linux-2.6-microblaze.git
3 years agonet: ethernet: ti: ale: switch to use tables for vlan entry description
Grygorii Strashko [Thu, 10 Sep 2020 20:28:06 +0000 (23:28 +0300)]
net: ethernet: ti: ale: switch to use tables for vlan entry description

The ALE VLAN entries are too much differ between different TI CPSW ALE
versions. So, handling them using flags, defines and get/set functions
became over-complicated.

This patch introduces tables to describe the ALE VLAN entries fields, which
are different between TI CPSW ALE versions, and new get/set access
functions. It also allows to detect incorrect access to not available ALL
entry fields.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ethernet: ti: am65-cpsw: enable hw auto ageing
Grygorii Strashko [Thu, 10 Sep 2020 20:28:05 +0000 (23:28 +0300)]
net: ethernet: ti: am65-cpsw: enable hw auto ageing

The AM65x ALE supports HW auto-ageing which can be enabled by programming
ageing interval in ALE_AGING_TIMER register. For this CPSW fck_clk
frequency has to be know by ALE.

This patch extends cpsw_ale_params with bus_freq field and enables ALE HW
auto ageing for AM65x CPSW2G ALE version.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ethernet: ti: ale: make usage of ale dev_id mandatory
Grygorii Strashko [Thu, 10 Sep 2020 20:28:04 +0000 (23:28 +0300)]
net: ethernet: ti: ale: make usage of ale dev_id mandatory

Hence all existing driver updated to use ALE dev_id the usage of ale dev_id
can be made mandatory and cpsw_ale_create() can be updated to use
"features" property from ALE static configuration.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ethernet: ti: am65-cpsw: use dev_id for ale configuration
Grygorii Strashko [Thu, 10 Sep 2020 20:28:03 +0000 (23:28 +0300)]
net: ethernet: ti: am65-cpsw: use dev_id for ale configuration

The previous patch has introduced possibility to select CPSW ALE by using
ALE dev_id identifier. Switch TI TI AM65x/J721E CPSW NUSS driver to use
dev_id.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: netcp: ethss: use dev_id for ale configuration
Grygorii Strashko [Thu, 10 Sep 2020 20:28:02 +0000 (23:28 +0300)]
net: netcp: ethss: use dev_id for ale configuration

The previous patch has introduced possibility to select CPSW ALE by using
ALE dev_id identifier. Switch TI Keystone 2 NETCP driver to use dev_id and
perform clean up by removing "ale_entries" configuration code.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ethernet: ti: cpsw: use dev_id for ale configuration
Grygorii Strashko [Thu, 10 Sep 2020 20:28:01 +0000 (23:28 +0300)]
net: ethernet: ti: cpsw: use dev_id for ale configuration

The previous patch has introduced possibility to select CPSW ALE by using
ALE dev_id identifier. Switch TI cpsw driver to use dev_id="cpsw" and
perform clean up by removing "ale_entries" configuration code.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ethernet: ti: ale: add static configuration
Grygorii Strashko [Thu, 10 Sep 2020 20:28:00 +0000 (23:28 +0300)]
net: ethernet: ti: ale: add static configuration

As existing, as newly introduced CPSW ALE versions have differences in
supported features and ALE table formats. Especially it's actual for the
recent AM65x/J721E/J7200 SoC and feature AM64x, which supports features
like: auto-aging, classifiers, Link aggregation, additional hw filtering,
etc.

Existing ALE configuration interface is not practical in terms of adding
new features and requires consumers to program a lot static parameters. Any
attempt to add new options will case endless adding and maintaining
different combination of flags and options.

Hence CPSW ALE configuration is static and fixed for SoC (or set of SoC) It
is reasonable to add support for static ALE configurations inside ALE
module. This patch adds static ALE configuration table for different ALE
versions and provides option for consumers to select required ALE
configuration by providing ALE const char *dev_id identifier.

This feature is not enabled by default until existing CPSW drivers will be
modified by follow up patches.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ethernet: ti: ale: add cpsw_ale_get_num_entries api
Grygorii Strashko [Thu, 10 Sep 2020 20:27:59 +0000 (23:27 +0300)]
net: ethernet: ti: ale: add cpsw_ale_get_num_entries api

Add cpsw_ale_get_num_entries() API to return number of ALE table entries
and update existing drivers to use it.

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'DSA-tag_8021q-cleanup'
David S. Miller [Sat, 12 Sep 2020 00:30:43 +0000 (17:30 -0700)]
Merge branch 'DSA-tag_8021q-cleanup'

Vladimir Oltean says:

====================
DSA tag_8021q cleanup

This small series tries to consolidate the VLAN handling in DSA a little
bit. It reworks tag_8021q to be minimally invasive to the dsa_switch_ops
structure. This makes the rest of the code a bit easier to follow.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoRevert "net: dsa: Add more convenient functions for installing port VLANs"
Vladimir Oltean [Thu, 10 Sep 2020 16:48:57 +0000 (19:48 +0300)]
Revert "net: dsa: Add more convenient functions for installing port VLANs"

This reverts commit 314f76d7a68bab0516aa52877944e6aacfa0fc3f.

Citing that commit message, the call graph was:

    dsa_slave_vlan_rx_add_vid   dsa_port_setup_8021q_tagging
                |                        |
                |                        |
                |          +-------------+
                |          |
                v          v
               dsa_port_vid_add      dsa_slave_port_obj_add
                      |                         |
                      +-------+         +-------+
                              |         |
                              v         v
                           dsa_port_vlan_add

Now that tag_8021q has its own ops structure, it no longer relies on
dsa_port_vid_add, and therefore on the dsa_switch_ops to install its
VLANs.

