 Apr 2022 01:10:04 +0000 (18:10 -0700)]
net_sched: make qdisc_reset() smaller

For some unknown reason qdisc_reset() is using
a convoluted way of freeing two lists of skbs.

Use __skb_queue_purge() instead.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://lore.kernel.org/r/20220414011004.2378350-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoocteon_ep: Remove custom driver version
Leon Romanovsky [Thu, 14 Apr 2022 07:52:42 +0000 (10:52 +0300)]
octeon_ep: Remove custom driver version

In review comment [1] was pointed that new code is not supposed
to set driver version and should rely on kernel version instead.

As an outcome of that comment all the dance around setting such
driver version to FW should be removed too, because in upstream
kernel whole driver will have same version so read/write from/to
FW will give same result.

[1] https://lore.kernel.org/all/YladGTmon1x3dfxI@unreal

Fixes: 862cd659a6fb ("octeon_ep: Add driver framework and device initialization")
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/5d76f3116ee795071ec044eabb815d6c2bdc7dbd.1649922731.git.leonro@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'ibmvnic-use-a-set-of-ltbs-per-pool'
Jakub Kicinski [Fri, 15 Apr 2022 21:02:09 +0000 (14:02 -0700)]
Merge branch 'ibmvnic-use-a-set-of-ltbs-per-pool'

Sukadev Bhattiprolu says:

====================
ibmvnic: Use a set of LTBs per pool

ibmvnic uses a single large long term buffer (LTB) per rx or tx
pool (queue). This has two limitations.

First, if we need to free/allocate an LTB (eg during a reset), under
low memory conditions, the allocation can fail.

Second, the kernel limits the size of single LTB (DMA buffer) to 16MB
(based on MAX_ORDER). With jumbo frames (mtu = 9000) we can only have
about 1763 buffers per LTB (16MB / 9588 bytes per frame) even though
the max supported buffers is 4096. (The 9588 instead of 9088 comes from
IBMVNIC_BUFFER_HLEN.)

To overcome these limitations, allow creating a set of LTBs per queue.
====================

Link: https://lore.kernel.org/r/20220413171026.1264294-1-drt@linux.ibm.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoibmvnic: Allow multiple ltbs in txpool ltb_set
Sukadev Bhattiprolu [Wed, 13 Apr 2022 17:10:26 +0000 (13:10 -0400)]
ibmvnic: Allow multiple ltbs in txpool ltb_set

Allow multiple LTBs in the txpool's ltb_set. i.e rather than using
a single large LTB, use several smaller LTBs.

The first n-1 LTBs will all be of the same size. The size of the last
LTB in the set depends on the number of buffers and buffer (mtu) size.
This strategy hopefully allows more reuse of the initial LTBs and also
reduces the chances of an allocation failure (of the large LTB) when
system is low in memory.

Suggested-by: Brian King <brking@linux.ibm.com>
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: Dany Madden <drt@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoibmvnic: Allow multiple ltbs in rxpool ltb_set
Sukadev Bhattiprolu [Wed, 13 Apr 2022 17:10:25 +0000 (13:10 -0400)]
ibmvnic: Allow multiple ltbs in rxpool ltb_set

Allow multiple LTBs in the rxpool's ltb_set. The first n-1 LTBs will all
be of the same size. The size of the last LTB in the set depends on the
number of buffers and buffer (mtu) size.

Having a set of LTBs per pool provides a couple of benefits.

First, with the current value of IBMVNIC_MAX_LTB_SIZE of 16MB, with an
MTU of 9000, we need a LTB (DMA buffer) of that size but the allocation
can fail in low memory conditions. With a set of LTBs per pool, we can
use several smaller (8MB) LTBs and hopefully have fewer allocation
failures. (See also comments in ibmvnic.h on the trade-off with smaller
LTBs)

Second since the kernel limits the size of the DMA buffer to 16MB (based
on MAX_ORDER), with a single DMA buffer per pool, the pool is also limited
to 16MB. This in turn limits the number of buffers per pool to 1763 when
MTU is 9000. With a set of LTBs per pool, we can have upto the max of 4096
buffers per pool even when MTU is 9000.

