OSDN Git Service

uclinux-h8/linux.git
4 years agonet/mlx5e: Move devlink port register and unregister calls
Vladyslav Tarasiuk [Fri, 10 Jul 2020 12:25:12 +0000 (15:25 +0300)]
net/mlx5e: Move devlink port register and unregister calls

Register devlink ports upon NIC init. TX and RX health reporters handle
errors which may occur early on at driver initialization. And because
these reporters are to be moved to port context, they require devlink
ports to be already registered.

Signed-off-by: Vladyslav Tarasiuk <vladyslavt@mellanox.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Add devlink health port reporters API
Vladyslav Tarasiuk [Fri, 10 Jul 2020 12:25:11 +0000 (15:25 +0300)]
devlink: Add devlink health port reporters API

In order to use new devlink port health reporters infrastructure, add
corresponding constructor and destructor functions.

Signed-off-by: Vladyslav Tarasiuk <vladyslavt@mellanox.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Implement devlink health reporters on per-port basis
Vladyslav Tarasiuk [Fri, 10 Jul 2020 12:25:10 +0000 (15:25 +0300)]
devlink: Implement devlink health reporters on per-port basis

Add devlink-health reporter support on per-port basis.
The main difference existing devlink-health is that port reporters are
stored in per-devlink_port lists. Upon creation of such health reporter the
reference to a port it belongs to is stored in reporter struct.

Fill the port index attribute in devlink-health response to
allow devlink userspace utility to distinguish between device and port
reporters.

Signed-off-by: Vladyslav Tarasiuk <vladyslavt@mellanox.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Create generic devlink health reporter search function
Vladyslav Tarasiuk [Fri, 10 Jul 2020 12:25:09 +0000 (15:25 +0300)]
devlink: Create generic devlink health reporter search function

Add a generic __devlink_health_reporter_find_by_name() that can be used
with arbitrary devlink health reporter list.

Signed-off-by: Vladyslav Tarasiuk <vladyslavt@mellanox.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Rework devlink health reporter destructor
Vladyslav Tarasiuk [Fri, 10 Jul 2020 12:25:08 +0000 (15:25 +0300)]
devlink: Rework devlink health reporter destructor

Devlink keeps its own reference to every reporter in a list and inits
refcount to 1 upon reporter's creation. Existing destructor waits to
free the memory indefinitely using msleep() until all references except
devlink's own are put.

Rework this mechanism by moving memory free routine to a separate
function, which is called when the last reporter reference is put.

Besides, it allows to call __devlink_health_reporter_destroy() while
locked on a reporters list mutex in symmetry to
__devlink_health_reporter_create(), which is required in follow-up
patch.

Signed-off-by: Vladyslav Tarasiuk <vladyslavt@mellanox.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Refactor devlink health reporter constructor
Vladyslav Tarasiuk [Fri, 10 Jul 2020 12:25:07 +0000 (15:25 +0300)]
devlink: Refactor devlink health reporter constructor

Prepare a common routine in devlink_health_reporter_create() for usage
in similar functions for devlink port health reporters.

Signed-off-by: Vladyslav Tarasiuk <vladyslavt@mellanox.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge tag 'mlx5-updates-2020-07-09' of git://git.kernel.org/pub/scm/linux/kernel...
David S. Miller [Fri, 10 Jul 2020 21:10:45 +0000 (14:10 -0700)]
Merge tag 'mlx5-updates-2020-07-09' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-updates-2020-07-09

This series provides updates to mlx5 CT (connection tracking) offloads
For more information please see tag log below.

Please pull and let me know if there is any problem.

The following conflict is expected when net is merged into net-next:
to resolve just use the hunks from net-next.

<<<<<<< HEAD (net-next)
mlx5_tc_ct_del_ft_entry(ct_priv, entry);
kfree(entry);
======= (net)
mlx5_tc_ct_entry_del_rules(ct_priv, entry);
kfree(entry);
>>>>>>> b1a7d5bdfe54c98eca46e2c997d4e3b1484a49af

mlx5 connection tracking offloads updates:

1)  Restore CT state from lookup in zone instead of tupleid

    On a miss, Use this zone + 5 tuple taken from the skb, to lookup the CT
    entry and restore it, instead of the driver allocated tuple id.

    This improves flow insertion rate by avoiding the allocation of a header
    rewrite context to maintain the tupleid.

2) Re-use modify header HW objects for identical modify actions.

3) Expand tunnel register mappings
   Reg_c1 is 32 bits wide. Before this patchset, 24 bit were allocated
   for the tuple_id,  6 bits for tunnel mapping and 2 bits for tunnel
   options mappings.

   Restoring the ct state from zone lookup instead of tuple id requires
   reg_c1 to store 8 bits mapping the ct zone, leaving 24 bits for tunnel
   mappings.

   Expand tunnel and tunnel options register mappings to 12 bit each.

4) Trivial cleanup and fixes.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'udp_tunnel-add-NIC-RX-port-offload-infrastructure'
David S. Miller [Fri, 10 Jul 2020 20:54:00 +0000 (13:54 -0700)]
Merge branch 'udp_tunnel-add-NIC-RX-port-offload-infrastructure'

Jakub Kicinski says:

====================
udp_tunnel: add NIC RX port offload infrastructure

Kernel has a facility to notify drivers about the UDP tunnel ports
so that devices can recognize tunneled packets. This is important
mostly for RX - devices which don't support CHECKSUM_COMPLETE can
report checksums of inner packets, and compute RSS over inner headers.
Some drivers also match the UDP tunnel ports also for TX, although
doing so may lead to false positives and negatives.

Unfortunately the user experience when trying to take adavantage
of these facilities is suboptimal. First of all there is no way
for users to check which ports are offloaded. Many drivers resort
to printing messages to aid debugging, other use debugfs. Even worse
the availability of the RX features (NETIF_F_RX_UDP_TUNNEL_PORT)
is established purely on the basis of the driver having the ndos
installed. For most drivers, however, the ability to perform offloads
is contingent on device capabilities (driver support multiple device
and firmware versions). Unless driver resorts to hackish clearing
of features set incorrectly by the core - users are left guessing
whether their device really supports UDP tunnel port offload or not.

There is currently no way to indicate or configure whether RX
features include just the checksum offload or checksum and using
inner headers for RSS. Many drivers default to not using inner
headers for RSS because most implementations populate the source
port with entropy from the inner headers. This, however, is not
always the case, for example certain switches are only able to
use a fixed source port during encapsulation.

We have also seen many driver authors get the intricacies of UDP
tunnel port offloads wrong. Most commonly the drivers forget to
perform reference counting, or take sleeping locks in the callbacks.

This work tries to improve the situation by pulling the UDP tunnel
port table maintenance out of the drivers. It turns out that almost
all drivers maintain a fixed size table of ports (in most cases one
per tunnel type), so we can take care of all the refcounting in the
core, and let the driver specify if they need to sleep in the
callbacks or not. The new common implementation will also support
replacing ports - when a port is removed from a full table it will
try to find a previously missing port to take its place.

This patch only implements the core functionality along with a few
drivers I was hoping to test manually [1] along with a test based
on a netdevsim implementation. Following patches will convert all
the drivers. Once that's complete we can remove the ndos, and rely
directly on the new infrastrucutre.

Then after RSS (RXFH) is converted to netlink we can add the ability
to configure the use of inner RSS headers for UDP tunnels.

