platform/kernel/linux-starfive.git
2 years agonet: ethernet: mtk_eth_soc: add check for allocation failure
Dan Carpenter [Thu, 21 Apr 2022 15:49:02 +0000 (18:49 +0300)]
net: ethernet: mtk_eth_soc: add check for allocation failure

Check if the kzalloc() failed.

Fixes: 804775dfc288 ("net: ethernet: mtk_eth_soc: add support for Wireless Ethernet Dispatch (WED)")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoethernet: broadcom/sb1250-mac: remove BUG_ON in sbmac_probe()
Yang Yingliang [Thu, 21 Apr 2022 13:51:48 +0000 (21:51 +0800)]
ethernet: broadcom/sb1250-mac: remove BUG_ON in sbmac_probe()

Replace the BUG_ON() with returning error code to handle
the fault more gracefully.

Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: mscc: ocelot: Remove useless code
Haowen Bai [Thu, 21 Apr 2022 10:34:41 +0000 (18:34 +0800)]
net: mscc: ocelot: Remove useless code

payload only memset but no use at all, so we drop them.

Signed-off-by: Haowen Bai <baihaowen@meizu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'mlxsw-line-card-model'
David S. Miller [Mon, 25 Apr 2022 09:42:29 +0000 (10:42 +0100)]
Merge branch 'mlxsw-line-card-model'

Ido Schimmel says:

====================
mlxsw: extend line card model by devices and info

Jiri says:

This patchset is extending the line card model by three items:
1) line card devices
2) line card info
3) line card device info

First three patches are introducing the necessary changes in devlink
core.

Then, all three extensions are implemented in mlxsw alongside with
selftest.

Examples:
$ devlink lc show pci/0000:01:00.0 lc 8
pci/0000:01:00.0:
  lc 8 state active type 16x100G
    supported_types:
      16x100G
    devices:
      device 0
      device 1
      device 2
      device 3
$ devlink lc info pci/0000:01:00.0 lc 8
pci/0000:01:00.0:
  lc 8
    versions:
        fixed:
          hw.revision 0
        running:
          ini.version 4
    devices:
      device 0
        versions:
            running:
              fw 19.2010.1310
      device 1
        versions:
            running:
              fw 19.2010.1310
      device 2
        versions:
            running:
              fw 19.2010.1310
      device 3
        versions:
            running:
              fw 19.2010.1310

Note that device FW flashing is going to be implemented in the follow-up
patchset.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: mlxsw: Check device info on activated line card
Jiri Pirko [Mon, 25 Apr 2022 03:44:31 +0000 (06:44 +0300)]
selftests: mlxsw: Check device info on activated line card

Once line card is activated, check the device FW version is exposed.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_linecards: Expose device FW version over device info
Jiri Pirko [Mon, 25 Apr 2022 03:44:30 +0000 (06:44 +0300)]
mlxsw: core_linecards: Expose device FW version over device info

Extend MDDQ to obtain FW version of line card device and implement
device_info_get() op to fill up the info with that.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: reg: Extend MDDQ device_info by FW version fields
Jiri Pirko [Mon, 25 Apr 2022 03:44:29 +0000 (06:44 +0300)]
mlxsw: reg: Extend MDDQ device_info by FW version fields

Add FW version fields to MDDQ device_info.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: mlxsw: Check line card info on provisioned line card
Jiri Pirko [Mon, 25 Apr 2022 03:44:28 +0000 (06:44 +0300)]
selftests: mlxsw: Check line card info on provisioned line card

Once line card is provisioned, check if HW revision and INI version
are exposed.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_linecards: Expose HW revision and INI version
Jiri Pirko [Mon, 25 Apr 2022 03:44:27 +0000 (06:44 +0300)]
mlxsw: core_linecards: Expose HW revision and INI version

Implement info_get() to expose HW revision of a linecard and loaded INI
version.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: mlxsw: Check devices on provisioned line card
Jiri Pirko [Mon, 25 Apr 2022 03:44:26 +0000 (06:44 +0300)]
selftests: mlxsw: Check devices on provisioned line card

Once line card is provisioned, check the count of devices on it and
print them out.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_linecards: Probe provisioned line cards for devices and attach them
Jiri Pirko [Mon, 25 Apr 2022 03:44:25 +0000 (06:44 +0300)]
mlxsw: core_linecards: Probe provisioned line cards for devices and attach them

In case the line card is provisioned, go over all possible existing
devices (gearboxes) on it and attach them, so devlink core is aware of
them.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: reg: Extend MDDQ by device_info
Jiri Pirko [Mon, 25 Apr 2022 03:44:24 +0000 (06:44 +0300)]
mlxsw: reg: Extend MDDQ by device_info

Extend existing MDDQ register by possibility to query information about
devices residing on a line card.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodevlink: introduce line card device info infrastructure
Jiri Pirko [Mon, 25 Apr 2022 03:44:23 +0000 (06:44 +0300)]
devlink: introduce line card device info infrastructure

Extend the line card info message with information (e.g., FW version)
about devices found on the line card.

Example:
$ devlink lc info pci/0000:01:00.0 lc 8
pci/0000:01:00.0:
  lc 8
    versions:
        fixed:
          hw.revision 0
        running:
          ini.version 4
    devices:
      device 0
        versions:
            running:
              fw 19.2010.1310
      device 1
        versions:
            running:
              fw 19.2010.1310
      device 2
        versions:
            running:
              fw 19.2010.1310
      device 3
        versions:
            running:
              fw 19.2010.1310

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodevlink: introduce line card info get message
Jiri Pirko [Mon, 25 Apr 2022 03:44:22 +0000 (06:44 +0300)]
devlink: introduce line card info get message

Allow the driver to provide per line card info get op to fill-up info,
similar to the "devlink dev info".

Example:

$ devlink lc info pci/0000:01:00.0 lc 8
pci/0000:01:00.0:
  lc 8
    versions:
        fixed:
          hw.revision 0
        running:
          ini.version 4

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodevlink: introduce line card devices support
Jiri Pirko [Mon, 25 Apr 2022 03:44:21 +0000 (06:44 +0300)]
devlink: introduce line card devices support

Line card can contain one or more devices that makes sense to make
visible to the user. For example, this can be a gearbox with
flash memory, which could be updated.

Provide the driver possibility to attach such devices to a line card
and expose those to user.

Example:
$ devlink lc show pci/0000:01:00.0 lc 8
pci/0000:01:00.0:
  lc 8 state active type 16x100G
    supported_types:
      16x100G
    devices:
      device 0
      device 1
      device 2
      device 3

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'dsa-selftests'
David S. Miller [Sat, 23 Apr 2022 11:18:16 +0000 (12:18 +0100)]
Merge branch 'dsa-selftests'

Vladimir Oltean says:

====================
DSA selftests

When working on complex new features or reworks it becomes increasingly
difficult to ensure there aren't regressions being introduced, and
therefore it would be nice if we could go over the functionality we
already have and write some tests for it.

Verbally I know from Tobias Waldekranz that he has been working on some
selftests for DSA, yet I have never seen them, so here I am adding some
tests I have written which have been useful for me. The list is by no
means complete (it only covers elementary functionality), but it's still
good to have as a starting point. I also borrowed some refactoring
changes from Joachim Wiberg that he submitted for his "net: bridge:
forwarding of unknown IPv4/IPv6/MAC BUM traffic" series, but not the
entirety of his selftests. I now think that his selftests have some
overlap with bridge_vlan_unaware.sh and bridge_vlan_aware.sh and they
should be more tightly integrated with each other - yet I didn't do that
either :). Another issue I had with his selftests was that they jumped
straight ahead to configure brport flags on br0 (a radical new idea
still at RFC status) while we have bigger problems, and we don't have
nearly enough coverage for the *existing* functionality.

One idea introduced here which I haven't seen before is the symlinking
of relevant forwarding selftests to the selftests/drivers/net/<my-driver>/
folder, plus a forwarding.config file. I think there's some value in
having things structured this way, since the forwarding dir has so many
selftests that aren't relevant to DSA that it is a bit difficult to find
the ones that are.

While searching for applications that I could use for multicast testing
(not my domain of interest/knowledge really), I found Joachim Wiberg's
mtools, mcjoin and omping, and I tried them all with various degrees of
success. In particular, I was going to use mcjoin, but I faced some
issues getting IPv6 multicast traffic to work in a VRF, and I bothered
David Ahern about it here:
https://lore.kernel.org/netdev/97eaffb8-2125-834e-641f-c99c097b6ee2@gmail.com/t/
It seems that the problem is that this application should use
SO_BINDTODEVICE, yet it doesn't.

So I ended up patching the bare-bones mtools (msend, mreceive) forked by
Joachim from the University of Virginia's Multimedia Networks Group to
include IPv6 support, and to use SO_BINDTODEVICE. This is what I'm using
now for IPv6.

Note that mausezahn doesn't appear to do a particularly good job of
supporting IPv6 really, and I needed a program to emit the actual
IP_ADD_MEMBERSHIP calls, for dev_mc_add(), so I could test RX filtering.
Crafting the IGMP/MLD reports by hand doesn't really do the trick.
While extremely bare-bones, the mreceive application now seems to do
what I need it to.

Feedback appreciated, it is very likely that I could have done things in
a better way.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: drivers: dsa: add a subset of forwarding selftests
Vladimir Oltean [Fri, 22 Apr 2022 10:15:04 +0000 (13:15 +0300)]
selftests: drivers: dsa: add a subset of forwarding selftests

This adds an initial subset of forwarding selftests which I considered
to be relevant for DSA drivers, along with a forwarding.config that
makes it easier to run them (disables veth pair creation, makes sure MAC
addresses are unique and stable).