Suggested-by: Brian King <brking@linux.ibm.com>
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: Dany Madden <drt@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoibmvnic: convert rxpool ltb to a set of ltbs
Sukadev Bhattiprolu [Wed, 13 Apr 2022 17:10:24 +0000 (13:10 -0400)]
ibmvnic: convert rxpool ltb to a set of ltbs

Define and use interfaces that treat the long term buffer (LTB) of an
rxpool as a set of LTBs rather than a single LTB. The set only has one
LTB for now.

Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: Dany Madden <drt@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoibmvnic: define map_txpool_buf_to_ltb()
Sukadev Bhattiprolu [Wed, 13 Apr 2022 17:10:23 +0000 (13:10 -0400)]
ibmvnic: define map_txpool_buf_to_ltb()

Define a helper to map a given txpool buffer into its corresponding long
term buffer (LTB) and offset. Currently there is just one LTB per txpool
so the mapping is trivial. When we add support for multiple LTBs per
txpool, this helper will be more useful.

Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: Dany Madden <drt@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoibmvnic: define map_rxpool_buf_to_ltb()
Sukadev Bhattiprolu [Wed, 13 Apr 2022 17:10:22 +0000 (13:10 -0400)]
ibmvnic: define map_rxpool_buf_to_ltb()

Define a helper to map a given rx pool buffer into its corresponding long
term buffer (LTB) and offset. Currently there is just one LTB per pool so
the mapping is trivial. When we add support for multiple LTBs per pool,
this helper will be more useful.

Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: Dany Madden <drt@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoibmvnic: rename local variable index to bufidx
Sukadev Bhattiprolu [Wed, 13 Apr 2022 17:10:21 +0000 (13:10 -0400)]
ibmvnic: rename local variable index to bufidx

The local variable 'index' is heavily used in some functions and is
confusing with the presence of other "index" fields like pool->index,
->consumer_index, etc. Rename it to bufidx to better reflect that its
the index of a buffer in the pool.

Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: Dany Madden <drt@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'net-ethool-add-support-to-get-set-tx-push-by-ethtool-g-g'
Jakub Kicinski [Fri, 15 Apr 2022 18:41:56 +0000 (11:41 -0700)]
Merge branch 'net-ethool-add-support-to-get-set-tx-push-by-ethtool-g-g'

Jie Wang says:

====================
net: ethool: add support to get/set tx push by ethtool -G/g

These three patches add tx push in ring params and adapt the set and get APIs
of ring params.
====================

Link: https://lore.kernel.org/r/20220412020121.14140-1-huangguangbin2@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: hns3: add tx push support in hns3 ring param process
Jie Wang [Tue, 12 Apr 2022 02:01:21 +0000 (10:01 +0800)]
net: hns3: add tx push support in hns3 ring param process

This patch adds tx push param to hns3 ring param and adapts the set and get
API of ring params. So users can set it by cmd ethtool -G and get it by cmd
ethtool -g.

Signed-off-by: Jie Wang <wangjie125@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: ethtool: move checks before rtnl_lock() in ethnl_set_rings
Jie Wang [Tue, 12 Apr 2022 02:01:20 +0000 (10:01 +0800)]
net: ethtool: move checks before rtnl_lock() in ethnl_set_rings

Currently these two checks in ethnl_set_rings are added after rtnl_lock()
which will do useless works if the request is invalid.

So this patch moves these checks before the rtnl_lock() to avoid these
costs.

Signed-off-by: Jie Wang <wangjie125@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: ethtool: extend ringparam set/get APIs for tx_push
Jie Wang [Tue, 12 Apr 2022 02:01:19 +0000 (10:01 +0800)]
net: ethtool: extend ringparam set/get APIs for tx_push

Currently tx push is a standard driver feature which controls use of a fast
path descriptor push. So this patch extends the ringparam APIs and data
structures to support set/get tx push by ethtool -G/g.

Signed-off-by: Jie Wang <wangjie125@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoocteon_ep: fix error return code in octep_probe()
Yang Yingliang [Fri, 15 Apr 2022 02:39:57 +0000 (10:39 +0800)]
octeon_ep: fix error return code in octep_probe()

If register_netdev() fails , it should return error
code in octep_probe().

