linux-2.6-microblaze.git
9 months agobnxt_en: Fix backing store V2 logic
Michael Chan [Fri, 1 Dec 2023 22:39:10 +0000 (14:39 -0800)]
bnxt_en: Fix backing store V2 logic

The current code determines the last backing store valid type during
bnxt_hwrm_func_backing_store_qcaps_v2().  In effect, the last type
is determined based on what firmware advertises.  The more correct
way is to determine it based on what the driver is configuring.  The
driver may not configure all the backing store types advertised by
firmware.

Move the logic to determine the last type to bnxt_backing_store_cfg_v2().
We need to pass the legacy enable flags to the function in case only
the legacy types are being configured.

Fixes: 236e237f8ffe ("bnxt_en: Add support for HWRM_FUNC_BACKING_STORE_CFG_V2 firmware calls")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20231201223924.26955-2-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agotcp: Dump bound-only sockets in inet_diag.
Guillaume Nault [Fri, 1 Dec 2023 14:49:52 +0000 (15:49 +0100)]
tcp: Dump bound-only sockets in inet_diag.

Walk the hashinfo->bhash2 table so that inet_diag can dump TCP sockets
that are bound but haven't yet called connect() or listen().

The code is inspired by the ->lhash2 loop. However there's no manual
test of the source port, since this kind of filtering is already
handled by inet_diag_bc_sk(). Also, a maximum of 16 sockets are dumped
at a time, to avoid running with bh disabled for too long.

There's no TCP state for bound but otherwise inactive sockets. Such
sockets normally map to TCP_CLOSE. However, "ss -l", which is supposed
to only dump listening sockets, actually requests the kernel to dump
sockets in either the TCP_LISTEN or TCP_CLOSE states. To avoid dumping
bound-only sockets with "ss -l", we therefore need to define a new
pseudo-state (TCP_BOUND_INACTIVE) that user space will be able to set
explicitly.

With an IPv4, an IPv6 and an IPv6-only socket, bound respectively to
40000, 64000, 60000, an updated version of iproute2 could work as
follow:

  $ ss -t state bound-inactive
  Recv-Q   Send-Q     Local Address:Port       Peer Address:Port   Process
  0        0                0.0.0.0:40000           0.0.0.0:*
  0        0                   [::]:60000              [::]:*
  0        0                      *:64000                 *:*

Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://lore.kernel.org/r/b3a84ae61e19c06806eea9c602b3b66e8f0cfc81.1701362867.git.gnault@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoMerge branch 'net-phy-micrel-additional-clock-handling'
Jakub Kicinski [Mon, 4 Dec 2023 22:43:56 +0000 (14:43 -0800)]
Merge branch 'net-phy-micrel-additional-clock-handling'

Heiko Stuebner says:

====================
net: phy: micrel: additional clock handling

Some Micrel phys define a specific rmii-ref clock (added in 2014) while
the generic phy binding specifies an unnamed clock for ethernet phys.

This allows Micrel phys to use both, so as to keep the phys not using
the named rmii-ref clock to conform to the generic binding while allowing
them to enable a supplying clock, when the phy is not supplied by a
dedicated oscillator.
====================

Link: https://lore.kernel.org/r/20231201150131.326766-1-heiko@sntech.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agonet: phy: micrel: allow usage of generic ethernet-phy clock
Heiko Stuebner [Fri, 1 Dec 2023 15:01:31 +0000 (16:01 +0100)]
net: phy: micrel: allow usage of generic ethernet-phy clock

The generic ethernet-phy binding allows describing an external clock since
commit 350b7a258f20 ("dt-bindings: net: phy: Document support for external PHY clk")
for cases where the phy is not supplied by an oscillator but instead
by a clock from the host system.

And the old named "rmii-ref" clock from 2014 is only specified for phys
of the KSZ8021, KSZ8031, KSZ8081, KSZ8091 types.

So allow retrieving and enabling the optional generic clock on phys that
do not provide a rmii-ref clock.

Signed-off-by: Heiko Stuebner <heiko.stuebner@cherry.de>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20231201150131.326766-3-heiko@sntech.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agonet: phy: micrel: use devm_clk_get_optional_enabled for the rmii-ref clock
Heiko Stuebner [Fri, 1 Dec 2023 15:01:30 +0000 (16:01 +0100)]
net: phy: micrel: use devm_clk_get_optional_enabled for the rmii-ref clock

While the external clock input will most likely be enabled, it's not
guaranteed and clk_get_rate in some suppliers will even just return
valid results when the clock is running.

So use devm_clk_get_optional_enabled to retrieve and enable the clock
in one go.

Signed-off-by: Heiko Stuebner <heiko.stuebner@cherry.de>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20231201150131.326766-2-heiko@sntech.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agodocs: netlink: add NLMSG_DONE message format for doit actions
Jiri Pirko [Fri, 1 Dec 2023 18:01:54 +0000 (19:01 +0100)]
docs: netlink: add NLMSG_DONE message format for doit actions

In case NLMSG_DONE message is sent as a reply to doit action, multiple
kernel implementation do not send anything else than struct nlmsghdr.
Add this note to the Netlink intro documentation.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Bagas Sanjaya <bagasdotme@gmail.com>
Link: https://lore.kernel.org/r/20231201180154.864007-1-jiri@resnulli.us
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoMerge branch 'octeontx2-multicast-mirror-offload'
David S. Miller [Mon, 4 Dec 2023 11:06:46 +0000 (11:06 +0000)]
Merge branch 'octeontx2-multicast-mirror-offload'

Suman Ghosh says:

====================
octeontx2: Multicast/mirror offload changes

This patchset includes changes to support TC multicast/mirror offload.

Patch #1: Adds changes to support new mailbox to offload multicast/mirror
offload.

Patch #2: Adds TC related changes which uses the newly added mailboxes to
offload multicast/mirror rules.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 months agoocteontx2-pf: TC flower offload support for mirror
Suman Ghosh [Thu, 30 Nov 2023 03:43:24 +0000 (09:13 +0530)]
octeontx2-pf: TC flower offload support for mirror

This patch extends TC flower offload support for mirroring ingress
traffic to a different PF/VF. Below is an example command,

'tc filter add dev eth1 ingress protocol ip flower src_ip <ip-addr>
skip_sw action mirred ingress mirror dev eth2'

Signed-off-by: Suman Ghosh <sumang@marvell.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 months agoocteontx2-af: Add new mbox to support multicast/mirror offload
Suman Ghosh [Thu, 30 Nov 2023 03:43:23 +0000 (09:13 +0530)]
octeontx2-af: Add new mbox to support multicast/mirror offload

A new mailbox is added to support offloading of multicast/mirror
functionality. The mailbox also supports dynamic updation of the
multicast/mirror list.

Signed-off-by: Suman Ghosh <sumang@marvell.com>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 months agoMerge branch 'net-cacheline-optimizations'
David S. Miller [Sat, 2 Dec 2023 22:24:37 +0000 (22:24 +0000)]
Merge branch 'net-cacheline-optimizations'

Coco Li says:

====================
Analyze and Reorganize core Networking Structs to optimize cacheline consumption

Currently, variable-heavy structs in the networking stack is organized
chronologically, logically and sometimes by cacheline access.

This patch series attempts to reorganize the core networking stack
variables to minimize cacheline consumption during the phase of data
transfer. Specifically, we looked at the TCP/IP stack and the fast
path definition in TCP.

For documentation purposes, we also added new files for each core data
structure we considered, although not all ended up being modified due
to the amount of existing cacheline they span in the fast path. In
the documentation, we recorded all variables we identified on the
fast path and the reasons. We also hope that in the future when
variables are added/modified, the document can be referred to and
updated accordingly to reflect the latest variable organization.

Tested:
Our tests were run with neper tcp_rr using tcp traffic. The tests have $cpu
number of threads and variable number of flows (see below).

Tests were run on 6.5-rc1

Efficiency is computed as cpu seconds / throughput (one tcp_rr round trip).
The following result shows efficiency delta before and after the patch
series is applied.

On AMD platforms with 100Gb/s NIC and 256Mb L3 cache:
IPv4
Flows   with patches    clean kernel      Percent reduction
30k     0.0001736538065 0.0002741191042 -36.65%
20k     0.0001583661752 0.0002712559158 -41.62%
10k     0.0001639148817 0.0002951800751 -44.47%
5k      0.0001859683866 0.0003320642536 -44.00%
1k      0.0002035190546 0.0003152056382 -35.43%

IPv6
Flows   with patches  clean kernel    Percent reduction
30k     0.000202535503  0.0003275329163 -38.16%
20k     0.0002020654777 0.0003411304786 -40.77%
10k     0.0002122427035 0.0003803674705 -44.20%
5k      0.0002348776729 0.0004030403953 -41.72%
1k      0.0002237384583 0.0002813646157 -20.48%

On Intel platforms with 200Gb/s NIC and 105Mb L3 cache:
IPv6
Flows   with patches    clean kernel    Percent reduction
30k     0.0006296537873 0.0006370427753 -1.16%
20k     0.0003451029365 0.0003628016076 -4.88%
10k     0.0003187646958 0.0003346835645 -4.76%
5k      0.0002954676348 0.000311807592  -5.24%
1k      0.0001909169342 0.0001848069709 3.31%

v8 changes:
1. Update net_device_read_txrx cache group maximum
2. Update MAINTAINERS for documentations
3. Skip __cache_group variables in scripts/kernel-doc
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 months agonetns-ipv4: reorganize netns_ipv4 fast path variables
Coco Li [Wed, 29 Nov 2023 07:27:54 +0000 (07:27 +0000)]
netns-ipv4: reorganize netns_ipv4 fast path variables

Reorganize fast path variables on tx-txrx-rx order.
Fastpath cacheline ends after sysctl_tcp_rmem.
There are only read-only variables here. (write is on the control path
and not considered in this case)

Below data generated with pahole on x86 architecture.
Fast path variables span cache lines before change: 4
Fast path variables span cache lines after change: 2

Suggested-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Wei Wang <weiwan@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: Coco Li <lixiaoyan@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 months agocache: enforce cache groups
Coco Li [Wed, 29 Nov 2023 07:27:53 +0000 (07:27 +0000)]
cache: enforce cache groups

Set up build time warnings to safeguard against future header changes of
organized structs.

Warning includes:

1) whether all variables are still in the same cache group
2) whether all the cache groups have the sum of the members size (in the
   maximum condition, including all members defined in configs)

The __cache_group* variables are ignored in kernel-doc check in the
various header files they appear in to enforce the cache groups.

Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Coco Li <lixiaoyan@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 months agoDocumentations: Analyze heavily used Networking related structs
Coco Li [Wed, 29 Nov 2023 07:27:52 +0000 (07:27 +0000)]
Documentations: Analyze heavily used Networking related structs

Analyzed a few structs in the networking stack by looking at variables
within them that are used in the TCP/IP fast path.