So dsa_port_vid_add now only has one single caller. So we can simplify
the call graph to what it was before, aka:

        dsa_slave_vlan_rx_add_vid     dsa_slave_port_obj_add
                      |                         |
                      +-------+         +-------+
                              |         |
                              v         v
                           dsa_port_vlan_add

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: tag_8021q: add a context structure
Vladimir Oltean [Thu, 10 Sep 2020 16:48:56 +0000 (19:48 +0300)]
net: dsa: tag_8021q: add a context structure

While working on another tag_8021q driver implementation, some things
became apparent:

- It is not mandatory for a DSA driver to offload the tag_8021q VLANs by
  using the VLAN table per se. For example, it can add custom TCAM rules
  that simply encapsulate RX traffic, and redirect & decapsulate rules
  for TX traffic. For such a driver, it makes no sense to receive the
  tag_8021q configuration through the same callback as it receives the
  VLAN configuration from the bridge and the 8021q modules.

- Currently, sja1105 (the only tag_8021q user) sets a
  priv->expect_dsa_8021q variable to distinguish between the bridge
  calling, and tag_8021q calling. That can be improved, to say the
  least.

- The crosschip bridging operations are, in fact, stateful already. The
  list of crosschip_links must be kept by the caller and passed to the
  relevant tag_8021q functions.

So it would be nice if the tag_8021q configuration was more
self-contained. This patch attempts to do that.

Create a struct dsa_8021q_context which encapsulates a struct
dsa_switch, and has 2 function pointers for adding and deleting a VLAN.
These will replace the previous channel to the driver, which was through
the .port_vlan_add and .port_vlan_del callbacks of dsa_switch_ops.

Also put the list of crosschip_links into this dsa_8021q_context.
Drivers that don't support cross-chip bridging can simply omit to
initialize this list, as long as they dont call any cross-chip function.

The sja1105_vlan_add and sja1105_vlan_del functions are refactored into
a smaller sja1105_vlan_add_one, which now has 2 entry points:
- sja1105_vlan_add, from struct dsa_switch_ops
- sja1105_dsa_8021q_vlan_add, from the tag_8021q ops
But even this change is fairly trivial. It just reflects the fact that
for sja1105, the VLANs from these 2 channels end up in the same hardware
table. However that is not necessarily true in the general sense (and
that's the reason for making this change).

The rest of the patch is mostly plain refactoring of "ds" -> "ctx". The
dsa_8021q_context structure needs to be propagated because adding a VLAN
is now done through the ops function pointers inside of it.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: tag_8021q: setup tagging via a single function call
Vladimir Oltean [Thu, 10 Sep 2020 16:48:55 +0000 (19:48 +0300)]
net: dsa: tag_8021q: setup tagging via a single function call

There is no point in calling dsa_port_setup_8021q_tagging for each
individual port. Additionally, it will become more difficult to do that
when we'll have a context structure to tag_8021q (next patch). So
refactor this now.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: tag_8021q: include missing refcount.h
Vladimir Oltean [Thu, 10 Sep 2020 16:48:54 +0000 (19:48 +0300)]
net: dsa: tag_8021q: include missing refcount.h

The previous assumption was that the caller would already have this
header file included.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agocrypto/chcr: move nic TLS functionality to drivers/net
Rohit Maheshwari [Thu, 10 Sep 2020 14:21:47 +0000 (19:51 +0530)]
crypto/chcr: move nic TLS functionality to drivers/net

This patch moves complete nic tls offload (kTLS) code from crypto
directory to drivers/net/ethernet/chelsio/inline_crypto/ch_ktls
directory. nic TLS is made a separate ULD of cxgb4.

Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'sfc-encap-offloads-on-EF10'
David S. Miller [Sat, 12 Sep 2020 00:15:22 +0000 (17:15 -0700)]
Merge branch 'sfc-encap-offloads-on-EF10'

Edward Cree says:

====================
sfc: encap offloads on EF10

EF10 NICs from the 8000 series onwards support TX offloads (checksumming,
 TSO) on VXLAN- and NVGRE-encapsulated packets.  This series adds driver
 support for these offloads.

Changes from v1:
 * Fix 'no TXQ of type' error handling in patch #1 (and clear up the
   misleading comment that inspired the wrong version)
 * Add comment in patch #5 explaining what the deal with TSOv3 is
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: advertise encapsulated offloads on EF10
Edward Cree [Fri, 11 Sep 2020 22:40:41 +0000 (23:40 +0100)]
sfc: advertise encapsulated offloads on EF10

Necessitates an .ndo_features_check, as the EF10 datapath has several
 limitations on what it can handle.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: implement encapsulated TSO on EF10
Edward Cree [Fri, 11 Sep 2020 22:40:14 +0000 (23:40 +0100)]
sfc: implement encapsulated TSO on EF10

>From the 8000 series onwards, EF10 NICs with suitable firmware are able
 to perform TSO within VXLAN or NVGRE encapsulation.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: de-indirect TSO handling
Edward Cree [Fri, 11 Sep 2020 22:40:03 +0000 (23:40 +0100)]
sfc: de-indirect TSO handling