[1] Unfortunately I wasn't able to, turns out 2 of the devices
I had access to were older generation or had old FW, and they
did not actually support UDP tunnel port notifications (see
the second paragraph). The thrid device appears to program
the UDP ports correctly but it generates bad UDP checksums with
or without these patches. Long story short - I'd appreciate
reviews and testing here..

v4:
 - better build fix (hopefully this one does it..)
v3:
 - fix build issue;
 - improve bnxt changes.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlx4: convert to new udp_tunnel_nic infra
Jakub Kicinski [Fri, 10 Jul 2020 00:42:53 +0000 (17:42 -0700)]
mlx4: convert to new udp_tunnel_nic infra

Convert to new infra, make use of the ability to sleep in the callback.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobnxt: convert to new udp_tunnel_nic infra
Jakub Kicinski [Fri, 10 Jul 2020 00:42:52 +0000 (17:42 -0700)]
bnxt: convert to new udp_tunnel_nic infra

Convert to new infra, taking advantage of sleeping in callbacks.

v2:
 - use bp->*_fw_dst_port_id != INVALID_HW_RING_ID as indication
   that the offload is active.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoixgbe: convert to new udp_tunnel_nic infra
Jakub Kicinski [Fri, 10 Jul 2020 00:42:51 +0000 (17:42 -0700)]
ixgbe: convert to new udp_tunnel_nic infra

Make use of new common udp_tunnel_nic infra. ixgbe supports
IPv4 only, and only single VxLAN and Geneve ports (one each).

v2:
 - split out the RXCSUM feature handling to separate change;
 - declare structs separately;
 - use ti.type instead of assuming table 0 is VxLAN;
 - move setting netdev->udp_tunnel_nic_info to its own switch.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoixgbe: don't clear UDP tunnel ports when RXCSUM is disabled
Jakub Kicinski [Fri, 10 Jul 2020 00:42:50 +0000 (17:42 -0700)]
ixgbe: don't clear UDP tunnel ports when RXCSUM is disabled

It appears the clearing of UDP tunnel ports when RXCSUM
is disabled is unnecessary. Driver will not pay attention
to checksum bits if RXCSUM is not set, so we can let
the hardware parse the packets.

Note that the UDP tunnel port NDO handlers don't pay attention
to the state of RXCSUM, so the ports could had been re-programmed,
anyway.

This cleanup simplifies later conversion patch.

v2:
 - break this out of the following patch.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests: net: add a test for UDP tunnel info infra
Jakub Kicinski [Fri, 10 Jul 2020 00:42:49 +0000 (17:42 -0700)]
selftests: net: add a test for UDP tunnel info infra

Add validating the UDP tunnel infra works.

$ ./udp_tunnel_nic.sh
PASSED all 383 checks

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonetdevsim: add UDP tunnel port offload support
Jakub Kicinski [Fri, 10 Jul 2020 00:42:48 +0000 (17:42 -0700)]
netdevsim: add UDP tunnel port offload support

Add UDP tunnel port handlers to our fake driver so we can test
the core infra.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoethtool: add tunnel info interface
Jakub Kicinski [Fri, 10 Jul 2020 00:42:47 +0000 (17:42 -0700)]
ethtool: add tunnel info interface

Add an interface to report offloaded UDP ports via ethtool netlink.

Now that core takes care of tracking which UDP tunnel ports the NICs
are aware of we can quite easily export this information out to
user space.

The responsibility of writing the netlink dumps is split between
ethtool code and udp_tunnel_nic.c - since udp_tunnel module may
not always be loaded, yet we should always report the capabilities
of the NIC.

$ ethtool --show-tunnels eth0
Tunnel information for eth0:
  UDP port table 0:
    Size: 4
    Types: vxlan
    No entries
  UDP port table 1:
    Size: 4
    Types: geneve, vxlan-gpe
    Entries (1):
        port 1230, vxlan-gpe

v4:
 - back to v2, build fix is now directly in udp_tunnel.h
v3:
 - don't compile ETHTOOL_MSG_TUNNEL_INFO_GET in if CONFIG_INET
   not set.
v2:
 - fix string set count,
 - reorder enums in the uAPI,
 - fix type of ETHTOOL_A_TUNNEL_UDP_TABLE_TYPES to bitset
   in docs and comments.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoudp_tunnel: add central NIC RX port offload infrastructure
Jakub Kicinski [Fri, 10 Jul 2020 00:42:46 +0000 (17:42 -0700)]
udp_tunnel: add central NIC RX port offload infrastructure

Cater to devices which:
 (a) may want to sleep in the callbacks;
 (b) only have IPv4 support;
 (c) need all the programming to happen while the netdev is up.

Drivers attach UDP tunnel offload info struct to their netdevs,
where they declare how many UDP ports of various tunnel types
they support. Core takes care of tracking which ports to offload.

Use a fixed-size array since this matches what almost all drivers
do, and avoids a complexity and uncertainty around memory allocations
in an atomic context.

Make sure that tunnel drivers don't try to replay the ports when
new NIC netdev is registered. Automatic replays would mess up
reference counting, and will be removed completely once all drivers
are converted.

v4:
 - use a #define NULL to avoid build issues with CONFIG_INET=n.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoudp_tunnel: re-number the offload tunnel types
Jakub Kicinski [Fri, 10 Jul 2020 00:42:45 +0000 (17:42 -0700)]
udp_tunnel: re-number the offload tunnel types

Make it possible to use tunnel types as flags more easily.
There doesn't appear to be any user using the type as an
array index, so this should make no difference.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodebugfs: make sure we can remove u32_array files cleanly
Jakub Kicinski [Fri, 10 Jul 2020 00:42:44 +0000 (17:42 -0700)]
debugfs: make sure we can remove u32_array files cleanly

debugfs_create_u32_array() allocates a small structure to wrap
the data and size information about the array. If users ever
try to remove the file this leads to a leak since nothing ever
frees this wrapper.

That said there are no upstream users of debugfs_create_u32_array()
that'd remove a u32 array file (we only have one u32 array user in
CMA), so there is no real bug here.

Make callers pass a wrapper they allocated. This way the lifetime
management of the wrapper is on the caller, and we can avoid the
potential leak in debugfs.

CC: Chucheng Luo <luochucheng@vivo.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet/mlx5e: CT: Fix releasing ft entries
Roi Dayan [Tue, 30 Jun 2020 12:40:37 +0000 (15:40 +0300)]
net/mlx5e: CT: Fix releasing ft entries

Before this commit, on ft flush, ft entries were not removed
from the ct_tuple hashtables. Fix it.

Fixes: ac991b48d43c ("net/mlx5e: CT: Offload established flows")
Signed-off-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Eli Britstein <elibr@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: CT: Remove unused function param
Saeed Mahameed [Mon, 4 May 2020 22:53:06 +0000 (15:53 -0700)]
net/mlx5e: CT: Remove unused function param

"flow" parameter is not used in __mlx5_tc_ct_flow_offload_clear(),
remove it.

Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
4 years agonet/mlx5e: CT: Return err_ptr from internal functions
Saeed Mahameed [Mon, 4 May 2020 22:52:14 +0000 (15:52 -0700)]
net/mlx5e: CT: Return err_ptr from internal functions

Instead of having to deal with converting between int and ERR_PTR for
return values in mlx5_tc_ct_flow_offload(), make the internal helper
functions return a ptr to mlx5_flow_handle instead of passing it as
output param, this will also avoid gcc confusion and false alarms,
thus we remove the redundant ERR_PTR rule initialization.

Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Suggested-by: Jason Gunthorpe <jgg@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
4 years agonet/mlx5e: CT: Expand tunnel register mappings
Paul Blakey [Tue, 5 May 2020 13:41:02 +0000 (16:41 +0300)]
net/mlx5e: CT: Expand tunnel register mappings

Reg_c1 is 32 bits wide. Originally, 24 bit were allocated for the tuple_id,
6 bits for tunnel mapping and 2 bits for tunnel options mappings.

Restoring the ct state from zone lookup instead of tuple id requires
reg_c1 to store 8 bits mapping the ct zone, leaving 24 bits for tunnel
mappings.