Fast path is defined as TCP path where data is transferred from sender to
receiver unidirectionally. It doesn't include phases other than
TCP_ESTABLISHED, nor does it look at error paths.

We hope to re-organizing variables that span many cachelines whose fast
path variables are also spread out, and this document can help future
developers keep networking fast path cachelines small.

Optimized_cacheline field is computed as
(Fastpath_Bytes/L3_cacheline_size_x86), and not the actual organized
results (see patches to come for these).

Investigation is done on 6.5

Name                 Struct_Cachelines  Cur_fastpath_cache Fastpath_Bytes Optimized_cacheline
tcp_sock         42 (2664 Bytes)    12    396 8
net_device         39 (2240 bytes)    12 234 4
inet_sock         15 (960 bytes)    14 922 14
Inet_connection_sock 22 (1368 bytes)    18 1166 18
Netns_ipv4 (sysctls) 12 (768 bytes)     4 77 2
linux_mib         16 (1060)    6 104 2

Note how there isn't much improvement space for inet_sock and
Inet_connection_sock because sk and icsk_inet respectively takes up so
much of the struct that rest of the variables become a small portion of
the struct size.

So, we decided to reorganize tcp_sock, net_device, netns_ipv4

Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Coco Li <lixiaoyan@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 months agonfp: ethtool: expose transmit SO_TIMESTAMPING capability
Yinjun Zhang [Wed, 29 Nov 2023 08:04:13 +0000 (10:04 +0200)]
nfp: ethtool: expose transmit SO_TIMESTAMPING capability

NFP always supports software time stamping of tx, now expose
the capability through ethtool ops.

Signed-off-by: Yinjun Zhang <yinjun.zhang@corigine.com>
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Link: https://lore.kernel.org/r/20231129080413.83789-1-louis.peens@corigine.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agonet: ethernet: renesas: rcar_gen4_ptp: Depend on PTP_1588_CLOCK
Niklas Söderlund [Wed, 29 Nov 2023 11:11:42 +0000 (12:11 +0100)]
net: ethernet: renesas: rcar_gen4_ptp: Depend on PTP_1588_CLOCK

When breaking out the Gen4 gPTP support to its own module the dependency
on the PTP_1588_CLOCK framework was left as optional and only stated for
the driver using the module. This leads to issues when doing
COMPILE_TEST of RENESAS_GEN4_PTP separately and PTP_1588_CLOCK is built
as a module and the other as a built-in. Add an explicit depend on
PTP_1588_CLOCK.

While at it remove the optional support for PTP_1588_CLOCK from
RENESAS_ETHER_SWITCH as the driver unconditionally calls the Gen4 gPTP
module and thus also requires the PTP_1588_CLOCK framework.

Reported-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 8c1c66235e03 ("net: ethernet: renesas: rcar_gen4_ptp: Break out to module")
Signed-off-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/r/20231129111142.3322667-1-niklas.soderlund+renesas@ragnatech.se
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agodt-bindings: net: qcom,ipa: document SM8650 compatible
Neil Armstrong [Wed, 29 Nov 2023 17:22:58 +0000 (18:22 +0100)]
dt-bindings: net: qcom,ipa: document SM8650 compatible

Document the IPA on the SM8650 Platform which uses version 5.5.1,
which is a minor revision of v5.5 found on SM8550, thus we can
use the SM8550 bindings as fallback since it shares the same
register mappings.

Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20231129-topic-sm8650-upstream-bindings-ipa-v1-1-ca21eb2dfb14@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoocteontx2-af: debugfs: update CQ context fields
Nithin Dabilpuram [Thu, 30 Nov 2023 06:07:03 +0000 (11:37 +0530)]
octeontx2-af: debugfs: update CQ context fields

This patch update the CQ structure fields to support the feature
added in new silicons and also dump these fields in debugfs.

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Link: https://lore.kernel.org/r/20231130060703.16769-1-gakula@marvell.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoDocumentation: networking: add missing PLCA messages from the message list
Maxime Chevallier [Thu, 30 Nov 2023 19:13:59 +0000 (20:13 +0100)]
Documentation: networking: add missing PLCA messages from the message list

Physical Layer Collision Avoidance messages are correctly documented but
were left-out of the global list of ethnl messages, add them to the
list.

Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Link: https://lore.kernel.org/r/20231130191400.817948-1-maxime.chevallier@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agotcp: tcp_gro_dev_warn() cleanup
Eric Dumazet [Thu, 30 Nov 2023 18:41:35 +0000 (18:41 +0000)]
tcp: tcp_gro_dev_warn() cleanup

Use DO_ONCE_LITE_IF() and __cold attribute to put tcp_gro_dev_warn()
out of line.

This also allows the message to be printed again after a
"echo 1 > /sys/kernel/debug/clear_warn_once"

Also add a READ_ONCE() when reading device mtu, as it could
be changed concurrently.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20231130184135.4130860-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agodoc/netlink: Add bitfield32, s8, s16 to the netlink-raw schema
Donald Hunter [Thu, 30 Nov 2023 21:49:53 +0000 (21:49 +0000)]
doc/netlink: Add bitfield32, s8, s16 to the netlink-raw schema

The netlink-raw schema was not updated when bitfield32 was added
to the genetlink-legacy schema. It is needed for rtnetlink families.

s8 and s16 were also missing.

Signed-off-by: Donald Hunter <donald.hunter@gmail.com>
Link: https://lore.kernel.org/r/20231130214959.27377-2-donald.hunter@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoocteon_ep: set backpressure watermark for RX queues
Shinas Rasheed [Wed, 29 Nov 2023 05:31:31 +0000 (21:31 -0800)]
octeon_ep: set backpressure watermark for RX queues

Set backpressure watermark for hardware RX queues. Backpressure
gets triggered when the available buffers of a hardware RX queue
falls below the set watermark. This backpressure will propagate
to packet processing pipeline in the OCTEON card, so that the host
receives fewer packets and prevents packet dropping at host.

Signed-off-by: Shinas Rasheed <srasheed@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 months agoocteon_ep: Fix error code in probe()
Dan Carpenter [Tue, 28 Nov 2023 13:13:19 +0000 (16:13 +0300)]
octeon_ep: Fix error code in probe()

Set the error code if octep_ctrl_net_get_mtu() fails.  Currently the code
returns success.

Fixes: 0a5f8534e398 ("octeon_ep: get max rx packet length from firmware")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Reviewed-by: Sathesh B Edara <sedara@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 months agoMerge branch 'selftests-tc-testing-more-tdc-updates'
Jakub Kicinski [Fri, 1 Dec 2023 07:22:19 +0000 (23:22 -0800)]
Merge branch 'selftests-tc-testing-more-tdc-updates'

Pedro Tammela says:

====================
selftests: tc-testing: more tdc updates

Follow-up on a feedback from Jakub and random cleanups from related
net/sched patches
====================

Link: https://lore.kernel.org/r/20231129222424.910148-1-pctammela@mojatatu.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: tc-testing: remove filters/tests.json
Pedro Tammela [Wed, 29 Nov 2023 22:24:24 +0000 (19:24 -0300)]
selftests: tc-testing: remove filters/tests.json

Remove this generic file and move the tests to their appropriate files

Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://lore.kernel.org/r/20231129222424.910148-5-pctammela@mojatatu.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: tc-testing: rename concurrency.json to flower.json
Pedro Tammela [Wed, 29 Nov 2023 22:24:23 +0000 (19:24 -0300)]
selftests: tc-testing: rename concurrency.json to flower.json

All tests in this file pertain to flower, so name it appropriately

Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://lore.kernel.org/r/20231129222424.910148-4-pctammela@mojatatu.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: tc-testing: remove spurious './' from Makefile
Pedro Tammela [Wed, 29 Nov 2023 22:24:22 +0000 (19:24 -0300)]
selftests: tc-testing: remove spurious './' from Makefile

Patchwork CI didn't like the extra './', so remove it.

Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://lore.kernel.org/r/20231129222424.910148-3-pctammela@mojatatu.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: tc-testing: remove spurious nsPlugin usage
Pedro Tammela [Wed, 29 Nov 2023 22:24:21 +0000 (19:24 -0300)]
selftests: tc-testing: remove spurious nsPlugin usage

Tests using DEV2 should not be run in a dedicated net namespace,
and in parallel, as this device cannot be shared.

Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://lore.kernel.org/r/20231129222424.910148-2-pctammela@mojatatu.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agodocs: netlink: link to family documentations from spec info
Jakub Kicinski [Wed, 29 Nov 2023 04:14:27 +0000 (20:14 -0800)]
docs: netlink: link to family documentations from spec info

To increase the chances of people finding the rendered docs
add a link to specs.rst and index.rst.

Add a label in the generated index.rst and while at it adjust
the title a little bit.

Reviewed-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Donald Hunter <donald.hunter@gmail.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/20231129041427.2763074-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoMerge branch 'support-octeon-cn98-devices'
Jakub Kicinski [Fri, 1 Dec 2023 07:17:06 +0000 (23:17 -0800)]
Merge branch 'support-octeon-cn98-devices'

Shinas Rasheed says:

====================
support OCTEON CN98 devices

Implement device unload control net API required for CN98
devices and add support in driver for the same.

V1: https://lore.kernel.org/all/20231127162135.2529363-1-srasheed@marvell.com/
====================

Link: https://lore.kernel.org/r/20231129045348.2538843-1-srasheed@marvell.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoocteon_ep: support OCTEON CN98 devices
Shinas Rasheed [Wed, 29 Nov 2023 04:53:48 +0000 (20:53 -0800)]
octeon_ep: support OCTEON CN98 devices

Add PCI Endpoint NIC support for Octeon CN98 devices.
CN98 devices are part of Octeon 9 family products with
similar PCI NIC characteristics to CN93, already supported
driver.

Add CN98 card to the device id table, as well
as support differences in the register fields and
certain usage scenarios such as unload.

Signed-off-by: Shinas Rasheed <srasheed@marvell.com>
Link: https://lore.kernel.org/r/20231129045348.2538843-3-srasheed@marvell.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoocteon_ep: implement device unload control net API
Shinas Rasheed [Wed, 29 Nov 2023 04:53:47 +0000 (20:53 -0800)]
octeon_ep: implement device unload control net API

Device unload control net function should inform firmware
of driver unload to let it take necessary actions to cleanup.