Remove the tx_queue->handle_tso function pointer, and just use
 tx_queue->tso_version to decide which function to call, thus removing
 an indirect call from the fast path.
Instead of passing a tso_v2 flag to efx_mcdi_tx_init(), set the desired
 tx_queue->tso_version before calling it.
In efx_mcdi_tx_init(), report back failure to obtain a TSOv2 context by
 setting tx_queue->tso_version to 0, which will cause the TX path to
 use the GSO-based fallback.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: select inner-csum-offload TX queues for skbs that need it
Edward Cree [Fri, 11 Sep 2020 22:39:39 +0000 (23:39 +0100)]
sfc: select inner-csum-offload TX queues for skbs that need it

Won't actually be exercised until we start advertising the corresponding
 offload features.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: create inner-csum queues on EF10 if supported
Edward Cree [Fri, 11 Sep 2020 22:39:28 +0000 (23:39 +0100)]
sfc: create inner-csum queues on EF10 if supported

If the MC reports the VXLAN_NVGRE datapath capability, then these queues
 can be used for checksum offload of encapsulated packets.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: define inner/outer csum offload TXQ types
Edward Cree [Fri, 11 Sep 2020 22:39:14 +0000 (23:39 +0100)]
sfc: define inner/outer csum offload TXQ types

Nothing yet creates inner csum TXQs; just change all references to
 EFX_TXQ_TYPE_OFFLOAD to the new EFX_TXQ_TYPE_OUTER_CSUM.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: decouple TXQ type from label
Edward Cree [Fri, 11 Sep 2020 22:39:02 +0000 (23:39 +0100)]
sfc: decouple TXQ type from label

Make it possible to have an arbitrary mapping from types to labels,
 because when we add inner-csum-offload TXQs there will no longer be a
 convenient nesting hierarchy of NIC types (EF10 will have inner-csum
 TXQs, while Siena will have HIGHPRI).
Correct a misleading comment on efx_hard_start_xmit().

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoocteontx2-af: Constify npc_kpu_profile_{action,cam}
Rikard Falkeborn [Fri, 11 Sep 2020 22:00:15 +0000 (00:00 +0200)]
octeontx2-af: Constify npc_kpu_profile_{action,cam}

These are never modified, so constify them to allow the compiler to
place them in read-only memory. This moves about 25kB to read-only
memory as seen by the output of the size command.

Before:
   text    data     bss     dec     hex filename
 296203   65464    1248  362915   589a3 drivers/net/ethernet/marvell/octeontx2/af/octeontx2_af.ko

After:
   text    data     bss     dec     hex filename
 321003   40664    1248  362915   589a3 drivers/net/ethernet/marvell/octeontx2/af/octeontx2_af.ko

Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'sfc-misc-cleanups'
David S. Miller [Fri, 11 Sep 2020 21:55:14 +0000 (14:55 -0700)]
Merge branch 'sfc-misc-cleanups'

Edward Cree says:

====================
sfc: misc cleanups

Clean up a few nits I noticed while working on TXQ stuff.
====================

Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: cleanups around efx_alloc_channel
Edward Cree [Fri, 11 Sep 2020 18:45:14 +0000 (19:45 +0100)]
sfc: cleanups around efx_alloc_channel

The old_channel argument is never used, so remove it.
The function is only called from elsewhere in efx_channels.c, so make
 it static and remove the declaration from the header file.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: remove spurious unreachable return statement
Edward Cree [Fri, 11 Sep 2020 18:44:56 +0000 (19:44 +0100)]
sfc: remove spurious unreachable return statement

The statement above it already returns, so there is no way to get here.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: remove duplicate call to efx_init_channels from EF100 probe
Edward Cree [Fri, 11 Sep 2020 18:44:44 +0000 (19:44 +0100)]
sfc: remove duplicate call to efx_init_channels from EF100 probe

efx_init_struct already calls this, we don't need to do it again.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agobridge: mcast: Fix incomplete MDB dump
Ido Schimmel [Fri, 11 Sep 2020 13:24:47 +0000 (16:24 +0300)]
bridge: mcast: Fix incomplete MDB dump

Each MDB entry is encoded in a nested netlink attribute called
'MDBA_MDB_ENTRY'. In turn, this attribute contains another nested
attributed called 'MDBA_MDB_ENTRY_INFO', which encodes a single port
group entry within the MDB entry.

The cited commit added the ability to restart a dump from a specific
port group entry. However, on failure to add a port group entry to the
dump the entire MDB entry (stored in 'nest2') is removed, resulting in
missing port group entries.

Fix this by finalizing the MDB entry with the partial list of already
encoded port group entries.

Fixes: 5205e919c9f0 ("net: bridge: mcast: add support for src list and filter mode dumping")
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoipv6: remove redundant assignment to variable err
Colin Ian King [Fri, 11 Sep 2020 10:35:09 +0000 (11:35 +0100)]
ipv6: remove redundant assignment to variable err

The variable err is being initialized with a value that is never read and
it is being updated later with a new value. The initialization is redundant
and can be removed.  Also re-order variable declarations in reverse
Christmas tree ordering.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'ag71xx-add-ethtool-and-flow-control-support'
David S. Miller [Fri, 11 Sep 2020 21:45:55 +0000 (14:45 -0700)]
Merge branch 'ag71xx-add-ethtool-and-flow-control-support'

Oleksij Rempel says:

====================
ag71xx: add ethtool and flow control support

The main target of this patches is to provide flow control support
for ag71xx driver. To be able to validate this functionality, I also
added ethtool support with HW counters. So, this patches was validated
with iperf3 and counters showing Pause frames send or received by this
NIC.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ag71xx: add flow control support
Oleksij Rempel [Fri, 11 Sep 2020 08:25:28 +0000 (10:25 +0200)]
net: ag71xx: add flow control support

Add flow control support. The functionality was tested on AR9331 SoC and
confirmed by iperf3 results and HW counters exported over ethtool.
Following test configurations was used:

iMX6S receiver <--- TL-SG1005D switch <---- AR9331 sender

The switch is supporting symmytric flow control:
Settings for eth0:
        Supported ports: [ MII ]
        Supported link modes:   10baseT/Half 10baseT/Full
                                100baseT/Half 100baseT/Full
        Supported pause frame use: Symmetric Receive-only
        Supports auto-negotiation: Yes
        Supported FEC modes: Not reported
        Advertised link modes:  10baseT/Half 10baseT/Full
                                100baseT/Half 100baseT/Full
        Advertised pause frame use: Symmetric
        Advertised auto-negotiation: Yes
        Advertised FEC modes: Not reported
        Link partner advertised link modes:  10baseT/Half 10baseT/Full
                                             100baseT/Half 100baseT/Full
--->>   Link partner advertised pause frame use: Symmetric
        Link partner advertised auto-negotiation: Yes
        Link partner advertised FEC modes: Not reported
        Speed: 100Mb/s
        Duplex: Full
        Auto-negotiation: on
        Port: MII
        PHYAD: 4
        Transceiver: external
        Link detected: yes

The iMX6S system was configured to 10Mbit, to let the switch use flow
control:
  - ethtool -s eth0 speed 10

With flow control disabled on AR9331:
  - ethtool -A eth0  rx off tx off
  - iperf3 -u -c 172.17.0.1 -b100M -l1472 -t10

[ ID] Interval           Transfer     Bitrate         Jitter    Lost/Total Datagrams
[  5]   0.00-10.00  sec  66.2 MBytes  55.5 Mbits/sec  0.000 ms  0/47155 (0%)  sender
[  5]   0.00-10.04  sec  11.5 MBytes  9.57 Mbits/sec  1.309 ms  38986/47146 (83%)  receiver

With flow control enabled on AR9331:
  - ethtool -A eth0  rx on tx on
  - iperf3 -u -c 172.17.0.1 -b100M -l1472 -t10

[ ID] Interval           Transfer     Bitrate         Jitter    Lost/Total Datagrams
[  5]   0.00-10.00  sec  15.1 MBytes  12.6 Mbits/sec  0.000 ms  0/10727 (0%)  sender
[  5]   0.00-10.05  sec  11.5 MBytes  9.57 Mbits/sec  1.371 ms  2525/10689 (24%)  receiver

Similar results are get in opposite direction by introducing extra CPU
load on AR9331:
  - chrt 40 dd if=/dev/zero of=/dev/null &

Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ag71xx: add ethtool support
Oleksij Rempel [Fri, 11 Sep 2020 08:25:27 +0000 (10:25 +0200)]
net: ag71xx: add ethtool support

Add basic ethtool support. The functionality was tested on AR9331 SoC.

Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agodrivers/net/wan/x25_asy: Remove an unused flag "SLF_OUTWAIT"
Xie He [Fri, 11 Sep 2020 06:35:03 +0000 (23:35 -0700)]
drivers/net/wan/x25_asy: Remove an unused flag "SLF_OUTWAIT"

The "SLF_OUTWAIT" flag defined in x25_asy.h is not actually used.
It is only cleared at one place in x25_asy.c but is never read or set.
So we can remove it.

Signed-off-by: Xie He <xie.he.0141@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: stmmac: set get_rx_header_len() as void for it didn't have any error code to...
Luo Jiaxing [Fri, 11 Sep 2020 03:55:58 +0000 (11:55 +0800)]
net: stmmac: set get_rx_header_len() as void for it didn't have any error code to return

We found the following warning when using W=1 to build kernel:

drivers/net/ethernet/stmicro/stmmac/stmmac_main.c:3634:6: warning: variable â€˜ret’ set but not used [-Wunused-but-set-variable]
int ret, coe = priv->hw->rx_csum;

When digging stmmac_get_rx_header_len(), dwmac4_get_rx_header_len() and
dwxgmac2_get_rx_header_len() return 0 only, without any error code to
report. Therefore, it's better to define get_rx_header_len() as void.

Signed-off-by: Luo Jiaxing <luojiaxing@huawei.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'Add-GVE-Features'
David S. Miller [Fri, 11 Sep 2020 21:31:54 +0000 (14:31 -0700)]
Merge branch 'Add-GVE-Features'

David Awogbemila says:

====================
Add GVE Features.

Note: Patch 4 in v3 was dropped.

Patch 4 (patch 5 in v3): Start/stop timer only when report stats is
enabled/disabled.
Patch 7 (patch 8 in v3): Use netdev_info, not dev_info, to log
device link status.
====================

Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agogve: Enable Link Speed Reporting in the driver.
David Awogbemila [Fri, 11 Sep 2020 17:38:51 +0000 (10:38 -0700)]
gve: Enable Link Speed Reporting in the driver.

This change allows the driver to report the device link speed
when the ethtool command:
ethtool <nic name>
is run.
Getting the link speed is done via a new admin queue command:
ReportLinkSpeed.