Expand tunnel and tunnel options register mappings to 12 bit each.

Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: CT: Use mapping for zone restore register
Paul Blakey [Tue, 5 May 2020 13:37:22 +0000 (16:37 +0300)]
net/mlx5e: CT: Use mapping for zone restore register

Use a single byte mapping for zone restore register (zone matching
remains 16 bit).

This makes room for using the freed 8 bits on register C1 for
mapping more tunnels and tunnel options.

Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: CT: Re-use tuple modify headers for identical modify actions
Paul Blakey [Tue, 18 Feb 2020 08:24:07 +0000 (10:24 +0200)]
net/mlx5e: CT: Re-use tuple modify headers for identical modify actions

After removing the tupleid register which changed per tuple,
tuple modify headers set the ct_state, zone, mark, and label registers.
For non-natted tuples going through the same tc rules path, their values
will be the same, and all their modify headers will be the same.

Re-use tuple modify header when possible, by adding each new modify
header to an hahstable, and looking up identical ones before creating
a new one.

Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: Export sharing of mod headers to a new file
Paul Blakey [Tue, 17 Mar 2020 14:32:21 +0000 (16:32 +0200)]
net/mlx5e: Export sharing of mod headers to a new file

Refactor sharing of mod headers to new file and while there,
remove spin lock and flows list, as this is only used for warn on.

Use the generic API in the next patch to re-use tuple modify headers
for identical modify actions,

Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: CT: Restore ct state from lookup in zone instead of tupleid
Paul Blakey [Sun, 29 Mar 2020 10:50:47 +0000 (13:50 +0300)]
net/mlx5e: CT: Restore ct state from lookup in zone instead of tupleid

Remove tupleid, and replace it with zone_restore, which is the zone an
established tuple sets after match. On miss, Use this zone + tuple
taken from the skb, to lookup the ct entry and restore it.

This improves flow insertion rate by avoiding the allocation of a header
rewrite context.

Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: CT: Don't offload tuple rewrites for established tuples
Paul Blakey [Wed, 22 Apr 2020 15:00:25 +0000 (18:00 +0300)]
net/mlx5e: CT: Don't offload tuple rewrites for established tuples

Next patches will remove the tupleid registers that is used
to restore the ct state on miss, and instead use the tuple on
the missed packet to lookup which state to restore.
Disable tuple rewrites after connection tracking.

For tuple rewrites, inject a ct_state=-trk match so it won't
change the tuple for established flows (+trk) that passed connection
tracking, and instead miss to software.

Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: Use netdev_info instead of pr_info
Oz Shlomo [Mon, 1 Jun 2020 17:08:55 +0000 (17:08 +0000)]
net/mlx5e: Use netdev_info instead of pr_info

The next patch will pass the mlx5e_priv struct to the
modify_header_match_supported method. Use this opportunity to refactor
the existing pr_info call to a netdev_info call.

Signed-off-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: CT: Allow header rewrite of 5-tuple and ct clear action
Paul Blakey [Sun, 3 May 2020 13:45:02 +0000 (16:45 +0300)]
net/mlx5e: CT: Allow header rewrite of 5-tuple and ct clear action

With ct clear we don't jump to the ct tables, so header rewrite
of 5-tuple can be done in place (and not moved to after the CT action).

Check for ct clear action, and if so, allow 5-tuple header
rewrite.

Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5e: CT: Save ct entries tuples in hashtables
Paul Blakey [Sun, 29 Mar 2020 10:07:49 +0000 (13:07 +0300)]
net/mlx5e: CT: Save ct entries tuples in hashtables

Save original tuple and natted tuple in two new hashtables.

This is a pre-step for restoring ct state after hw miss by performing a
5-tuple lookup on the hash tables.

Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Oz Shlomo <ozsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agonet/mlx5: E-switch, When eswitch is unsupported, return -EOPNOTSUPP
Parav Pandit [Wed, 24 Jun 2020 10:56:25 +0000 (05:56 -0500)]
net/mlx5: E-switch, When eswitch is unsupported, return -EOPNOTSUPP

When eswitch is unsupported, currently -EPERM error code is returned
instead of -EOPNOTSUPP.

Due to this VF device's devlink virtual port is not enumerated because
port_function_get() callback returned -EPERM instead of -EOPNOTSUPP.

Hence, return the error code -EOPNOTSUPP when eswitch is unsupported.

Fixes: bd93975353d5 ("net/mlx5: E-switch, Introduce and use eswitch support check helper")
Signed-off-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
4 years agoMerge branch 'Expose-port-split-attributes'
David S. Miller [Thu, 9 Jul 2020 20:15:30 +0000 (13:15 -0700)]
Merge branch 'Expose-port-split-attributes'

Ido Schimmel says:

====================
Expose port split attributes

Danielle says:

Currently, user space has no way of knowing if a port can be split and
into how many ports. Among other things, this makes it impossible to
write generic tests for port split functionality.

Therefore, this set exposes two new devlink port attributes to user
space: Number of lanes and whether the port can be split or not.

Patch set overview:

Patches #1-#4 cleanup 'struct devlink_port_attrs' and reduce the number
of parameters passed between drivers and devlink via
devlink_port_attrs_set()

Patch #5 adds devlink port lanes attributes

Patches #6-#7 add devlink port splittable attribute

Patch #8 exploits the fact that devlink is now aware of port's number of
lanes and whether the port can be split or not and moves some checks
from drivers to devlink

Patch #9 adds a port split test

Changes since v2:
* Remove some local variables from patch #3
* Reword function description in patch #5
* Fix a bug in patch #8
* Add a test for the splittable attribute in patch #9

Changes since v1:
* Rename 'width' attribute to 'lanes'
* Add 'splittable' attribute
* Move checks from drivers to devlink
====================

Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests: net: Add port split test
Danielle Ratson [Thu, 9 Jul 2020 13:18:22 +0000 (16:18 +0300)]
selftests: net: Add port split test

Test port split configuration using previously added number of port lanes
attribute.

Check that all the splittable ports are successfully split to their maximum
number of lanes and below, and that those which are not splittable fail to
be split.

Test output example:

TEST: swp4 is unsplittable                                         [ OK ]
TEST: split port swp53 into 4                                      [ OK ]
TEST: Unsplit port pci/0000:03:00.0/25                             [ OK ]
TEST: split port swp53 into 2                                      [ OK ]
TEST: Unsplit port pci/0000:03:00.0/25                             [ OK ]

Signed-off-by: Danielle Ratson <danieller@mellanox.com>
Reviewed-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Move input checks from driver to devlink
Danielle Ratson [Thu, 9 Jul 2020 13:18:21 +0000 (16:18 +0300)]
devlink: Move input checks from driver to devlink

Currently, all the input checks are done in driver.

After adding the split capability to devlink port, move the checks to
devlink.

Signed-off-by: Danielle Ratson <danieller@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Add a new devlink port split ability attribute and pass to netlink
Danielle Ratson [Thu, 9 Jul 2020 13:18:20 +0000 (16:18 +0300)]
devlink: Add a new devlink port split ability attribute and pass to netlink

Add a new attribute that indicates the split ability of devlink port.

Drivers are expected to set it via devlink_port_attrs_set(), before
registering the port.

Signed-off-by: Danielle Ratson <danieller@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: Set port split ability attribute in driver
Danielle Ratson [Thu, 9 Jul 2020 13:18:19 +0000 (16:18 +0300)]
mlxsw: Set port split ability attribute in driver

Currently, port attributes like flavour, port number and whether the port
was split are set when initializing a port.

Set the split ability of the port as well, based on port_mapping->width
field and split attribute of devlink port in spectrum, so that it could be
easily passed to devlink in the next patch.