The intention is to request driver writers to run these selftests during
review and make sure that the tests pass, or at least that the problems
are known.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: forwarding: add a test for local_termination.sh
Vladimir Oltean [Fri, 22 Apr 2022 10:15:03 +0000 (13:15 +0300)]
selftests: forwarding: add a test for local_termination.sh

This tests the capability of switch ports to filter out undesired
traffic. Different drivers are expected to have different capabilities
here (so some may fail and some may pass), yet the test still has some
value, for example to check for regressions.

There are 2 kinds of failures, one is when a packet which should have
been accepted isn't (and that should be fixed), and the other "failure"
(as reported by the test) is when a packet could have been filtered out
(for being unnecessary) yet it was received.

The bridge driver fares particularly badly at this test:

TEST: br0: Unicast IPv4 to primary MAC address                      [ OK ]
TEST: br0: Unicast IPv4 to macvlan MAC address                      [ OK ]
TEST: br0: Unicast IPv4 to unknown MAC address                      [FAIL]
        reception succeeded, but should have failed
TEST: br0: Unicast IPv4 to unknown MAC address, promisc             [ OK ]
TEST: br0: Unicast IPv4 to unknown MAC address, allmulti            [FAIL]
        reception succeeded, but should have failed
TEST: br0: Multicast IPv4 to joined group                           [ OK ]
TEST: br0: Multicast IPv4 to unknown group                          [FAIL]
        reception succeeded, but should have failed
TEST: br0: Multicast IPv4 to unknown group, promisc                 [ OK ]
TEST: br0: Multicast IPv4 to unknown group, allmulti                [ OK ]
TEST: br0: Multicast IPv6 to joined group                           [ OK ]
TEST: br0: Multicast IPv6 to unknown group                          [FAIL]
        reception succeeded, but should have failed
TEST: br0: Multicast IPv6 to unknown group, promisc                 [ OK ]
TEST: br0: Multicast IPv6 to unknown group, allmulti                [ OK ]

mainly because it does not implement IFF_UNICAST_FLT. Yet I still think
having the test (with the failures) is useful in case somebody wants to
tackle that problem in the future, to make an easy before-and-after
comparison.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: forwarding: add a no_forwarding.sh test
Vladimir Oltean [Fri, 22 Apr 2022 10:15:02 +0000 (13:15 +0300)]
selftests: forwarding: add a no_forwarding.sh test

Bombard a standalone switch port with various kinds of traffic to ensure
it is really standalone and doesn't leak packets to other switch ports.
Also check for switch ports in different bridges, and switch ports in a
VLAN-aware bridge but having different pvids. No forwarding should take
place in either case.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: forwarding: add helper for retrieving IPv6 link-local address of interface
Vladimir Oltean [Fri, 22 Apr 2022 10:15:01 +0000 (13:15 +0300)]
selftests: forwarding: add helper for retrieving IPv6 link-local address of interface

Pinging an IPv6 link-local multicast address selects the link-local
unicast address of the interface as source, and we'd like to monitor for
that in tcpdump.

Add a helper to the forwarding library which retrieves the link-local
IPv6 address of an interface, to make that task easier.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: forwarding: add helpers for IP multicast group joins/leaves
Vladimir Oltean [Fri, 22 Apr 2022 10:15:00 +0000 (13:15 +0300)]
selftests: forwarding: add helpers for IP multicast group joins/leaves

Extend the forwarding library with calls to some small C programs which
join an IP multicast group and send some packets to it. Both IPv4 and
IPv6 groups are supported. Use cases range from testing IGMP/MLD
snooping, to RX filtering, to multicast routing.

Testing multicast traffic using msend/mreceive is intended to be done
using tcpdump.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: forwarding: multiple instances in tcpdump helper
Joachim Wiberg [Fri, 22 Apr 2022 10:14:59 +0000 (13:14 +0300)]
selftests: forwarding: multiple instances in tcpdump helper

Extend tcpdump_start() & C:o to handle multiple instances.  Useful when
observing bridge operation, e.g., unicast learning/flooding, and any
case of multicast distribution (to these ports but not that one ...).

This means the interface argument is now a mandatory argument to all
tcpdump_*() functions, hence the changes to the ocelot flower test.

Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: forwarding: add TCPDUMP_EXTRA_FLAGS to lib.sh
Joachim Wiberg [Fri, 22 Apr 2022 10:14:58 +0000 (13:14 +0300)]
selftests: forwarding: add TCPDUMP_EXTRA_FLAGS to lib.sh

For some use-cases we may want to change the tcpdump flags used in
tcpdump_start().  For instance, observing interfaces without the PROMISC
flag, e.g. to see what's really being forwarded to the bridge interface.

Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: forwarding: add option to run tests with stable MAC addresses
Vladimir Oltean [Fri, 22 Apr 2022 10:14:57 +0000 (13:14 +0300)]
selftests: forwarding: add option to run tests with stable MAC addresses

By default, DSA switch ports inherit their MAC address from the DSA
master.

This works well for practical situations, but some selftests like
bridge_vlan_unaware.sh loop back 2 standalone DSA ports with 2 bridged
DSA ports, and require the bridge to forward packets between the
standalone ports.

Due to the bridge seeing that the MAC DA it needs to forward is present
as a local FDB entry (it coincides with the MAC address of the bridge
ports), the test packets are not forwarded, but terminated locally on
br0. In turn, this makes the ping and ping6 tests fail.

Address this by introducing an option to have stable MAC addresses.
When mac_addr_prepare is called, the current addresses of the netifs are
saved and replaced with 00:01:02:03:04:${netif number}. Then when
mac_addr_restore is called at the end of the test, the original MAC
addresses are restored. This ensures that the MAC addresses are unique,
which makes the test pass even for DSA ports.

The usage model is for the behavior to be opt-in via STABLE_MAC_ADDRS,
which DSA should set to true, all others behave as before. By hooking
the calls to mac_addr_prepare and mac_addr_restore within the forwarding
lib itself, we do not need to patch each individual selftest, the only
requirement is that pre_cleanup is called.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'mptcp-tcp-fallback'
David S. Miller [Sat, 23 Apr 2022 10:51:06 +0000 (11:51 +0100)]
Merge branch 'mptcp-tcp-fallback'

Mat Martineau says:

====================
mptcp: TCP fallback for established connections

RFC 8684 allows some MPTCP connections to fall back to regular TCP when
the MPTCP DSS checksum detects middlebox interference, there is only a
single subflow, and there is no unacknowledged out-of-sequence
data. When this condition is detected, the stack sends a MPTCP DSS
option with an "infinite mapping" to signal that a fallback is
happening, and the peers will stop sending MPTCP options in their TCP
headers. The Linux MPTCP stack has not yet supported this type of
fallback, instead closing the connection when the MPTCP checksum fails.

This series adds support for fallback to regular TCP in a more limited
scenario, for only MPTCP connections that have never connected
additional subflows or transmitted out-of-sequence data. The selftests
are also updated to check new MIBs that track infinite mappings.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: mptcp: add infinite map mibs check
Geliang Tang [Fri, 22 Apr 2022 21:55:43 +0000 (14:55 -0700)]
selftests: mptcp: add infinite map mibs check

This patch adds a function chk_infi_nr() to check the mibs for the
infinite mapping. Invoke it in chk_join_nr() when validate_checksum
is set.

Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomptcp: dump infinite_map field in mptcp_dump_mpext
Geliang Tang [Fri, 22 Apr 2022 21:55:42 +0000 (14:55 -0700)]
mptcp: dump infinite_map field in mptcp_dump_mpext

In trace event class mptcp_dump_mpext, dump the newly added infinite_map
field of struct mptcp_dump_mpext too.

Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomptcp: add mib for infinite map sending
Geliang Tang [Fri, 22 Apr 2022 21:55:41 +0000 (14:55 -0700)]
mptcp: add mib for infinite map sending

This patch adds a new mib named MPTCP_MIB_INFINITEMAPTX, increase it
when a infinite mapping has been sent out.

Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomptcp: infinite mapping receiving
Geliang Tang [Fri, 22 Apr 2022 21:55:40 +0000 (14:55 -0700)]
mptcp: infinite mapping receiving

This patch adds the infinite mapping receiving logic. When the infinite
mapping is received, set the map_data_len of the subflow to 0.

In subflow_check_data_avail(), only reset the subflow when the map_data_len
of the subflow is non-zero.

Suggested-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomptcp: infinite mapping sending
Geliang Tang [Fri, 22 Apr 2022 21:55:39 +0000 (14:55 -0700)]
mptcp: infinite mapping sending

This patch adds the infinite mapping sending logic.

Add a new flag send_infinite_map in struct mptcp_subflow_context. Set
it true when a single contiguous subflow is in use and the
allow_infinite_fallback flag is true in mptcp_pm_mp_fail_received().

In mptcp_sendmsg_frag(), if this flag is true, call the new function
mptcp_update_infinite_map() to set the infinite mapping.

Add a new flag infinite_map in struct mptcp_ext, set it true in
mptcp_update_infinite_map(), and check this flag in a new helper
mptcp_check_infinite_map().

In mptcp_update_infinite_map(), set data_len to 0, and clear the
send_infinite_map flag, then do fallback.

In mptcp_established_options(), use the helper mptcp_check_infinite_map()
to let the infinite mapping DSS can be sent out in the fallback mode.