Signed-off-by: Shinas Rasheed <srasheed@marvell.com>
Link: https://lore.kernel.org/r/20231129045348.2538843-2-srasheed@marvell.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agonet/sched: cbs: Use units.h instead of the copy of a definition
Andy Shevchenko [Tue, 28 Nov 2023 17:48:13 +0000 (19:48 +0200)]
net/sched: cbs: Use units.h instead of the copy of a definition

BYTES_PER_KBIT is defined in units.h, use that definition.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20231128174813.394462-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agonet: phy: mdio_device: Reset device only when necessary
Andrew Halaney [Mon, 27 Nov 2023 21:41:10 +0000 (15:41 -0600)]
net: phy: mdio_device: Reset device only when necessary

Currently the phy reset sequence is as shown below for a
devicetree described mdio phy on boot:

1. Assert the phy_device's reset as part of registering
2. Deassert the phy_device's reset as part of registering
3. Deassert the phy_device's reset as part of phy_probe
4. Deassert the phy_device's reset as part of phy_hw_init

The extra two deasserts include waiting the deassert delay afterwards,
which is adding unnecessary delay.

This applies to both possible types of resets (reset controller
reference and a reset gpio) that can be used.

Here's some snipped tracing output using the following command line
params "trace_event=gpio:* trace_options=stacktrace" illustrating
the reset handling and where its coming from:

    /* Assert */
       systemd-udevd-283     [002] .....     6.780434: gpio_value: 544 set 0
       systemd-udevd-283     [002] .....     6.783849: <stack trace>
     => gpiod_set_raw_value_commit
     => gpiod_set_value_nocheck
     => gpiod_set_value_cansleep
     => mdio_device_reset
     => mdiobus_register_device
     => phy_device_register
     => fwnode_mdiobus_phy_device_register
     => fwnode_mdiobus_register_phy
     => __of_mdiobus_register
     => stmmac_mdio_register
     => stmmac_dvr_probe
     => stmmac_pltfr_probe
     => devm_stmmac_pltfr_probe
     => qcom_ethqos_probe
     => platform_probe

    /* Deassert */
       systemd-udevd-283     [002] .....     6.802480: gpio_value: 544 set 1
       systemd-udevd-283     [002] .....     6.805886: <stack trace>
     => gpiod_set_raw_value_commit
     => gpiod_set_value_nocheck
     => gpiod_set_value_cansleep
     => mdio_device_reset
     => phy_device_register
     => fwnode_mdiobus_phy_device_register
     => fwnode_mdiobus_register_phy
     => __of_mdiobus_register
     => stmmac_mdio_register
     => stmmac_dvr_probe
     => stmmac_pltfr_probe
     => devm_stmmac_pltfr_probe
     => qcom_ethqos_probe
     => platform_probe

    /* Deassert */
       systemd-udevd-283     [002] .....     6.882601: gpio_value: 544 set 1
       systemd-udevd-283     [002] .....     6.886014: <stack trace>
     => gpiod_set_raw_value_commit
     => gpiod_set_value_nocheck
     => gpiod_set_value_cansleep
     => mdio_device_reset
     => phy_probe
     => really_probe
     => __driver_probe_device
     => driver_probe_device
     => __device_attach_driver
     => bus_for_each_drv
     => __device_attach
     => device_initial_probe
     => bus_probe_device
     => device_add
     => phy_device_register
     => fwnode_mdiobus_phy_device_register
     => fwnode_mdiobus_register_phy
     => __of_mdiobus_register
     => stmmac_mdio_register
     => stmmac_dvr_probe
     => stmmac_pltfr_probe
     => devm_stmmac_pltfr_probe
     => qcom_ethqos_probe
     => platform_probe

    /* Deassert */
      NetworkManager-477     [000] .....     7.023144: gpio_value: 544 set 1
      NetworkManager-477     [000] .....     7.026596: <stack trace>
     => gpiod_set_raw_value_commit
     => gpiod_set_value_nocheck
     => gpiod_set_value_cansleep
     => mdio_device_reset
     => phy_init_hw
     => phy_attach_direct
     => phylink_fwnode_phy_connect
     => __stmmac_open
     => stmmac_open

There's a lot of paths where the device is getting its reset
asserted and deasserted. Let's track the state and only actually
do the assert/deassert when it changes.

Reported-by: Sagar Cheluvegowda <quic_scheluve@quicinc.com>
Signed-off-by: Andrew Halaney <ahalaney@redhat.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20231127-net-phy-reset-once-v2-1-448e8658779e@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoMerge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf...
Jakub Kicinski [Fri, 1 Dec 2023 00:56:09 +0000 (16:56 -0800)]
Merge tag 'for-netdev' of https://git./linux/kernel/git/bpf/bpf-next

Daniel Borkmann says:

====================
pull-request: bpf-next 2023-11-30

We've added 30 non-merge commits during the last 7 day(s) which contain
a total of 58 files changed, 1598 insertions(+), 154 deletions(-).

The main changes are:

1) Add initial TX metadata implementation for AF_XDP with support in mlx5
   and stmmac drivers. Two types of offloads are supported right now, that
   is, TX timestamp and TX checksum offload, from Stanislav Fomichev with
   stmmac implementation from Song Yoong Siang.

2) Change BPF verifier logic to validate global subprograms lazily instead
   of unconditionally before the main program, so they can be guarded using
   BPF CO-RE techniques, from Andrii Nakryiko.

3) Add BPF link_info support for uprobe multi link along with bpftool
   integration for the latter, from Jiri Olsa.

4) Use pkg-config in BPF selftests to determine ld flags which is
   in particular needed for linking statically, from Akihiko Odaki.

5) Fix a few BPF selftest failures to adapt to the upcoming LLVM18,
   from Yonghong Song.

* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (30 commits)
  bpf/tests: Remove duplicate JSGT tests
  selftests/bpf: Add TX side to xdp_hw_metadata
  selftests/bpf: Convert xdp_hw_metadata to XDP_USE_NEED_WAKEUP
  selftests/bpf: Add TX side to xdp_metadata
  selftests/bpf: Add csum helpers
  selftests/xsk: Support tx_metadata_len
  xsk: Add option to calculate TX checksum in SW
  xsk: Validate xsk_tx_metadata flags
  xsk: Document tx_metadata_len layout
  net: stmmac: Add Tx HWTS support to XDP ZC
  net/mlx5e: Implement AF_XDP TX timestamp and checksum offload
  tools: ynl: Print xsk-features from the sample
  xsk: Add TX timestamp and TX checksum offload support
  xsk: Support tx_metadata_len
  selftests/bpf: Use pkg-config for libelf
  selftests/bpf: Override PKG_CONFIG for static builds
  selftests/bpf: Choose pkg-config for the target
  bpftool: Add support to display uprobe_multi links
  selftests/bpf: Add link_info test for uprobe_multi link
  selftests/bpf: Use bpf_link__destroy in fill_link_info tests
  ...
====================

Conflicts:

Documentation/netlink/specs/netdev.yaml:
  839ff60df3ab ("net: page_pool: add nlspec for basic access to page pools")
  48eb03dd2630 ("xsk: Add TX timestamp and TX checksum offload support")
https://lore.kernel.org/all/20231201094705.1ee3cab8@canb.auug.org.au/

While at it also regen, tree is dirty after:
  48eb03dd2630 ("xsk: Add TX timestamp and TX checksum offload support")
looks like code wasn't re-rendered after "render-max" was removed.

Link: https://lore.kernel.org/r/20231130145708.32573-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Jakub Kicinski [Fri, 1 Dec 2023 00:10:40 +0000 (16:10 -0800)]
Merge git://git./linux/kernel/git/netdev/net

Cross-merge networking fixes after downstream PR.

No conflicts.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoMerge tag 'net-6.7-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 30 Nov 2023 23:24:46 +0000 (08:24 +0900)]
Merge tag 'net-6.7-rc4' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Paolo Abeni:
 "Including fixes from bpf and wifi.

  Current release - regressions:

   - neighbour: fix __randomize_layout crash in struct neighbour

   - r8169: fix deadlock on RTL8125 in jumbo mtu mode

  Previous releases - regressions:

   - wifi:
       - mac80211: fix warning at station removal time
       - cfg80211: fix CQM for non-range use

   - tools: ynl-gen: fix unexpected response handling

   - octeontx2-af: fix possible buffer overflow

   - dpaa2: recycle the RX buffer only after all processing done

   - rswitch: fix missing dev_kfree_skb_any() in error path

  Previous releases - always broken:

   - ipv4: fix uaf issue when receiving igmp query packet

   - wifi: mac80211: fix debugfs deadlock at device removal time

   - bpf:
       - sockmap: af_unix stream sockets need to hold ref for pair sock
       - netdevsim: don't accept device bound programs

   - selftests: fix a char signedness issue

   - dsa: mv88e6xxx: fix marvell 6350 probe crash

   - octeontx2-pf: restore TC ingress police rules when interface is up

   - wangxun: fix memory leak on msix entry

   - ravb: keep reverse order of operations in ravb_remove()"

* tag 'net-6.7-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (51 commits)
  net: ravb: Keep reverse order of operations in ravb_remove()
  net: ravb: Stop DMA in case of failures on ravb_open()
  net: ravb: Start TX queues after HW initialization succeeded
  net: ravb: Make write access to CXR35 first before accessing other EMAC registers
  net: ravb: Use pm_runtime_resume_and_get()
  net: ravb: Check return value of reset_control_deassert()
  net: libwx: fix memory leak on msix entry
  ice: Fix VF Reset paths when interface in a failed over aggregate
  bpf, sockmap: Add af_unix test with both sockets in map
  bpf, sockmap: af_unix stream sockets need to hold ref for pair sock
  tools: ynl-gen: always construct struct ynl_req_state
  ethtool: don't propagate EOPNOTSUPP from dumps
  ravb: Fix races between ravb_tx_timeout_work() and net related ops
  r8169: prevent potential deadlock in rtl8169_close
  r8169: fix deadlock on RTL8125 in jumbo mtu mode
  neighbour: Fix __randomize_layout crash in struct neighbour
  octeontx2-pf: Restore TC ingress police rules when interface is up
  octeontx2-pf: Fix adding mbox work queue entry when num_vfs > 64
  net: stmmac: xgmac: Disable FPE MMC interrupts
  octeontx2-af: Fix possible buffer overflow
  ...