Signed-off-by: Danielle Ratson <danieller@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Add a new devlink port lanes attribute and pass to netlink
Danielle Ratson [Thu, 9 Jul 2020 13:18:18 +0000 (16:18 +0300)]
devlink: Add a new devlink port lanes attribute and pass to netlink

Add a new devlink port attribute that indicates the port's number of lanes.

Drivers are expected to set it via devlink_port_attrs_set(), before
registering the port.

The attribute is not passed to user space in case the number of lanes is
invalid (0).

Signed-off-by: Danielle Ratson <danieller@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: Set number of port lanes attribute in driver
Danielle Ratson [Thu, 9 Jul 2020 13:18:17 +0000 (16:18 +0300)]
mlxsw: Set number of port lanes attribute in driver

Currently, port attributes like flavour, port number and whether the
port was split are set when initializing a port.

Set the number of lanes of the port as well so that it could be easily
passed to devlink in the next patch.

Signed-off-by: Danielle Ratson <danieller@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Replace devlink_port_attrs_set parameters with a struct
Danielle Ratson [Thu, 9 Jul 2020 13:18:16 +0000 (16:18 +0300)]
devlink: Replace devlink_port_attrs_set parameters with a struct

Currently, devlink_port_attrs_set accepts a long list of parameters,
that most of them are devlink port's attributes.

Use the devlink_port_attrs struct to replace the relevant parameters.

Signed-off-by: Danielle Ratson <danieller@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Move switch_port attribute of devlink_port_attrs to devlink_port
Danielle Ratson [Thu, 9 Jul 2020 13:18:15 +0000 (16:18 +0300)]
devlink: Move switch_port attribute of devlink_port_attrs to devlink_port

The struct devlink_port_attrs holds the attributes of devlink_port.

Similarly to the previous patch, 'switch_port' attribute is another
exception.

Move 'switch_port' to be devlink_port's field.

Signed-off-by: Danielle Ratson <danieller@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Move set attribute of devlink_port_attrs to devlink_port
Danielle Ratson [Thu, 9 Jul 2020 13:18:14 +0000 (16:18 +0300)]
devlink: Move set attribute of devlink_port_attrs to devlink_port

The struct devlink_port_attrs holds the attributes of devlink_port.

The 'set' field is not devlink_port's attribute as opposed to most of the
others.

Move 'set' to be devlink_port's field called 'attrs_set'.

Signed-off-by: Danielle Ratson <danieller@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: mscc: fix ptr_ret.cocci warnings
kernel test robot [Thu, 9 Jul 2020 16:06:18 +0000 (00:06 +0800)]
net: phy: mscc: fix ptr_ret.cocci warnings

drivers/net/phy/mscc/mscc_ptp.c:1496:1-3: WARNING: PTR_ERR_OR_ZERO can be used

 Use PTR_ERR_OR_ZERO rather than if(IS_ERR(...)) + PTR_ERR

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

Fixes: 7d272e63e097 ("net: phy: mscc: timestamping and PHC support")
CC: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: kernel test robot <lkp@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: systemport: fix double shift of a vlan_tci by VLAN_PRIO_SHIFT
Colin Ian King [Wed, 8 Jul 2020 18:37:23 +0000 (19:37 +0100)]
net: systemport: fix double shift of a vlan_tci by VLAN_PRIO_SHIFT

Currently the u16 skb->vlan_tci is being right  shifted twice by
VLAN_PRIO_SHIFT, once in the macro skb_vlan_tag_get_pri and explicitly
by VLAN_PRIO_SHIFT afterwards. The combined shift amount is larger than
the u16 so the end result is always zero.  Remove the second explicit
shift as this is extraneous.

Fixes: 6e9fdb60d362 ("net: systemport: Add support for VLAN transmit acceleration")
Addresses-Coverity: ("Operands don't affect result")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'mptcp-introduce-msk-diag-interface'
David S. Miller [Thu, 9 Jul 2020 19:38:41 +0000 (12:38 -0700)]
Merge branch 'mptcp-introduce-msk-diag-interface'

Paolo Abeni says:

====================
mptcp: introduce msk diag interface

This series implements the diag interface for the MPTCP sockets.

Since the MPTCP protocol value can't be represented with the
current diag uAPI, the first patch introduces an extended attribute
allowing user-space to specify lager protocol values.

The token APIs are then extended to allow traversing the
whole token container.

Patch 3 carries the actual diag interface implementation, and
later patch bring-in some functional self-tests.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests/mptcp: add diag interface tests
Paolo Abeni [Thu, 9 Jul 2020 13:12:42 +0000 (15:12 +0200)]
selftests/mptcp: add diag interface tests

basic functional test, triggering the msk diag interface
code. Require appropriate iproute2 support, skip elsewhere.

Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomptcp: add MPTCP socket diag interface
Paolo Abeni [Thu, 9 Jul 2020 13:12:41 +0000 (15:12 +0200)]
mptcp: add MPTCP socket diag interface

exposes basic inet socket attribute, plus some MPTCP socket
fields comprising PM status and MPTCP-level sequence numbers.

Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomptcp: add msk interations helper
Paolo Abeni [Thu, 9 Jul 2020 13:12:40 +0000 (15:12 +0200)]
mptcp: add msk interations helper

mptcp_token_iter_next() allow traversing all the MPTCP
sockets inside the token container belonging to the given
network namespace with a quite standard iterator semantic.

That will be used by the next patch, but keep the API generic,
as we plan to use this later for PM's sake.

Additionally export mptcp_token_get_sock(), as it also
will be used by the diag module.

Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoinet_diag: support for wider protocol numbers
Paolo Abeni [Thu, 9 Jul 2020 13:12:39 +0000 (15:12 +0200)]
inet_diag: support for wider protocol numbers

After commit bf9765145b85 ("sock: Make sk_protocol a 16-bit value")
the current size of 'sdiag_protocol' is not sufficient to represent
the possible protocol values.

This change introduces a new inet diag request attribute to let
user space specify the relevant protocol number using u32 values.

The attribute is parsed by inet diag core on get/dump command
and the extended protocol value, if available, is preferred to
'sdiag_protocol' to lookup the diag handler.

The parse attributed are exposed to all the diag handlers via
the cb->data.

Note that inet_diag_dump_one_icsk() is left unmodified, as it
will not be used by protocol using the extended attribute.

Suggested-by: David S. Miller <davem@davemloft.net>
Co-developed-by: Christoph Paasch <cpaasch@apple.com>
Signed-off-by: Christoph Paasch <cpaasch@apple.com>
Acked-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: enetc: use eth_broadcast_addr() to assign broadcast
Xu Wang [Thu, 9 Jul 2020 06:48:55 +0000 (06:48 +0000)]
net: enetc: use eth_broadcast_addr() to assign broadcast

This patch is to use eth_broadcast_addr() to assign broadcast address
insetad of memset().

Signed-off-by: Xu Wang <vulab@iscas.ac.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobonding: don't need RTNL for ipsec helpers
Jarod Wilson [Wed, 8 Jul 2020 22:58:49 +0000 (18:58 -0400)]
bonding: don't need RTNL for ipsec helpers

The bond_ipsec_* helpers don't need RTNL, and can potentially get called
without it being held, so switch from rtnl_dereference() to
rcu_dereference() to access bond struct data.

Lightly tested with xfrm bonding, no problems found, should address the
syzkaller bug referenced below.

Reported-by: syzbot+582c98032903dcc04816@syzkaller.appspotmail.com
CC: Huy Nguyen <huyn@mellanox.com>
CC: Saeed Mahameed <saeedm@mellanox.com>
CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Veaceslav Falico <vfalico@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
CC: "David S. Miller" <davem@davemloft.net>
CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: Jakub Kicinski <kuba@kernel.org>
CC: Steffen Klassert <steffen.klassert@secunet.com>
CC: Herbert Xu <herbert@gondor.apana.org.au>
CC: netdev@vger.kernel.org
CC: intel-wired-lan@lists.osuosl.org
Signed-off-by: Jarod Wilson <jarod@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodt-bindings: dp83869: Fix the type of device
Fabio Estevam [Wed, 8 Jul 2020 21:24:22 +0000 (18:24 -0300)]
dt-bindings: dp83869: Fix the type of device

DP83869 is an Ethernet PHY, not a charger, so fix the documentation
accordingly.