Fixes: 862cd659a6fb ("octeon_ep: Add driver framework and device initialization")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'emaclite-cleanups'
David S. Miller [Fri, 15 Apr 2022 10:46:29 +0000 (11:46 +0100)]
Merge branch 'emaclite-cleanups'

Radhey Shyam Pandey says:

====================
net: emaclite: Trivial code cleanup

This patchset fix coding style issues, remove BUFFER_ALIGN
macro and also update copyright text.

I have to resend as earlier series didn't reach mailing list
due to some configuration issue.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: emaclite: Remove custom BUFFER_ALIGN macro
Shravya Kumbham [Thu, 14 Apr 2022 12:37:11 +0000 (18:07 +0530)]
net: emaclite: Remove custom BUFFER_ALIGN macro

BUFFER_ALIGN macro is used to calculate the number of bytes
required for the next alignment. Instead of this, we can directly
use the skb_reserve(skb, NET_IP_ALIGN) to make the protocol header
buffer aligned on at least a 4-byte boundary, where the NET_IP_ALIGN
is by default defined as 2. So removing the BUFFER_ALIGN and its
related defines which it can be done by the skb_reserve() itself.

Signed-off-by: Shravya Kumbham <shravya.kumbham@xilinx.com>
Signed-off-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: emaclite: Update copyright text to correct format
Michal Simek [Thu, 14 Apr 2022 12:37:10 +0000 (18:07 +0530)]
net: emaclite: Update copyright text to correct format

Based on recommended guidance Copyright term should be also present in
front of (c). That's why aligned driver to match this pattern.
It helps automated tools with source code scanning.

Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: emaclite: Fix coding style
Radhey Shyam Pandey [Thu, 14 Apr 2022 12:37:09 +0000 (18:07 +0530)]
net: emaclite: Fix coding style

Make coding style changes to fix checkpatch script warnings.
There is no functional change. Fixes below check and warnings-

CHECK: Blank lines aren't necessary after an open brace '{'
CHECK: spinlock_t definition without comment
CHECK: Please don't use multiple blank lines
WARNING: Prefer 'unsigned int' to bare use of 'unsigned'
CHECK: braces {} should be used on all arms of this statement
CHECK: Unbalanced braces around else statement
CHECK: Alignment should match open parenthesis
WARNING: Missing a blank line after declarations

Signed-off-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: ethernet: ti: davinci_emac: using pm_runtime_resume_and_get instead of pm_runtim...
Minghao Chi [Thu, 14 Apr 2022 09:08:00 +0000 (09:08 +0000)]
net: ethernet: ti: davinci_emac: using pm_runtime_resume_and_get instead of pm_runtime_get_sync

Using pm_runtime_resume_and_get() to replace pm_runtime_get_sync and
pm_runtime_put_noidle. This change is just to simplify the code, no
actual functional changes.

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoocteon_ep: Fix spelling mistake "inerrupts" -> "interrupts"
Colin Ian King [Thu, 14 Apr 2022 08:08:34 +0000 (09:08 +0100)]
octeon_ep: Fix spelling mistake "inerrupts" -> "interrupts"

There is a spelling mistake in a dev_info message. Fix it.

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'mlxsw-line-card-prep'
David S. Miller [Fri, 15 Apr 2022 10:06:13 +0000 (11:06 +0100)]
Merge branch 'mlxsw-line-card-prep'

Ido Schimmel says:

====================
mlxsw: Preparations for line cards support

Currently, mlxsw registers thermal zones as well as hwmon entries for
objects such as transceiver modules and gearboxes. In upcoming modular
systems, these objects are no longer found on the main board (i.e., slot
0), but on plug-able line cards. This patchset prepares mlxsw for such
systems in terms of hwmon, thermal and cable access support.

Patches #1-#3 gradually prepare mlxsw for transceiver modules access
support for line cards by splitting some of the internal structures and
some APIs.

Patches #4-#5 gradually prepare mlxsw for hwmon support for line cards
by splitting some of the internal structures and augmenting them with a
slot index.