Suggested-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomptcp: track and update contiguous data status
Geliang Tang [Fri, 22 Apr 2022 21:55:38 +0000 (14:55 -0700)]
mptcp: track and update contiguous data status

This patch adds a new member allow_infinite_fallback in mptcp_sock,
which is initialized to 'true' when the connection begins and is set
to 'false' on any retransmit or successful MP_JOIN. Only do infinite
mapping fallback if there is a single subflow AND there have been no
retransmissions AND there have never been any MP_JOINs.

Suggested-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomptcp: add the fallback check
Geliang Tang [Fri, 22 Apr 2022 21:55:37 +0000 (14:55 -0700)]
mptcp: add the fallback check

This patch adds the fallback check in subflow_check_data_avail(). Only
do the fallback when the msk hasn't fallen back yet.

Suggested-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomptcp: don't send RST for single subflow
Geliang Tang [Fri, 22 Apr 2022 21:55:36 +0000 (14:55 -0700)]
mptcp: don't send RST for single subflow

When a bad checksum is detected and a single subflow is in use, don't
send RST + MP_FAIL, send data_ack + MP_FAIL instead.

So invoke tcp_send_active_reset() only when mptcp_has_another_subflow()
is true.

Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: hns3: Fix spelling mistake "actvie" -> "active"
Colin Ian King [Thu, 21 Apr 2022 08:55:46 +0000 (09:55 +0100)]
net: hns3: Fix spelling mistake "actvie" -> "active"

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

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Link: https://lore.kernel.org/r/20220421085546.321792-1-colin.i.king@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotsnep: Remove useless null check before call of_node_put()
Haowen Bai [Thu, 21 Apr 2022 02:48:03 +0000 (10:48 +0800)]
tsnep: Remove useless null check before call of_node_put()

No need to add null check before call of_node_put(), since the
implementation of of_node_put() has done it.

Signed-off-by: Haowen Bai <baihaowen@meizu.com>
Link: https://lore.kernel.org/r/1650509283-26168-1-git-send-email-baihaowen@meizu.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'add-ethtool-sqi-support-for-lan87xx-t1-phy'
Jakub Kicinski [Fri, 22 Apr 2022 23:30:07 +0000 (16:30 -0700)]
Merge branch 'add-ethtool-sqi-support-for-lan87xx-t1-phy'

Arun Ramadoss says:

====================
add ethtool SQI support for LAN87xx T1 Phy

This patch series add the Signal Quality Index measurement for the LAN87xx and
LAN937x T1 phy. Updated the maintainers file for microchip_t1.c.
====================

Link: https://lore.kernel.org/r/20220420152016.9680-1-arun.ramadoss@microchip.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMAINTAINERS: Add maintainers for Microchip T1 Phy driver
Arun Ramadoss [Wed, 20 Apr 2022 15:20:16 +0000 (20:50 +0530)]
MAINTAINERS: Add maintainers for Microchip T1 Phy driver

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: phy: LAN87xx: add ethtool SQI support
Arun Ramadoss [Wed, 20 Apr 2022 15:20:15 +0000 (20:50 +0530)]
net: phy: LAN87xx: add ethtool SQI support

This patch add the support for measuring Signal Quality Index for
LAN87xx and LAN937x T1 Phy. It uses the SQI Method 5 for obtaining the
values.

Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agomlxsw: core_linecards: Fix size of array element during ini_files allocation
Jiri Pirko [Wed, 20 Apr 2022 14:20:07 +0000 (17:20 +0300)]
mlxsw: core_linecards: Fix size of array element during ini_files allocation

types_info->ini_files is an array of pointers
to struct mlxsw_linecard_ini_file.

Fix the kmalloc_array() argument to be of a size of a pointer.

Addresses-Coverity: ("Incorrect expression  (SIZEOF_MISMATCH)")
Fixes: b217127e5e4e ("mlxsw: core_linecards: Add line card objects and implement provisioning")
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/20220420142007.3041173-1-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoqed: Remove IP services API.
Guillaume Nault [Thu, 21 Apr 2022 12:47:26 +0000 (14:47 +0200)]
qed: Remove IP services API.

qed_nvmetcp_ip_services.c and its corresponding header file were
introduced in commit 806ee7f81a2b ("qed: Add IP services APIs support")
but there's still no users for any of the functions they declare.
Since these files are effectively unused, let's just drop them.

Found by code inspection. Compile-tested only.

Signed-off-by: Guillaume Nault <gnault@redhat.com>
Link: https://lore.kernel.org/r/351ac8c847980e22850eb390553f8cc0e1ccd0ce.1650545051.git.gnault@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agodt-bindings: net: mediatek,net: convert to the json-schema
Lorenzo Bianconi [Wed, 20 Apr 2022 14:07:07 +0000 (16:07 +0200)]
dt-bindings: net: mediatek,net: convert to the json-schema

This patch converts the existing mediatek-net.txt binding file
in yaml format.

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'ipv6-RT_ONLINK-remove-prep'
David S. Miller [Fri, 22 Apr 2022 12:06:03 +0000 (13:06 +0100)]
Merge branch 'ipv6-RT_ONLINK-remove-prep'

Guillaume Nault says:

====================
ipv4: First steps toward removing RTO_ONLINK

RTO_ONLINK is a flag that allows to reduce the scope of route lookups.
It's stored in a normally unused bit of the ->flowi4_tos field, in
struct flowi4. However it has several problems:

 * This bit is also used by ECN. Although ECN bits are supposed to be
   cleared before doing a route lookup, it happened that some code
   paths didn't properly sanitise their ->flowi4_tos. So this mechanism
   is fragile and we had bugs in the past where ECN bits slipped in and
   could end up being erroneously interpreted as RTO_ONLINK.

 * A dscp_t type was recently introduced to ensure ECN bits are cleared
   during route lookups. ->flowi4_tos is the most important structure
   field to convert, but RTO_ONLINK prevents such conversion, as dscp_t
   mandates that ECN bits (where RTO_ONLINK is stored) be zero.

Therefore we need to stop using RTO_ONLINK altogether. Fortunately
RTO_ONLINK isn't a necessity. Instead of passing a flag in ->flowi4_tos
to tell the route lookup function to restrict the scope, we can simply
initialise the scope correctly.

Patch 1 does some preparatory work: it stops resetting ->flowi4_scope
automatically before a route lookup, thus allowing callers to set their
desired scope without having to rely on the RTO_ONLINK flag.

Patch 2-3 convert a few code paths to avoid relying on RTO_ONLINK.

More conversions will have to take place before we can eventually
remove this flag.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoipv4: Initialise ->flowi4_scope properly in ICMP handlers.
Guillaume Nault [Wed, 20 Apr 2022 23:21:37 +0000 (01:21 +0200)]
ipv4: Initialise ->flowi4_scope properly in ICMP handlers.

All the *_redirect() and *_update_pmtu() functions initialise their
struct flowi4 variable with either __build_flow_key() or
build_sk_flow_key(). When sk is provided, these functions use
RT_CONN_FLAGS() to set ->flowi4_tos and always use RT_SCOPE_UNIVERSE
for ->flowi4_scope. Then they rely on ip_rt_fix_tos() to adjust the
scope based on the RTO_ONLINK bit and to mask the tos with
IPTOS_RT_MASK.

This patch modifies __build_flow_key() and build_sk_flow_key() to
properly initialise ->flowi4_tos and ->flowi4_scope, so that the
ICMP redirects and PMTU handlers don't need an extra call to
ip_rt_fix_tos() before doing a fib lookup. That is, we:

  * Drop RT_CONN_FLAGS(): use ip_sock_rt_tos() and ip_sock_rt_scope()
    instead, so that we don't have to rely on ip_rt_fix_tos() to adjust
    the scope anymore.

  * Apply IPTOS_RT_MASK to the tos, so that we don't need
    ip_rt_fix_tos() to do it for us.

  * Drop the ip_rt_fix_tos() calls that now become useless.

The only remaining ip_rt_fix_tos() caller is ip_route_output_key_hash()
which needs it as long as external callers still use the RTO_ONLINK
flag.

Note:
  This patch also drops some useless RT_TOS() calls as IPTOS_RT_MASK is
  a stronger mask.

Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoipv4: Avoid using RTO_ONLINK with ip_route_connect().
Guillaume Nault [Wed, 20 Apr 2022 23:21:33 +0000 (01:21 +0200)]
ipv4: Avoid using RTO_ONLINK with ip_route_connect().

Now that ip_rt_fix_tos() doesn't reset ->flowi4_scope unconditionally,
we don't have to rely on the RTO_ONLINK bit to properly set the scope
of a flowi4 structure. We can just set ->flowi4_scope explicitly and
avoid using RTO_ONLINK in ->flowi4_tos.

This patch converts callers of ip_route_connect(). Instead of setting
the tos parameter with RT_CONN_FLAGS(sk), as all callers do, we can:

  1- Drop the tos parameter from ip_route_connect(): its value was
     entirely based on sk, which is also passed as parameter.

  2- Set ->flowi4_scope depending on the SOCK_LOCALROUTE socket option
     instead of always initialising it with RT_SCOPE_UNIVERSE (let's
     define ip_sock_rt_scope() for this purpose).

  3- Avoid overloading ->flowi4_tos with RTO_ONLINK: since the scope is
     now properly initialised, we don't need to tell ip_rt_fix_tos() to
     adjust ->flowi4_scope for us. So let's define ip_sock_rt_tos(),
     which is the same as RT_CONN_FLAGS() but without the RTO_ONLINK
     bit overload.