Fixes: 4d66c56f7efe ("dt-bindings: net: dp83869: Add TI dp83869 phy")
Signed-off-by: Fabio Estevam <festevam@gmail.com>
Acked-by: Dan Murphy <dmurphy@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodt-bindings: dp83867: Fix the type of device
Fabio Estevam [Wed, 8 Jul 2020 21:24:21 +0000 (18:24 -0300)]
dt-bindings: dp83867: Fix the type of device

DP83867 is an Ethernet PHY, not a charger, so fix the documentation
accordingly.

Fixes: 74ac28f16486 ("dt-bindings: dp83867: Convert DP83867 to yaml")
Signed-off-by: Fabio Estevam <festevam@gmail.com>
Acked-by: Dan Murphy <dmurphy@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobonding: deal with xfrm state in all modes and add more error-checking
Jarod Wilson [Wed, 8 Jul 2020 17:46:31 +0000 (13:46 -0400)]
bonding: deal with xfrm state in all modes and add more error-checking

It's possible that device removal happens when the bond is in non-AB mode,
and addition happens in AB mode, so bond_ipsec_del_sa() never gets called,
which leaves security associations in an odd state if bond_ipsec_add_sa()
then gets called after switching the bond into AB. Just call add and
delete universally for all modes to keep things consistent.

However, it's also possible that this code gets called when the system is
shutting down, and the xfrm subsystem has already been disconnected from
the bond device, so we need to do some error-checking and bail, lest we
hit a null ptr deref.

Fixes: a3b658cfb664 ("bonding: allow xfrm offload setup post-module-load")
CC: Huy Nguyen <huyn@mellanox.com>
CC: Saeed Mahameed <saeedm@mellanox.com>
CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Veaceslav Falico <vfalico@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
CC: "David S. Miller" <davem@davemloft.net>
CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: Jakub Kicinski <kuba@kernel.org>
CC: Steffen Klassert <steffen.klassert@secunet.com>
CC: Herbert Xu <herbert@gondor.apana.org.au>
CC: netdev@vger.kernel.org
CC: intel-wired-lan@lists.osuosl.org
Signed-off-by: Jarod Wilson <jarod@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'RTL8366RB-tagging-support'
David S. Miller [Wed, 8 Jul 2020 22:36:19 +0000 (15:36 -0700)]
Merge branch 'RTL8366RB-tagging-support'

Linus Walleij says:

====================
RTL8366RB tagging support

This patch set adds DSA tagging support to the RTL8366RB
DSA driver.

There is a minor performance improvement in the tag parser
compared to the previous patch set and the review tags
have been collected.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: rtl8366rb: Support the CPU DSA tag
Linus Walleij [Wed, 8 Jul 2020 12:25:37 +0000 (14:25 +0200)]
net: dsa: rtl8366rb: Support the CPU DSA tag

This activates the support to use the CPU tag to properly
direct ingress traffic to the right port.

Bit 15 in register RTL8368RB_CPU_CTRL_REG can be set to
1 to disable the insertion of the CPU tag which is what
the code currently does. The bit 15 define calls this
setting RTL8368RB_CPU_INSTAG which is confusing since the
inverse meaning is implied: programmers may think that
setting this bit to 1 will *enable* inserting the tag
rather than disabling it, so rename this setting in
bit 15 to RTL8368RB_CPU_NO_TAG which is more to the
point.

After this e.g. ping works out-of-the-box with the
RTL8366RB.

Cc: DENG Qingfang <dqfext@gmail.com>
Cc: Mauri Sandberg <sandberg@mailfence.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: tag_rtl4_a: Implement Realtek 4 byte A tag
Linus Walleij [Wed, 8 Jul 2020 12:25:36 +0000 (14:25 +0200)]
net: dsa: tag_rtl4_a: Implement Realtek 4 byte A tag

This implements the known parts of the Realtek 4 byte
tag protocol version 0xA, as found in the RTL8366RB
DSA switch.

It is designated as protocol version 0xA as a
different Realtek 4 byte tag format with protocol
version 0x9 is known to exist in the Realtek RTL8306
chips.

The tag and switch chip lacks public documentation, so
the tag format has been reverse-engineered from
packet dumps. As only ingress traffic has been available
for analysis an egress tag has not been possible to
develop (even using educated guesses about bit fields)
so this is as far as it gets. It is not known if the
switch even supports egress tagging.

Excessive attempts to figure out the egress tag format
was made. When nothing else worked, I just tried all bit
combinations with 0xannp where a is protocol and p is
port. I looped through all values several times trying
to get a response from ping, without any positive
result.

Using just these ingress tags however, the switch
functionality is vastly improved and the packets find
their way into the destination port without any
tricky VLAN configuration. On the D-Link DIR-685 the
LAN ports now come up and respond to ping without
any command line configuration so this is a real
improvement for users.

Egress packets need to be restricted to the proper
target ports using VLAN, which the RTL8366RB DSA
switch driver already sets up.

Cc: DENG Qingfang <dqfext@gmail.com>
Cc: Mauri Sandberg <sandberg@mailfence.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet/mlx5: Added support for 100Gbps per lane link modes
Meir Lichtinger [Tue, 7 Jul 2020 03:42:33 +0000 (20:42 -0700)]
net/mlx5: Added support for 100Gbps per lane link modes

This patch exposes new link modes using 100Gbps per lane, including 100G,
200G and 400G modes.

Signed-off-by: Meir Lichtinger <meirl@mellanox.com>
Reviewed-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoethtool: Add support for 100Gbps per lane link modes
Meir Lichtinger [Tue, 7 Jul 2020 03:42:32 +0000 (20:42 -0700)]
ethtool: Add support for 100Gbps per lane link modes

Define 100G, 200G and 400G link modes using 100Gbps per lane

LR, ER and FR are defined as a single link mode because they are
using same technology and by design are fully interoperable.
EEPROM content indicates if the module is LR, ER, or FR, and the
user space ethtool decoder is planned to support decoding these
modes in the EEPROM.

Signed-off-by: Meir Lichtinger <meirl@mellanox.com>
CC: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'bnxt_en-Driver-update-for-net-next'
David S. Miller [Wed, 8 Jul 2020 22:21:14 +0000 (15:21 -0700)]
Merge branch 'bnxt_en-Driver-update-for-net-next'

Michael Chan says:

====================
bnxt_en: Driver update for net-next.

This patchset implements ethtool -X to setup user-defined RSS indirection
table.  The new infrastructure also allows the proper logical ring index
to be used to populate the RSS indirection when queried by ethtool -x.
Prior to these patches, we were incorrectly populating the output of
ethtool -x with internal ring IDs which would make no sense to the user.

The last 2 patches add some cleanups to the VLAN acceleration logic
and check the firmware capabilities before allowing VLAN acceleration
offloads.

v4: Move bnxt_get_rxfh_indir_size() fix to a new patch #2.
    Modify patch #7 to revert RSS map to default only when necessary.

v3: Use ALIGN() in patch 5.
    Add warning messages in patch 6.

v2: Some RSS indirection table changes requested by Jakub Kicinski.
====================

Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobnxt_en: allow firmware to disable VLAN offloads
Edwin Peer [Wed, 8 Jul 2020 11:54:01 +0000 (07:54 -0400)]
bnxt_en: allow firmware to disable VLAN offloads

Bare-metal use cases require giving firmware and the embedded
application processor control over VLAN offloads. The driver should
not attempt to override or utilize this feature in such scenarios
since it will not work as expected.

