platform/kernel/linux-starfive.git
3 years agonet: dlink: remove leading spaces before tabs
Hui Tang [Wed, 19 May 2021 05:30:41 +0000 (13:30 +0800)]
net: dlink: remove leading spaces before tabs

There are a few leading spaces before tabs and remove it by running the
following commard:

$ find . -name '*.c' | xargs sed -r -i 's/^[ ]+\t/\t/'
$ find . -name '*.h' | xargs sed -r -i 's/^[ ]+\t/\t/'
Cc: "Alexander A. Klimov" <grandmaster@al2klimov.de>
Signed-off-by: Hui Tang <tanghui20@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dec: remove leading spaces before tabs
Hui Tang [Wed, 19 May 2021 05:30:40 +0000 (13:30 +0800)]
net: dec: remove leading spaces before tabs

There are a few leading spaces before tabs and remove it by running the
following commard:

$ find . -name '*.c' | xargs sed -r -i 's/^[ ]+\t/\t/'
$ find . -name '*.h' | xargs sed -r -i 's/^[ ]+\t/\t/'

Signed-off-by: Hui Tang <tanghui20@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: chelsio: remove leading spaces before tabs
Hui Tang [Wed, 19 May 2021 05:30:39 +0000 (13:30 +0800)]
net: chelsio: remove leading spaces before tabs

There are a few leading spaces before tabs and remove it by running the
following commard:

$ find . -name '*.c' | xargs sed -r -i 's/^[ ]+\t/\t/'
$ find . -name '*.h' | xargs sed -r -i 's/^[ ]+\t/\t/'

Cc: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: Hui Tang <tanghui20@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: broadcom: remove leading spaces before tabs
Hui Tang [Wed, 19 May 2021 05:30:38 +0000 (13:30 +0800)]
net: broadcom: remove leading spaces before tabs

There are a few leading spaces before tabs and remove it by running the
following commard:

$ find . -name '*.c' | xargs sed -r -i 's/^[ ]+\t/\t/'
$ find . -name '*.h' | xargs sed -r -i 's/^[ ]+\t/\t/'

Cc: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Hui Tang <tanghui20@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: apple: remove leading spaces before tabs
Hui Tang [Wed, 19 May 2021 05:30:37 +0000 (13:30 +0800)]
net: apple: remove leading spaces before tabs

There are a few leading spaces before tabs and remove it by running the
following commard:

$ find . -name '*.c' | xargs sed -r -i 's/^[ ]+\t/\t/'
$ find . -name '*.h' | xargs sed -r -i 's/^[ ]+\t/\t/'

Signed-off-by: Hui Tang <tanghui20@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: amd: remove leading spaces before tabs
Hui Tang [Wed, 19 May 2021 05:30:36 +0000 (13:30 +0800)]
net: amd: remove leading spaces before tabs

There are a few leading spaces before tabs and remove it by running the
following commard:

$ find . -name '*.c' | xargs sed -r -i 's/^[ ]+\t/\t/'
$ find . -name '*.h' | xargs sed -r -i 's/^[ ]+\t/\t/'
Signed-off-by: Hui Tang <tanghui20@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: alteon: remove leading spaces before tabs
Hui Tang [Wed, 19 May 2021 05:30:35 +0000 (13:30 +0800)]
net: alteon: remove leading spaces before tabs

There are a few leading spaces before tabs and remove it by running the
following commard:

$ find . -name '*.c' | xargs sed -r -i 's/^[ ]+\t/\t/'
$ find . -name '*.h' | xargs sed -r -i 's/^[ ]+\t/\t/'

Cc: Jes Sorensen <jes@trained-monkey.org>
Signed-off-by: Hui Tang <tanghui20@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: 3com: remove leading spaces before tabs
Hui Tang [Wed, 19 May 2021 05:30:34 +0000 (13:30 +0800)]
net: 3com: remove leading spaces before tabs

There are a few leading space before tabs and remove it by running the
following commard:

$ find . -name '*.c' | xargs sed -r -i 's/^[ ]+\t/\t/'
$ find . -name '*.h' | xargs sed -r -i 's/^[ ]+\t/\t/'

Cc: Steffen Klassert <klassert@kernel.org>
Signed-off-by: Hui Tang <tanghui20@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agotun: use DEVICE_ATTR_RO macro
YueHaibing [Wed, 19 May 2021 02:38:50 +0000 (10:38 +0800)]
tun: use DEVICE_ATTR_RO macro

Use DEVICE_ATTR_RO helper instead of plain DEVICE_ATTR,
which makes the code a bit shorter and easier to read.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoibmveth: fix kobj_to_dev.cocci warnings
YueHaibing [Wed, 19 May 2021 02:28:49 +0000 (10:28 +0800)]
ibmveth: fix kobj_to_dev.cocci warnings

Use kobj_to_dev() instead of container_of()

Generated by: scripts/coccinelle/api/kobj_to_dev.cocci

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Lijun Pan <lijunp213@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: mdio: provide shim implementation of devm_of_mdiobus_register
Vladimir Oltean [Tue, 18 May 2021 17:49:24 +0000 (20:49 +0300)]
net: mdio: provide shim implementation of devm_of_mdiobus_register

Similar to the way in which of_mdiobus_register() has a fallback to the
non-DT based mdiobus_register() when CONFIG_OF is not set, we can create
a shim for the device-managed devm_of_mdiobus_register() which calls
devm_mdiobus_register() and discards the struct device_node *.

In particular, this solves a build issue with the qca8k DSA driver which
uses devm_of_mdiobus_register and can be compiled without CONFIG_OF.

Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dcb: Remove unnecessary INIT_LIST_HEAD()
Yang Yingliang [Tue, 18 May 2021 13:03:58 +0000 (21:03 +0800)]
net: dcb: Remove unnecessary INIT_LIST_HEAD()

The list_head dcb_app_list is initialized statically.
It is unnecessary to initialize by INIT_LIST_HEAD().

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agocxgb4: clip_tbl: use list_del_init instead of list_del/INIT_LIST_HEAD
Yang Yingliang [Tue, 18 May 2021 13:01:35 +0000 (21:01 +0800)]
cxgb4: clip_tbl: use list_del_init instead of list_del/INIT_LIST_HEAD

Using list_del_init() instead of list_del() + INIT_LIST_HEAD()
to simpify the code.

Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'wan-cleanups'
David S. Miller [Tue, 18 May 2021 20:42:42 +0000 (13:42 -0700)]
Merge branch 'wan-cleanups'

Guangbin Huang says:

====================
net: wan: clean up some code style issues

This patchset clean up some code style issues.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: wan: fix variable definition style
Peng Li [Tue, 18 May 2021 12:29:54 +0000 (20:29 +0800)]
net: wan: fix variable definition style

Fix the checkpatch error: "foo* bar" should be "foo *bar".

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: wan: remove redundant space
Peng Li [Tue, 18 May 2021 12:29:53 +0000 (20:29 +0800)]
net: wan: remove redundant space

Space prohibited before that close parenthesis ')',
so removes the redundant space.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: wan: remove redundant braces {}
Peng Li [Tue, 18 May 2021 12:29:52 +0000 (20:29 +0800)]
net: wan: remove redundant braces {}

Braces {} are not necessary for single statement blocks,
this patch removes redundant braces {}.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: wan: add some required spaces
Peng Li [Tue, 18 May 2021 12:29:51 +0000 (20:29 +0800)]
net: wan: add some required spaces

Add space required before the open parenthesis '(',
and add spaces required around that '<', '>' and '!='.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: wan: remove redundant blank lines
Peng Li [Tue, 18 May 2021 12:29:50 +0000 (20:29 +0800)]
net: wan: remove redundant blank lines

This patch removes some redundant blank lines.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: qca8k: fix missing unlock on error in qca8k_vlan_(add|del)
Wei Yongjun [Tue, 18 May 2021 11:24:13 +0000 (11:24 +0000)]
net: dsa: qca8k: fix missing unlock on error in qca8k_vlan_(add|del)

Add the missing unlock before return from function qca8k_vlan_add()
and qca8k_vlan_del() in the error handling case.

Fixes: 028f5f8ef44f ("net: dsa: qca8k: handle error with qca8k_read operation")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agocipso: correct comments of cipso_v4_cache_invalidate()
Zheng Yejian [Tue, 18 May 2021 09:11:41 +0000 (17:11 +0800)]
cipso: correct comments of cipso_v4_cache_invalidate()

Since cipso_v4_cache_invalidate() has no return value, so drop
related descriptions in its comments.

Fixes: 446fda4f2682 ("[NetLabel]: CIPSOv4 engine")
Signed-off-by: Zheng Yejian <zhengyejian1@huawei.com>
Acked-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'custom-multipath-hash'
David S. Miller [Tue, 18 May 2021 20:27:32 +0000 (13:27 -0700)]
Merge branch 'custom-multipath-hash'

Ido Schimmel says:

====================
Add support for custom multipath hash

This patchset adds support for custom multipath hash policy for both
IPv4 and IPv6 traffic. The new policy allows user space to control the
outer and inner packet fields used for the hash computation.

Motivation
==========

Linux currently supports different multipath hash policies for IPv4 and
IPv6 traffic:

* Layer 3
* Layer 4
* Layer 3 or inner layer 3, if present

These policies hash on a fixed set of fields, which is inflexible and
against operators' requirements to control the hash input: "The ability
to control the inputs to the hash function should be a consideration in
any load-balancing RFP" [1].