9 months agoMerge tag 'pmdomain-v6.7-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh...
Linus Torvalds [Thu, 30 Nov 2023 23:17:08 +0000 (08:17 +0900)]
Merge tag 'pmdomain-v6.7-rc2' of git://git./linux/kernel/git/ulfh/linux-pm

Pull pmdomain fix from Ulf Hansson:

 - Avoid polling for the scmi_perf_domain on arm

* tag 'pmdomain-v6.7-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/linux-pm:
  pmdomain: arm: Avoid polling for scmi_perf_domain

9 months agoMerge tag 'mmc-v6.7-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Linus Torvalds [Thu, 30 Nov 2023 23:15:05 +0000 (08:15 +0900)]
Merge tag 'mmc-v6.7-rc1' of git://git./linux/kernel/git/ulfh/mmc

Pull MMC fixes from Ulf Hansson:
 "MMC core:
   - Fix CQE error recovery path

  MMC host:
   - cqhci: Fix CQE error recovery path
   - sdhci-pci-gli: Fix initialization of LPM
   - sdhci-sprd: Fix enabling/disabling of the vqmmc regulator"

* tag 'mmc-v6.7-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  mmc: sdhci-sprd: Fix vqmmc not shutting down after the card was pulled
  mmc: sdhci-pci-gli: Disable LPM during initialization
  mmc: cqhci: Fix task clearing in CQE error recovery
  mmc: cqhci: Warn of halt or task clear failure
  mmc: block: Retry commands in CQE error recovery
  mmc: block: Be sure to wait while busy in CQE error recovery
  mmc: cqhci: Increase recovery halt timeout
  mmc: block: Do not lose cache flush during CQE error recovery

9 months agoMerge tag 'leds-fixes-6.7' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/leds
Linus Torvalds [Thu, 30 Nov 2023 23:00:02 +0000 (08:00 +0900)]
Merge tag 'leds-fixes-6.7' of git://git./linux/kernel/git/lee/leds

Pull LED fix from Lee Jones:

 - Remove duplicate sysfs entry 'color' from LEDs class

* tag 'leds-fixes-6.7' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/leds:
  leds: class: Don't expose color sysfs entry

9 months agoMerge tag 'efi-urgent-for-v6.7-1' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 30 Nov 2023 22:57:08 +0000 (07:57 +0900)]
Merge tag 'efi-urgent-for-v6.7-1' of git://git./linux/kernel/git/efi/efi

Pull EFI fix from Ard Biesheuvel:

 - Fix for EFI unaccepted memory handling

* tag 'efi-urgent-for-v6.7-1' of git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi:
  efi/unaccepted: Fix off-by-one when checking for overlapping ranges

9 months agonet: page_pool: fix general protection fault in page_pool_unlist
Eric Dumazet [Thu, 30 Nov 2023 09:22:59 +0000 (09:22 +0000)]
net: page_pool: fix general protection fault in page_pool_unlist

syzbot was able to trigger a crash [1] in page_pool_unlist()

page_pool_list() only inserts a page pool into a netdev page pool list
if a netdev was set in params.

Even if the kzalloc() call in page_pool_create happens to initialize
pool->user.list, I chose to be more explicit in page_pool_list()
adding one INIT_HLIST_NODE().

We could test in page_pool_unlist() if netdev was set,
but since netdev can be changed to lo, it seems more robust to
check if pool->user.list is hashed  before calling hlist_del().

[1]

Illegal XDP return value 4294946546 on prog  (id 2) dev N/A, expect packet loss!
general protection fault, probably for non-canonical address 0xdffffc0000000000: 0000 [#1] PREEMPT SMP KASAN
KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007]
CPU: 0 PID: 5064 Comm: syz-executor391 Not tainted 6.7.0-rc2-syzkaller-00533-ga379972973a8 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 11/10/2023
RIP: 0010:__hlist_del include/linux/list.h:988 [inline]
RIP: 0010:hlist_del include/linux/list.h:1002 [inline]
RIP: 0010:page_pool_unlist+0xd1/0x170 net/core/page_pool_user.c:342
Code: df 48 89 fa 48 c1 ea 03 80 3c 02 00 0f 85 90 00 00 00 4c 8b a3 f0 06 00 00 48 b8 00 00 00 00 00 fc ff df 4c 89 e2 48 c1 ea 03 <80> 3c 02 00 75 68 48 85 ed 49 89 2c 24 74 24 e8 1b ca 07 f9 48 8d
RSP: 0018:ffffc900039ff768 EFLAGS: 00010246
RAX: dffffc0000000000 RBX: ffff88814ae02000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 0000000000000004 RDI: ffff88814ae026f0
RBP: 0000000000000000 R08: 0000000000000000 R09: fffffbfff1d57fdc
R10: ffffffff8eabfee3 R11: ffffffff8aa0008b R12: 0000000000000000
R13: ffff88814ae02000 R14: dffffc0000000000 R15: 0000000000000001
FS:  000055555717a380(0000) GS:ffff8880b9800000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000002555398 CR3: 0000000025044000 CR4: 00000000003506f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
 <TASK>
 __page_pool_destroy net/core/page_pool.c:851 [inline]
 page_pool_release+0x507/0x6b0 net/core/page_pool.c:891
 page_pool_destroy+0x1ac/0x4c0 net/core/page_pool.c:956
 xdp_test_run_teardown net/bpf/test_run.c:216 [inline]
 bpf_test_run_xdp_live+0x1578/0x1af0 net/bpf/test_run.c:388
 bpf_prog_test_run_xdp+0x827/0x1530 net/bpf/test_run.c:1254
 bpf_prog_test_run kernel/bpf/syscall.c:4041 [inline]
 __sys_bpf+0x11bf/0x4920 kernel/bpf/syscall.c:5402
 __do_sys_bpf kernel/bpf/syscall.c:5488 [inline]
 __se_sys_bpf kernel/bpf/syscall.c:5486 [inline]
 __x64_sys_bpf+0x78/0xc0 kernel/bpf/syscall.c:5486

Fixes: 083772c9f972 ("net: page_pool: record pools per netdev")
Reported-and-tested-by: syzbot+f9f8efb58a4db2ca98d0@syzkaller.appspotmail.com
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20231130092259.3797753-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoMerge branch 'net-ethernet-convert-to-platform-remove-callback-returning-void'
Paolo Abeni [Thu, 30 Nov 2023 12:04:15 +0000 (13:04 +0100)]
Merge branch 'net-ethernet-convert-to-platform-remove-callback-returning-void'

Uwe Kleine-König says:

====================
net: ethernet: Convert to platform remove callback returning void

in (implicit) v1 of this series
(https://lore.kernel.org/netdev/20231117091655.872426-1-u.kleine-koenig@pengutronix.de)
I tried to address the resource leaks in the three cpsw drivers. However
this is hard to get right without being able to test the changes. So
here comes a series that just converts all drivers below
drivers/net/ethernet to use .remove_new() and adds a comment about the
potential leaks for someone else to fix the problem.

See commit 5c5a7680e67b ("platform: Provide a remove callback that
returns no value") for an extended explanation and the eventual goal.
The TL;DR; is to prevent bugs like the three noticed here.

Note this series results in no change of behaviour apart from improving
the error message for the three cpsw drivers from

remove callback returned a non-zero value. This will be ignored.

to

Failed to resume device (-ESOMETHING)
====================

Link: https://lore.kernel.org/r/20231128173823.867512-1-u.kleine-koenig@pengutronix.de
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agonet: ethernet: ezchip: Convert to platform remove callback returning void
Uwe Kleine-König [Tue, 28 Nov 2023 17:38:28 +0000 (18:38 +0100)]
net: ethernet: ezchip: Convert to platform remove callback returning void

The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is ignored (apart
from emitting a warning) and this typically results in resource leaks.

To improve here there is a quest to make the remove callback return
void. In the first step of this quest all drivers are converted to
.remove_new(), which already returns void. Eventually after all drivers
are converted, .remove_new() will be renamed to .remove().

Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agonet: ethernet: ti: cpsw-new: Convert to platform remove callback returning void
Uwe Kleine-König [Tue, 28 Nov 2023 17:38:27 +0000 (18:38 +0100)]
net: ethernet: ti: cpsw-new: Convert to platform remove callback returning void

The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is ignored (apart
from emitting a warning) and this typically results in resource leaks.

To improve here there is a quest to make the remove callback return
void. In the first step of this quest all drivers are converted to
.remove_new(), which already returns void. Eventually after all drivers
are converted, .remove_new() will be renamed to .remove().

Replace the error path returning a non-zero value by an error message
and a comment that there is more to do. With that this patch results in
no change of behaviour in this driver apart from improving the error
message.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Roger Quadros <rogerq@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agonet: ethernet: ti: cpsw: Convert to platform remove callback returning void
Uwe Kleine-König [Tue, 28 Nov 2023 17:38:26 +0000 (18:38 +0100)]
net: ethernet: ti: cpsw: Convert to platform remove callback returning void

The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is ignored (apart
from emitting a warning) and this typically results in resource leaks.

To improve here there is a quest to make the remove callback return
void. In the first step of this quest all drivers are converted to
.remove_new(), which already returns void. Eventually after all drivers
are converted, .remove_new() will be renamed to .remove().

Replace the error path returning a non-zero value by an error message
and a comment that there is more to do. With that this patch results in
no change of behaviour in this driver apart from improving the error
message.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Roger Quadros <rogerq@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agonet: ethernet: ti: am65-cpsw: Convert to platform remove callback returning void
Uwe Kleine-König [Tue, 28 Nov 2023 17:38:25 +0000 (18:38 +0100)]
net: ethernet: ti: am65-cpsw: Convert to platform remove callback returning void

The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is ignored (apart
from emitting a warning) and this typically results in resource leaks.

To improve here there is a quest to make the remove callback return
void. In the first step of this quest all drivers are converted to
.remove_new(), which already returns void. Eventually after all drivers
are converted, .remove_new() will be renamed to .remove().

Replace the error path returning a non-zero value by an error message
and a comment that there is more to do. With that this patch results in
no change of behaviour in this driver apart from improving the error
message.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Roger Quadros <rogerq@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agoMerge branch 'devlink-warn-about-existing-entities-during-reload-reinit'
Paolo Abeni [Thu, 30 Nov 2023 11:31:39 +0000 (12:31 +0100)]
Merge branch 'devlink-warn-about-existing-entities-during-reload-reinit'

Jiri Pirko says:

====================
devlink: warn about existing entities during reload-reinit

Recently there has been a couple of attempts from drivers to block
devlink reload in certain situations. Turned out, the drivers do not
properly tear down ports and related netdevs during reload.

To address this, add couple of checks to be done during devlink reload
reinit action. Also, extend documentation to be more explicit.
====================

Link: https://lore.kernel.org/r/20231128115255.773377-1-jiri@resnulli.us
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agodevlink: warn about existing entities during reload-reinit
Jiri Pirko [Tue, 28 Nov 2023 11:52:55 +0000 (12:52 +0100)]
devlink: warn about existing entities during reload-reinit

During reload-reinit, all entities except for params, resources, regions
and health reporter should be removed and re-added. Add a warning to
be triggered in case the driver behaves differently.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agoDocumentation: devlink: extend reload-reinit description
Jiri Pirko [Tue, 28 Nov 2023 11:52:54 +0000 (12:52 +0100)]
Documentation: devlink: extend reload-reinit description

Be more explicit about devlink entities that may stay and that have to
be removed during reload reinit action.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agobpf/tests: Remove duplicate JSGT tests
Yujie Liu [Thu, 30 Nov 2023 03:40:18 +0000 (11:40 +0800)]
bpf/tests: Remove duplicate JSGT tests

It seems unnecessary that JSGT is tested twice (one before JSGE and one
after JSGE) since others are tested only once. Remove the duplicate JSGT
tests.