Signed-off-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobnxt_en: clean up VLAN feature bit handling
Edwin Peer [Wed, 8 Jul 2020 11:54:00 +0000 (07:54 -0400)]
bnxt_en: clean up VLAN feature bit handling

The hardware VLAN offload feature on our NIC does not have separate
knobs for handling customer and service tags on RX. Either offloading
of both must be enabled or both must be disabled. Introduce definitions
for the combined feature set in order to clean up the code and make
this constraint more clear. Technically these features can be separately
enabled on TX, however, since the default is to turn both on, the
combined TX feature set is also introduced for code consistency.

Signed-off-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobnxt_en: Implement ethtool -X to set indirection table.
Michael Chan [Wed, 8 Jul 2020 11:53:59 +0000 (07:53 -0400)]
bnxt_en: Implement ethtool -X to set indirection table.

With the new infrastructure in place, we can now support the setting of
the indirection table from ethtool.

When changing channels, in a rare case that firmware cannot reserve the
rings that were promised, we will still try to keep the RSS map and only
revert to default when absolutely necessary.

v4: Revert RSS map to default during ring change only when absolutely
    necessary.

v3: Add warning messages when firmware cannot reserve the requested RX
    rings, and when the RSS table entries have to change to default.

v2: When changing channels, if the RSS table size changes and RSS map
    is non-default, return error.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobnxt_en: Return correct RSS indirection table entries to ethtool -x.
Michael Chan [Wed, 8 Jul 2020 11:53:58 +0000 (07:53 -0400)]
bnxt_en: Return correct RSS indirection table entries to ethtool -x.

Now that we have the logical indirection table, we can return these
proper logical indices directly to ethtool -x instead of the physical
IDs.

Reported-by: Jakub Kicinski <kicinski@fb.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobnxt_en: Fill HW RSS table from the RSS logical indirection table.
Michael Chan [Wed, 8 Jul 2020 11:53:57 +0000 (07:53 -0400)]
bnxt_en: Fill HW RSS table from the RSS logical indirection table.

Now that we have the logical table, we can fill the HW RSS table
using the logical table's entries and converting them to the HW
specific format.  Re-initialize the logical table to standard
distribution if the number of RX rings changes during ring reservation.

v4: Use bnxt_get_rxfh_indir_size() to get the RSS table size.

v2: Use ALIGN() to roundup the RSS table size.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobnxt_en: Add helper function to return the number of RSS contexts.
Michael Chan [Wed, 8 Jul 2020 11:53:56 +0000 (07:53 -0400)]
bnxt_en: Add helper function to return the number of RSS contexts.

On some chips, this varies based on the number of RX rings.  Add this
helper function and refactor the existing code to use it.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobnxt_en: Add logical RSS indirection table structure.
Michael Chan [Wed, 8 Jul 2020 11:53:55 +0000 (07:53 -0400)]
bnxt_en: Add logical RSS indirection table structure.

The driver currently does not keep track of the logical RSS indirection
table.  The hardware RSS table is set up with standard default ring
distribution when initializing the chip.  This makes it difficult to
support user sepcified indirection table entries.  As a first step, add
the logical table in the main bnxt structure and allocate it according
to chip specific table size.  Add a function that sets up default
RSS distribution based on the number of RX rings.

v4: Use bnxt_get_rxfh_indir_size() for the current RSS table size.

v2: Use kmalloc_array() since we init. all entries afterwards.
    Use ALIGN() to roundup the RSS table size.
    Use ethtool_rxfh_indir_default() to init. the default entries.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobnxt_en: Fix up bnxt_get_rxfh_indir_size().
Michael Chan [Wed, 8 Jul 2020 11:53:54 +0000 (07:53 -0400)]
bnxt_en: Fix up bnxt_get_rxfh_indir_size().

Fix up bnxt_get_rxfh_indir_size() to return the proper current RSS
table size for P5 chips.  Change it to non-static so that bnxt.c
can use it to get the table size.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobnxt_en: Set up the chip specific RSS table size.
Michael Chan [Wed, 8 Jul 2020 11:53:53 +0000 (07:53 -0400)]
bnxt_en: Set up the chip specific RSS table size.

Currently, we allocate one page for the hardware DMA RSS indirection
table.  While the size is currently big enough for all chips, future
chip variations may support bigger sizes, so it is better to calculate
and store the chip specific size and allocate accordingly.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf-next
David S. Miller [Wed, 8 Jul 2020 19:42:40 +0000 (12:42 -0700)]
Merge git://git./linux/kernel/git/pablo/nf-next

Pablo Neira Ayuso says:

====================
Netfilter/IPVS updates for net-next

The following patchset contains Netfilter updates for net-next:

1) Support for rejecting packets from the prerouting chain, from
   Laura Garcia Liebana.

2) Remove useless assignment in pipapo, from Stefano Brivio.

3) On demand hook registration in IPVS, from Julian Anastasov.

4) Expire IPVS connection from process context to not overload
   timers, also from Julian.

5) Fallback to conntrack TCP tracker to handle connection reuse
   in IPVS, from Julian Anastasov.

6) Several patches to support for chain bindings.

7) Expose enum nft_chain_flags through UAPI.

8) Reject unsupported chain flags from the netlink control plane.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'net-phy-Uninline-PHY-ethtool-statistics'
David S. Miller [Wed, 8 Jul 2020 19:39:05 +0000 (12:39 -0700)]
Merge branch 'net-phy-Uninline-PHY-ethtool-statistics'

Florian Fainelli says:

====================
net: phy: Uninline PHY ethtool statistics

Now that we have introduced ethtool_phy_ops we can uninline those
operations and move them back into phy.c where they belong. Since those
functions are used by DSA, we need to continue exporting those symbols.

It might be possible to remove ndo_get_ethtool_phy_stats in a subsequent
patch since we could have DSA register its own ethtool_phy_ops instance
instead of overloading the ethtool_ops.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: Uninline PHY ethtool statistics operations
Florian Fainelli [Wed, 8 Jul 2020 16:46:25 +0000 (09:46 -0700)]
net: phy: Uninline PHY ethtool statistics operations

Now that we have moved the PHY ethtool statistics to be dynamically
registered, we no longer need to inline those for ethtool. This used to
be done to avoid cross symbol referencing and allow ethtool to be
decoupled from PHYLIB entirely.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: Define PHY statistics ethtool_phy_ops
Florian Fainelli [Wed, 8 Jul 2020 16:46:24 +0000 (09:46 -0700)]
net: phy: Define PHY statistics ethtool_phy_ops

Extend ethtool_phy_ops to include the 3 function pointers necessary for
implementing PHY statistics. In a subsequent change we will uninline
those functions.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoReplace HTTP links with HTTPS ones: ATMEL MACB ETHERNET DRIVER
Alexander A. Klimov [Wed, 8 Jul 2020 10:35:19 +0000 (12:35 +0200)]
Replace HTTP links with HTTPS ones: ATMEL MACB ETHERNET DRIVER

Rationale:
Reduces attack surface on kernel devs opening the links for MITM
as HTTPS traffic is much harder to manipulate.

Deterministic algorithm:
For each file:
  If not .svg:
    For each line:
      If doesn't contain `\bxmlns\b`:
        For each link, `\bhttp://[^# \t\r\n]*(?:\w|/)`:
  If neither `\bgnu\.org/license`, nor `\bmozilla\.org/MPL\b`:
            If both the HTTP and HTTPS versions
            return 200 OK and serve the same content:
              Replace HTTP with HTTPS.