An example of this inflexibility can be seen by the fact that none of
the current policies allows operators to use the standard 5-tuple and
the flow label for multipath hash computation. Such a policy is useful
in the following real-world example of a data center with the following
types of traffic:

* Anycast IPv6 TCP traffic towards layer 4 load balancers. Flow label is
constant (zero) to avoid breaking established connections

* Non-encapsulated IPv6 traffic. Flow label is used to re-route flows
around problematic (congested / failed) paths [2]

* IPv6 encapsulated traffic (IPv4-in-IPv6 or IPv6-in-IPv6). Outer flow
label is generated from encapsulated packet

* UDP encapsulated traffic. Outer source port is generated from
encapsulated packet

In the above example, using the inner flow information for hash
computation in addition to the outer flow information is useful during
failures of the BPF agent that selectively generates the flow label
based on the traffic type. In such cases, the self-healing properties of
the flow label are lost, but encapsulated flows are still load balanced.

Control over the inner fields is even more critical when encapsulation
is performed by hardware routers. For example, the Spectrum ASIC can
only encode 8 bits of entropy in the outer flow label / outer UDP source
port when performing IP / UDP encapsulation. In the case of IPv4 GRE
encapsulation there is no outer field to encode the inner hash in.

User interface
==============

In accordance with existing multipath hash configuration, the new custom
policy is added as a new option (3) to the
net.ipv{4,6}.fib_multipath_hash_policy sysctls. When the new policy is
used, the packet fields used for hash computation are determined by the
net.ipv{4,6}.fib_multipath_hash_fields sysctls. These sysctls accept a
bitmask according to the following table (from ip-sysctl.rst):

====== ============================
0x0001 Source IP address
0x0002 Destination IP address
0x0004 IP protocol
0x0008 Flow Label
0x0010 Source port
0x0020 Destination port
0x0040 Inner source IP address
0x0080 Inner destination IP address
0x0100 Inner IP protocol
0x0200 Inner Flow Label
0x0400 Inner source port
0x0800 Inner destination port
====== ============================

For example, to allow IPv6 traffic to be hashed based on standard
5-tuple and flow label:

 # sysctl -wq net.ipv6.fib_multipath_hash_fields=0x0037
 # sysctl -wq net.ipv6.fib_multipath_hash_policy=3

Implementation
==============

As with existing policies, the new policy relies on the flow dissector
to extract the packet fields for the hash computation. However, unlike
existing policies that either use the outer or inner flow, the new
policy might require both flows to be dissected.

To avoid unnecessary invocations of the flow dissector, the data path
skips dissection of the outer or inner flows if none of the outer or
inner fields are required.

In addition, inner flow dissection is not performed when no
encapsulation was encountered (i.e., 'FLOW_DIS_ENCAPSULATION' not set by
flow dissector) during dissection of the outer flow.

Testing
=======

Three new selftests are added with three different topologies that allow
testing of following traffic combinations:

* Non-encapsulated IPv4 / IPv6 traffic
* IPv4 / IPv6 overlay over IPv4 underlay
* IPv4 / IPv6 overlay over IPv6 underlay

All three tests follow the same pattern. Each time a different packet
field is used for hash computation. When the field changes in the packet
stream, traffic is expected to be balanced across the two paths. When
the field does not change, traffic is expected to be unbalanced across
the two paths.

Patchset overview
=================

Patches #1-#3 add custom multipath hash support for IPv4 traffic
Patches #4-#7 do the same for IPv6
Patches #8-#10 add selftests

Future work
===========

mlxsw support can be found here [3].

Changes since RFC v2 [4]:

* Patch #2: Document that 0x0008 is used for Flow Label
* Patch #2: Do not allow the bitmask to be zero
* Patch #6: Do not allow the bitmask to be zero

Changes since RFC v1 [5]:

* Use a bitmask instead of a bitmap

[1] https://blog.apnic.net/2018/01/11/ipv6-flow-label-misuse-hashing/
[2] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=3acf3ec3f4b0fd4263989f2e4227bbd1c42b5fe1
[3] https://github.com/idosch/linux/tree/submit/custom_hash_mlxsw_v2
[4] https://lore.kernel.org/netdev/20210509151615.200608-1-idosch@idosch.org/
[5] https://lore.kernel.org/netdev/20210502162257.3472453-1-idosch@idosch.org/
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: forwarding: Add test for custom multipath hash with IPv6 GRE
Ido Schimmel [Mon, 17 May 2021 18:15:26 +0000 (21:15 +0300)]
selftests: forwarding: Add test for custom multipath hash with IPv6 GRE

Test that when the hash policy is set to custom, traffic is distributed
only according to the inner fields set in the fib_multipath_hash_fields
sysctl.

Each time set a different field and make sure traffic is only
distributed when the field is changed in the packet stream.

The test only verifies the behavior of IPv4/IPv6 overlays on top of an
IPv6 underlay network. The previous patch verified the same with an IPv4
underlay network.

Example output:

 # ./ip6gre_custom_multipath_hash.sh
 TEST: ping                                                          [ OK ]
 TEST: ping6                                                         [ OK ]
 INFO: Running IPv4 overlay custom multipath hash tests
 TEST: Multipath hash field: Inner source IP (balanced)              [ OK ]
 INFO: Packets sent on path1 / path2: 6602 / 6002
 TEST: Multipath hash field: Inner source IP (unbalanced)            [ OK ]
 INFO: Packets sent on path1 / path2: 1 / 12601
 TEST: Multipath hash field: Inner destination IP (balanced)         [ OK ]
 INFO: Packets sent on path1 / path2: 6802 / 5801
 TEST: Multipath hash field: Inner destination IP (unbalanced)       [ OK ]
 INFO: Packets sent on path1 / path2: 12602 / 3
 TEST: Multipath hash field: Inner source port (balanced)            [ OK ]
 INFO: Packets sent on path1 / path2: 16431 / 16344
 TEST: Multipath hash field: Inner source port (unbalanced)          [ OK ]
 INFO: Packets sent on path1 / path2: 0 / 32773
 TEST: Multipath hash field: Inner destination port (balanced)       [ OK ]
 INFO: Packets sent on path1 / path2: 16431 / 16344
 TEST: Multipath hash field: Inner destination port (unbalanced)     [ OK ]
 INFO: Packets sent on path1 / path2: 2 / 32772
 INFO: Running IPv6 overlay custom multipath hash tests
 TEST: Multipath hash field: Inner source IP (balanced)              [ OK ]
 INFO: Packets sent on path1 / path2: 6704 / 5902
 TEST: Multipath hash field: Inner source IP (unbalanced)            [ OK ]
 INFO: Packets sent on path1 / path2: 1 / 12600
 TEST: Multipath hash field: Inner destination IP (balanced)         [ OK ]
 INFO: Packets sent on path1 / path2: 5751 / 6852
 TEST: Multipath hash field: Inner destination IP (unbalanced)       [ OK ]
 INFO: Packets sent on path1 / path2: 12602 / 0
 TEST: Multipath hash field: Inner flowlabel (balanced)              [ OK ]
 INFO: Packets sent on path1 / path2: 8272 / 8181
 TEST: Multipath hash field: Inner flowlabel (unbalanced)            [ OK ]
 INFO: Packets sent on path1 / path2: 3 / 12602
 TEST: Multipath hash field: Inner source port (balanced)            [ OK ]
 INFO: Packets sent on path1 / path2: 16424 / 16351
 TEST: Multipath hash field: Inner source port (unbalanced)          [ OK ]
 INFO: Packets sent on path1 / path2: 3 / 32774
 TEST: Multipath hash field: Inner destination port (balanced)       [ OK ]
 INFO: Packets sent on path1 / path2: 16425 / 16350
 TEST: Multipath hash field: Inner destination port (unbalanced)     [ OK ]
 INFO: Packets sent on path1 / path2: 2 / 32773

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: forwarding: Add test for custom multipath hash with IPv4 GRE
Ido Schimmel [Mon, 17 May 2021 18:15:25 +0000 (21:15 +0300)]
selftests: forwarding: Add test for custom multipath hash with IPv4 GRE

Test that when the hash policy is set to custom, traffic is distributed
only according to the inner fields set in the fib_multipath_hash_fields
sysctl.

Each time set a different field and make sure traffic is only
distributed when the field is changed in the packet stream.

The test only verifies the behavior of IPv4/IPv6 overlays on top of an
IPv4 underlay network. A subsequent patch will do the same with an IPv6
underlay network.