Note:
  In the original ip_route_connect() code, __ip_route_output_key()
  might clear the RTO_ONLINK bit of fl4->flowi4_tos (because of
  ip_rt_fix_tos()). Therefore flowi4_update_output() had to reuse the
  original tos variable. Now that we don't set RTO_ONLINK any more,
  this is not a problem and we can use fl4->flowi4_tos in
  flowi4_update_output().

Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoipv4: Don't reset ->flowi4_scope in ip_rt_fix_tos().
Guillaume Nault [Wed, 20 Apr 2022 23:21:24 +0000 (01:21 +0200)]
ipv4: Don't reset ->flowi4_scope in ip_rt_fix_tos().

All callers already initialise ->flowi4_scope with RT_SCOPE_UNIVERSE,
either by manual field assignment, memset(0) of the whole structure or
implicit structure initialisation of on-stack variables
(RT_SCOPE_UNIVERSE actually equals 0).

Therefore, we don't need to always initialise ->flowi4_scope in
ip_rt_fix_tos(). We only need to reduce the scope to RT_SCOPE_LINK when
the special RTO_ONLINK flag is present in the tos.

This will allow some code simplification, like removing
ip_rt_fix_tos(). Also, the long term idea is to remove RTO_ONLINK
entirely by properly initialising ->flowi4_scope, instead of
overloading ->flowi4_tos with a special flag. Eventually, this will
allow to convert ->flowi4_tos to dscp_t.

Signed-off-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'zynqmp-phy-config-optional'
David S. Miller [Fri, 22 Apr 2022 11:52:27 +0000 (12:52 +0100)]
Merge branch 'zynqmp-phy-config-optional'

Radhey Shyam Pandey says:

====================
net: macb: Make ZynqMP SGMII phy configuration optional

This patchset drop phy-names property from MACB node and also make
SGMII Phy configuration optional. The motivation for this change
is to support traditional usescase in which first stage bootloader
does PS-GT configuration, and should still be supported in macb
driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: macb: In ZynqMP initialization make SGMII phy configuration optional
Radhey Shyam Pandey [Wed, 20 Apr 2022 11:03:10 +0000 (16:33 +0530)]
net: macb: In ZynqMP initialization make SGMII phy configuration optional

In the macb binding documentation "phys" is an optional property. Make
implementation in line with it. This change allows the traditional flow
in which first stage bootloader does PS-GT configuration to work along
with newer use cases in which PS-GT configuration is managed by the
phy-zynqmp driver.

It fixes below macb probe failure when macb DT node doesn't have SGMII
phys handle.
"macb ff0b0000.ethernet: error -ENODEV: failed to get PS-GTR PHY"

Signed-off-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Reviewed-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodt-bindings: net: cdns,macb: Drop phy-names property for ZynqMP SGMII PHY
Radhey Shyam Pandey [Wed, 20 Apr 2022 11:03:09 +0000 (16:33 +0530)]
dt-bindings: net: cdns,macb: Drop phy-names property for ZynqMP SGMII PHY

In zynqmp SGMII initialization, there is a single PHY so remove phy-names
property as there is no real need of having it.

Signed-off-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'ipv6-only-sock'
David S. Miller [Fri, 22 Apr 2022 11:47:51 +0000 (12:47 +0100)]
Merge branch 'ipv6-only-sock'

Kuniyuki Iwashima says:

====================
ipv6: Use ipv6_only_sock helper function.

The first patch removes __ipv6_only_sock(), and the second replaces
ipv6only tests with ipv6_only_sock().
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoipv6: Use ipv6_only_sock() helper in condition.
Kuniyuki Iwashima [Wed, 20 Apr 2022 01:58:51 +0000 (10:58 +0900)]
ipv6: Use ipv6_only_sock() helper in condition.

This patch replaces some sk_ipv6only tests with ipv6_only_sock().

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.co.jp>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoipv6: Remove __ipv6_only_sock().
Kuniyuki Iwashima [Wed, 20 Apr 2022 01:58:50 +0000 (10:58 +0900)]
ipv6: Remove __ipv6_only_sock().

Since commit 9fe516ba3fb2 ("inet: move ipv6only in sock_common"),
ipv6_only_sock() and __ipv6_only_sock() are the same macro.  Let's
remove the one.

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.co.jp>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoUSB2NET : SR9800 : change SR9800_BULKIN_SIZE from global to static
Tom Rix [Tue, 19 Apr 2022 14:06:25 +0000 (10:06 -0400)]
USB2NET : SR9800 : change SR9800_BULKIN_SIZE from global to static

Smatch reports this issue
sr9800.h:166:53: warning: symbol 'SR9800_BULKIN_SIZE' was not declared. Should it be static?

Global variables should not be defined in header files.
This only works because sr9800.h in only included by sr9800.c
Change the storage-class specifier to static.
And since it does not change add type qualifier const.

Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoRevert "rtnetlink: return EINVAL when request cannot succeed"
Florent Fourcot [Tue, 19 Apr 2022 12:51:51 +0000 (14:51 +0200)]
Revert "rtnetlink: return EINVAL when request cannot succeed"

This reverts commit b6177d3240a4

ip-link command is testing kernel capability by sending a RTM_NEWLINK
request, without any argument. It accepts everything in reply, except
EOPNOTSUPP and EINVAL (functions iplink_have_newlink / accept_msg)

So we must keep compatiblity here, invalid empty message should not
return EINVAL

Signed-off-by: Florent Fourcot <florent.fourcot@wifirst.fr>
Tested-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonfp: support 802.1ad VLAN assingment to VF
Baowen Zheng [Tue, 19 Apr 2022 12:44:43 +0000 (14:44 +0200)]
nfp: support 802.1ad VLAN assingment to VF

The NFP driver already supports assignment of 802.1Q VLANs to VFs

e.g.
 # ip link set $DEV vf $VF_NUM vlan $VLAN_ID [proto 802.1Q]

This patch enhances the NFP driver to also allow assingment of
802.1ad VLANs to VFs.

e.g.
 # ip link set $DEV vf $VF_NUM vlan $VLAN_ID proto 802.1ad

Signed-off-by: Bin Chen <bin.chen@corigine.com>
Signed-off-by: Baowen Zheng <baowen.zheng@corigine.com>
Signed-off-by: Yinjun Zhang <yunjin.zhang@corigine.com>
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: ethernet: ti: am65-cpsw-ethtool: use pm_runtime_resume_and_get
Minghao Chi [Tue, 19 Apr 2022 11:03:52 +0000 (11:03 +0000)]
net: ethernet: ti: am65-cpsw-ethtool: use pm_runtime_resume_and_get

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

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/ipv6: Enforce limits for accept_unsolicited_na sysctl
Arun Ajith S [Tue, 19 Apr 2022 10:59:10 +0000 (10:59 +0000)]
net/ipv6: Enforce limits for accept_unsolicited_na sysctl

Fix mistake in the original patch where limits were specified but the
handler didn't take care of the limits.

Signed-off-by: Arun Ajith S <aajith@arista.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomyri10ge: remove redundant assignment to variable status
Colin Ian King [Mon, 18 Apr 2022 14:37:59 +0000 (15:37 +0100)]
myri10ge: remove redundant assignment to variable status

Variable status is being assigned a value that is never read, it is
being re-assigned again later on. The assignment is redundant and can
be removed.

Cleans up clang scan build warning:
drivers/net/ethernet/myricom/myri10ge/myri10ge.c:582:7: warning: Although
the value stored to 'status' is used in the enclosing expression, the
value is never actually read from 'status' [deadcode.DeadStores]

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: cdc-ncm: Move spin_lock_bh() to spin_lock()
Yunbo Yu [Mon, 18 Apr 2022 14:18:12 +0000 (22:18 +0800)]
net: cdc-ncm: Move spin_lock_bh() to spin_lock()

It is unnecessary to call spin_lock_bh() for you are already in a tasklet.

Signed-off-by: Yunbo Yu <yuyunbo519@gmail.com>
Acked-by: Oliver Neukum <oneukum@suse.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Paolo Abeni [Fri, 22 Apr 2022 07:56:00 +0000 (09:56 +0200)]
Merge git://git./linux/kernel/git/netdev/net

drivers/net/ethernet/microchip/lan966x/lan966x_main.c
  d08ed852560e ("net: lan966x: Make sure to release ptp interrupt")
  c8349639324a ("net: lan966x: Add FDMA functionality")

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoMerge tag 'net-5.18-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 21 Apr 2022 19:29:08 +0000 (12:29 -0700)]
Merge tag 'net-5.18-rc4' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Paolo Abeni:
 "Including fixes from xfrm and can.