Signed-off-by: Alexander A. Klimov <grandmaster@al2klimov.de>
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: loop: Print when registration is successful
Florian Fainelli [Wed, 8 Jul 2020 04:45:13 +0000 (21:45 -0700)]
net: dsa: loop: Print when registration is successful

We have a number of error conditions that can lead to the driver not
probing successfully, move the print when we are sure
dsa_register_switch() has suceeded. This avoids repeated prints in case
of probe deferral for instance.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Vivien Didelot <vivien.didelot@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: sfp: add error checking with sfp_irq_name
Chris Healy [Tue, 7 Jul 2020 20:32:05 +0000 (13:32 -0700)]
net: sfp: add error checking with sfp_irq_name

Add error checking with sfp_irq_name before use.

Signed-off-by: Chris Healy <cphealy@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet/sched: Use fallthrough pseudo-keyword
Gustavo A. R. Silva [Tue, 7 Jul 2020 17:21:38 +0000 (12:21 -0500)]
net/sched: Use fallthrough pseudo-keyword

Replace the existing /* fall through */ comments and its variants with
the new pseudo-keyword macro fallthrough[1]. Also, remove unnecessary
fall-through markings when it is the case.

[1] https://www.kernel.org/doc/html/latest/process/deprecated.html?highlight=fallthrough#implicit-switch-case-fall-through

Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoReplace HTTP links with HTTPS ones: X.25 network layer
Alexander A. Klimov [Mon, 6 Jul 2020 20:50:30 +0000 (22:50 +0200)]
Replace HTTP links with HTTPS ones: X.25 network layer

Rationale:
Reduces attack surface on kernel devs opening the links for MITM
as HTTPS traffic is much harder to manipulate.

Deterministic algorithm:
For each file:
  If not .svg:
    For each line:
      If doesn't contain `\bxmlns\b`:
        For each link, `\bhttp://[^# \t\r\n]*(?:\w|/)`:
          If both the HTTP and HTTPS versions
          return 200 OK and serve the same content:
            Replace HTTP with HTTPS.

Signed-off-by: Alexander A. Klimov <grandmaster@al2klimov.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoReplace HTTP links with HTTPS ones: GRETH 10/100/1G Ethernet MAC device driver
Alexander A. Klimov [Mon, 6 Jul 2020 20:50:18 +0000 (22:50 +0200)]
Replace HTTP links with HTTPS ones: GRETH 10/100/1G Ethernet MAC device driver

Rationale:
Reduces attack surface on kernel devs opening the links for MITM
as HTTPS traffic is much harder to manipulate.

Deterministic algorithm:
For each file:
  If not .svg:
    For each line:
      If doesn't contain `\bxmlns\b`:
        For each link, `\bhttp://[^# \t\r\n]*(?:\w|/)`:
          If both the HTTP and HTTPS versions
          return 200 OK and serve the same content:
            Replace HTTP with HTTPS.

Signed-off-by: Alexander A. Klimov <grandmaster@al2klimov.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agobnx2x: fix spelling mistake "occurd" -> "occurred"
Colin Ian King [Mon, 6 Jul 2020 12:12:40 +0000 (13:12 +0100)]
bnx2x: fix spelling mistake "occurd" -> "occurred"

There are spelling mistakes in various literal strings. Fix these.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'net-ethtool-Untangle-PHYLIB-dependency'
David S. Miller [Tue, 7 Jul 2020 22:41:05 +0000 (15:41 -0700)]
Merge branch 'net-ethtool-Untangle-PHYLIB-dependency'

Florian Fainelli says:

====================
net: ethtool: Untangle PHYLIB dependency

This patch series untangles the ethtool netlink dependency with PHYLIB
which exists because the cable test feature calls directly into PHY
library functions. The approach taken here is to introduce
ethtool_phy_ops function pointers which can be dynamically registered
when PHYLIB loads.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ethtool: Remove PHYLIB direct dependency
Florian Fainelli [Mon, 6 Jul 2020 04:27:58 +0000 (21:27 -0700)]
net: ethtool: Remove PHYLIB direct dependency

Now that we have introduced ethtool_phy_ops and the PHY library
dynamically registers its operations with that function pointer, we can
remove the direct PHYLIB dependency in favor of using dynamic
operations.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: Register ethtool PHY operations
Florian Fainelli [Mon, 6 Jul 2020 04:27:57 +0000 (21:27 -0700)]
net: phy: Register ethtool PHY operations

Utilize ethtool_set_ethtool_phy_ops to register a suitable set of PHY
ethtool operations in a dynamic fashion such that ethtool will no longer
directy reference PHY library symbols.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ethtool: Introduce ethtool_phy_ops
Florian Fainelli [Mon, 6 Jul 2020 04:27:56 +0000 (21:27 -0700)]
net: ethtool: Introduce ethtool_phy_ops

In order to decouple ethtool from its PHY library dependency, define an
ethtool_phy_ops singleton which can be overriden by the PHY library when
it loads with an appropriate set of function pointers.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoDocumentation: networking: fix ethtool-netlink table formats
Randy Dunlap [Mon, 6 Jul 2020 02:55:44 +0000 (19:55 -0700)]
Documentation: networking: fix ethtool-netlink table formats

Fix table formatting to eliminate warnings.

Documentation/networking/ethtool-netlink.rst:509: WARNING: Malformed table.
Documentation/networking/ethtool-netlink.rst:522: WARNING: Malformed table.
Documentation/networking/ethtool-netlink.rst:543: WARNING: Malformed table.
Documentation/networking/ethtool-netlink.rst:555: WARNING: Malformed table.
Documentation/networking/ethtool-netlink.rst:591: WARNING: Malformed table.

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: add a Kconfig option for mdio_devres
Bartosz Golaszewski [Sun, 5 Jul 2020 09:55:47 +0000 (11:55 +0200)]
net: phy: add a Kconfig option for mdio_devres

If phylib is built as a module and CONFIG_MDIO_DEVICE is 'y', the
mdio_device and mdio_bus code will be in the phylib module, not in the
kernel image. Meanwhile we build mdio_devres depending on the
CONFIG_MDIO_DEVICE symbol, so if it's 'y', it will go into the kernel
and we'll hit the following linker error:

   ld: drivers/net/phy/mdio_devres.o: in function `devm_mdiobus_alloc_size':
>> drivers/net/phy/mdio_devres.c:38: undefined reference to `mdiobus_alloc_size'
   ld: drivers/net/phy/mdio_devres.o: in function `devm_mdiobus_free':
>> drivers/net/phy/mdio_devres.c:16: undefined reference to `mdiobus_free'
   ld: drivers/net/phy/mdio_devres.o: in function `__devm_mdiobus_register':
>> drivers/net/phy/mdio_devres.c:87: undefined reference to `__mdiobus_register'
   ld: drivers/net/phy/mdio_devres.o: in function `devm_mdiobus_unregister':
>> drivers/net/phy/mdio_devres.c:53: undefined reference to `mdiobus_unregister'
   ld: drivers/net/phy/mdio_devres.o: in function `devm_of_mdiobus_register':
>> drivers/net/phy/mdio_devres.c:120: undefined reference to `of_mdiobus_register'

Add a hidden Kconfig option for MDIO_DEVRES which will be currently
selected by CONFIG_PHYLIB as there are no non-phylib users of these
helpers.

Reported-by: kernel test robot <lkp@intel.com>
Fixes: ac3a68d56651 ("net: phy: don't abuse devres in devm_mdiobus_register()")
Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'smsc-use-generic-power-management'
David S. Miller [Tue, 7 Jul 2020 22:20:49 +0000 (15:20 -0700)]
Merge branch 'smsc-use-generic-power-management'

Vaibhav Gupta says:

====================
smsc: use generic power management

Linux Kernel Mentee: Remove Legacy Power Management.

The purpose of this patch series is to remove legacy power management callbacks
from smsc ethernet drivers.