Example output:

 # ./gre_custom_multipath_hash.sh
 TEST: ping                                                          [ OK ]
 TEST: ping6                                                         [ OK ]
 INFO: Running IPv4 overlay custom multipath hash tests
 TEST: Multipath hash field: Inner source IP (balanced)              [ OK ]
 INFO: Packets sent on path1 / path2: 6601 / 6001
 TEST: Multipath hash field: Inner source IP (unbalanced)            [ OK ]
 INFO: Packets sent on path1 / path2: 0 / 12600
 TEST: Multipath hash field: Inner destination IP (balanced)         [ OK ]
 INFO: Packets sent on path1 / path2: 6802 / 5802
 TEST: Multipath hash field: Inner destination IP (unbalanced)       [ OK ]
 INFO: Packets sent on path1 / path2: 12601 / 1
 TEST: Multipath hash field: Inner source port (balanced)            [ OK ]
 INFO: Packets sent on path1 / path2: 16430 / 16344
 TEST: Multipath hash field: Inner source port (unbalanced)          [ OK ]
 INFO: Packets sent on path1 / path2: 0 / 32772
 TEST: Multipath hash field: Inner destination port (balanced)       [ OK ]
 INFO: Packets sent on path1 / path2: 16430 / 16343
 TEST: Multipath hash field: Inner destination port (unbalanced)     [ OK ]
 INFO: Packets sent on path1 / path2: 0 / 32772
 INFO: Running IPv6 overlay custom multipath hash tests
 TEST: Multipath hash field: Inner source IP (balanced)              [ OK ]
 INFO: Packets sent on path1 / path2: 6702 / 5900
 TEST: Multipath hash field: Inner source IP (unbalanced)            [ OK ]
 INFO: Packets sent on path1 / path2: 0 / 12601
 TEST: Multipath hash field: Inner destination IP (balanced)         [ OK ]
 INFO: Packets sent on path1 / path2: 5751 / 6851
 TEST: Multipath hash field: Inner destination IP (unbalanced)       [ OK ]
 INFO: Packets sent on path1 / path2: 12602 / 1
 TEST: Multipath hash field: Inner flowlabel (balanced)              [ OK ]
 INFO: Packets sent on path1 / path2: 8364 / 8065
 TEST: Multipath hash field: Inner flowlabel (unbalanced)            [ OK ]
 INFO: Packets sent on path1 / path2: 12601 / 0
 TEST: Multipath hash field: Inner source port (balanced)            [ OK ]
 INFO: Packets sent on path1 / path2: 16425 / 16349
 TEST: Multipath hash field: Inner source port (unbalanced)          [ OK ]
 INFO: Packets sent on path1 / path2: 1 / 32770
 TEST: Multipath hash field: Inner destination port (balanced)       [ OK ]
 INFO: Packets sent on path1 / path2: 16425 / 16349
 TEST: Multipath hash field: Inner destination port (unbalanced)     [ OK ]
 INFO: Packets sent on path1 / path2: 2 / 32770

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: forwarding: Add test for custom multipath hash
Ido Schimmel [Mon, 17 May 2021 18:15:24 +0000 (21:15 +0300)]
selftests: forwarding: Add test for custom multipath hash

Test that when the hash policy is set to custom, traffic is distributed
only according to the outer fields set in the fib_multipath_hash_fields
sysctl.

Each time set a different field and make sure traffic is only
distributed when the field is changed in the packet stream.

The test only verifies the behavior with non-encapsulated IPv4 and IPv6
packets. Subsequent patches will add tests for IPv4/IPv6 overlays on top
of IPv4/IPv6 underlay networks.

Example output:

 # ./custom_multipath_hash.sh
 TEST: ping                                                          [ OK ]
 TEST: ping6                                                         [ OK ]
 INFO: Running IPv4 custom multipath hash tests
 TEST: Multipath hash field: Source IP (balanced)                    [ OK ]
 INFO: Packets sent on path1 / path2: 6353 / 6254
 TEST: Multipath hash field: Source IP (unbalanced)                  [ OK ]
 INFO: Packets sent on path1 / path2: 0 / 12600
 TEST: Multipath hash field: Destination IP (balanced)               [ OK ]
 INFO: Packets sent on path1 / path2: 6102 / 6502
 TEST: Multipath hash field: Destination IP (unbalanced)             [ OK ]
 INFO: Packets sent on path1 / path2: 1 / 12601
 TEST: Multipath hash field: Source port (balanced)                  [ OK ]
 INFO: Packets sent on path1 / path2: 16428 / 16345
 TEST: Multipath hash field: Source port (unbalanced)                [ OK ]
 INFO: Packets sent on path1 / path2: 32770 / 2
 TEST: Multipath hash field: Destination port (balanced)             [ OK ]
 INFO: Packets sent on path1 / path2: 16428 / 16345
 TEST: Multipath hash field: Destination port (unbalanced)           [ OK ]
 INFO: Packets sent on path1 / path2: 32770 / 2
 INFO: Running IPv6 custom multipath hash tests
 TEST: Multipath hash field: Source IP (balanced)                    [ OK ]
 INFO: Packets sent on path1 / path2: 6704 / 5903
 TEST: Multipath hash field: Source IP (unbalanced)                  [ OK ]
 INFO: Packets sent on path1 / path2: 12600 / 0
 TEST: Multipath hash field: Destination IP (balanced)               [ OK ]
 INFO: Packets sent on path1 / path2: 5551 / 7052
 TEST: Multipath hash field: Destination IP (unbalanced)             [ OK ]
 INFO: Packets sent on path1 / path2: 12603 / 0
 TEST: Multipath hash field: Flowlabel (balanced)                    [ OK ]
 INFO: Packets sent on path1 / path2: 8378 / 8080
 TEST: Multipath hash field: Flowlabel (unbalanced)                  [ OK ]
 INFO: Packets sent on path1 / path2: 2 / 12603
 TEST: Multipath hash field: Source port (balanced)                  [ OK ]
 INFO: Packets sent on path1 / path2: 16385 / 16388
 TEST: Multipath hash field: Source port (unbalanced)                [ OK ]
 INFO: Packets sent on path1 / path2: 0 / 32774
 TEST: Multipath hash field: Destination port (balanced)             [ OK ]
 INFO: Packets sent on path1 / path2: 16386 / 16390
 TEST: Multipath hash field: Destination port (unbalanced)           [ OK ]
 INFO: Packets sent on path1 / path2: 32771 / 2

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoipv6: Add custom multipath hash policy
Ido Schimmel [Mon, 17 May 2021 18:15:23 +0000 (21:15 +0300)]
ipv6: Add custom multipath hash policy

Add a new multipath hash policy where the packet fields used for hash
calculation are determined by user space via the
fib_multipath_hash_fields sysctl that was introduced in the previous
patch.

The current set of available packet fields includes both outer and inner
fields, which requires two invocations of the flow dissector. Avoid
unnecessary dissection of the outer or inner flows by skipping
dissection if none of the outer or inner fields are required.

In accordance with the existing policies, when an skb is not available,
packet fields are extracted from the provided flow key. In which case,
only outer fields are considered.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoipv6: Add a sysctl to control multipath hash fields
Ido Schimmel [Mon, 17 May 2021 18:15:22 +0000 (21:15 +0300)]
ipv6: Add a sysctl to control multipath hash fields

A subsequent patch will add a new multipath hash policy where the packet
fields used for multipath hash calculation are determined by user space.
This patch adds a sysctl that allows user space to set these fields.

The packet fields are represented using a bitmask and are common between
IPv4 and IPv6 to allow user space to use the same numbering across both
protocols. For example, to hash based on standard 5-tuple:

 # sysctl -w net.ipv6.fib_multipath_hash_fields=0x0037
 net.ipv6.fib_multipath_hash_fields = 0x0037

To avoid introducing holes in 'struct netns_sysctl_ipv6', move the
'bindv6only' field after the multipath hash fields.

The kernel rejects unknown fields, for example:

 # sysctl -w net.ipv6.fib_multipath_hash_fields=0x1000
 sysctl: setting key "net.ipv6.fib_multipath_hash_fields": Invalid argument

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoipv6: Calculate multipath hash inside switch statement
Ido Schimmel [Mon, 17 May 2021 18:15:21 +0000 (21:15 +0300)]
ipv6: Calculate multipath hash inside switch statement

A subsequent patch will add another multipath hash policy where the
multipath hash is calculated directly by the policy specific code and
not outside of the switch statement.

Prepare for this change by moving the multipath hash calculation inside
the switch statement.

No functional changes intended.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoipv6: Use a more suitable label name
Ido Schimmel [Mon, 17 May 2021 18:15:20 +0000 (21:15 +0300)]
ipv6: Use a more suitable label name

The 'out_timer' label was added in commit 63152fc0de4d ("[NETNS][IPV6]
ip6_fib - gc timer per namespace") when the timer was allocated on the
heap.

Commit 417f28bb3407 ("netns: dont alloc ipv6 fib timer list") removed
the allocation, but kept the label name.

Rename it to a more suitable name.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoipv4: Add custom multipath hash policy
Ido Schimmel [Mon, 17 May 2021 18:15:19 +0000 (21:15 +0300)]
ipv4: Add custom multipath hash policy

Add a new multipath hash policy where the packet fields used for hash
calculation are determined by user space via the
fib_multipath_hash_fields sysctl that was introduced in the previous
patch.

The current set of available packet fields includes both outer and inner
fields, which requires two invocations of the flow dissector. Avoid
unnecessary dissection of the outer or inner flows by skipping
dissection if none of the outer or inner fields are required.

In accordance with the existing policies, when an skb is not available,
packet fields are extracted from the provided flow key. In which case,
only outer fields are considered.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoipv4: Add a sysctl to control multipath hash fields
Ido Schimmel [Mon, 17 May 2021 18:15:18 +0000 (21:15 +0300)]
ipv4: Add a sysctl to control multipath hash fields

A subsequent patch will add a new multipath hash policy where the packet
fields used for multipath hash calculation are determined by user space.
This patch adds a sysctl that allows user space to set these fields.

The packet fields are represented using a bitmask and are common between
IPv4 and IPv6 to allow user space to use the same numbering across both
protocols. For example, to hash based on standard 5-tuple:

 # sysctl -w net.ipv4.fib_multipath_hash_fields=0x0037
 net.ipv4.fib_multipath_hash_fields = 0x0037

The kernel rejects unknown fields, for example:

 # sysctl -w net.ipv4.fib_multipath_hash_fields=0x1000
 sysctl: setting key "net.ipv4.fib_multipath_hash_fields": Invalid argument

More fields can be added in the future, if needed.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoipv4: Calculate multipath hash inside switch statement
Ido Schimmel [Mon, 17 May 2021 18:15:17 +0000 (21:15 +0300)]
ipv4: Calculate multipath hash inside switch statement

A subsequent patch will add another multipath hash policy where the
multipath hash is calculated directly by the policy specific code and
not outside of the switch statement.