Patches #6-#7 do the same for thermal zones.

Patch #8 selects cooling device for binding to a thermal zone by exact
name match to prevent binding to non-relevant devices.

Patch #9 replaces internal define for thermal zone name length with a
common define.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_thermal: Use common define for thermal zone name length
Vadim Pasternak [Wed, 13 Apr 2022 15:17:33 +0000 (18:17 +0300)]
mlxsw: core_thermal: Use common define for thermal zone name length

Replace internal define 'MLXSW_THERMAL_ZONE_MAX_NAME' by common
'THERMAL_NAME_LENGTH'.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_thermal: Use exact name of cooling devices for binding
Vadim Pasternak [Wed, 13 Apr 2022 15:17:32 +0000 (18:17 +0300)]
mlxsw: core_thermal: Use exact name of cooling devices for binding

Modular system supports additional cooling devices "mlxreg_fan1",
"mlxreg_fan2", etcetera. Thermal zones in "mlxsw" driver should be
bound to the same device as before called "mlxreg_fan". Used exact
match for cooling device name to avoid binding to new additional
cooling devices.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_thermal: Add line card id prefix to line card thermal zone name
Vadim Pasternak [Wed, 13 Apr 2022 15:17:31 +0000 (18:17 +0300)]
mlxsw: core_thermal: Add line card id prefix to line card thermal zone name

Add prefix "lc#n" to thermal zones associated with the thermal objects
found on line cards.

For example thermal zone for module #9 located at line card #7 will
have type:
mlxsw-lc7-module9.
And thermal zone for gearbox #3 located at line card #5 will have type:
mlxsw-lc5-gearbox3.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_thermal: Extend internal structures to support multi thermal areas
Vadim Pasternak [Wed, 13 Apr 2022 15:17:30 +0000 (18:17 +0300)]
mlxsw: core_thermal: Extend internal structures to support multi thermal areas

Introduce intermediate level for thermal zones areas.
Currently all thermal zones are associated with thermal objects located
within the main board. Such objects are created during driver
initialization and removed during driver de-initialization.

For line cards in modular system the thermal zones are to be associated
with the specific line card. They should be created whenever new line
card is available (inserted, validated, powered and enabled) and
removed, when line card is getting unavailable.
The thermal objects found on the line card #n are accessed by setting
slot index to #n, while for access to objects found on the main board
slot index should be set to default value zero.

Each thermal area contains the set of thermal zones associated with
particular slot index.
Thus introduction of thermal zone areas allows to use the same APIs for
the main board and line cards, by adding slot index argument.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_hwmon: Introduce slot parameter in hwmon interfaces
Vadim Pasternak [Wed, 13 Apr 2022 15:17:29 +0000 (18:17 +0300)]
mlxsw: core_hwmon: Introduce slot parameter in hwmon interfaces

Add 'slot' parameter to 'mlxsw_hwmon_dev' structure. Use this parameter
in mlxsw_reg_mtmp_pack(), mlxsw_reg_mtbr_pack(), mlxsw_reg_mgpir_pack()
and mlxsw_reg_mtmp_slot_index_set() routines.
For main board it'll always be zero, for line cards it'll be set to
the physical slot number in modular systems.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_hwmon: Extend internal structures to support multi hwmon objects
Vadim Pasternak [Wed, 13 Apr 2022 15:17:28 +0000 (18:17 +0300)]
mlxsw: core_hwmon: Extend internal structures to support multi hwmon objects

Currently, mlxsw supports a single hwmon device and registers it with
attributes corresponding to the various objects found on the main
board such as fans and gearboxes.

Line cards can have the same objects, but unlike the main board they
can be added and removed while the system is running. The various
hwmon objects found on these line cards should be created when the
line card becomes available and destroyed when the line card becomes
unavailable.

The above can be achieved by representing each line card as a
different hwmon device and registering / unregistering it when the
line card becomes available / unavailable.