Reviewed-by: Yangchun Fu <yangchun@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agogve: Use link status register to report link status
Patricio Noyola [Fri, 11 Sep 2020 17:38:50 +0000 (10:38 -0700)]
gve: Use link status register to report link status

This makes the driver better aware of the connectivity status of the
device. Based on the device's status register, the driver can call
netif_carrier_{on,off}.

Reviewed-by: Yangchun Fu <yangchun@google.com>
Signed-off-by: Patricio Noyola <patricion@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agogve: Batch AQ commands for creating and destroying queues.
Sagi Shahar [Fri, 11 Sep 2020 17:38:49 +0000 (10:38 -0700)]
gve: Batch AQ commands for creating and destroying queues.

Adds support for batching AQ commands and uses it for creating and
destroying queues.

Reviewed-by: Yangchun Fu <yangchun@google.com>
Signed-off-by: Sagi Shahar <sagis@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agogve: NIC stats for report-stats and for ethtool
David Awogbemila [Fri, 11 Sep 2020 17:38:48 +0000 (10:38 -0700)]
gve: NIC stats for report-stats and for ethtool

This adds per queue NIC stats to ethtool stats and to report-stats.
These stats are always exposed to guest whether or not the
report-stats flag is turned on.

Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agogve: Add Gvnic stats AQ command and ethtool show/set-priv-flags.
Kuo Zhao [Fri, 11 Sep 2020 17:38:47 +0000 (10:38 -0700)]
gve: Add Gvnic stats AQ command and ethtool show/set-priv-flags.

This adds functionality to report driver stats to Hypervisor.
(Users may want to turn this feature off as a matter of privacy
so a "report-stats" flag is added as an ethtool priv option.
It is also disabled by default.)
The hypervisor would trigger a stats report in case "too many"
packets dropped; the stats would be useful in debugging stuck
queues.
A "stats_report_trigger_cnt" stat is added to count the number of times
the hypervisor attempts to trigger stats report.

A timer is also added so that when report-stats is enabled, stat are
updated once every 20 seconds.

Reviewed-by: Yangchun Fu <yangchun@google.com>
Signed-off-by: Kuo Zhao <kuozhao@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agogve: Use dev_info/err instead of netif_info/err.
Catherine Sullivan [Fri, 11 Sep 2020 17:38:46 +0000 (10:38 -0700)]
gve: Use dev_info/err instead of netif_info/err.

Update the driver to use dev_info/err instead of netif_info/err.

Reviewed-by: Yangchun Fu <yangchun@google.com>
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agogve: Add stats for gve.
Kuo Zhao [Fri, 11 Sep 2020 17:38:45 +0000 (10:38 -0700)]
gve: Add stats for gve.

Sample output of "ethtool -S <interface-name>" with 1 RX queue and 1 TX
queue:
NIC statistics:
     rx_packets: 1039
     tx_packets: 37
     rx_bytes: 822071
     tx_bytes: 4100
     rx_dropped: 0
     tx_dropped: 0
     tx_timeouts: 0
     rx_skb_alloc_fail: 0
     rx_buf_alloc_fail: 0
     rx_desc_err_dropped_pkt: 0
     interface_up_cnt: 1
     interface_down_cnt: 0
     reset_cnt: 0
     page_alloc_fail: 0
     dma_mapping_error: 0
     rx_posted_desc[0]: 1365
     rx_completed_desc[0]: 341
     rx_bytes[0]: 215094
     rx_dropped_pkt[0]: 0
     rx_copybreak_pkt[0]: 3
     rx_copied_pkt[0]: 3
     tx_posted_desc[0]: 6
     tx_completed_desc[0]: 6
     tx_bytes[0]: 420
     tx_wake[0]: 0
     tx_stop[0]: 0
     tx_event_counter[0]: 6
     adminq_prod_cnt: 34
     adminq_cmd_fail: 0
     adminq_timeouts: 0
     adminq_describe_device_cnt: 1
     adminq_cfg_device_resources_cnt: 1
     adminq_register_page_list_cnt: 16
     adminq_unregister_page_list_cnt: 0
     adminq_create_tx_queue_cnt: 8
     adminq_create_rx_queue_cnt: 8
     adminq_destroy_tx_queue_cnt: 0
     adminq_destroy_rx_queue_cnt: 0
     adminq_dcfg_device_resources_cnt: 0
     adminq_set_driver_parameter_cnt: 0

Reviewed-by: Yangchun Fu <yangchun@google.com>
Signed-off-by: Kuo Zhao <kuozhao@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agogve: Get and set Rx copybreak via ethtool
Kuo Zhao [Fri, 11 Sep 2020 17:38:44 +0000 (10:38 -0700)]
gve: Get and set Rx copybreak via ethtool

This adds support for getting and setting the RX copybreak
value via ethtool.

Reviewed-by: Yangchun Fu <yangchun@google.com>
Signed-off-by: Kuo Zhao <kuozhao@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge tag 'wireless-drivers-next-2020-09-11' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Fri, 11 Sep 2020 20:32:31 +0000 (13:32 -0700)]
Merge tag 'wireless-drivers-next-2020-09-11' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
wireless-drivers-next patches for v5.10

First set of patches for v5.10. Most noteworthy here is ath11k getting
initial support for QCA6390 and IPQ6018 devices. But most of the
patches are cleanup: W=1 warning fixes, fallthrough keywords, DMA API
changes and tasklet API changes.