Fixes: 0bbaa02b4816 ("bpf/tests: Add tests to check source register zero-extension")
Signed-off-by: Yujie Liu <yujie.liu@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Johan Almbladh <johan.almbladh@anyfinetworks.com>
Link: https://lore.kernel.org/bpf/20231130034018.2144963-1-yujie.liu@intel.com
9 months agoMerge branch 'net-ravb-fixes-for-the-ravb-driver'
Paolo Abeni [Thu, 30 Nov 2023 09:59:10 +0000 (10:59 +0100)]
Merge branch 'net-ravb-fixes-for-the-ravb-driver'

Claudiu Beznea says:

====================
net: ravb: Fixes for the ravb driver

This series adds some fixes for ravb driver. Patches in this series
were initilly part of series at [1].

Changes in v2:
- in description of patch 1/6 documented the addition of
  out_free_netdev goto label
- collected tags
- s/out_runtime_disable/out_rpm_disable in patch 2/6
- fixed typos in description of patch 6/6

Changes since [1]:
- addressed review comments
- added patch 6/6

[1] https://lore.kernel.org/all/20231120084606.4083194-1-claudiu.beznea.uj@bp.renesas.com/
====================

Link: https://lore.kernel.org/r/20231128080439.852467-1-claudiu.beznea.uj@bp.renesas.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agonet: ravb: Keep reverse order of operations in ravb_remove()
Claudiu Beznea [Tue, 28 Nov 2023 08:04:39 +0000 (10:04 +0200)]
net: ravb: Keep reverse order of operations in ravb_remove()

On RZ/G3S SMARC Carrier II board having RGMII connections b/w Ethernet
MACs and PHYs it has been discovered that doing unbind/bind for ravb
driver in a loop leads to wrong speed and duplex for Ethernet links and
broken connectivity (the connectivity cannot be restored even with
bringing interface down/up). Before doing unbind/bind the Ethernet
interfaces were configured though systemd. The sh instructions used to
do unbind/bind were:

$ cd /sys/bus/platform/drivers/ravb/
$ while :; do echo 11c30000.ethernet > unbind ; \
  echo 11c30000.ethernet > bind; done

It has been discovered that there is a race b/w IOCTLs initialized by
systemd at the response of success binding and the
"ravb_write(ndev, CCC_OPC_RESET, CCC)" call in ravb_remove() as
follows:

1/ as a result of bind success the user space open/configures the
   interfaces tough an IOCTL; the following stack trace has been
   identified on RZ/G3S:

Call trace:
dump_backtrace+0x9c/0x100
show_stack+0x20/0x38
dump_stack_lvl+0x48/0x60
dump_stack+0x18/0x28
ravb_open+0x70/0xa58
__dev_open+0xf4/0x1e8
__dev_change_flags+0x198/0x218
dev_change_flags+0x2c/0x80
devinet_ioctl+0x640/0x708
inet_ioctl+0x1e4/0x200
sock_do_ioctl+0x50/0x108
sock_ioctl+0x240/0x358
__arm64_sys_ioctl+0xb0/0x100
invoke_syscall+0x50/0x128
el0_svc_common.constprop.0+0xc8/0xf0
do_el0_svc+0x24/0x38
el0_svc+0x34/0xb8
el0t_64_sync_handler+0xc0/0xc8
el0t_64_sync+0x190/0x198

2/ this call may execute concurrently with ravb_remove() as the
   unbind/bind operation was executed in a loop
3/ if the operation mode is changed to RESET (through
   ravb_write(ndev, CCC_OPC_RESET, CCC) call in ravb_remove())
   while the above ravb_open() is in progress it may lead to MAC
   (or PHY, or MAC-PHY connection, the right point hasn't been identified
   at the moment) to be broken, thus the Ethernet connectivity fails to
   restore.

The simple fix for this is to move ravb_write(ndev, CCC_OPC_RESET, CCC))
after unregister_netdev() to avoid resetting the controller while the
netdev interface is still registered.

To avoid future issues in ravb_remove(), the patch follows the proper order
of operations in ravb_remove(): reverse order compared with ravb_probe().
This avoids described races as the IOCTLs as well as unregister_netdev()
(called now at the beginning of ravb_remove()) calls rtnl_lock() before
continuing and IOCTLs check (though devinet_ioctl()) if device is still
registered just after taking the lock:

int devinet_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr)
{
// ...

        rtnl_lock();

        ret = -ENODEV;
        dev = __dev_get_by_name(net, ifr->ifr_name);
        if (!dev)
                goto done;

// ...
done:
        rtnl_unlock();
out:
        return ret;
}

Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agonet: ravb: Stop DMA in case of failures on ravb_open()
Claudiu Beznea [Tue, 28 Nov 2023 08:04:38 +0000 (10:04 +0200)]
net: ravb: Stop DMA in case of failures on ravb_open()

In case ravb_phy_start() returns with error the settings applied in
ravb_dmac_init() are not reverted (e.g. config mode). For this call
ravb_stop_dma() on failure path of ravb_open().

Fixes: a0d2f20650e8 ("Renesas Ethernet AVB PTP clock driver")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agonet: ravb: Start TX queues after HW initialization succeeded
Claudiu Beznea [Tue, 28 Nov 2023 08:04:37 +0000 (10:04 +0200)]
net: ravb: Start TX queues after HW initialization succeeded

ravb_phy_start() may fail. If that happens, the TX queues will remain
started. Thus, move the netif_tx_start_all_queues() after PHY is
successfully initialized.

Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agonet: ravb: Make write access to CXR35 first before accessing other EMAC registers
Claudiu Beznea [Tue, 28 Nov 2023 08:04:36 +0000 (10:04 +0200)]
net: ravb: Make write access to CXR35 first before accessing other EMAC registers

Hardware manual of RZ/G3S (and RZ/G2L) specifies the following on the
description of CXR35 register (chapter "PHY interface select register
(CXR35)"): "After release reset, make write-access to this register before
making write-access to other registers (except MDIOMOD). Even if not need
to change the value of this register, make write-access to this register
at least one time. Because RGMII/MII MODE is recognized by accessing this
register".

The setup procedure for EMAC module (chapter "Setup procedure" of RZ/G3S,
RZ/G2L manuals) specifies the E-MAC.CXR35 register is the first EMAC
register that is to be configured.

Note [A] from chapter "PHY interface select register (CXR35)" specifies
the following:
[A] The case which CXR35 SEL_XMII is used for the selection of RGMII/MII
in APB Clock 100 MHz.
(1) To use RGMII interface, Set ‘H’03E8_0000’ to this register.
(2) To use MII interface, Set ‘H’03E8_0002’ to this register.

Take into account these indication.

Fixes: 1089877ada8d ("ravb: Add RZ/G2L MII interface support")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agonet: ravb: Use pm_runtime_resume_and_get()
Claudiu Beznea [Tue, 28 Nov 2023 08:04:35 +0000 (10:04 +0200)]
net: ravb: Use pm_runtime_resume_and_get()

pm_runtime_get_sync() may return an error. In case it returns with an error
dev->power.usage_count needs to be decremented. pm_runtime_resume_and_get()
takes care of this. Thus use it.

Fixes: c156633f1353 ("Renesas Ethernet AVB driver proper")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agonet: ravb: Check return value of reset_control_deassert()
Claudiu Beznea [Tue, 28 Nov 2023 08:04:34 +0000 (10:04 +0200)]
net: ravb: Check return value of reset_control_deassert()

reset_control_deassert() could return an error. Some devices cannot work
if reset signal de-assert operation fails. To avoid this check the return
code of reset_control_deassert() in ravb_probe() and take proper action.

Along with it, the free_netdev() call from the error path was moved after
reset_control_assert() on its own label (out_free_netdev) to free
netdev in case reset_control_deassert() fails.

Fixes: 0d13a1a464a0 ("ravb: Add reset support")
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
9 months agoMerge branch 'clean-up-and-refactor-cookie_v46_check'
Jakub Kicinski [Thu, 30 Nov 2023 04:16:42 +0000 (20:16 -0800)]
Merge branch 'clean-up-and-refactor-cookie_v46_check'

Kuniyuki Iwashima says:

====================
tcp: Clean up and refactor cookie_v[46]_check().

This is a preparation series for upcoming arbitrary SYN Cookie
support with BPF. [0]

There are slight differences between cookie_v[46]_check().  Such a
discrepancy caused an issue in the past, and BPF SYN Cookie support
will add more churn.

The primary purpose of this series is to clean up and refactor
cookie_v[46]_check() to minimise such discrepancies and make the
BPF series easier to review.

[0]: https://lore.kernel.org/netdev/20231121184245.69569-1-kuniyu@amazon.com/
v2: https://lore.kernel.org/netdev/20231125011638.72056-1-kuniyu@amazon.com/
v1: https://lore.kernel.org/netdev/20231123012521.62841-1-kuniyu@amazon.com/
====================

Link: https://lore.kernel.org/r/20231129022924.96156-1-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agotcp: Factorise cookie-dependent fields initialisation in cookie_v[46]_check()
Kuniyuki Iwashima [Wed, 29 Nov 2023 02:29:24 +0000 (18:29 -0800)]
tcp: Factorise cookie-dependent fields initialisation in cookie_v[46]_check()

We will support arbitrary SYN Cookie with BPF, and then kfunc at
TC will preallocate reqsk and initialise some fields that should
not be overwritten later by cookie_v[46]_check().

To simplify the flow in cookie_v[46]_check(), we move such fields'
initialisation to cookie_tcp_reqsk_alloc() and factorise non-BPF
SYN Cookie handling into cookie_tcp_check(), where we validate the
cookie and allocate reqsk, as done by kfunc later.

Note that we set ireq->ecn_ok in two steps, the latter of which will
be shared by the BPF case.  As cookie_ecn_ok() is one-liner, now
it's inlined.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20231129022924.96156-9-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agotcp: Factorise cookie-independent fields initialisation in cookie_v[46]_check().
Kuniyuki Iwashima [Wed, 29 Nov 2023 02:29:23 +0000 (18:29 -0800)]
tcp: Factorise cookie-independent fields initialisation in cookie_v[46]_check().

We will support arbitrary SYN Cookie with BPF, and then some reqsk fields
are initialised in kfunc, and others are done in cookie_v[46]_check().