Prepare for this change by moving the multipath hash calculation inside
the switch statement.

No functional changes intended.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agodrivers/net: Remove leading spaces in Kconfig
Juerg Haefliger [Mon, 17 May 2021 09:58:33 +0000 (11:58 +0200)]
drivers/net: Remove leading spaces in Kconfig

Remove leading spaces before tabs in Kconfig file(s) by running the
following command:

  $ find drivers/net -name 'Kconfig*' | xargs sed -r -i 's/^[ ]+\t/\t/'

Signed-off-by: Juerg Haefliger <juergh@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/packet: Remove redundant assignment to ret
Jiapeng Chong [Mon, 17 May 2021 10:15:25 +0000 (18:15 +0800)]
net/packet: Remove redundant assignment to ret

Variable ret is set to '0' or '-EBUSY', but this value is never read
as it is not used later on, hence it is a redundant assignment and
can be removed.

Clean up the following clang-analyzer warning:

net/packet/af_packet.c:3936:4: warning: Value stored to 'ret' is never
read [clang-analyzer-deadcode.DeadStores].

net/packet/af_packet.c:3933:4: warning: Value stored to 'ret' is never
read [clang-analyzer-deadcode.DeadStores].

No functional change.

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'stmmac-xpcs-eee'
David S. Miller [Mon, 17 May 2021 22:53:59 +0000 (15:53 -0700)]
Merge branch 'stmmac-xpcs-eee'

Michael Sit Wei Hong says:

====================
Introducing support for DWC xpcs Energy Efficient Ethernet

The goal of this patch set is to enable EEE in the xpcs so that when
EEE is enabled, the MAC-->xpcs-->PHY have all the EEE related
configurations enabled.

Patch 1 adds the functions to enable EEE in the xpcs and sets it to
transparent mode.
Patch 2 adds the callbacks to configure the xpcs EEE mode.

The results are tested by checking the lpi counters of the tx and rx
path of the interface. When EEE is enabled, the lpi counters should
increament as it enters and exits lpi states.

host@EHL$ ethtool --show-eee enp0s30f4
EEE Settings for enp0s30f4:
        EEE status: disabled
        Tx LPI: disabled
        Supported EEE link modes:  100baseT/Full
                                   1000baseT/Full
        Advertised EEE link modes:  Not reported
        Link partner advertised EEE link modes:  100baseT/Full
                                                 1000baseT/Full
host@EHL$ ethtool -S enp0s30f4 | grep lpi
     irq_tx_path_in_lpi_mode_n: 0
     irq_tx_path_exit_lpi_mode_n: 0
     irq_rx_path_in_lpi_mode_n: 0
     irq_rx_path_exit_lpi_mode_n: 0
host@EHL$ ethtool --set-eee enp0s30f4 eee on
host@EHL$ [  110.265154] intel-eth-pci 0000:00:1e.4 enp0s30f4: Link is Down
[  112.315155] intel-eth-pci 0000:00:1e.4 enp0s30f4: Link is Up - 1Gbps/Full - flow control off
[  112.324612] IPv6: ADDRCONF(NETDEV_CHANGE): enp0s30f4: link becomes ready
host@EHL$ ethtool --show-eee enp0s30f4
EEE Settings for enp0s30f4:
        EEE status: enabled - active
        Tx LPI: 1000000 (us)
        Supported EEE link modes:  100baseT/Full
                                   1000baseT/Full
        Advertised EEE link modes:  100baseT/Full
                                    1000baseT/Full
        Link partner advertised EEE link modes:  100baseT/Full
                                                 1000baseT/Full
host@EHL$ ethtool -S enp0s30f4 | grep lpi
     irq_tx_path_in_lpi_mode_n: 6
     irq_tx_path_exit_lpi_mode_n: 5
     irq_rx_path_in_lpi_mode_n: 7
     irq_rx_path_exit_lpi_mode_n: 6
host@EHL$ ping 192.168.1.1
PING 192.168.1.1 (192.168.1.1) 56(84) bytes of data.
64 bytes from 192.168.1.1: icmp_seq=1 ttl=64 time=1.02 ms
64 bytes from 192.168.1.1: icmp_seq=2 ttl=64 time=0.510 ms
64 bytes from 192.168.1.1: icmp_seq=3 ttl=64 time=0.489 ms
64 bytes from 192.168.1.1: icmp_seq=4 ttl=64 time=0.484 ms
64 bytes from 192.168.1.1: icmp_seq=5 ttl=64 time=0.504 ms
64 bytes from 192.168.1.1: icmp_seq=6 ttl=64 time=0.466 ms
64 bytes from 192.168.1.1: icmp_seq=7 ttl=64 time=0.529 ms
64 bytes from 192.168.1.1: icmp_seq=8 ttl=64 time=0.519 ms
64 bytes from 192.168.1.1: icmp_seq=9 ttl=64 time=0.518 ms
64 bytes from 192.168.1.1: icmp_seq=10 ttl=64 time=0.501 ms

--- 192.168.1.1 ping statistics ---
10 packets transmitted, 10 received, 0% packet loss, time 9216ms
rtt min/avg/max/mdev = 0.466/0.553/1.018/0.155 ms
host@EHL$ ethtool -S enp0s30f4 | grep lpi
     irq_tx_path_in_lpi_mode_n: 22
     irq_tx_path_exit_lpi_mode_n: 21
     irq_rx_path_in_lpi_mode_n: 21
     irq_rx_path_exit_lpi_mode_n: 20
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: stmmac: Add callbacks for DWC xpcs Energy Efficient Ethernet
Michael Sit Wei Hong [Mon, 17 May 2021 09:43:32 +0000 (17:43 +0800)]
net: stmmac: Add callbacks for DWC xpcs Energy Efficient Ethernet

Link xpcs callback functions for MAC to configure the xpcs EEE feature.

The clk_eee frequency is used to calculate the MULT_FACT_100NS. This is
to adjust the clock tic closer to 100ns.

Signed-off-by: Michael Sit Wei Hong <michael.wei.hong.sit@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: pcs: Introducing support for DWC xpcs Energy Efficient Ethernet
Michael Sit Wei Hong [Mon, 17 May 2021 09:43:31 +0000 (17:43 +0800)]
net: pcs: Introducing support for DWC xpcs Energy Efficient Ethernet

Add DWC xpcs EEE support callbacks.The callback function is used to
set EEE registers on xpcs.

xpcs transparent mode is enabled to allow PHY to detect MAC EEE status.

Signed-off-by: Michael Sit Wei Hong <michael.wei.hong.sit@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoalx: fix a double unlock in alx_probe()
Dan Carpenter [Mon, 17 May 2021 08:57:56 +0000 (11:57 +0300)]
alx: fix a double unlock in alx_probe()

We're not holding the lock at this point so "goto unlock;" should be
"goto unmap;"

Fixes: 4a5fe57e7751 ("alx: use fine-grained locking instead of RTNL")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: wwan: Add WWAN port type attribute
Loic Poulain [Mon, 17 May 2021 09:53:34 +0000 (11:53 +0200)]
net: wwan: Add WWAN port type attribute

The port type is by default part of the WWAN port device name.
However device name can not be considered as a 'stable' API and
may be subject to change in the future. This change adds a proper
device attribute that can be used to determine the WWAN protocol/
type.

Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'stmmac-RK3568'
David S. Miller [Mon, 17 May 2021 22:34:43 +0000 (15:34 -0700)]
Merge branch 'stmmac-RK3568'

Ezequiel Garcia says:

====================
stmmmac: RK3568

Here's the third version of this patchset, taking
the feedback from Heiko and Chen-Yu Tsai.

Although this solution is a tad ugly as it hardcodes
the register addresses, we believe it's the most robust approach.

See:

https://lore.kernel.org/netdev/CAGb2v67ZBR=XDFPeXQc429HNu_dbY__-KN50tvBW44fXMs78_w@mail.gmail.com/

This is tested on RK3566 EVB2 and seems to work well.
Once the RK3568 devicetree lands upstream, we'll post
patches to add network support for RK3566 and RK3568.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: stmmac: Add RK3566/RK3568 SoC support
David Wu [Mon, 17 May 2021 15:40:37 +0000 (12:40 -0300)]
net: stmmac: Add RK3566/RK3568 SoC support

Add constants and callback functions for the dwmac present
on RK3566/RK3568 SoCs.

RK3568 has two MACs, and RK3566 just one, but it's otherwise
the same IP core.

Signed-off-by: David Wu <david.wu@rock-chips.com>
[Ezequiel: Separate rk3566-gmac support]
Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agodt-bindings: net: rockchip-dwmac: add rk3568 compatible string
Ezequiel Garcia [Mon, 17 May 2021 15:40:36 +0000 (12:40 -0300)]
dt-bindings: net: rockchip-dwmac: add rk3568 compatible string

Add compatible string for RK3568 gmac, and constrain it to
be compatible with Synopsys dwmac 4.20a.

Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: stmmac: dwmac-rk: Check platform-specific ops
David Wu [Mon, 17 May 2021 15:40:35 +0000 (12:40 -0300)]
net: stmmac: dwmac-rk: Check platform-specific ops

Add a check for non-null struct rk_gmac_ops for the
configured PHY interface mode, failing if unsupported.

Signed-off-by: David Wu <david.wu@rock-chips.com>
[Ezequiel: Refactor so it fails if unsupported]
Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: stmmac: Don't set has_gmac if has_gmac4 is set
Ezequiel Garcia [Mon, 17 May 2021 15:40:34 +0000 (12:40 -0300)]
net: stmmac: Don't set has_gmac if has_gmac4 is set

Some Rockchip platforms have a GMAC4 core, and therefore
'plat_stmmacenet_data.has_gmac' shouldn't be set if
'plat_stmmacenet_data.has_gmac4' is set.

Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: Remove the member netns_ok
Yejune Deng [Mon, 17 May 2021 12:22:05 +0000 (20:22 +0800)]
net: Remove the member netns_ok

Every protocol has the 'netns_ok' member and it is euqal to 1. The
'if (!prot->netns_ok)' always false in inet_add_protocol().

Signed-off-by: Yejune Deng <yejunedeng@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'mlxsw-next'
David S. Miller [Mon, 17 May 2021 22:15:47 +0000 (15:15 -0700)]
Merge branch 'mlxsw-next'

Ido Schimmel says:

====================
mlxsw: Various updates

This patchset contains various updates to the mlxsw driver and related
selftests.

Patches #1-#5 contain various updates to mlxsw selftests. The most
significant change is the conversion of the DCB selftests to use the new
iproute2 DCB support.

Patches #6-#9 contain mostly trivial changes to the driver itself. No
user facing changes.

Patches #10-#11 remove support for SwitchX-2 and SwitchIB ASICs that did
not see any updates in the last 4-5 years and will not see any in the
future. See individual commit messages for detailed explanation as to
why it is OK to remove these drivers from the kernel.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agomlxsw: Remove Mellanox SwitchX-2 ASIC support
Amit Cohen [Mon, 17 May 2021 17:04:01 +0000 (20:04 +0300)]
mlxsw: Remove Mellanox SwitchX-2 ASIC support

Initial support for the Mellanox SwitchX-2 ASIC was added in July 2015.
Since then all development efforts shifted towards the Mellanox Spectrum
ASICs and development of this driver stopped beside trivial fixes and
refactoring. Therefore, the driver does not support any switch offloads
and simply traps all traffic to the CPU, rendering it irrelevant for
deployment.

In addition, support for this ASIC was dropped by Mellanox a few years
ago.

Given the driver is not used by any users and that there is no
intention of investing in its development, remove it from the kernel.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agomlxsw: Remove Mellanox SwitchIB ASIC support
Amit Cohen [Mon, 17 May 2021 17:04:00 +0000 (20:04 +0300)]
mlxsw: Remove Mellanox SwitchIB ASIC support

Initial support for the Mellanox SwitchIB and SwitchIB-2 ASICs was added
in October 2016, but since then development of this driver stopped.
Therefore, the driver does not support any offloads and simply registers
devlink ports for its front panel ports, rendering it irrelevant for
deployment.

Given the driver is not used by any users and that there is no intention
of investing in its development, remove it from the kernel.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agomlxsw: spectrum_router: Avoid missing error code warning
Ido Schimmel [Mon, 17 May 2021 17:03:59 +0000 (20:03 +0300)]
mlxsw: spectrum_router: Avoid missing error code warning

Explicitly set the error code to zero before the goto statement to avoid
the following smatch warning:

drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c:3598 mlxsw_sp_nexthop_group_refresh() warn: missing error code 'err'

The warning is a false positive, but the change both suppresses the
warning and makes it clear to future readers that this is not an error
path.

The original report and discussion can be found here [1].

[1] https://lore.kernel.org/lkml/202105141823.Td2h3Mbi-lkp@intel.com/

Cc: Dan Carpenter <dan.carpenter@oracle.com>
Suggested-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agomlxsw: core: Avoid unnecessary EMAD buffer copy
Ido Schimmel [Mon, 17 May 2021 17:03:58 +0000 (20:03 +0300)]
mlxsw: core: Avoid unnecessary EMAD buffer copy

mlxsw_emad_transmit() takes care of sending EMAD transactions to the
device. Since these transactions can time out, the driver performs up to
5 retransmissions, each time copying the skb with the original request.

The data of the skb does not change throughout the process, so there is
no need to copy it each time. Instead, only the skb itself can be
copied. Therefore, use skb_clone() instead of skb_copy().

This reduces the latency of the function by about 16%.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agomlxsw: Verify the accessed index doesn't exceed the array length
Danielle Ratson [Mon, 17 May 2021 17:03:57 +0000 (20:03 +0300)]
mlxsw: Verify the accessed index doesn't exceed the array length

There are few cases in which an array index queried from a fw register,
is accessed without any validation that it doesn't exceed the array
length.

Add a proper length validation, so accessing memory past the end of an
array will be forbidden.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agomlxsw: spectrum_buffers: Switch function arguments
Danielle Ratson [Mon, 17 May 2021 17:03:56 +0000 (20:03 +0300)]
mlxsw: spectrum_buffers: Switch function arguments

In the call path:

mlxsw_sp_hdroom_bufs_reset_sizes()
    mlxsw_sp_hdroom_int_buf_size_get()
        ->int_buf_size_get()

The 'speed' and 'mtu' arguments were mistakenly switched twice. The two
bugs thus canceled each other.

Clean this up by switching the arguments in both call sites, so that
they are passed in the right order.

Found during manual code inspection.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: mlxsw: qos_lib: Drop __mlnx_qos
Petr Machata [Mon, 17 May 2021 17:03:55 +0000 (20:03 +0300)]
selftests: mlxsw: qos_lib: Drop __mlnx_qos

Now that the two users of this helper have been converted to iproute2 dcb,
it is not necessary anymore. Drop it.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: mlxsw: qos_pfc: Convert to iproute2 dcb
Petr Machata [Mon, 17 May 2021 17:03:54 +0000 (20:03 +0300)]
selftests: mlxsw: qos_pfc: Convert to iproute2 dcb

There is a dedicated tool for configuration of DCB in iproute2 now. Use it
in the selftest instead of mlnx_qos.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: mlxsw: qos_headroom: Convert to iproute2 dcb
Petr Machata [Mon, 17 May 2021 17:03:53 +0000 (20:03 +0300)]
selftests: mlxsw: qos_headroom: Convert to iproute2 dcb

There is a dedicated tool for configuration of DCB in iproute2 now. Use it
in the selftest instead of mlnx_qos.

Signed-off-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: mlxsw: Make sampling test more robust
Ido Schimmel [Mon, 17 May 2021 17:03:52 +0000 (20:03 +0300)]
selftests: mlxsw: Make sampling test more robust

The test sometimes fails with an error message such as:

TEST: tc sample (w/ flower) rate (egress)                           [FAIL]
Expected 100 packets, got 70 packets, which is -30% off. Required accuracy is +-25%

Make the test more robust by generating more packets, therefore
increasing the number of expected samples. Decrease the transmission
delay in order not to needlessly prolong the test.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: mlxsw: Make the unsplit array global in port_scale test
Danielle Ratson [Mon, 17 May 2021 17:03:51 +0000 (20:03 +0300)]
selftests: mlxsw: Make the unsplit array global in port_scale test

Currently, the array of the ports that were split in the port_scale test
is local, so the port_cleanup() unsplits an empty array.

Make the array global so the cleanup will be preformed properly.

Suggested-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'func-names-comment'
David S. Miller [Mon, 17 May 2021 21:12:39 +0000 (14:12 -0700)]
Merge branch 'func-names-comment'

Yang Shen says:

====================
Rid W=1 warnings in net

This is a set to fully clean drivers/net.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hisilicon: hns: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:35 +0000 (12:45 +0800)]
net: hisilicon: hns: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/hisilicon/hns/hns_dsaf_mac.c:121: warning: expecting prototype for hns_mac_is_adjust_link(). Prototype was for hns_mac_need_adjust_link() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_mac.c:386: warning: expecting prototype for hns_mac_queue_config_bc_en(). Prototype was for hns_mac_port_config_bc_en() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_mac.c:607: warning: expecting prototype for hns_mac_set_autoneg(). Prototype was for hns_mac_set_pauseparam() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_main.c:236: warning: expecting prototype for hns_ppe_qid_cfg(). Prototype was for hns_dsaf_ppe_qid_cfg() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_main.c:623: warning: expecting prototype for dsaf_tbl_tcam_mcast_cfg(). Prototype was for hns_dsaf_tbl_tcam_mcast_cfg() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_main.c:1220: warning: expecting prototype for hns_dsaf_tbl_tcam_init(). Prototype was for hns_dsaf_comm_init() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_main.c:2121: warning: expecting prototype for dsaf_pfc_unit_cnt(). Prototype was for hns_dsaf_pfc_unit_cnt() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_main.c:2153: warning: expecting prototype for dsaf_port_work_rate_cfg(). Prototype was for hns_dsaf_port_work_rate_cfg() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_main.c:2745: warning: expecting prototype for hns_dsaf_get_sset_count(). Prototype was for hns_dsaf_get_regs_count() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_main.c:2957: warning: expecting prototype for dsaf_probe(). Prototype was for hns_dsaf_probe() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_main.c:3011: warning: expecting prototype for dsaf_remove(). Prototype was for hns_dsaf_remove() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_misc.c:366: warning: expecting prototype for hns_dsaf_srst_chns(). Prototype was for hns_dsaf_srst_chns_acpi() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_misc.c:509: warning: expecting prototype for hns_mac_get_sds_mode(). Prototype was for hns_mac_get_phy_if() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_ppe.c:303: warning: expecting prototype for ppe_init_hw(). Prototype was for hns_ppe_init_hw() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_ppe.c:350: warning: expecting prototype for ppe_uninit_hw(). Prototype was for hns_ppe_uninit_hw() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_ppe.c:391: warning: expecting prototype for hns_ppe_reset(). Prototype was for hns_ppe_reset_common() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_ppe.c:464: warning: expecting prototype for ppe_get_strings(). Prototype was for hns_ppe_get_strings() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_rcb.c:920: warning: expecting prototype for rcb_get_sset_count(). Prototype was for hns_rcb_get_ring_regs_count() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_xgmac.c:112: warning: expecting prototype for hns_xgmac_tx_lf_rf_insert(). Prototype was for hns_xgmac_lf_rf_insert() instead
 drivers/net/ethernet/hisilicon/hns/hns_dsaf_xgmac.c:122: warning: expecting prototype for hns_xgmac__lf_rf_control_init(). Prototype was for hns_xgmac_lf_rf_control_init() instead
 drivers/net/ethernet/hisilicon/hns/hns_enet.c:777: warning: expecting prototype for hns_nic_adp_coalesce(). Prototype was for hns_nic_adpt_coalesce() instead
 drivers/net/ethernet/hisilicon/hns/hns_ethtool.c:202: warning: expecting prototype for hns_nic_set_link_settings(). Prototype was for hns_nic_set_link_ksettings() instead
 drivers/net/ethernet/hisilicon/hns/hns_ethtool.c:837: warning: expecting prototype for get_ethtool_stats(). Prototype was for hns_get_ethtool_stats() instead
 drivers/net/ethernet/hisilicon/hns/hns_ethtool.c:894: warning: expecting prototype for get_strings(). Prototype was for hns_get_strings() instead