Prepare for multi hwmon device support by splitting
'struct mlxsw_hwmon' into 'struct mlxsw_hwmon' and
'struct mlxsw_hwmon_dev'. The first will hold information relevant to
all hwmon devices, whereas the second will hold per-hwmon device
information.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core: Move port module events enablement to a separate function
Vadim Pasternak [Wed, 13 Apr 2022 15:17:27 +0000 (18:17 +0300)]
mlxsw: core: Move port module events enablement to a separate function

Use a separate function for enablement of port module events such
plug/unplug and temperature threshold crossing. The motivation is to
reuse the function for line cards.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core: Extend port module data structures for line cards
Vadim Pasternak [Wed, 13 Apr 2022 15:17:26 +0000 (18:17 +0300)]
mlxsw: core: Extend port module data structures for line cards

The port module core is tasked with module operations such as setting
power mode policy and reset. The per-module information is currently
stored in one large array suited for non-modular systems where only the
main board is present (i.e., slot index 0).

As a preparation for line cards support, allocate a per line card array
according to the queried number of slots in the system. For each line
card, allocate a module array according to the queried maximum number of
modules per-slot.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core: Extend interfaces for cable info access with slot argument
Vadim Pasternak [Wed, 13 Apr 2022 15:17:25 +0000 (18:17 +0300)]
mlxsw: core: Extend interfaces for cable info access with slot argument

Extend all cable info APIs with 'slot_index' argument.

For main board, slot will always be set to zero and these APIs will work
as before. If reading cable information is required from cages located
on line cards, slot should be set to the physical slot number, where
line card is located in modular systems.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: ethernet: ti: cpsw_priv: using pm_runtime_resume_and_get instead of pm_runtime_g...
Minghao Chi [Wed, 13 Apr 2022 09:38:36 +0000 (09:38 +0000)]
net: ethernet: ti: cpsw_priv: using pm_runtime_resume_and_get instead of pm_runtime_get_sync

Using pm_runtime_resume_and_get is more appropriate
for simplifing code

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: stmmac: stmmac_main: using pm_runtime_resume_and_get instead of pm_runtime_get_sync
Minghao Chi [Wed, 13 Apr 2022 09:38:01 +0000 (09:38 +0000)]
net: stmmac: stmmac_main: using pm_runtime_resume_and_get instead of pm_runtime_get_sync

Using pm_runtime_resume_and_get is more appropriate
for simplifing code

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: ethernet: ti: cpsw_new: use pm_runtime_resume_and_get() instead of pm_runtime_ge...
Minghao Chi [Wed, 13 Apr 2022 09:35:29 +0000 (09:35 +0000)]
net: ethernet: ti: cpsw_new: use pm_runtime_resume_and_get() instead of pm_runtime_get_sync()

Using pm_runtime_resume_and_get is more appropriate
for simplifing code

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agogeneve: avoid indirect calls in GRO path, when possible
Paolo Abeni [Wed, 13 Apr 2022 08:44:40 +0000 (10:44 +0200)]
geneve: avoid indirect calls in GRO path, when possible

In the most common setups, the geneve tunnels use an inner
ethernet encapsulation. In the GRO path, when such condition is
true, we can call directly the relevant GRO helper and avoid
a few indirect calls.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'mneta-page_pool_get_stats'
David S. Miller [Fri, 15 Apr 2022 09:43:48 +0000 (10:43 +0100)]
Merge branch 'mneta-page_pool_get_stats'

Lorenzo Bianconi says:

====================
net: mvneta: add support for page_pool_get_stats

Introduce page_pool stats ethtool APIs in order to avoid driver duplicated
code.

Changes since v4:
- rebase on top of net-next

Changes since v3:
- get rid of wrong for loop in page_pool_ethtool_stats_get()
- add API stubs when page_pool_stats are not compiled in

Changes since v2:
- remove enum list of page_pool stats in page_pool.h
- remove leftover change in mvneta.c for ethtool_stats array allocation

Changes since v1:
- move stats accounting to page_pool code
- move stats string management to page_pool code
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: mvneta: add support for page_pool_get_stats
Lorenzo Bianconi [Tue, 12 Apr 2022 16:31:59 +0000 (18:31 +0200)]
net: mvneta: add support for page_pool_get_stats

Introduce support for the page_pool stats API into mvneta driver.
Report page_pool stats through ethtool.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>