This patch factorises the common part as cookie_tcp_reqsk_init() and
calls it in cookie_tcp_reqsk_alloc() to minimise the discrepancy between
cookie_v[46]_check().

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20231129022924.96156-8-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agotcp: Move TCP-AO bits from cookie_v[46]_check() to tcp_ao_syncookie().
Kuniyuki Iwashima [Wed, 29 Nov 2023 02:29:22 +0000 (18:29 -0800)]
tcp: Move TCP-AO bits from cookie_v[46]_check() to tcp_ao_syncookie().

We initialise treq->af_specific in cookie_tcp_reqsk_alloc() so that
we can look up a key later in tcp_create_openreq_child().

Initially, that change was added for MD5 by commit ba5a4fdd63ae ("tcp:
make sure treq->af_specific is initialized"), but it has not been used
since commit d0f2b7a9ca0a ("tcp: Disable header prediction for MD5
flow.").

Now, treq->af_specific is used only by TCP-AO, so, we can move that
initialisation into tcp_ao_syncookie().

In addition to that, l3index in cookie_v[46]_check() is only used for
tcp_ao_syncookie(), so let's move it as well.

While at it, we move down tcp_ao_syncookie() in cookie_v4_check() so
that it will be called after security_inet_conn_request() to make
functions order consistent with cookie_v6_check().

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20231129022924.96156-7-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agotcp: Don't initialise tp->tsoffset in tcp_get_cookie_sock().
Kuniyuki Iwashima [Wed, 29 Nov 2023 02:29:21 +0000 (18:29 -0800)]
tcp: Don't initialise tp->tsoffset in tcp_get_cookie_sock().

When we create a full socket from SYN Cookie, we initialise
tcp_sk(sk)->tsoffset redundantly in tcp_get_cookie_sock() as
the field is inherited from tcp_rsk(req)->ts_off.

  cookie_v[46]_check
  |- treq->ts_off = 0
  `- tcp_get_cookie_sock
     |- tcp_v[46]_syn_recv_sock
     |  `- tcp_create_openreq_child
     |    `- newtp->tsoffset = treq->ts_off
     `- tcp_sk(child)->tsoffset = tsoff

Let's initialise tcp_rsk(req)->ts_off with the correct offset
and remove the second initialisation of tcp_sk(sk)->tsoffset.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20231129022924.96156-6-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agotcp: Don't pass cookie to __cookie_v[46]_check().
Kuniyuki Iwashima [Wed, 29 Nov 2023 02:29:20 +0000 (18:29 -0800)]
tcp: Don't pass cookie to __cookie_v[46]_check().

tcp_hdr(skb) and SYN Cookie are passed to __cookie_v[46]_check(), but
none of the callers passes cookie other than ntohl(th->ack_seq) - 1.

Let's fetch it in __cookie_v[46]_check() instead of passing the cookie
over and over.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20231129022924.96156-5-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agotcp: Clean up goto labels in cookie_v[46]_check().
Kuniyuki Iwashima [Wed, 29 Nov 2023 02:29:19 +0000 (18:29 -0800)]
tcp: Clean up goto labels in cookie_v[46]_check().

We will support arbitrary SYN Cookie with BPF, and then reqsk
will be preallocated before cookie_v[46]_check().

Depending on how validation fails, we send RST or just drop skb.

To make the error handling easier, let's clean up goto labels.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20231129022924.96156-4-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agotcp: Cache sock_net(sk) in cookie_v[46]_check().
Kuniyuki Iwashima [Wed, 29 Nov 2023 02:29:18 +0000 (18:29 -0800)]
tcp: Cache sock_net(sk) in cookie_v[46]_check().

sock_net(sk) is used repeatedly in cookie_v[46]_check().
Let's cache it in a variable.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20231129022924.96156-3-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agotcp: Clean up reverse xmas tree in cookie_v[46]_check().
Kuniyuki Iwashima [Wed, 29 Nov 2023 02:29:17 +0000 (18:29 -0800)]
tcp: Clean up reverse xmas tree in cookie_v[46]_check().

We will grow and cut the xmas tree in cookie_v[46]_check().
This patch cleans it up to make later patches tidy.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20231129022924.96156-2-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agonet: mana: Fix spelling mistake "enforecement" -> "enforcement"
Colin Ian King [Tue, 28 Nov 2023 09:53:04 +0000 (09:53 +0000)]
net: mana: Fix spelling mistake "enforecement" -> "enforcement"

There is a spelling mistake in struct field hc_tx_err_sqpdid_enforecement.
Fix it.

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: Shradha Gupta <shradhagupta@linux.microsoft.com>
Link: https://lore.kernel.org/r/20231128095304.515492-1-colin.i.king@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agonet: libwx: fix memory leak on msix entry
Jiawen Wu [Tue, 28 Nov 2023 09:59:28 +0000 (17:59 +0800)]
net: libwx: fix memory leak on msix entry

Since pci_free_irq_vectors() set pdev->msix_enabled as 0 in the
calling of pci_msix_shutdown(), wx->msix_entries is never freed.
Reordering the lines to fix the memory leak.

Cc: stable@vger.kernel.org
Fixes: 3f703186113f ("net: libwx: Add irq flow functions")
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Link: https://lore.kernel.org/r/20231128095928.1083292-1-jiawenwu@trustnetic.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agonet: dsa: sja1105: Use units.h instead of the copy of a definition
Andy Shevchenko [Tue, 28 Nov 2023 17:50:27 +0000 (19:50 +0200)]
net: dsa: sja1105: Use units.h instead of the copy of a definition

BYTES_PER_KBIT is defined in units.h, use that definition.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/20231128175027.394754-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoMerge branch 'mptcp-more-selftest-coverage-and-code-cleanup-for-net-next'
Jakub Kicinski [Thu, 30 Nov 2023 04:10:27 +0000 (20:10 -0800)]
Merge branch 'mptcp-more-selftest-coverage-and-code-cleanup-for-net-next'

Mat Martineau says:

====================
mptcp: More selftest coverage and code cleanup for net-next

Patches 1-5 and 7-8 add selftest coverage (and an associated subflow
counter in the kernel) to validate the recently-updated handling of
subflows with ID 0.

Patch 6 renames a label in the userspace path manager for clarity.

Patches 9-11 and 13-15 factor out common selftest code by moving certain
functions to mptcp_lib.sh

Patch 12 makes sure the random data file generated for selftest
payloads has the intended size.

v3: https://lore.kernel.org/r/20231115-send-net-next-2023107-v3-0-1ef58145a882@kernel.org
v2: https://lore.kernel.org/r/20231114-send-net-next-2023107-v2-0-b650a477362c@kernel.org
v1: https://lore.kernel.org/r/20231027-send-net-next-2023107-v1-0-03eff9452957@kernel.org
====================

Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-0-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: add mptcp_lib_wait_local_port_listen
Geliang Tang [Tue, 28 Nov 2023 23:18:59 +0000 (15:18 -0800)]
selftests: mptcp: add mptcp_lib_wait_local_port_listen

To avoid duplicated code in different MPTCP selftests, we can add
and use helpers defined in mptcp_lib.sh.

wait_local_port_listen() helper is defined in diag.sh, mptcp_connect.sh,
mptcp_join.sh and simult_flows.sh, export it into mptcp_lib.sh and
rename it with mptcp_lib_ prefix. Use this new helper in all these
scripts.

Note: We only have IPv4 connections in this helper, not looking at IPv6
(tcp6) but that's OK because we only have IPv4 connections here in diag.sh.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-15-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: add mptcp_lib_check_transfer
Geliang Tang [Tue, 28 Nov 2023 23:18:58 +0000 (15:18 -0800)]
selftests: mptcp: add mptcp_lib_check_transfer

To avoid duplicated code in different MPTCP selftests, we can add
and use helpers defined in mptcp_lib.sh.

check_transfer() and print_file_err() helpers are defined both in
mptcp_connect.sh and mptcp_sockopt.sh, export them into mptcp_lib.sh
and rename them with mptcp_lib_ prefix. And use them in all scripts.

Note: In mptcp_sockopt.sh it is OK to drop 'ret=1' in check_transfer()
because it will be set in run_tests() anyway.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-14-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: add mptcp_lib_make_file
Geliang Tang [Tue, 28 Nov 2023 23:18:57 +0000 (15:18 -0800)]
selftests: mptcp: add mptcp_lib_make_file

To avoid duplicated code in different MPTCP selftests, we can add
and use helpers defined in mptcp_lib.sh.

make_file() helper in mptcp_sockopt.sh and userspace_pm.sh are the same.
Export it into mptcp_lib.sh and rename it as mptcp_lib_kill_wait(). Use
it in both mptcp_connect.sh and mptcp_join.sh.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-13-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: add missing oflag=append
Geliang Tang [Tue, 28 Nov 2023 23:18:56 +0000 (15:18 -0800)]
selftests: mptcp: add missing oflag=append

In mptcp_connect.sh we are missing something like "oflag=append"
because this will write "${rem}" bytes at the beginning of the file
where there is already some random bytes. It should write that at
the end.

This patch adds this missing 'oflag=append' flag for 'dd' command in
make_file().

Suggested-by: Matthieu Baerts <matttbe@kernel.org>
Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-12-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: add mptcp_lib_get_counter
Geliang Tang [Tue, 28 Nov 2023 23:18:55 +0000 (15:18 -0800)]
selftests: mptcp: add mptcp_lib_get_counter

To avoid duplicated code in different MPTCP selftests, we can add
and use helpers defined in mptcp_lib.sh.

The helper get_counter() in mptcp_join.sh and get_mib_counter() in
mptcp_connect.sh have the same functionality, export get_counter() into
mptcp_lib.sh and rename it as mptcp_lib_get_counter(). Use this new
helper instead of get_counter() and get_mib_counter().

Use this helper in test_prio() in userspace_pm.sh too instead of
open-coding.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-11-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: add mptcp_lib_is_v6
Geliang Tang [Tue, 28 Nov 2023 23:18:54 +0000 (15:18 -0800)]
selftests: mptcp: add mptcp_lib_is_v6

To avoid duplicated code in different MPTCP selftests, we can add
and use helpers defined in mptcp_lib.sh.

is_v6() helper is defined in mptcp_connect.sh, mptcp_join.sh and
mptcp_sockopt.sh, so export it into mptcp_lib.sh and rename it as
mptcp_lib_is_v6(). Use this new helper in all scripts.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-10-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: add mptcp_lib_kill_wait
Geliang Tang [Tue, 28 Nov 2023 23:18:53 +0000 (15:18 -0800)]
selftests: mptcp: add mptcp_lib_kill_wait

To avoid duplicated code in different MPTCP selftests, we can add
and use helpers defined in mptcp_lib.sh.