Cc: Yisen Zhuang <yisen.zhuang@huawei.com>
Cc: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: phy: Demote non-compliant kernel-doc headers
Yang Shen [Mon, 17 May 2021 04:45:34 +0000 (12:45 +0800)]
net: phy: Demote non-compliant kernel-doc headers

Fixes the following W=1 kernel build warning(s):

 drivers/net/phy/adin.c:3: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst
 drivers/net/phy/rockchip.c:3: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst

Cc: Heiner Kallweit <hkallweit1@gmail.com>
Cc: Russell King <linux@armlinux.org.uk>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: via: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:33 +0000 (12:45 +0800)]
net: via: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/via/via-velocity.c:1908: warning: expecting prototype for tx_srv(). Prototype was for velocity_tx_srv() instead
 drivers/net/ethernet/via/via-velocity.c:2466: warning: expecting prototype for velocity_get_status(). Prototype was for velocity_get_stats() instead
 drivers/net/ethernet/via/via-velocity.c:3734: warning: expecting prototype for velocity_cleanup(). Prototype was for velocity_cleanup_module() instead

Cc: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ti: Fix wrong struct name in comments
Yang Shen [Mon, 17 May 2021 04:45:32 +0000 (12:45 +0800)]
net: ti: Fix wrong struct name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/ti/cpsw_ale.c:88: warning: expecting prototype for struct ale_dev_id. Prototype was for struct cpsw_ale_dev_id instead

Cc: Cyril Chemparathy <cyril@ti.com>
Cc: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: socionext: Demote non-compliant kernel-doc headers
Yang Shen [Mon, 17 May 2021 04:45:31 +0000 (12:45 +0800)]
net: socionext: Demote non-compliant kernel-doc headers

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/socionext/sni_ave.c:28: warning: expecting prototype for sni_ave.c(). Prototype was for AVE_IDR() instead

Signed-off-by: Yang Shen <shenyang39@huawei.com>
Reviewed-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: samsung: sxgbe: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:30 +0000 (12:45 +0800)]
net: samsung: sxgbe: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c:797: warning: expecting prototype for sxgbe_tx_clean(). Prototype was for sxgbe_tx_all_clean() instead
 drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c:1026: warning: expecting prototype for sxgbe_init_tx_coalesce(). Prototype was for sxgbe_tx_init_coalesce() instead

Cc: Byungho An <bh74.an@samsung.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: calxeda: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:29 +0000 (12:45 +0800)]
net: calxeda: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/qlogic/qlcnic/qlcnic_83xx_init.c:761: warning: expecting prototype for qlcnic_83xx_idc_cold_state(). Prototype was for qlcnic_83xx_idc_cold_state_handler() instead
 drivers/net/ethernet/qlogic/qlcnic/qlcnic_83xx_vnic.c:192: warning: expecting prototype for qlcnic_83xx_vnic_opmode(). Prototype was for qlcnic_83xx_config_vnic_opmode() instead

Cc: Shahed Shaikh <shshaikh@marvell.com>
Cc: Manish Chopra <manishc@marvell.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: netronome: nfp: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:28 +0000 (12:45 +0800)]
net: netronome: nfp: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/netronome/nfp/ccm_mbox.c:52: warning: expecting prototype for struct nfp_ccm_mbox_skb_cb. Prototype was for struct nfp_ccm_mbox_cmsg_cb instead
 drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c:35: warning: expecting prototype for struct nfp_tun_pre_run_rule. Prototype was for struct nfp_tun_pre_tun_rule instead
 drivers/net/ethernet/netronome/nfp/nfpcore/nfp_nffw.c:38: warning: expecting prototype for NFFW_INFO_VERSION history(). Prototype was for NFFW_INFO_VERSION_CURRENT() instead

Cc: Simon Horman <simon.horman@netronome.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: neterion: vxge: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:27 +0000 (12:45 +0800)]
net: neterion: vxge: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/neterion/vxge/vxge-config.c:4895: warning: expecting prototype for vxge_hw_vpath_rx_doorbell_post(). Prototype was for vxge_hw_vpath_rx_doorbell_init() instead
 drivers/net/ethernet/neterion/vxge/vxge-main.c:1814: warning: expecting prototype for vxge_poll(). Prototype was for vxge_poll_msix() instead
 drivers/net/ethernet/neterion/vxge/vxge-main.c:4761: warning: expecting prototype for vxge_rem_nic(). Prototype was for vxge_remove() instead

Cc: Jon Mason <jdmason@kudzu.us>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: neterion: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:26 +0000 (12:45 +0800)]
net: neterion: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/neterion/s2io.c:2759: warning: expecting prototype for s2io_poll(). Prototype was for s2io_poll_msix() instead
 drivers/net/ethernet/neterion/s2io.c:5304: warning: expecting prototype for s2io_ethtol_get_link_ksettings(). Prototype was for s2io_ethtool_get_link_ksettings() instead

Cc: Jon Mason <jdmason@kudzu.us>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: microchip: Demote non-compliant kernel-doc headers
Yang Shen [Mon, 17 May 2021 04:45:25 +0000 (12:45 +0800)]
net: microchip: Demote non-compliant kernel-doc headers

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/microchip/encx24j600.c:3: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst

Cc: Jon Ringle <jringle@gridpoint.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: micrel: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:24 +0000 (12:45 +0800)]
net: micrel: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/micrel/ksz884x.c:2163: warning: expecting prototype for sw_get_board_storm(). Prototype was for sw_get_broad_storm() instead
 drivers/net/ethernet/micrel/ksz884x.c:2985: warning: expecting prototype for port_w_phy(). Prototype was for hw_w_phy() instead
 drivers/net/ethernet/micrel/ksz884x.c:4792: warning: expecting prototype for transmit_done(). Prototype was for tx_done() instead

Cc: Tristram Ha <Tristram.Ha@micrel.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: huawei: hinic: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:23 +0000 (12:45 +0800)]
net: huawei: hinic: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):
 drivers/net/ethernet/huawei/hinic/hinic_hw_cmdq.c:604: warning: expecting prototype for cmdq_arm_ceq_handler(). Prototype was for cmdq_sync_cmd_handler() instead
 drivers/net/ethernet/huawei/hinic/hinic_hw_dev.c:59: warning: expecting prototype for get_capability(). Prototype was for parse_capability() instead
 drivers/net/ethernet/huawei/hinic/hinic_hw_dev.c:101: warning: expecting prototype for get_cap_from_fw(). Prototype was for get_capability() instead
 drivers/net/ethernet/huawei/hinic/hinic_hw_dev.c:355: warning: expecting prototype for clear_io_resource(). Prototype was for clear_io_resources() instead
 drivers/net/ethernet/huawei/hinic/hinic_hw_dev.c:1100: warning: expecting prototype for hinic_hwdev_get_sq(). Prototype was for hinic_hwdev_get_rq() instead
 drivers/net/ethernet/huawei/hinic/hinic_hw_if.c:341: warning: expecting prototype for dma_attr_table_init(). Prototype was for dma_attr_init() instead
 drivers/net/ethernet/huawei/hinic/hinic_hw_qp.c:904: warning: expecting prototype for hinic_put_wqe(). Prototype was for hinic_rq_put_wqe() instead
 drivers/net/ethernet/huawei/hinic/hinic_main.c:241: warning: expecting prototype for create_txqs(). Prototype was for create_rxqs() instead
 drivers/net/ethernet/huawei/hinic/hinic_main.c:295: warning: expecting prototype for free_txqs(). Prototype was for free_rxqs() instead
 drivers/net/ethernet/huawei/hinic/hinic_tx.c:667: warning: expecting prototype for free_all_rx_skbs(). Prototype was for free_all_tx_skbs() instead

Cc: Bin Luo <luobin9@huawei.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: chelsio: cxgb4vf: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:22 +0000 (12:45 +0800)]
net: chelsio: cxgb4vf: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/chelsio/cxgb4vf/sge.c:966: warning: expecting prototype for check_ring_tx_db(). Prototype was for ring_tx_db() instead

Cc: Raju Rangoju <rajur@chelsio.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: chelsio: cxgb4: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:21 +0000 (12:45 +0800)]
net: chelsio: cxgb4: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/chelsio/cxgb3/sge.c:677: warning: expecting prototype for free_qset(). Prototype was for t3_free_qset() instead
 drivers/net/ethernet/chelsio/cxgb3/sge.c:1266: warning: expecting prototype for eth_xmit(). Prototype was for t3_eth_xmit() instead