The callbacks performing suspend() and resume() operations are still calling
pci_save_state(), pci_set_power_state(), etc. and handling the power management
themselves, which is not recommended.

The conversion requires the removal of the those function calls and change the
callback definition accordingly and make use of dev_pm_ops structure.

All patches are compile-tested only.

V2: Kbuild in V1, warning: variable 'err' is used uninitialized whenever 'if'
conditio is false in funcution .resume() .
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agosmsc9420: use generic power management
Vaibhav Gupta [Fri, 3 Jul 2020 03:01:38 +0000 (08:31 +0530)]
smsc9420: use generic power management

Drivers should not use legacy power management as they have to manage power
states and related operations, for the device, themselves. This driver was
handling them with the help of PCI helper functions.

With generic PM, all essentials will be handled by the PCI core. Driver
needs to do only device-specific operations.

Compile-tested only.

Signed-off-by: Vaibhav Gupta <vaibhavgupta40@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoepic100: use generic power management
Vaibhav Gupta [Fri, 3 Jul 2020 03:01:37 +0000 (08:31 +0530)]
epic100: use generic power management

Drivers should not use legacy power management as they have to manage power
states and related operations, for the device, themselves.

With generic PM, all essentials will be handled by the PCI core. Driver
needs to do only device-specific operations.

Compile-tested only.

Signed-off-by: Vaibhav Gupta <vaibhavgupta40@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomvpp2: fix pointer check
Matteo Croce [Tue, 7 Jul 2020 13:19:13 +0000 (15:19 +0200)]
mvpp2: fix pointer check

priv->page_pool is an array, so comparing against it will always return true.
Do a meaningful check by checking priv->page_pool[0] instead.
While at it, clear the page_pool pointers on deallocation, or when an
allocation error happens during init.

Reported-by: Colin Ian King <colin.king@canonical.com>
Fixes: c2d6fe6163de ("mvpp2: XDP TX support")
Signed-off-by: Matteo Croce <mcroce@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomptcp: use mptcp worker for path management
Florian Westphal [Tue, 7 Jul 2020 12:40:48 +0000 (14:40 +0200)]
mptcp: use mptcp worker for path management

We can re-use the existing work queue to handle path management
instead of a dedicated work queue.  Just move pm_worker to protocol.c,
call it from the mptcp worker and get rid of the msk lock (already held).

Signed-off-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agosun/cassini: mark cas_resume() as __maybe_unused
Wei Yongjun [Tue, 7 Jul 2020 10:55:43 +0000 (18:55 +0800)]
sun/cassini: mark cas_resume() as __maybe_unused

In certain configurations without power management support, gcc report
the following warning:

drivers/net/ethernet/sun/cassini.c:5206:12: warning:
 'cas_resume' defined but not used [-Wunused-function]
 5206 | static int cas_resume(struct device *dev_d)
      |            ^~~~~~~~~~

Mark cas_resume() as __maybe_unused to make it clear.

Fixes: f193f4ebde3d ("sun/cassini: use generic power management")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agosun/niu: add __maybe_unused attribute to PM functions
Vaibhav Gupta [Tue, 7 Jul 2020 07:41:22 +0000 (13:11 +0530)]
sun/niu: add __maybe_unused attribute to PM functions

The upgraded .suspend() and .resume() throw
"defined but not used [-Wunused-function]" warning for certain
configurations.

Mark them with "__maybe_unused" attribute.

Compile-tested only.

Fixes: b0db0cc2f695 ("sun/niu: use generic power management")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Vaibhav Gupta <vaibhavgupta40@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'phy-warn'
David S. Miller [Tue, 7 Jul 2020 19:47:11 +0000 (12:47 -0700)]
Merge branch 'phy-warn'

Andrew Lunn says:

====================
drivers/net/phy C=1 W=1 fixes

This fixes most of the Sparse and W=1 warnings in drivers/net/phy. The
Cavium code is still not fully clean, but it might actually be the
strange code is confusing Sparse.

v2
--
Added RB, TB, AB.
s/case/cause
Reverse Christmas tree
Module soft dependencies
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: mdio-octeon: Cleanup module loading dependencies
Andrew Lunn [Tue, 7 Jul 2020 01:49:39 +0000 (03:49 +0200)]
net: phy: mdio-octeon: Cleanup module loading dependencies

To ensure that the octeon MDIO driver has been loaded, the Cavium
ethernet drivers reference a dummy symbol in the MDIO driver. This
forces it to be loaded first. And this symbol has not been cleanly
implemented, resulting in warnings when build W=1 C=1.

Since device tree is being used, and a phandle points to the PHY on
the MDIO bus, we can make use of deferred probing. If the PHY fails to
connect, it should be because the MDIO bus driver has not loaded
yet. Return -EPROBE_DEFER so it will be tried again later.

Additionally, add a MODULE_SOFTDEP() to give user space a hint as to
what order it should load the modules.

v2:
s/octoen/octeon/
Add MODULE_SOFTDEP()

Cc: Sunil Goutham <sgoutham@marvell.com>
Cc: Robert Richter <rrichter@marvell.com>
Cc: Chris Packham <chris.packham@alliedtelesis.co.nz>
Tested-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: cavium: Improve __iomem mess
Andrew Lunn [Tue, 7 Jul 2020 01:49:38 +0000 (03:49 +0200)]
net: phy: cavium: Improve __iomem mess

The MIPS low level register access functions seem to be missing
__iomem annotation. This causes lots of sparse warnings, when code
casts off the __iomem. Make the Cavium MDIO drivers cleaner by pushing
the casts lower down into the helpers, allow the drivers to work as
normal, with __iomem.

bus->register_base is now an void *, rather than a u64. So forming the
mii_bus->id string cannot use %llx any more. Use %px, so this kernel
address is still exposed to user space, as it was before.

v2: s/cases/causes/g

Cc: Sunil Goutham <sgoutham@marvell.com>
Cc: Robert Richter <rrichter@marvell.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: dp83640: Fixup cast to restricted __be16 warning
Andrew Lunn [Tue, 7 Jul 2020 01:49:37 +0000 (03:49 +0200)]
net: phy: dp83640: Fixup cast to restricted __be16 warning

ntohs() expects to be passed a __be16. Correct the type of the
variable holding the sequence ID.

Cc: Richard Cochran <richardcochran@gmail.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: Make phy_10gbit_fec_features_array static
Andrew Lunn [Tue, 7 Jul 2020 01:49:36 +0000 (03:49 +0200)]
net: phy: Make phy_10gbit_fec_features_array static

This array is not used outside of phy_device.c, so make it static.

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: Properly define genphy_c45_driver
Andrew Lunn [Tue, 7 Jul 2020 01:49:35 +0000 (03:49 +0200)]
net: phy: Properly define genphy_c45_driver

Avoid the W=1 warning that symbol 'genphy_c45_driver' was not
declared. Should it be static?

Declare it on the phy header file.

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: Fixup parameters in kerneldoc
Andrew Lunn [Tue, 7 Jul 2020 01:49:34 +0000 (03:49 +0200)]
net: phy: Fixup parameters in kerneldoc

Correct the kerneldoc for a few structure and function calls,
as reported by C=1 W=1.

Cc: Alexandru Ardelean <alexaundru.ardelean@analog.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: at803x: Avoid comparison is always false warning
Andrew Lunn [Tue, 7 Jul 2020 01:49:33 +0000 (03:49 +0200)]
net: phy: at803x: Avoid comparison is always false warning

By placing the GENMASK value into an unsigned int and then passing it
to PREF_FIELD, the type is reduces down from ULL. Given the reduced
size of the type, the range checks in PREP_FAIL() are always true, and
-Wtype-limits then gives a warning.

By skipping the intermediate variable, the warning can be avoided.

Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>