Export kill_wait() helper in userspace_pm.sh into mptcp_lib.sh and
rename it as mptcp_lib_kill_wait(). It can be used to instead of
kill_wait() in mptcp_join.sh. Use the new helper in both scripts.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-9-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: userspace pm send RM_ADDR for ID 0
Geliang Tang [Tue, 28 Nov 2023 23:18:52 +0000 (15:18 -0800)]
selftests: mptcp: userspace pm send RM_ADDR for ID 0

This patch adds a selftest for userspace PM to remove id 0 address.

Use userspace_pm_add_addr() helper to add an id 10 address, then use
userspace_pm_rm_addr() helper to remove id 0 address.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-8-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: userspace pm remove initial subflow
Geliang Tang [Tue, 28 Nov 2023 23:18:51 +0000 (15:18 -0800)]
selftests: mptcp: userspace pm remove initial subflow

This patch adds a selftest for userspace PM to remove the initial
subflow.

Use userspace_pm_add_sf() to add a subflow, and pass initial IP address
to userspace_pm_rm_sf() to remove the initial subflow.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-7-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomptcp: userspace pm rename remove_err to out
Geliang Tang [Tue, 28 Nov 2023 23:18:50 +0000 (15:18 -0800)]
mptcp: userspace pm rename remove_err to out

The value of 'err' will not be only '-EINVAL', but can be '0' in some
cases.

So it's better to rename the label 'remove_err' to 'out' to avoid
confusions.

Suggested-by: Matthieu Baerts <matttbe@kernel.org>
Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-6-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: userspace pm create id 0 subflow
Geliang Tang [Tue, 28 Nov 2023 23:18:49 +0000 (15:18 -0800)]
selftests: mptcp: userspace pm create id 0 subflow

This patch adds a selftest to create id 0 subflow. Pass id 0 to the
helper userspace_pm_add_sf() to create id 0 subflow. chk_mptcp_info
shows one subflow but chk_subflows_total shows two subflows in each
namespace.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-5-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: update userspace pm test helpers
Geliang Tang [Tue, 28 Nov 2023 23:18:48 +0000 (15:18 -0800)]
selftests: mptcp: update userspace pm test helpers

This patch adds a new argument namespace to userspace_pm_add_addr() and
userspace_pm_add_sf() to make these two helper more versatile.

Add two more versatile helpers for userspace pm remove subflow or address:
userspace_pm_rm_addr() and userspace_pm_rm_sf(). The original test helpers
userspace_pm_rm_sf_addr_ns1() and userspace_pm_rm_sf_addr_ns2() can be
replaced by these new helpers.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-4-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: add chk_subflows_total helper
Geliang Tang [Tue, 28 Nov 2023 23:18:47 +0000 (15:18 -0800)]
selftests: mptcp: add chk_subflows_total helper

This patch adds a new helper chk_subflows_total(), in it use the newly
added counter mptcpi_subflows_total to get the "correct" amount of
subflows, including the initial one.

To be compatible with old 'ss' or kernel versions not supporting this
counter, get the total subflows by listing TCP connections that are
MPTCP subflows:

    ss -ti state state established state syn-sent state syn-recv |
        grep -c tcp-ulp-mptcp.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-3-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoselftests: mptcp: add evts_get_info helper
Geliang Tang [Tue, 28 Nov 2023 23:18:46 +0000 (15:18 -0800)]
selftests: mptcp: add evts_get_info helper

This patch adds a new helper get_info_value(), using 'sed' command to
parse the value of the given item name in the line with the given keyword,
to make chk_mptcp_info() and pedit_action_pkts() more readable.

Also add another helper evts_get_info() to use get_info_value() to parse
the output of 'pm_nl_ctl events' command, to make all the userspace pm
selftests more readable, both in mptcp_join.sh and userspace_pm.sh.

Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-2-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomptcp: add mptcpi_subflows_total counter
Geliang Tang [Tue, 28 Nov 2023 23:18:45 +0000 (15:18 -0800)]
mptcp: add mptcpi_subflows_total counter

If the initial subflow has been removed, we cannot know without checking
other counters, e.g. ss -ti <filter> | grep -c tcp-ulp-mptcp or
getsockopt(SOL_MPTCP, MPTCP_FULL_INFO, ...) (or others except MPTCP_INFO
of course) and then check mptcp_subflow_data->num_subflows to get the
total amount of subflows.

This patch adds a new counter mptcpi_subflows_total in mptcpi_flags to
store the total amount of subflows, including the initial one. A new
helper __mptcp_has_initial_subflow() is added to check whether the
initial subflow has been removed or not. With this helper, we can then
compute the total amount of subflows from mptcp_info by doing something
like:

    mptcpi_subflows_total = mptcpi_subflows +
            __mptcp_has_initial_subflow(msk).

Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/428
Reviewed-by: Matthieu Baerts <matttbe@kernel.org>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <martineau@kernel.org>
Link: https://lore.kernel.org/r/20231128-send-net-next-2023107-v4-1-8d6b94150f6b@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agoMerge branch 'mlxsw-support-cff-flood-mode'
Jakub Kicinski [Thu, 30 Nov 2023 04:03:27 +0000 (20:03 -0800)]
Merge branch 'mlxsw-support-cff-flood-mode'

Petr Machata says:

====================
mlxsw: Support CFF flood mode

The registers to configure to initialize a flood table differ between the
controlled and CFF flood modes. In therefore needs to be an op. Add it,
hook up the current init to the existing families, and invoke the op.

PGT is an in-HW table that maps addresses to sets of ports. Then when some
HW process needs a set of ports as an argument, instead of embedding the
actual set in the dynamic configuration, what gets configured is the
address referencing the set. The HW then works with the appropriate PGT
entry.

Among other allocations, the PGT currently contains two large blocks for
bridge flooding: one for 802.1q and one for 802.1d. Within each of these
blocks are three tables, for unknown-unicast, multicast and broadcast
flooding:

      . . . |    802.1q    |    802.1d    | . . .
            | UC | MC | BC | UC | MC | BC |
             \______ _____/ \_____ ______/
                    v             v
                   FID flood vectors

Thus each FID (which corresponds to an 802.1d bridge or one VLAN in an
802.1q bridge) uses three flood vectors spread across a fairly large region
of PGT.

This way of organizing the flood table (called "controlled") is not very
flexible. E.g. to decrease a bridge scale and store more IP MC vectors, one
would need to completely rewrite the bridge PGT blocks, or resort to hacks
such as storing individual MC flood vectors into unused part of the bridge
table.

In order to address these shortcomings, Spectrum-2 and above support what
is called CFF flood mode, for Compressed FID Flooding. In CFF flood mode,
each FID has a little table of its own, with three entries adjacent to each
other, one for unknown-UC, one for MC, one for BC. This allows for a much
more fine-grained approach to PGT management, where bits of it are
allocated on demand.

      . . . | FID | FID | FID | FID | FID | . . .
            |U|M|B|U|M|B|U|M|B|U|M|B|U|M|B|
             \_____________ _____________/
                           v
                   FID flood vectors

Besides the FID table organization, the CFF flood mode also impacts Router
Subport (RSP) table. This table contains flood vectors for rFIDs, which are
FIDs that reference front panel ports or LAGs. The RSP table contains two
entries per front panel port and LAG, one for unknown-UC traffic, and one
for everything else. Currently, the FW allocates and manages the table in
its own part of PGT. rFIDs are marked with flood_rsp bit and managed
specially. In CFF mode, rFIDs are managed as all other FIDs. The driver
therefore has to allocate and maintain the flood vectors. Like with bridge
FIDs, this is more work, but increases flexibility of the system.

The FW currently supports both the controlled and CFF flood modes. To shed
complexity, in the future it should only support CFF flood mode. Hence this
patchset, which adds CFF flood mode support to mlxsw.

Since mlxsw needs to maintain both the controlled mode as well as CFF mode
support, we will keep the layout as compatible as possible. The bridge
tables will stay in the same overall shape, just their inner organization
will change from flood mode -> FID to FID -> flood mode. Likewise will RSP
be kept as a contiguous block of PGT memory, as was the case when the FW
maintained it.

- The way FIDs get configured under the CFF flood mode differs from the
  currently used controlled mode. The simple approach of having several
  globally visible arrays for spectrum.c to statically choose from no
  longer works.

  Patch #1 thus privatizes all FID initialization and finalization logic,
  and exposes it as ops instead.

- Patch #2 renames the ops that are specific to the controlled mode, to
  make room in the namespace for the CFF variants.

  Patch #3 extracts a helper to compute flood table base out of
  mlxsw_sp_fid_flood_table_mid().

- The op fid_setup configured fid_offset, i.e. the number of this FID
  within its family. For rFIDs in CFF mode, to determine this number, the
  driver will need to do fallible queries.

  Thus in patch #4, make the FID setup operation fallible as well.

- Flood mode initialization routine differs between the controlled and CFF
  flood modes. The controlled mode needs to configure flood table layout,
  which the CFF mode does not need to do.

  In patch #5, move mlxsw_sp_fid_flood_table_init() up so that the
  following patch can make use of it.

  In patch #6, add an op to be invoked per table (if defined).

- The current way of determining PGT allocation size depends on the number
  of FIDs and number of flood tables. RFIDs however have PGT footprint
  depending not on number of FIDs, but on number of ports and LAGs, because
  which ports an rFID should flood to does not depend on the FID itself,
  but on the port or LAG that it references.

  Therefore in patch #7, add FID family ops for determining PGT allocation
  size.

- As elaborated above, layout of PGT will differ between controlled and CFF
  flood modes. In CFF mode, it will further differ between rFIDs and other
  FIDs (as described at previous patch). The way to pack the SFMR register
  to configure a FID will likewise differ from controlled to CFF.

  Thus in patches #8 and #9 add FID family ops to determine PGT base
  address for a FID and to pack SFMR.

- Patches #10 and #11 add more bits for RSP support. In patch #10, add a
  new traffic type enumerator, for non-UC traffic. This is a combination of
  BC and MC traffic, but the way that mlxsw maps these mnemonic names to
  actual traffic type configurations requires that we have a new name to
  describe this class of traffic.

  Patch #11 then adds hooks necessary for RSP table maintenance. As ports
  come and go, and join and leave LAGs, it is necessary to update flood
  vectors that the rFIDs use. These new hooks will make that possible.

- Patches #12, #13 and #14 introduce flood profiles. These have been
  implicit so far, but the way that CFF flood mode works with profile IDs
  requires that we make them explicit.

  Thus in patch #12, introduce flood profile objects as a set of flood
  tables that FID families then refer to. The FID code currently only
  uses a single flood profile.

  In patch #13, add a flood profile ID to flood profile objects.

  In patch #14, when in CFF mode, configure SFFP according to the existing
  flood profiles (or the one that exists as of that point).