Cc: Raju Rangoju <rajur@chelsio.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: chelsio: cxgb3: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:20 +0000 (12:45 +0800)]
net: chelsio: cxgb3: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/chelsio/cxgb3/sge.c:677: warning: expecting prototype for free_qset(). Prototype was for t3_free_qset() instead
 drivers/net/ethernet/chelsio/cxgb3/sge.c:1266: warning: expecting prototype for eth_xmit(). Prototype was for t3_eth_xmit() instead

Cc: Raju Rangoju <rajur@chelsio.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: calxeda: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:19 +0000 (12:45 +0800)]
net: calxeda: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/calxeda/xgmac.c:720: warning: expecting prototype for init_xgmac_dma_desc_rings(). Prototype was for xgmac_dma_desc_rings_init() instead
 drivers/net/ethernet/calxeda/xgmac.c:867: warning: expecting prototype for xgmac_tx(). Prototype was for xgmac_tx_complete() instead
 drivers/net/ethernet/calxeda/xgmac.c:1049: warning: expecting prototype for xgmac_release(). Prototype was for xgmac_stop() instead
 drivers/net/ethernet/calxeda/xgmac.c:1822: warning: expecting prototype for xgmac_dvr_remove(). Prototype was for xgmac_remove() instead

Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: cadence: Demote non-compliant kernel-doc headers
Yang Shen [Mon, 17 May 2021 04:45:18 +0000 (12:45 +0800)]
net: cadence: Demote non-compliant kernel-doc headers

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/cadence/macb_pci.c:3: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst
 drivers/net/ethernet/cadence/macb_ptp.c:3: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst

Cc: Nicolas Ferre <nicolas.ferre@microchip.com>
Cc: Claudiu Beznea <claudiu.beznea@microchip.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: brocade: bna: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:17 +0000 (12:45 +0800)]
net: brocade: bna: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/brocade/bna/bfa_cee.c:91: warning: expecting prototype for bfa_cee_get_attr_isr(). Prototype was for bfa_cee_get_stats_isr() instead

Cc: Rasesh Mody <rmody@marvell.com>
Cc: Sudarsana Kalluru <skalluru@marvell.com>
Cc: GR-Linux-NIC-Dev@marvell.com
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: broadcom: bnx2x: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:16 +0000 (12:45 +0800)]
net: broadcom: bnx2x: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c:13595: warning: expecting prototype for bnx2x_get_num_none_def_sbs(). Prototype was for bnx2x_get_num_non_def_sbs() instead
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_sp.c:4165: warning: expecting prototype for atomic_add_ifless(). Prototype was for __atomic_add_ifless() instead
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_sp.c:4193: warning: expecting prototype for atomic_dec_ifmoe(). Prototype was for __atomic_dec_ifmoe() instead

Cc: Ariel Elior <aelior@marvell.com>
Cc: Sudarsana Kalluru <skalluru@marvell.com>
Cc: GR-everest-linux-l2@marvell.com
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: atheros: atl1x: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:15 +0000 (12:45 +0800)]
net: atheros: atl1x: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/atheros/atlx/atl1.c:1020: warning: expecting prototype for atl1_setup_mem_resources(). Prototype was for atl1_setup_ring_resources() instead

Cc: Chris Snook <chris.snook@gmail.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: atheros: atl1e: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:14 +0000 (12:45 +0800)]
net: atheros: atl1e: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/atheros/atl1e/atl1e_main.c:367: warning: expecting prototype for atl1e_set_mac(). Prototype was for atl1e_set_mac_addr() instead
 drivers/net/ethernet/atheros/atl1e/atl1e_main.c:796: warning: expecting prototype for atl1e_setup_mem_resources(). Prototype was for atl1e_setup_ring_resources() instead

Cc: Chris Snook <chris.snook@gmail.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: atheros: atl1c: Fix wrong function name in comments
Yang Shen [Mon, 17 May 2021 04:45:13 +0000 (12:45 +0800)]
net: atheros: atl1c: Fix wrong function name in comments

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/atheros/atl1c/atl1c_main.c:442: warning: expecting prototype for atl1c_set_mac(). Prototype was for atl1c_set_mac_addr() instead
 drivers/net/ethernet/atheros/atl1c/atl1c_main.c:969: warning: expecting prototype for atl1c_setup_mem_resources(). Prototype was for atl1c_setup_ring_resources() instead
 drivers/net/ethernet/atheros/atl1c/atl1c_main.c:1375: warning: expecting prototype for atl1c_configure(). Prototype was for atl1c_configure_mac() instead

Cc: Chris Snook <chris.snook@gmail.com>
Signed-off-by: Yang Shen <shenyang39@huawei.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: arc: Demote non-compliant kernel-doc headers
Yang Shen [Mon, 17 May 2021 04:45:12 +0000 (12:45 +0800)]
net: arc: Demote non-compliant kernel-doc headers

Fixes the following W=1 kernel build warning(s):

 drivers/net/ethernet/arc/emac_rockchip.c:18: warning: expecting prototype for emac(). Prototype was for DRV_NAME() instead

Signed-off-by: Yang Shen <shenyang39@huawei.com>
Reviewed-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agor8169: use KBUILD_MODNAME instead of own module name definition
Heiner Kallweit [Sat, 15 May 2021 11:53:21 +0000 (13:53 +0200)]
r8169: use KBUILD_MODNAME instead of own module name definition

Remove own module name definition and use KBUILD_MODNAME instead.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'ipv4-unicast'
David S. Miller [Mon, 17 May 2021 20:47:58 +0000 (13:47 -0700)]
Merge branch 'ipv4-unicast'

Seth David Schoen says:

====================
Treat IPv4 lowest address as ordinary unicast address

Treat the lowest address in a subnet (the address within the subnet
which contains all 0 bits) as an ordinary unicast address instead
of as a potential second broadcast address.  For example, in subnet
192.168.17.24/29, which contains 8 addresses, make address 192.168.17.24
usable as a normal unicast address (while continuing to support
192.168.17.31 as a broadcast address).

Since EVERY network number or subnet formerly had its host number 0
reserved, this patchset adds 1 more usable host address to every network
and subnet (i.e., 2^(32-n)-1 instead of 2^(32-n)-2 addresses available
for assignment on each IPv4 /n subnet).  For small subnets, this is a
significant gain; instead of 6 usable host addresses, a /29 would now
contain 7, a 16% increase.

The reserving of host number 0 for broadcast came about in RFC 1122 from
1989 (page 31, "IP addresses are not permitted to have the value 0 or -1
for any of the <Host-number>, <Network-number>, or <Subnet-number>
fields (except in the special cases listed above)" and page 66, "There
is a class of hosts [4.2BSD Unix and its derivatives, but not 4.3BSD]
that use non-standard broadcast address forms, substituting 0 for -1.
All hosts SHOULD recognize and accept any of these non-standard
broadcast addresses as the destination address of an incoming
datagram.").  This has been repeated in subsequent RFCs, always with
backwards-compatibility rationales.  Network troubles (broadcast storms)
ensued when some early hosts on a LAN treated the lowest address as
unicast and others treated it as broadcast.  Multiple 1989 changes to IP
successfully prevented these.  The key was adding the layering violation
rule requiring hosts to ignore all IP datagrams with unicast destination
addresses that were received in low-level (Ethernet) broadcasts.  That
change is still in effect, and this patchset does not alter it.  All
operating systems since 4.3BSD, including all the current BSD OSes, now
use the standard IP broadcast address.  4.2BSD has been obsolete for
more than 30 years, and all modern hosts ignore hardware broadcasts
containing unicast IP addresses, so there is no modern likelihood of
broadcast storms even when hosts disagree on the unicast vs. broadcast
status of a given address.

Tests with this patchset show that other Linux hosts on the local segment
simply ignore a host numbered with the lowest address, both for incoming
and outgoing packet purposes.  They don't interoperate with it, but they
also don't cause broadcast storms or any other malfunction.  If patched,
they have no trouble interoperating with a host at the lowest address.

Unmodified "distant" hosts that are not on the same segment successfully
interoperate, as long as the gateway on the local segment, and the local
host itself using the lowest address, have this patch.  (Distant hosts
have no way of knowing whether a given address is the lowest address
in a faraway network segment, so they treat it no differently than any
other unicast address.)  This means that each local site can change this
behavior locally, resulting immediately in global interoperability with
the newly usable lowest local address.

Modern software and documentation continues to use the definition of the
directed, or "net-directed", broadcast address as "a host ID of all one
bits".  The Internet no longer gets any benefit from having two different
broadcast addresses usable on every Ethernet segment.  I have not been
able to find any documentation that suggests that users or software should
ever intentionally use the all-zero form, or that justifies it other than
as a historic Berkeleyism.  RFCs 1112, 1812, and 3021 state that hosts and
routers need to maintain compatibility with the old form -- but they give
no rationale other than the past existence of the 4.2BSD behavior.

We're happy to provide more historical details or information about
behavior of other systems in this regard by e-mail or as future patches
to kernel documentation files.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: Lowest IPv4 address in a subnet is valid
Seth David Schoen [Thu, 13 May 2021 04:38:25 +0000 (21:38 -0700)]
selftests: Lowest IPv4 address in a subnet is valid

Expect the lowest IPv4 address in a subnet to be assignable
and addressable as a unicast (non-broadcast) address on a
local network segment.