  Current release - regressions:

   - rxrpc: restore removed timer deletion

  Current release - new code bugs:

   - gre: fix device lookup for l3mdev use-case

   - xfrm: fix egress device lookup for l3mdev use-case

  Previous releases - regressions:

   - sched: cls_u32: fix netns refcount changes in u32_change()

   - smc: fix sock leak when release after smc_shutdown()

   - xfrm: limit skb_page_frag_refill use to a single page

   - eth: atlantic: invert deep par in pm functions, preventing null
     derefs

   - eth: stmmac: use readl_poll_timeout_atomic() in atomic state

  Previous releases - always broken:

   - gre: fix skb_under_panic on xmit

   - openvswitch: fix OOB access in reserve_sfa_size()

   - dsa: hellcreek: calculate checksums in tagger

   - eth: ice: fix crash in switchdev mode

   - eth: igc:
      - fix infinite loop in release_swfw_sync
      - fix scheduling while atomic"

* tag 'net-5.18-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (37 commits)
  drivers: net: hippi: Fix deadlock in rr_close()
  selftests: mlxsw: vxlan_flooding_ipv6: Prevent flooding of unwanted packets
  selftests: mlxsw: vxlan_flooding: Prevent flooding of unwanted packets
  nfc: MAINTAINERS: add Bug entry
  net: stmmac: Use readl_poll_timeout_atomic() in atomic state
  doc/ip-sysctl: add bc_forwarding
  netlink: reset network and mac headers in netlink_dump()
  net: mscc: ocelot: fix broken IP multicast flooding
  net: dsa: hellcreek: Calculate checksums in tagger
  net: atlantic: invert deep par in pm functions, preventing null derefs
  can: isotp: stop timeout monitoring when no first frame was sent
  bonding: do not discard lowest hash bit for non layer3+4 hashing
  net: lan966x: Make sure to release ptp interrupt
  ipv6: make ip6_rt_gc_expire an atomic_t
  net: Handle l3mdev in ip_tunnel_init_flow
  l3mdev: l3mdev_master_upper_ifindex_by_index_rcu should be using netdev_master_upper_dev_get_rcu
  net/sched: cls_u32: fix possible leak in u32_init_knode()
  net/sched: cls_u32: fix netns refcount changes in u32_change()
  powerpc: Update MAINTAINERS for ibmvnic and VAS
  net: restore alpha order to Ethernet devices in config
  ...

2 years agonet: eql: Use kzalloc instead of kmalloc/memset
Haowen Bai [Mon, 18 Apr 2022 10:22:13 +0000 (18:22 +0800)]
net: eql: Use kzalloc instead of kmalloc/memset

Use kzalloc rather than duplicating its implementation, which
makes code simple and easy to understand.

Signed-off-by: Haowen Bai <baihaowen@meizu.com>
Link: https://lore.kernel.org/r/1650277333-31090-1-git-send-email-baihaowen@meizu.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agodrivers: net: davinci_mdio: using pm_runtime_resume_and_get instead of pm_runtime_get...
Minghao Chi [Mon, 18 Apr 2022 06:29:21 +0000 (06:29 +0000)]
drivers: net: davinci_mdio: using pm_runtime_resume_and_get instead of pm_runtime_get_sync

Using pm_runtime_resume_and_get is more appropriate
for simplifing code

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: Minghao Chi <chi.minghao@zte.com.cn>
Link: https://lore.kernel.org/r/20220418062921.2557884-1-chi.minghao@zte.com.cn
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agodrivers: net: hippi: Fix deadlock in rr_close()
Duoming Zhou [Sun, 17 Apr 2022 12:55:19 +0000 (20:55 +0800)]
drivers: net: hippi: Fix deadlock in rr_close()

There is a deadlock in rr_close(), which is shown below:

   (Thread 1)                |      (Thread 2)
                             | rr_open()
rr_close()                   |  add_timer()
 spin_lock_irqsave() //(1)   |  (wait a time)
 ...                         | rr_timer()
 del_timer_sync()            |  spin_lock_irqsave() //(2)
 (wait timer to stop)        |  ...

We hold rrpriv->lock in position (1) of thread 1 and
use del_timer_sync() to wait timer to stop, but timer handler
also need rrpriv->lock in position (2) of thread 2.
As a result, rr_close() will block forever.

This patch extracts del_timer_sync() from the protection of
spin_lock_irqsave(), which could let timer handler to obtain
the needed lock.

Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Link: https://lore.kernel.org/r/20220417125519.82618-1-duoming@zju.edu.cn
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoMerge tag 'xtensa-20220416' of https://github.com/jcmvbkbc/linux-xtensa
Linus Torvalds [Wed, 20 Apr 2022 19:43:27 +0000 (12:43 -0700)]
Merge tag 'xtensa-20220416' of https://github.com/jcmvbkbc/linux-xtensa

Pull xtensa fixes from Max Filippov:

 - fix patching CPU selection in patch_text

 - fix potential deadlock in ISS platform serial driver

 - fix potential register clobbering in coprocessor exception handler

* tag 'xtensa-20220416' of https://github.com/jcmvbkbc/linux-xtensa:
  xtensa: fix a7 clobbering in coprocessor context load/store
  arch: xtensa: platforms: Fix deadlock in rs_close()
  xtensa: patch_text: Fixup last cpu should be master

2 years agoMerge tag 'erofs-for-5.18-rc4-fixes' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 20 Apr 2022 19:35:20 +0000 (12:35 -0700)]
Merge tag 'erofs-for-5.18-rc4-fixes' of git://git./linux/kernel/git/xiang/erofs

Pull erofs fixes from Gao Xiang:
 "One patch to fix a use-after-free race related to the on-stack
  z_erofs_decompressqueue, which happens very rarely but needs to be
  fixed properly soon.

  The other patch fixes some sysfs Sphinx warnings"

* tag 'erofs-for-5.18-rc4-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs:
  Documentation/ABI: sysfs-fs-erofs: Fix Sphinx errors
  erofs: fix use-after-free of on-stack io[]

2 years agoRevert "fs/pipe: use kvcalloc to allocate a pipe_buffer array"
Linus Torvalds [Wed, 20 Apr 2022 19:07:53 +0000 (12:07 -0700)]
Revert "fs/pipe: use kvcalloc to allocate a pipe_buffer array"

This reverts commit 5a519c8fe4d620912385f94372fc8472fa98c662.

It turns out that making the pipe almost arbitrarily large has some
rather unexpected downsides.  The kernel test robot reports a kernel
warning that is due to pipe->max_usage now growing to the point where
the iter_file_splice_write() buffer allocation can no longer be
satisfied as a slab allocation, and the

        int nbufs = pipe->max_usage;
        struct bio_vec *array = kcalloc(nbufs, sizeof(struct bio_vec),
                                        GFP_KERNEL);

code sequence there will now always fail as a result.

That code could be modified to use kvcalloc() too, but I feel very
uncomfortable making those kinds of changes for a very niche use case
that really should have other options than make these kinds of
fundamental changes to pipe behavior.

Maybe the CRIU process dumping should be multi-threaded, and use
multiple pipes and multiple cores, rather than try to use one larger
pipe to minimize splice() calls.

Reported-by: kernel test robot <oliver.sang@intel.com>
Link: https://lore.kernel.org/all/20220420073717.GD16310@xsang-OptiPlex-9020/
Cc: Andrei Vagin <avagin@gmail.com>
Cc: Dmitry Safonov <0x7f454c46@gmail.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agox86: __memcpy_flushcache: fix wrong alignment if size > 2^32
Mikulas Patocka [Tue, 19 Apr 2022 13:56:23 +0000 (09:56 -0400)]
x86: __memcpy_flushcache: fix wrong alignment if size > 2^32

The first "if" condition in __memcpy_flushcache is supposed to align the
"dest" variable to 8 bytes and copy data up to this alignment.  However,
this condition may misbehave if "size" is greater than 4GiB.

The statement min_t(unsigned, size, ALIGN(dest, 8) - dest); casts both
arguments to unsigned int and selects the smaller one.  However, the
cast truncates high bits in "size" and it results in misbehavior.

For example:

suppose that size == 0x100000001, dest == 0x200000002
min_t(unsigned, size, ALIGN(dest, 8) - dest) == min_t(0x1, 0xe) == 0x1;
...
dest += 0x1;

so we copy just one byte "and" dest remains unaligned.

This patch fixes the bug by replacing unsigned with size_t.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoselftests: mlxsw: vxlan_flooding_ipv6: Prevent flooding of unwanted packets
Ido Schimmel [Tue, 19 Apr 2022 13:51:55 +0000 (16:51 +0300)]
selftests: mlxsw: vxlan_flooding_ipv6: Prevent flooding of unwanted packets

The test verifies that packets are correctly flooded by the bridge and
the VXLAN device by matching on the encapsulated packets at the other
end. However, if packets other than those generated by the test also
ingress the bridge (e.g., MLD packets), they will be flooded as well and
interfere with the expected count.

Make the test more robust by making sure that only the packets generated
by the test can ingress the bridge. Drop all the rest using tc filters
on the egress of 'br0' and 'h1'.

In the software data path, the problem can be solved by matching on the
inner destination MAC or dropping unwanted packets at the egress of the
VXLAN device, but this is not currently supported by mlxsw.

Fixes: d01724dd2a66 ("selftests: mlxsw: spectrum-2: Add a test for VxLAN flooding with IPv6")
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: mlxsw: vxlan_flooding: Prevent flooding of unwanted packets
Ido Schimmel [Tue, 19 Apr 2022 13:51:54 +0000 (16:51 +0300)]
selftests: mlxsw: vxlan_flooding: Prevent flooding of unwanted packets

The test verifies that packets are correctly flooded by the bridge and
the VXLAN device by matching on the encapsulated packets at the other
end. However, if packets other than those generated by the test also
ingress the bridge (e.g., MLD packets), they will be flooded as well and
interfere with the expected count.

Make the test more robust by making sure that only the packets generated
by the test can ingress the bridge. Drop all the rest using tc filters
on the egress of 'br0' and 'h1'.

In the software data path, the problem can be solved by matching on the
inner destination MAC or dropping unwanted packets at the egress of the
VXLAN device, but this is not currently supported by mlxsw.

Fixes: 94d302deae25 ("selftests: mlxsw: Add a test for VxLAN flooding")
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Amit Cohen <amcohen@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'mlxsw-line-card-status-tracking'
David S. Miller [Wed, 20 Apr 2022 14:03:22 +0000 (15:03 +0100)]
Merge branch 'mlxsw-line-card-status-tracking'

Ido Schimmel says:

====================
mlxsw: Line cards status tracking

When a line card is provisioned, netdevs corresponding to the ports
found on the line card are registered. User space can then perform
various logical configurations (e.g., splitting, setting MTU) on these
netdevs.