- Patches #15 and #16 add code to implement, respectively, bridge FIDs and
  RSP FIDs in CFF mode.

- In patch #17, toggle flood_mode_prefer_cff on Spectrum-2 and above, which
  makes the newly-added code live.
====================

Link: https://lore.kernel.org/r/cover.1701183891.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum: Use CFF mode where available
Petr Machata [Tue, 28 Nov 2023 15:50:50 +0000 (16:50 +0100)]
mlxsw: spectrum: Use CFF mode where available

Mark all Spectrum>2 systems as preferring CFF flood mode if supported by
the firmware.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/8a3d2ad96b943f7e3f53f998bd333a14e19cd641.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Add support for rFID family in CFF flood mode
Petr Machata [Tue, 28 Nov 2023 15:50:49 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Add support for rFID family in CFF flood mode

In this patch, add the artifacts for the rFID family that works in CFF
flood mode.

The same that was said about PGT organization and lookup in bridge FID
families applies for the rFID family as well. The main difference lies in
the fact that in the controlled flood mode, the FW was taking care of
maintaining the PGT tables for rFIDs. In CFF mode, the responsibility
shifts to the driver.

All rFIDs are based off either a front panel port, or a LAG port. For those
based off ports, we need to maintain at worst one PGT block for each port,
for those based off LAGs, one PGT block per LAG. This reflects in the
pgt_size callback, which determines the PGT footprint based on number of
ports and the LAG capacity.

A number of FIDs may end up using the same PGT base. Unlike with bridges,
where membership of a port in a given FID is highly dynamic, an rFID based
of a port will just always need to flood to that port.

Both the port and the LAG subtables need to be actively maintained. To that
end, the CFF rFID family implements fid_port_init and fid_port_fini
callbacks, which toggle the necessary bits.

Both FID-MID translation and SFMR packing then point into either the port
or the LAG subtable, to the block that corresponds to a given port or a
given LAG, depending on what port the RIF bound to the rFID uses.

As in the previous patch, the way CFF flood mode organizes PGT accesses
allows for much more smarts and dynamism. As in the previous patch, we
rather aim to keep things simple and static.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/962deb4367585d38250e80c685a34735c0c7f3ad.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Add a family for bridge FIDs in CFF flood mode
Petr Machata [Tue, 28 Nov 2023 15:50:48 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Add a family for bridge FIDs in CFF flood mode

In this patch, add the artifacts for 802.1d and 802.1q FID families that
work in CFF flood mode.

In CFF flood mode, the way flood vectors are looked up changes: there's a
per-FID PGT base, to which a small offset is added depending on type of
traffic. Thus each FID occupies a small contiguous block of PGT memory,
whereas in the controlled flood mode, flood vectors for a given FID were
spread across the PGT.

The term "flood table" as used by the spectrum_fid module, borrows from
controlled flood mode way of organizing the PGT table. There flood tables
were actual tables, contiguous in the PGT. In the CFF flood mode, they are
more abstract: a flood table becomes a collection of e.g. all first rows of
the per-FID PGT blocks. Nonetheless we retain the nomenclature.

FIDs are still configured through the SFMR register, but there are
different fields to set under CFF mode: PGT base and profile. Thus register
packing gets a dedicated op overload as well.

The new organization of PGT makes it possible to treat the PGT as a block
of an ordinary memory, allocate and deallocate on demand, and achieve
better flexibility. Here instead, we aim to keep the code as close as
possible to the previous controlled flood mode, support for which we need
to retain for Spectrum-1 and older FW versions anyway. Thus the PGT
footprint of the individual families is the same as before, just the
internal organization of the per-family PGT region differs. Hence the
pgt_size callback is reused between the controlled and CFF flood modes.

Since the dummy family has no flood tables in either the CTL mode or in
CFF mode, the existing one can be reused for the CFF family array.

Users should not notice any changes between the controlled and CFF flood
modes.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/ca40b8163e6d6a21f63ef299619acee953cf9519.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Initialize flood profiles in CFF mode
Petr Machata [Tue, 28 Nov 2023 15:50:47 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Initialize flood profiles in CFF mode

In CFF flood mode, the way flood vectors are looked up changes: there's a
per-FID PGT base, to which a small offset is added depending on type of
traffic. Thus each FID occupies a small contiguous block of PGT memory,
whereas in the controlled flood mode, flood vectors for a given FID were
spread across the PGT.

Each FID is associated with one of a handful of profiles. The profile and
the traffic type are then used as keys to look up the PGT offset. This
offset is then added to the per-FID PGT base. The profile / type / offset
mapping needs to be configured by the driver, and is only relevant in CFF
flood mode.

In this patch, add the SFFP initialization code. Only initialize the one
profile currently explicitly used. As follow-up patch add more profiles,
this code will pick them up and initialize as well.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/2c4733ed72d439444218969c032acad22cd4ed88.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Add profile_id to flood profile
Petr Machata [Tue, 28 Nov 2023 15:50:46 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Add profile_id to flood profile

In the CFF mode, flood profiles are identified by a unique numerical
identifier. This is used for configuration of FIDs and for configuration of
traffic-type to PGT offset rules. In both cases, the numerical identifier
serves as a handle for the flood profile. Add the identifier to the flood
profile structure.

There is currently only one flood profile in use explicitly, the one used
for all bridging. Eventually three will be necessary in total: one for
bridges, one for rFIDs, one for NVE underlay. A total of four profiles
are supported by the HW. Start allocating at 1, because 0 is currently
used for underlay NVE flood.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/19ea9c35ba8b522fa5f7eb6fd7bc1b68f0f66b41.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Add an object to keep flood profiles
Petr Machata [Tue, 28 Nov 2023 15:50:45 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Add an object to keep flood profiles

A flood profile is a mapping from traffic type to an offset at which a
flood vector should be looked up. In mlxsw so far, a flood profile was
somewhat implicitly represented by flood table array. When the CFF flood
mode will be introduced, the flood profile will become more explicit: each
will get a number and the profile ID / traffic-type / offset mapping will
actually need to be initialized in the hardware.

Therefore it is going to be handy to have a structure that keeps all the
components that compose a flood profile. Add this structure, currently with
just the flood table array bits. In the FID families that flood at all,
reference the flood profile instead of just the table array.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/15e113de114d3f41ce3fd2a14a2fa6a1b1d7e8f2.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Add hooks for RSP table maintenance
Petr Machata [Tue, 28 Nov 2023 15:50:44 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Add hooks for RSP table maintenance

In the CFF flood mode, the driver has to allocate a table within PGT, which
holds flood vectors for router subport FIDs. For LAGs, these flood vectors
have to obviously be maintained dynamically as port membership in a LAG
changes. But even for physical ports, the flood vectors have to be kept
valid, and may not contain enabled bits corresponding to non-existent
ports. It is therefore not possible to precompute the port part of the RSP
table, it has to be maintained as ports come and go due to splits.

To support the RSP table maintenance, add to FID ops two new ops:
fid_port_init and fid_port_fini, for when a port comes to existence, or
joins a lag, and vice versa. Invoke these ops from
mlxsw_sp_port_fids_init() and mlxsw_sp_port_fids_fini(), which are called
when port is added and removed, respectively. Also add two new hooks for
LAG maintenance, mlxsw_sp_fid_port_join_lag() / _leave_lag() which
transitively call into the same ops.

Later patches will actually add the op implementations themselves, this
just adds the scaffolding.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/234398a23540317abb25f74f920a5c8121faecf0.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Add a not-UC packet type
Petr Machata [Tue, 28 Nov 2023 15:50:43 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Add a not-UC packet type

In CFF flood mode, the rFID family will allocate two tables. One for
unknown UC traffic, one for everything else. Add a traffic type for the
everything else traffic.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/8fb968b2d1cc37137cd0110c98cdeb625b03ca99.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Add an op for packing SFMR
Petr Machata [Tue, 28 Nov 2023 15:50:42 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Add an op for packing SFMR

The way SFMR is packed differs between the controlled and CFF flood modes.
Add an op to dispatch it dynamically.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/f12fe7879a7086ee86343ee4db02c859f78f0534.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Add an op to get PGT address of a FID
Petr Machata [Tue, 28 Nov 2023 15:50:41 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Add an op to get PGT address of a FID

In the CFF flood mode, the way to determine a PGT address where a given FID
/ flood table resides is different from the controlled flood mode, which
mlxsw currently uses. Furthermore, this will differ between rFID family and
bridge families. The operation therefore needs to be dynamically
dispatched. To that end, add an op to FID-family ops.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Amit Cohen <amcohen@nvidia.com>
Link: https://lore.kernel.org/r/00e8f6ad79009a9a77a5c95d596ea9574776dc95.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Add an op to get PGT allocation size
Petr Machata [Tue, 28 Nov 2023 15:50:40 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Add an op to get PGT allocation size

In the CFF flood mode, the PGT allocation size of RFID family will not
depend on number of FIDs, but rather number of ports and LAGs. Therefore
introduce a FID family operation to calculate the PGT allocation size.

The way that size is calculated in the CFF mode depends on calling fallible
functions. Thus express the op as returning an int, with the size returned
via a pointer argument.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/1174651b7160fcedbef50010ae4b68201112fe6f.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Add an op for flood table initialization
Petr Machata [Tue, 28 Nov 2023 15:50:39 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Add an op for flood table initialization

In controlled flood mode, for each bridge FID family (i.e., 802.1Q and
802.1D) and packet type (i.e., UUC/MC/BC), the hardware needs to be told
which PGT address to use as the base address for the flood table and how
to determine the offset from the base for each FID.

The above is not needed in CFF mode where each FID has its own flood
table instead of the FID family itself.

Therefore, create a new FID family operation for the above configuration
and only implement it for the 802.1Q and 802.1D families in controlled
flood mode.

No functional changes intended.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/06f71415eec75811585ec597e1dd101b6dff77e7.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Move mlxsw_sp_fid_flood_table_init() up
Petr Machata [Tue, 28 Nov 2023 15:50:38 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Move mlxsw_sp_fid_flood_table_init() up

Move the function to the point where it will need to be to be visible for
the 802.1d ops.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/aef09e26b0c2dd077531e665d7135b300bdaf0a8.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
9 months agomlxsw: spectrum_fid: Make mlxsw_sp_fid_ops.setup return an int
Petr Machata [Tue, 28 Nov 2023 15:50:37 +0000 (16:50 +0100)]
mlxsw: spectrum_fid: Make mlxsw_sp_fid_ops.setup return an int

This operation will be fallible for rFIDs in CFF mode, which will be
introduced in follow-up patches. Have it return an int, and handle
the failures in the caller.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/75f1b85c0cb86bea5501fcc8657042f221a78b32.1701183892.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>