Major changes:

ath10k

* support SDIO firmware codedumps

* support station specific TID configurations

ath11k

* add support for IPQ6018

* add support for QCA6390 PCI devices

ath9k

* add support for NL80211_EXT_FEATURE_CAN_REPLACE_PTK0 to improve PTK0
  rekeying

wcn36xx

* add support for TX ack
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge ath-next from git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git
Kalle Valo [Fri, 11 Sep 2020 15:03:00 +0000 (18:03 +0300)]
Merge ath-next from git://git./linux/kernel/git/kvalo/ath.git

ath.git patches for v5.10. Major changes:

ath10k

* support SDIO firmware codedumps

* support station specific TID configurations

ath11k

* add support for IPQ6018

3 years agoath10k: Remove unused macro ATH10K_ROC_TIMEOUT_HZ
YueHaibing [Wed, 9 Sep 2020 13:58:34 +0000 (21:58 +0800)]
ath10k: Remove unused macro ATH10K_ROC_TIMEOUT_HZ

There is no caller in tree, so can remove it.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200909135834.38448-1-yuehaibing@huawei.com
3 years agoath11k: Remove unused inline function htt_htt_stats_debug_dump()
YueHaibing [Wed, 9 Sep 2020 13:45:33 +0000 (21:45 +0800)]
ath11k: Remove unused inline function htt_htt_stats_debug_dump()

There is no caller in tree, so can remove it.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200909134533.19604-1-yuehaibing@huawei.com
3 years agoath11k: fix link error when CONFIG_REMOTEPROC is disabled
Kalle Valo [Thu, 10 Sep 2020 07:20:59 +0000 (07:20 +0000)]
ath11k: fix link error when CONFIG_REMOTEPROC is disabled

If CONFIG_REMOTEPROC was disabled the linking failed with:

ERROR: modpost: "rproc_get_by_phandle" [drivers/net/wireless/ath/ath11k/ath11k.ko] undefined!

Compile tested only.

Reported-by: Randy Dunlap <rdunlap@infradead.org>
Fixes: 1ff8ed786d5d ("ath11k: use remoteproc only with AHB devices")
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/0101017476e38f40-c4168ac4-c00a-4220-a032-fe17e4a157cb-000000@us-west-2.amazonses.com
3 years agoath11k: remove calling ath11k_init_hw_params() second time
Anilkumar Kolli [Tue, 8 Sep 2020 10:01:59 +0000 (10:01 +0000)]
ath11k: remove calling ath11k_init_hw_params() second time

During probe ath11k_init_hw_params() is called from ath11k_core_pre_init()
and is not needed agian in ath11k_core_init().

Tested on: IPQ8074 hw2.0 AHB WLAN.HK.2.4.0.1-00009-QCAHKSWPL_SILICONZ-1

Fixes: 1ff8ed786d5d (ath11k: use remoteproc only with AHB devices)
Signed-off-by: Anilkumar Kolli <akolli@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/010101746d2a40d3-25cd7dbe-c0dd-4fdf-8735-366d7fb40207-000000@us-west-2.amazonses.com
3 years agoath11k: add raw mode and software crypto support
Venkateswara Naralasetty [Tue, 8 Sep 2020 06:32:21 +0000 (06:32 +0000)]
ath11k: add raw mode and software crypto support

Adding raw mode tx/rx support. Also, adding support
for software crypto which depends on raw mode.

To enable raw mode tx/rx:
insmod ath11k.ko frame_mode=0

To enable software crypto:
insmod ath11k.ko crypto_mode=1

These modes could be helpful in debugging crypto related issues.

Tested-on: IPQ8074 WLAN.HK.2.1.0.1-01228-QCAHKSWPL_SILICONZ-1

Signed-off-by: Manikanta Pubbisetty <mpubbise@codeaurora.org>
Signed-off-by: Venkateswara Naralasetty <vnaralas@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/010101746c6a52d9-18302a2c-0d6d-4057-aa4b-95960c809646-000000@us-west-2.amazonses.com
3 years agoath11k: add ipq6018 support
Anilkumar Kolli [Tue, 8 Sep 2020 07:55:36 +0000 (07:55 +0000)]
ath11k: add ipq6018 support

IPQ6018 has one 5G and one 2G radio with 2x2,
shares ipq8074 configurations.

Tested on: IPQ6018 hw1.0 AHB WLAN.HK.2.2-02134-QCAHKSWPL_SILICONZ-1
Tested on: IPQ8074 hw2.0 AHB WLAN.HK.2.4.0.1-00009-QCAHKSWPL_SILICONZ-1

Signed-off-by: Anilkumar Kolli <akolli@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/010101746cb68b63-c2bc31ec-a31e-442e-a572-26f4c045c06b-000000@us-west-2.amazonses.com
3 years agoath11k: move target ce configs to hw_params
Anilkumar Kolli [Tue, 8 Sep 2020 07:55:35 +0000 (07:55 +0000)]
ath11k: move target ce configs to hw_params

Move target CE config and target CE service config to hw_params.
No functional changes.