However, since the line card is not present / powered on (i.e., it is
not in 'active' state), user space cannot access the various components
found on the line card. For example, user space cannot read the
temperature of gearboxes or transceiver modules found on the line card
via hwmon / thermal. Similarly, it cannot dump the EEPROM contents of
these transceiver modules. The above is only possible when the line card
becomes active.

This patchset solves the problem by tracking the status of each line
card and invoking callbacks from interested parties when a line card
becomes active / inactive.

Patchset overview:

Patch #1 adds the infrastructure in the line cards core that allows
users to registers a set of callbacks that are invoked when a line card
becomes active / inactive. To avoid races, if a line card is already
active during registration, the got_active() callback is invoked.

Patches #2-#3 are preparations.

Patch #4 changes the port module core to register a set of callbacks
with the line cards core. See detailed description with examples in the
commit message.

Patches #5-#6 do the same with regards to thermal / hwmon support, so
that user space will be able to monitor the temperature of various
components on the line card when it becomes active.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_hwmon: Add interfaces for line card initialization and de-initialization
Vadim Pasternak [Tue, 19 Apr 2022 14:54:31 +0000 (17:54 +0300)]
mlxsw: core_hwmon: Add interfaces for line card initialization and de-initialization

Add callback functions for line card 'hwmon' initialization and
de-initialization. Each line card is associated with the relevant
'hwmon' device, which may contain thermal attributes for the cages
and gearboxes found on this line card.

The line card 'hwmon' initialization / de-initialization APIs are to be
called when line card is set to active / inactive state by
got_active() / got_inactive() callbacks from line card state machine.

For example cage temperature for module #9 located at line card #7 will
be exposed by utility 'sensors' like:
linecard#07
front panel 009: +32.0C  (crit = +70.0C, emerg = +80.0C)
And temperature for gearbox #3 located at line card #5 will be exposed
like:
linecard#05
gearbox 003: +41.0C  (highest = +41.0C)

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_thermal: Add interfaces for line card initialization and de-initialization
Vadim Pasternak [Tue, 19 Apr 2022 14:54:30 +0000 (17:54 +0300)]
mlxsw: core_thermal: Add interfaces for line card initialization and de-initialization

Add callback functions for line card thermal area initialization and
de-initialization. Each line card is associated with the relevant
thermal area, which may contain thermal zones for cages and gearboxes
found on this line card.

The line card thermal initialization / de-initialization APIs are to be
called when line card is set to active / inactive state by
got_active() / got_inactive() callbacks from line card state machine.

For example thermal zone for module #9 located at line card #7 will
have type:
mlxsw-lc7-module9.
And thermal zone for gearbox #2 located at line card #5 will have type:
mlxsw-lc5-gearbox2.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_env: Add interfaces for line card initialization and de-initialization
Vadim Pasternak [Tue, 19 Apr 2022 14:54:29 +0000 (17:54 +0300)]
mlxsw: core_env: Add interfaces for line card initialization and de-initialization

Netdevs for ports found on line cards are registered upon provisioning.
However, user space is not allowed to access the transceiver modules
found on a line card until the line card becomes active.

Therefore, register event operations with the line card core to get
notifications whenever a line card becomes active or inactive.

When user space tries to dump the EEPROM of a transceiver module or reset
it and the corresponding line card is inactive, emit an error
message:
ethtool -m enp1s0nl7p9
netlink error: mlxsw_core: Cannot read EEPROM of module on an inactive line card
netlink error: Input/output error

When user space tries to set the power mode policy of such a transceiver,
cache the configuration and apply it when the line card becomes active. This
is consistent with other port configuration (e.g., MTU setting) that user space
is able to perform while the line card is provisioned, but inactive.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_env: Split module power mode setting to a separate function
Vadim Pasternak [Tue, 19 Apr 2022 14:54:28 +0000 (17:54 +0300)]
mlxsw: core_env: Split module power mode setting to a separate function

Move the code that applies the module power mode to the device to a
separate function. This function will be invoked by the next patch to
set the power mode on transceiver modules found on a line card when the
line card becomes active.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core: Add bus argument to environment init API
Vadim Pasternak [Tue, 19 Apr 2022 14:54:27 +0000 (17:54 +0300)]
mlxsw: core: Add bus argument to environment init API

Pass bus argument to mlxsw_env_init(). The purpose is to get access to
device handle, which is to be provided to error message in case of line
card activation failure.

Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: core_linecards: Introduce ops for linecards status change tracking
Jiri Pirko [Tue, 19 Apr 2022 14:54:26 +0000 (17:54 +0300)]
mlxsw: core_linecards: Introduce ops for linecards status change tracking

Introduce an infrastructure allowing users to register a set
of operations which are to be called whenever a line card gets
active/inactive.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Vadim Pasternak <vadimp@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonfc: MAINTAINERS: add Bug entry
Krzysztof Kozlowski [Wed, 20 Apr 2022 10:46:01 +0000 (12:46 +0200)]
nfc: MAINTAINERS: add Bug entry

Add a Bug section, indicating preferred mailing method for bug reports,
to NFC Subsystem entry.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge tag 'linux-can-next-for-5.19-20220419' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Wed, 20 Apr 2022 10:14:36 +0000 (11:14 +0100)]
Merge tag 'linux-can-next-for-5.19-20220419' of git://git./linux/kernel/git/mkl/linux-can-next

Marc Kleine-Budde says:

====================
pull-request: can-next 2022-04-19

this is a pull request of 17 patches for net-next/master.

The first 2 patches are by me and target the CAN driver
infrastructure. One patch renames a function in the rx_offload helper
the other one updates the CAN bitrate calculation to prefer small bit
rate pre-scalers over larger ones, which is encouraged by the CAN in
Automation.

Kris Bahnsen contributes a patch to fix the links to Technologic
Systems web resources in the sja1000 driver.

Christophe Leroy's patch prepares the mpc5xxx_can driver for upcoming
powerpc header cleanup.

Minghao Chi's patch converts the flexcan driver to use
pm_runtime_resume_and_get().

The next 2 patches target the Xilinx CAN driver. Lukas Bulwahn's patch
fixes an entry in the MAINTAINERS file. A patch by me marks the bit
timing constants as const.

Wolfram Sang's patch documents r8a77961 support on the
renesas,rcar-canfd bindings document.

The next 2 patches are by me and add support for the mcp251863 chip to
the mcp251xfd driver.

The last 7 patches are by Pavel Pisa, Martin Jerabek et al. and add
the ctucanfd driver for the CTU CAN FD IP Core.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: stmmac: Use readl_poll_timeout_atomic() in atomic state
Kevin Hao [Tue, 19 Apr 2022 08:42:26 +0000 (16:42 +0800)]
net: stmmac: Use readl_poll_timeout_atomic() in atomic state

The init_systime() may be invoked in atomic state. We have observed the
following call trace when running "phc_ctl /dev/ptp0 set" on a Intel
Agilex board.
  BUG: sleeping function called from invalid context at drivers/net/ethernet/stmicro/stmmac/stmmac_hwtstamp.c:74
  in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 381, name: phc_ctl
  preempt_count: 1, expected: 0
  RCU nest depth: 0, expected: 0
  Preemption disabled at:
  [<ffff80000892ef78>] stmmac_set_time+0x34/0x8c
  CPU: 2 PID: 381 Comm: phc_ctl Not tainted 5.18.0-rc2-next-20220414-yocto-standard+ #567
  Hardware name: SoCFPGA Agilex SoCDK (DT)
  Call trace:
   dump_backtrace.part.0+0xc4/0xd0
   show_stack+0x24/0x40
   dump_stack_lvl+0x7c/0xa0
   dump_stack+0x18/0x34
   __might_resched+0x154/0x1c0
   __might_sleep+0x58/0x90
   init_systime+0x78/0x120
   stmmac_set_time+0x64/0x8c
   ptp_clock_settime+0x60/0x9c
   pc_clock_settime+0x6c/0xc0
   __arm64_sys_clock_settime+0x88/0xf0
   invoke_syscall+0x5c/0x130
   el0_svc_common.constprop.0+0x4c/0x100
   do_el0_svc+0x7c/0xa0
   el0_svc+0x58/0xcc
   el0t_64_sync_handler+0xa4/0x130
   el0t_64_sync+0x18c/0x190

So we should use readl_poll_timeout_atomic() here instead of
readl_poll_timeout().

Also adjust the delay time to 10us to fix a "__bad_udelay" build error
reported by "kernel test robot <lkp@intel.com>". I have tested this on
Intel Agilex and NXP S32G boards, there is no delay needed at all.
So the 10us delay should be long enough for most cases.

Fixes: ff8ed737860e ("net: stmmac: use readl_poll_timeout() function in init_systime()")
Signed-off-by: Kevin Hao <haokexin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'net-sched-flower-num-vlan-tags'
David S. Miller [Wed, 20 Apr 2022 10:09:13 +0000 (11:09 +0100)]
Merge branch 'net-sched-flower-num-vlan-tags'

Boris Sukholitko says:

====================
net/sched: flower: match on the number of vlan tags

Our customers in the fiber telecom world have network configurations
where they would like to control their traffic according to the number
of tags appearing in the packet.

For example, TR247 GPON conformance test suite specification mostly
talks about untagged, single, double tagged packets and gives lax
guidelines on the vlan protocol vs. number of vlan tags.