Signed-off-by: Seth David Schoen <schoen@loyalty.org>
Suggested-by: John Gilmore <gnu@toad.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoip: Treat IPv4 segment's lowest address as unicast
Seth David Schoen [Thu, 13 May 2021 04:37:49 +0000 (21:37 -0700)]
ip: Treat IPv4 segment's lowest address as unicast

Treat only the highest, not the lowest, IPv4 address within a local
subnet as a broadcast address.

Signed-off-by: Seth David Schoen <schoen@loyalty.org>
Suggested-by: John Gilmore <gnu@toad.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: mdio: ipq8064: enlarge sleep after read/write operation
Ansuel Smith [Fri, 14 May 2021 21:03:51 +0000 (23:03 +0200)]
net: mdio: ipq8064: enlarge sleep after read/write operation

With the use of the qca8k dsa driver, some problem arised related to
port status detection. With a load on a specific port (for example a
simple speed test), the driver starts to behave in a strange way and
garbage data is produced. To address this, enlarge the sleep delay and
address a bug for the reg offset 31 that require additional delay for
this specific reg.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: mdio: ipq8064: add regmap config to disable REGCACHE
Ansuel Smith [Fri, 14 May 2021 21:03:50 +0000 (23:03 +0200)]
net: mdio: ipq8064: add regmap config to disable REGCACHE

mdio drivers should not use REGCHACHE. Also disable locking since it's
handled by the mdio users and regmap is always accessed atomically.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: mdio: ipq8064: clean whitespaces in define
Ansuel Smith [Fri, 14 May 2021 21:03:49 +0000 (23:03 +0200)]
net: mdio: ipq8064: clean whitespaces in define

Fix mixed whitespace and tab for define spacing.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'qca8k-improvements'
David S. Miller [Fri, 14 May 2021 22:30:23 +0000 (15:30 -0700)]
Merge branch 'qca8k-improvements'

Ansuel Smith says:

====================
Multiple improvement to qca8k stability

Currently qca8337 switch are widely used on ipq8064 based router.
On these particular router it was notice a very unstable switch with
port not link detected as link with unknown speed, port dropping
randomly and general unreliability. Lots of testing and comparison
between this dsa driver and the original qsdk driver showed lack of some
additional delay and values. A main difference arised from the original
driver and the dsa one. The original driver didn't use MASTER regs to
read phy status and the dedicated mdio driver worked correctly. Now that
the dsa driver actually use these regs, it was found that these special
read/write operation required mutual exclusion to normal
qca8k_read/write operation. The add of mutex for these operation fixed
the random port dropping and now only the actual linked port randomly
dropped. Adding additional delay for set_page operation and fixing a bug
in the mdio dedicated driver fixed also this problem. The current driver
requires also more time to apply vlan switch. All of these changes and
tweak permit a now very stable and reliable dsa driver and 0 port
dropping. This series is currently tested by at least 5 user with
different routers and all reports positive results and no problems.

Changes v6:
- Fix spelling mistake
- Change ms to ns (confirmed by datasheet)
Changes v5:
- Removed mdio patch (sent separetly to try to reduce the series)
  I know it was asked to reduced this series since it big, but rework
  the new changes to skip and error check looks wrong. Since half of them
  are actually already reviewed I think it's better to keep this series as is.
- Improve rgmii configurable patch
- Move qca8k phy dedicated driver to at803x phy driver
- Add support for dedicated internal mdio driver for qca8k
Changes v4:
- Use iopoll for busy_wait function
- Better describe and split some confusing commits
- Fix bad rgmii delay configurable patch
- Drop phy generic patch to pass flags with phylink_connect_phy
- Add dsa2 patch to declare mdio node in the switch node
- Add dsa patch to permit dsa driver to declare custom get_phys_mii_mask
    Some background about the last 2 patch.
    The qca8k switch doesn't have a 1:1 map between port reg and phy reg.
    Currently it's used a function to convert port to the internal phy reg.
    I added some patch to fix this.
    - The dsa driver now check if the mdio node is present and use the of variant
      of the mdiobus_register
    - A custom phy_mii_mask is required as currently the mask is generated from
      the port reg, but in our case the mask would be different as it should be
      generated from the phy reg. To generalize this I added an extra function
      that driver can provide to pass custom phy_mii_mask.
Changes v3:
- Revert mdio writel changes (use regmap with REGCACHE disabled)
- Split propagate error patch to 4 different patch
Changes v2:
- Implemented phy driver for internal PHYs
  I'm testing cable test functions as I found some documentation that
  actually declare regs about it. Problem is that it doesn't actually
  work. It seems that the value set are ignored by the phy.
- Made the rgmii delay configurable
- Reordered patch
- Split mdio patches to more specific ones
- Reworked mdio driver to use readl/writel instead of regmap
- Reworked the entire driver to make it aware of any read/write error.
- Added phy generic patch to pass flags with phylink_connect_phy
  function
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: phy: add support for qca8k switch internal PHY in at803x
Ansuel Smith [Fri, 14 May 2021 21:00:15 +0000 (23:00 +0200)]
net: phy: add support for qca8k switch internal PHY in at803x

Since the at803x share the same regs, it's assumed they are based on the
same implementation. Make it part of the at803x PHY driver to skip
having redudant code.
Add initial support for qca8k internal PHYs. The internal PHYs requires
special mmd and debug values to be set based on the switch revision
passwd using the dev_flags. Supports output of idle, receive and eee_wake
errors stats.
Some debug values sets can't be translated as the documentation lacks any
reference about them.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: phy: at803x: clean whitespace errors
Ansuel Smith [Fri, 14 May 2021 21:00:14 +0000 (23:00 +0200)]
net: phy: at803x: clean whitespace errors

Clean any whitespace errors and fix not aligned define.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: qca8k: pass switch_revision info to phy dev_flags
Ansuel Smith [Fri, 14 May 2021 21:00:13 +0000 (23:00 +0200)]
net: dsa: qca8k: pass switch_revision info to phy dev_flags

Define get_phy_flags to pass switch_Revision needed to tweak the
internal PHY with debug values based on the revision.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: qca8k: improve internal mdio read/write bus access
Ansuel Smith [Fri, 14 May 2021 21:00:12 +0000 (23:00 +0200)]
net: dsa: qca8k: improve internal mdio read/write bus access

Improve the internal mdio read/write bus access by caching the value
without accessing it for every read/write.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agodevicetree: bindings: dsa: qca8k: Document internal mdio definition
Ansuel Smith [Fri, 14 May 2021 21:00:11 +0000 (23:00 +0200)]
devicetree: bindings: dsa: qca8k: Document internal mdio definition

Document new way of declare mapping of internal PHY to port.
The new implementation directly declare the PHY connected to the port
by adding a node in the switch node. The driver detect this and register
an internal mdiobus using the mapping defined in the mdio node.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: qca8k: add support for internal phy and internal mdio
Ansuel Smith [Fri, 14 May 2021 21:00:10 +0000 (23:00 +0200)]
net: dsa: qca8k: add support for internal phy and internal mdio

Add support to setup_mdio_bus for internal phy declaration. Introduce a
flag to use the legacy port phy mapping by default and use the direct
mapping if a mdio node is detected in the switch node. Register a
dedicated mdio internal mdio bus to address the different mapping
between port and phy if the mdio node is detected.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: qca8k: enlarge mdio delay and timeout
Ansuel Smith [Fri, 14 May 2021 21:00:09 +0000 (23:00 +0200)]
net: dsa: qca8k: enlarge mdio delay and timeout

The witch require some extra delay after setting page or the next
read/write can use still use the old page. Add a delay after the
set_page function to address this as it's done in QSDK legacy driver.
Some timeouts were notice with VLAN and phy function, enlarge the
mdio busy wait timeout to fix these problems.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: qca8k: dsa: qca8k: protect MASTER busy_wait with mdio mutex
Ansuel Smith [Fri, 14 May 2021 21:00:08 +0000 (23:00 +0200)]
net: dsa: qca8k: dsa: qca8k: protect MASTER busy_wait with mdio mutex

MDIO_MASTER operation have a dedicated busy wait that is not protected
by the mdio mutex. This can cause situation where the MASTER operation
is done and a normal operation is executed between the MASTER read/write
and the MASTER busy_wait. Rework the qca8k_mdio_read/write function to
address this issue by binding the lock for the whole MASTER operation
and not only the mdio read/write common operation.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: qca8k: clear MASTER_EN after phy read/write
Ansuel Smith [Fri, 14 May 2021 21:00:07 +0000 (23:00 +0200)]
net: dsa: qca8k: clear MASTER_EN after phy read/write

Clear MDIO_MASTER_EN bit from MDIO_MASTER_CTRL after read/write
operation. The MDIO_MASTER_EN bit is not reset after read/write
operation and the next operation can be wrongly interpreted by the
switch as a mdio operation. This cause a production of wrong/garbage
data from the switch and underfined bheavior. (random port drop,
unplugged port flagged with link up, wrong port speed)
Also on driver remove the MASTER_CTRL can be left set and cause the
malfunction of any next driver using the mdio device.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: qca8k: make rgmii delay configurable
Ansuel Smith [Fri, 14 May 2021 21:00:06 +0000 (23:00 +0200)]
net: dsa: qca8k: make rgmii delay configurable

The legacy qsdk code used a different delay instead of the max value.
Qsdk use 1 ns for rx and 2 ns for tx. Make these values configurable
using the standard rx/tx-internal-delay-ps ethernet binding and apply
qsdk values by default. The connected gmac doesn't add any delay so no
additional delay is added to tx/rx.
On this switch the delay is actually in ns so value should be in the
1000 order. Any value converted from ps to ns by dividing it by 1000
as the switch max value for delay is 3ns.

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>