This is different from the common IT networks where 802.1Q and 802.1ad
protocols are usually describe single and double tagged packet. GPON
configurations that we work with have arbitrary mix the above protocols
and number of vlan tags in the packet.

The following patch series implement number of vlans flower filter. They
add num_of_vlans flower filter as an alternative to vlan ethtype protocol
matching. The end result is that the following command becomes possible:

tc filter add dev eth1 ingress flower \
  num_of_vlans 1 vlan_prio 5 action drop

Also, from our logs, we have redirect rules such that:

tc filter add dev $GPON ingress flower num_of_vlans $N \
     action mirred egress redirect dev $DEV

where N can range from 0 to 3 and $DEV is the function of $N.

Also there are rules setting skb mark based on the number of vlans:

tc filter add dev $GPON ingress flower num_of_vlans $N vlan_prio \
    $P action skbedit mark $M

More about the patch series:
  - patches 1-2 remove duplicate code by introducing is_key_vlan
    helper.
  - patch 3, 4 implement num_of_vlans in the dissector and in the
    flower.
  - patch 5 uses the num_of_vlans filter to allow further matching on
    vlan attributes.

Complementary iproute2 patches are being sent separately.

Thanks,
Boris.

- v4: rebased to the latest net-next
- v3:
    - more example commands in patch 3 description (request by Jamal)
    - patch 5 description made clearer (thanks to Jiri)
- v2:
    - add suitable subject prefixes
    - more evolved patch 5 description
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/sched: flower: Consider the number of tags for vlan filters
Boris Sukholitko [Tue, 19 Apr 2022 08:14:34 +0000 (11:14 +0300)]
net/sched: flower: Consider the number of tags for vlan filters

Before this patch the existence of vlan filters was conditional on the vlan
protocol being matched in the tc rule. For example, the following rule:

tc filter add dev eth1 ingress flower vlan_prio 5

was illegal because vlan protocol (e.g. 802.1q) does not appear in the rule.

Remove the above restriction by looking at the num_of_vlans filter to
allow further matching on vlan attributes. The following rule becomes
legal as a result of this commit:

tc filter add dev eth1 ingress flower num_of_vlans 1 vlan_prio 5

because having num_of_vlans==1 implies that the packet is single tagged.

Change is_vlan_key helper to look at the number of vlans in addition to
the vlan ethertype. The outcome of this change is that outer (e.g. vlan_prio)
and inner (e.g. cvlan_prio) tag vlan filters require the number of vlan
tags to be greater then 0 and 1 accordingly.

As a result of is_vlan_key change, the ethertype may be set to 0 when
matching on the number of vlans. Update fl_set_key_vlan to avoid setting
key, mask vlan_tpid for the 0 ethertype.

Signed-off-by: Boris Sukholitko <boris.sukholitko@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/sched: flower: Add number of vlan tags filter
Boris Sukholitko [Tue, 19 Apr 2022 08:14:33 +0000 (11:14 +0300)]
net/sched: flower: Add number of vlan tags filter

These are bookkeeping parts of the new num_of_vlans filter.
Defines, dump, load and set are being done here.

Signed-off-by: Boris Sukholitko <boris.sukholitko@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoflow_dissector: Add number of vlan tags dissector
Boris Sukholitko [Tue, 19 Apr 2022 08:14:32 +0000 (11:14 +0300)]
flow_dissector: Add number of vlan tags dissector

Our customers in the fiber telecom world have network configurations
where they would like to control their traffic according to the number
of tags appearing in the packet.

For example, TR247 GPON conformance test suite specification mostly
talks about untagged, single, double tagged packets and gives lax
guidelines on the vlan protocol vs. number of vlan tags.

This is different from the common IT networks where 802.1Q and 802.1ad
protocols are usually describe single and double tagged packet. GPON
configurations that we work with have arbitrary mix the above protocols
and number of vlan tags in the packet.

The goal is to make the following TC commands possible:

tc filter add dev eth1 ingress flower \
  num_of_vlans 1 vlan_prio 5 action drop

From our logs, we have redirect rules such that:

tc filter add dev $GPON ingress flower num_of_vlans $N \
     action mirred egress redirect dev $DEV

where N can range from 0 to 3 and $DEV is the function of $N.

Also there are rules setting skb mark based on the number of vlans:

tc filter add dev $GPON ingress flower num_of_vlans $N vlan_prio \
    $P action skbedit mark $M

This new dissector allows extracting the number of vlan tags existing in
the packet.

Signed-off-by: Boris Sukholitko <boris.sukholitko@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/sched: flower: Reduce identation after is_key_vlan refactoring
Boris Sukholitko [Tue, 19 Apr 2022 08:14:31 +0000 (11:14 +0300)]
net/sched: flower: Reduce identation after is_key_vlan refactoring

Whitespace only.

Signed-off-by: Boris Sukholitko <boris.sukholitko@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/sched: flower: Helper function for vlan ethtype checks
Boris Sukholitko [Tue, 19 Apr 2022 08:14:30 +0000 (11:14 +0300)]
net/sched: flower: Helper function for vlan ethtype checks

There are somewhat repetitive ethertype checks in fl_set_key. Refactor
them into is_vlan_key helper function.

To make the changes clearer, avoid touching identation levels. This is
the job for the next patch in the series.

Signed-off-by: Boris Sukholitko <boris.sukholitko@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoar5523: Use kzalloc instead of kmalloc/memset
Haowen Bai [Tue, 19 Apr 2022 01:37:31 +0000 (09:37 +0800)]
ar5523: Use kzalloc instead of kmalloc/memset

Use kzalloc rather than duplicating its implementation, which
makes code simple and easy to understand.

Signed-off-by: Haowen Bai <baihaowen@meizu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: realtek: remove realtek,rtl8367s string
Luiz Angelo Daros de Luca [Mon, 18 Apr 2022 23:35:58 +0000 (20:35 -0300)]
net: dsa: realtek: remove realtek,rtl8367s string

There is no need to add new compatible strings for each new supported
chip version. The compatible string is used only to select the subdriver
(rtl8365mb.c or rtl8366rb.c). Once in the subdriver, it will detect the
chip model by itself, ignoring which compatible string was used.

Link: https://lore.kernel.org/netdev/20220414014055.m4wbmr7tdz6hsa3m@bang-olufsen.dk/
Signed-off-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Arınç ÜNAL <arinc.unal@arinc9.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodt-bindings: net: dsa: realtek: cleanup compatible strings
Luiz Angelo Daros de Luca [Mon, 18 Apr 2022 23:35:57 +0000 (20:35 -0300)]
dt-bindings: net: dsa: realtek: cleanup compatible strings

Compatible strings are used to help the driver find the chip ID/version
register for each chip family. After that, the driver can setup the
switch accordingly. Keep only the first supported model for each family
as a compatible string and reference other chip models in the
description.

The removed compatible strings have never been used in a released kernel.

CC: devicetree@vger.kernel.org
Link: https://lore.kernel.org/netdev/20220414014055.m4wbmr7tdz6hsa3m@bang-olufsen.dk/
Signed-off-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Arınç ÜNAL <arinc.unal@arinc9.com>
Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'hns3-next'
David S. Miller [Wed, 20 Apr 2022 09:45:51 +0000 (10:45 +0100)]
Merge branch 'hns3-next'

Guangbin Huang says:

====================
net: hns3: updates for -next

This series includes some updates for the HNS3 ethernet driver.

Change logs:
V1 -> V2:
 - Fix failed to apply to net-next problem.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: hns3: remove unnecessary line wrap for hns3_set_tunable
Hao Chen [Tue, 19 Apr 2022 03:27:09 +0000 (11:27 +0800)]
net: hns3: remove unnecessary line wrap for hns3_set_tunable

Remove unnecessary line wrap for hns3_set_tunable to improve
function readability.

Signed-off-by: Hao Chen <chenhao288@hisilicon.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: hns3: replace magic value by HCLGE_RING_REG_OFFSET
Peng Li [Tue, 19 Apr 2022 03:27:08 +0000 (11:27 +0800)]
net: hns3: replace magic value by HCLGE_RING_REG_OFFSET

Magic values are not recommended.

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>
2 years agonet: hns3: fix the wrong words in comments
Peng Li [Tue, 19 Apr 2022 03:27:07 +0000 (11:27 +0800)]
net: hns3: fix the wrong words in comments

This patch fixes wrong words in comments.

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>
2 years agonet: hns3: update the comment of function hclgevf_get_mbx_resp
Peng Li [Tue, 19 Apr 2022 03:27:06 +0000 (11:27 +0800)]
net: hns3: update the comment of function hclgevf_get_mbx_resp

The param of function hclgevf_get_mbx_resp has been changed but the
comments not upodated. This patch updates it.

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>
2 years agonet: hns3: add log for setting tx spare buf size
Hao Chen [Tue, 19 Apr 2022 03:27:05 +0000 (11:27 +0800)]
net: hns3: add log for setting tx spare buf size

For the active tx spare buffer size maybe changed according
to the page size, so add log to notice it.

Signed-off-by: Hao Chen <chenhao288@hisilicon.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: hns3: add failure logs in hclge_set_vport_mtu
Jie Wang [Tue, 19 Apr 2022 03:27:04 +0000 (11:27 +0800)]
net: hns3: add failure logs in hclge_set_vport_mtu

Currently, There is a low probability that pf mtu configuration fails, but
the information in logs is insufficient for problem locating when the VF
mtu value is illegally modified.

So record the vf index and vf mtu value at the failure scenario.

Signed-off-by: Jie Wang <wangjie125@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: hns3: refine the definition for struct hclge_pf_to_vf_msg
Jian Shen [Tue, 19 Apr 2022 03:27:03 +0000 (11:27 +0800)]
net: hns3: refine the definition for struct hclge_pf_to_vf_msg

The struct hclge_pf_to_vf_msg is used for mailbox message from
PF to VF, including both response and request. But its definition
can only indicate respone, which makes the message data copy in
function hclge_send_mbx_msg() unreadable. So refine it by edding
a general message definition into it.

Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: hns3: refactor hns3_set_ringparam()
Hao Chen [Tue, 19 Apr 2022 03:27:02 +0000 (11:27 +0800)]
net: hns3: refactor hns3_set_ringparam()

Use struct hns3_ring_param to replace variable new/old_xxx and
add hns3_is_ringparam_changed() to judge them if is changed to
improve code readability.

Signed-off-by: Hao Chen <chenhao288@hisilicon.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: hns3: add ethtool parameter check for CQE/EQE mode
Yufeng Mo [Tue, 19 Apr 2022 03:27:01 +0000 (11:27 +0800)]
net: hns3: add ethtool parameter check for CQE/EQE mode

For DEVICE_VERSION_V2, the hardware does not support the CQE mode.
So add capability bit for coalesce CQE mode and add parameter check
for it in ethtool.

Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'atlantic-xdp-multi-buffer'
David S. Miller [Wed, 20 Apr 2022 09:42:57 +0000 (10:42 +0100)]
Merge branch 'atlantic-xdp-multi-buffer'

[PATCH net-next v5 0/3] net: atlantic: Add XDP support
@ 2022-04-17 10:12 Taehee Yoo
  2022-04-17 10:12 ` [PATCH net-next v5 1/3] net: atlantic: Implement xdp control plane Taehee Yoo
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: Taehee Yoo @ 2022-04-17 10:12 UTC (permalink / raw)
  To: davem, kuba, pabeni, netdev, irusskikh, ast, daniel, hawk,
john.fastabend, andrii, kafai, songliubraving, yhs, kpsingh, bpf
Cc: ap420073
This patchset is to make atlantic to support multi-buffer XDP.

The first patch implement control plane of xdp.
The aq_xdp(), callback of .xdp_bpf is added.

The second patch implements data plane of xdp.
XDP_TX, XDP_DROP, and XDP_PASS is supported.
__aq_ring_xdp_clean() is added to receive and execute xdp program.
aq_nic_xmit_xdpf() is added to send packet by XDP.

The third patch implements callback of .ndo_xdp_xmit.
aq_xdp_xmit() is added to send redirected packets and it internally
calls aq_nic_xmit_xdpf().

Memory model is MEM_TYPE_PAGE_SHARED.

Order-2 page allocation is used when XDP is enabled.

LRO will be disabled if XDP program doesn't supports multi buffer.

AQC chip supports 32 multi-queues and 8 vectors(irq).
There are two options.
1. under 8 cores and maximum 4 tx queues per core.
2. under 4 cores and maximum 8 tx queues per core.

Like other drivers, these tx queues can be used only for XDP_TX,
XDP_REDIRECT queue. If so, no tx_lock is needed.
But this patchset doesn't use this strategy because getting hardware tx
queue index cost is too high.
So, tx_lock is used in the aq_nic_xmit_xdpf().

single-core, single queue, 80% cpu utilization.

  32.30%  [kernel]                  [k] aq_get_rxpages_xdp
  10.44%  [kernel]                  [k] aq_hw_read_reg <---------- here
   9.86%  bpf_prog_xxx_xdp_prog_tx  [k] bpf_prog_xxx_xdp_prog_tx
   5.51%  [kernel]                  [k] aq_ring_rx_clean

single-core, 8 queues, 100% cpu utilization, half PPS.

  52.03%  [kernel]                  [k] aq_hw_read_reg <---------- here
  18.24%  [kernel]                  [k] aq_get_rxpages_xdp
   4.30%  [kernel]                  [k] hw_atl_b0_hw_ring_rx_receive
   4.24%  bpf_prog_xxx_xdp_prog_tx  [k] bpf_prog_xxx_xdp_prog_tx
   2.79%  [kernel]                  [k] aq_ring_rx_clean

Performance result(64 Byte)
1. XDP_TX
  a. xdp_geieric, single core
    - 2.5Mpps, 100% cpu
  b. xdp_driver, single core
    - 4.5Mpps, 80% cpu
  c. xdp_generic, 8 core(hyper thread)
    - 6.3Mpps, 40% cpu
  d. xdp_driver, 8 core(hyper thread)
    - 6.3Mpps, 30% cpu

2. XDP_REDIRECT
  a. xdp_generic, single core
    - 2.3Mpps
  b. xdp_driver, single core
    - 4.5Mpps

v5:
 - Use MEM_TYPE_PAGE_SHARED instead of MEM_TYPE_PAGE_ORDER0
 - Use 2K frame size instead of 3K
 - Use order-2 page allocation instead of order-0
 - Rename aq_get_rxpage() to aq_alloc_rxpages()
 - Add missing PageFree stats for ethtool
 - Remove aq_unset_rxpage_xdp(), introduced by v2 patch due to
   change of memory model
 - Fix wrong last parameter value of xdp_prepare_buff()
 - Add aq_get_rxpages_xdp() to increase page reference count

v4:
 - Fix compile warning

v3:
 - Change wrong PPS performance result 40% -> 80% in single
   core(Intel i3-12100)
 - Separate aq_nic_map_xdp() from aq_nic_map_skb()
 - Drop multi buffer packets if single buffer XDP is attached
 - Disable LRO when single buffer XDP is attached
 - Use xdp_get_{frame/buff}_len()

v2:
 - Do not use inline in C file

Taehee Yoo (3):
  net: atlantic: Implement xdp control plane
  net: atlantic: Implement xdp data plane
  net: atlantic: Implement .ndo_xdp_xmit handler

 .../net/ethernet/aquantia/atlantic/aq_cfg.h   |   1 +
 .../ethernet/aquantia/atlantic/aq_ethtool.c   |   9 +
 .../net/ethernet/aquantia/atlantic/aq_main.c  |  87 ++++
 .../net/ethernet/aquantia/atlantic/aq_main.h  |   2 +
 .../net/ethernet/aquantia/atlantic/aq_nic.c   | 136 ++++++
 .../net/ethernet/aquantia/atlantic/aq_nic.h   |   5 +
 .../net/ethernet/aquantia/atlantic/aq_ring.c  | 409 ++++++++++++++++--
 .../net/ethernet/aquantia/atlantic/aq_ring.h  |  21 +-
 .../net/ethernet/aquantia/atlantic/aq_vec.c   |  23 +-
 .../net/ethernet/aquantia/atlantic/aq_vec.h   |   6 +
 .../aquantia/atlantic/hw_atl/hw_atl_a0.c      |   6 +-
 .../aquantia/atlantic/hw_atl/hw_atl_b0.c      |  10 +-
 12 files changed, 670 insertions(+), 45 deletions(-)

--
2.17.1

^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH net-next v5 1/3] net: atlantic: Implement xdp control plane
  2022-04-17 10:12 [PATCH net-next v5 0/3] net: atlantic: Add XDP support Taehee Yoo
@ 2022-04-17 10:12 ` Taehee Yoo
  2022-04-17 10:12 ` [PATCH net-next v5 2/3] net: atlantic: Implement xdp data plane Taehee Yoo
  2022-04-17 10:12 ` [PATCH net-next v5 3/3] net: atlantic: Implement .ndo_xdp_xmit handler Taehee Yoo
  2 siblings, 0 replies; 4+ messages in thread
From: Taehee Yoo @ 2022-04-17 10:12 UTC (permalink / raw)
  To: davem, kuba, pabeni, netdev, irusskikh, ast, daniel, hawk,
john.fastabend, andrii, kafai, songliubraving, yhs, kpsingh, bpf
Cc: ap420073
aq_xdp() is a xdp setup callback function for Atlantic driver.
When XDP is attached or detached, the device will be restarted because
it uses different headroom, tailroom, and page order value.

If XDP enabled, it switches default page order value from 0 to 2.
Because the default maximum frame size is still 2K and it needs
additional area for headroom and tailroom.
The total size(headroom + frame size + tailroom) is 2624.
So, 1472Bytes will be always wasted for every frame.
But when order-2 is used, these pages can be used 6 times
with flip strategy.
It means only about 106Bytes per frame will be wasted.

Also, It supports xdp fragment feature.
MTU can be 16K if xdp prog supports xdp fragment.
If not, MTU can not exceed 2K - ETH_HLEN - ETH_FCS.

And a static key is added and It will be used to call the xdp_clean
handler in ->poll(). data plane implementation will be contained
the followed patch.

Signed-off-by: Taehee Yoo <ap420073@gmail.com>
---

v5:
 - Use MEM_TYPE_PAGE_SHARED instead of MEM_TYPE_PAGE_ORDER0
 - Use 2K frame size instead of 3K
 - Use order-2 page allocation instead of order-0
 - Rename aq_get_rxpage() to aq_alloc_rxpages()

v4:
 - No changed

v3:
 - Disable LRO when single buffer XDP is attached

v2:
 - No changed

2 years agonet: atlantic: Implement .ndo_xdp_xmit handler
Taehee Yoo [Sun, 17 Apr 2022 10:12:47 +0000 (10:12 +0000)]
net: atlantic: Implement .ndo_xdp_xmit handler

aq_xdp_xmit() is the callback function of .ndo_xdp_xmit.
It internally calls aq_nic_xmit_xdpf() to send packet.

Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>