Tested on: IPQ8074 hw2.0 AHB WLAN.HK.2.4.0.1-00009-QCAHKSWPL_SILICONZ-1

Signed-off-by: Anilkumar Kolli <akolli@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/010101746cb685d9-6bedeccb-29a1-4d32-8664-fcfe7d105f4a-000000@us-west-2.amazonses.com
3 years agodt: bindings: net: update compatible for ath11k
Anilkumar Kolli [Tue, 8 Sep 2020 07:55:31 +0000 (07:55 +0000)]
dt: bindings: net: update compatible for ath11k

Add IPQ6018 wireless driver support,
its based on ath11k driver.

Signed-off-by: Anilkumar Kolli <akolli@codeaurora.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/010101746cb6751a-ca300933-1174-4534-a01b-b1dbf1c1f305-000000@us-west-2.amazonses.com
3 years agonet: smc91x: Remove set but not used variable 'status' in smc_phy_configure()
Luo Jiaxing [Thu, 10 Sep 2020 13:12:16 +0000 (21:12 +0800)]
net: smc91x: Remove set but not used variable 'status' in smc_phy_configure()

Fixes the following warning when using W=1 to build kernel:

drivers/net/ethernet/smsc/smc91x.c: In function â€˜smc_phy_configure’:
drivers/net/ethernet/smsc/smc91x.c:1039:6: warning: variable â€˜status’ set but not used [-Wunused-but-set-variable]
  int status;

Signed-off-by: Luo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'smc-next'
David S. Miller [Thu, 10 Sep 2020 22:24:27 +0000 (15:24 -0700)]
Merge branch 'smc-next'

Karsten Graul says:

====================
net/smc: updates 2020-09-10

Please apply the following patch series for smc to netdev's net-next tree.

This patch series is a mix of various improvements and cleanups.
The patches 1 and 10 improve the handling of large parallel workloads.
Patch 8 corrects a kernel config default for config CCWGROUP on s390.
Patch 9 allows userspace tools to retrieve socket information for more
sockets.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/smc: use separate work queues for different worker types
Karsten Graul [Thu, 10 Sep 2020 16:48:29 +0000 (18:48 +0200)]
net/smc: use separate work queues for different worker types

There are 6 types of workers which exist per smc connection. 3 of them
are used for listen and handshake processing, another 2 are used for
close and abort processing and 1 is the tx worker that moves calls to
sleeping functions into a worker.
To prevent flooding of the system work queue when many connections are
opened or closed at the same time (some pattern uperf implements), move
those workers to one of 3 smc-specific work queues. Two work queues are
module-global and used for handshake and close workers. The third work
queue is defined per link group and used by the tx workers that may
sleep waiting for resources of this link group.
And in smc_llc_enqueue() queue the llc_event_work work to the system
prio work queue because its critical that this work is started fast.

Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/smc: use the retry mechanism for netlink messages
Guvenc Gulce [Thu, 10 Sep 2020 16:48:28 +0000 (18:48 +0200)]
net/smc: use the retry mechanism for netlink messages

When the netlink messages to be sent to the userspace
are too big for a single netlink message, send them in
chunks using the netlink_dump infrastructure. Modify the
smc diag dump code so that it can signal to the netlink_dump
infrastructure that it needs to send more data.

Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agos390/net: add SMC config as one of the defaults of CCWGROUP
Guvenc Gulce [Thu, 10 Sep 2020 16:48:27 +0000 (18:48 +0200)]
s390/net: add SMC config as one of the defaults of CCWGROUP

arch/s390/net/pnet.c uses ccwgroup function dev_is_ccwgroup()
in pnetid_by_dev_port().
For s390 the net/smc code makes use of function pnetid_by_dev_port().
Make sure ccwgroup is built into the kernel, if smc is to be built
into the kernel.

Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/smc: immediate freeing in smc_lgr_cleanup_early()
Ursula Braun [Thu, 10 Sep 2020 16:48:26 +0000 (18:48 +0200)]
net/smc: immediate freeing in smc_lgr_cleanup_early()

smc_lgr_cleanup_early() schedules the free worker with delay. DMB
unregistering occurs in this delayed worker increasing the risk
to reach the SMCD SBA limit without need. Terminate the
linkgroup immediately, since termination means early DMB unregistering.

For SMCD the global smc_server_lgr_pending lock is given up early.
A linkgroup to be given up with smc_lgr_cleanup_early() may already
contain more than one connection. Using __smc_lgr_terminate() in
smc_lgr_cleanup_early() covers this.

And consolidate smc_ism_put_vlan() and smc_put_device() into smc_lgr_free()
only.

Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/smc: reduce smc_listen_decline() calls
Ursula Braun [Thu, 10 Sep 2020 16:48:25 +0000 (18:48 +0200)]
net/smc: reduce smc_listen_decline() calls

smc_listen_work() contains already an smc_listen_decline() exit.
Use this exit for smc_listen_rdma_finish() problems as well.

Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/smc: improve server ISM device determination
Ursula Braun [Thu, 10 Sep 2020 16:48:24 +0000 (18:48 +0200)]
net/smc: improve server ISM device determination

Move check whether peer can be reached into smc_pnet_find_ism_by_pnetid().
Thus searching continues for another ism device, if check fails.

Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/smc: common routine for CLC accept and confirm
Ursula Braun [Thu, 10 Sep 2020 16:48:23 +0000 (18:48 +0200)]
net/smc: common routine for CLC accept and confirm

smc_clc_send_accept() and smc_clc_send_confirm() are quite similar.
Move common code into a separate function smc_clc_send_confirm_accept().
And introduce separate SMCD and SMCR struct definitions for CLC accept
resp. confirm.
No functional change.

Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/smc: dynamic allocation of CLC proposal buffer
Ursula Braun [Thu, 10 Sep 2020 16:48:22 +0000 (18:48 +0200)]
net/smc: dynamic allocation of CLC proposal buffer

Reduce stack size for smc_listen_work() and smc_clc_send_proposal()
by dynamic allocation of the CLC buffer to be received or sent.

Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/smc: introduce better field names
Ursula Braun [Thu, 10 Sep 2020 16:48:21 +0000 (18:48 +0200)]
net/smc: introduce better field names

Field names "srv_first_contact" and "cln_first_contact" are misleading,
since they apply to both, server and client. Rename them to
"first_contact_peer" and "first_contact_local".
Rename "ism_gid" by the more precise name "ism_peer_gid".
Rename version constant "SMC_CLC_V1" into "SMC_V1".
No functional change.

Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/smc: reduce active tcp_listen workers
Ursula Braun [Thu, 10 Sep 2020 16:48:20 +0000 (18:48 +0200)]
net/smc: reduce active tcp_listen workers

SMC starts a separate tcp_listen worker for every SMC socket in
state SMC_LISTEN, and can accept an incoming connection request only,
if this worker is really running and waiting in kernel_accept(). But
the number of running workers is limited.
This patch reworks the listening SMC code and starts a tcp_listen worker
after the SYN-ACK handshake on the internal clc-socket only.

Suggested-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Reviewed-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'nfc-s3fwrn5-Few-cleanups'
David S. Miller [Thu, 10 Sep 2020 22:22:17 +0000 (15:22 -0700)]
Merge branch 'nfc-s3fwrn5-Few-cleanups'

Krzysztof Kozlowski says:

====================
nfc: s3fwrn5: Few cleanups

Changes since v2:
1. Fix dtschema ID after rename (patch 1/8).
2. Apply patch 9/9 (defconfig change).

Changes since v1:
1. Rename dtschema file and add additionalProperties:false, as Rob
   suggested,
2. Add Marek's tested-by,
3. New patches: #4, #5, #6, #7 and #9.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoarm64: dts: exynos: Use newer S3FWRN5 GPIO properties in Exynos5433 TM2
Krzysztof Kozlowski [Thu, 10 Sep 2020 16:12:19 +0000 (18:12 +0200)]
arm64: dts: exynos: Use newer S3FWRN5 GPIO properties in Exynos5433 TM2

Since "s3fwrn5" is not a valid vendor prefix, use new GPIO properties
instead of the deprecated.

Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMAINTAINERS: Add Krzysztof Kozlowski to Samsung S3FWRN5 and remove Robert
Krzysztof Kozlowski [Thu, 10 Sep 2020 16:12:18 +0000 (18:12 +0200)]
MAINTAINERS: Add Krzysztof Kozlowski to Samsung S3FWRN5 and remove Robert

Robert BaƂdyga's email does not work (bounces) since 2016 so remove it.
Additionally there are no review/ack/tested tags from Krzysztof Opasiak
so it looks like the driver is not supported.

As a maintainer of Samsung ARM/ARM64 SoC, I can take care about this
driver and provide some review.  However clearly driver is not in
supported mode as I do not work in Samsung anymore.

Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonfc: s3fwrn5: Constify s3fwrn5_fw_info when not modified
Krzysztof Kozlowski [Thu, 10 Sep 2020 16:12:17 +0000 (18:12 +0200)]
nfc: s3fwrn5: Constify s3fwrn5_fw_info when not modified

Two functions accept pointer to struct s3fwrn5_fw_info but do not
modify the contents.  Make them const so the code is a little bit safer.

Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonfc: s3fwrn5: Add missing CRYPTO_HASH dependency
Krzysztof Kozlowski [Thu, 10 Sep 2020 16:12:16 +0000 (18:12 +0200)]
nfc: s3fwrn5: Add missing CRYPTO_HASH dependency

The driver uses crypto hash functions so it needs to select CRYPTO_HASH.
This fixes build errors:

  arc-linux-ld: drivers/nfc/s3fwrn5/firmware.o: in function `s3fwrn5_fw_download':
  firmware.c:(.text+0x152): undefined reference to `crypto_alloc_shash'

Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonfc: s3fwrn5: Remove unneeded 'ret' variable
Krzysztof Kozlowski [Thu, 10 Sep 2020 16:12:15 +0000 (18:12 +0200)]
nfc: s3fwrn5: Remove unneeded 'ret' variable

The local variable 'ret' can be removed:

  drivers/nfc/s3fwrn5/i2c.c:167:6: warning: variable 'ret' set but not used [-Wunused-but-set-variable]

Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonfc: s3fwrn5: Remove wrong vendor prefix from GPIOs
Krzysztof Kozlowski [Thu, 10 Sep 2020 16:12:14 +0000 (18:12 +0200)]
nfc: s3fwrn5: Remove wrong vendor prefix from GPIOs

The device tree property prefix describes the vendor, which in case of
S3FWRN5 chip is Samsung.  Therefore the "s3fwrn5" prefix for "en-gpios"
and "fw-gpios" is not correct and should be deprecated.  Introduce
properly named properties for these GPIOs but still support deprecated
ones.

Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agodt-bindings: net: nfc: s3fwrn5: Remove wrong vendor prefix from GPIOs
Krzysztof Kozlowski [Thu, 10 Sep 2020 16:12:13 +0000 (18:12 +0200)]
dt-bindings: net: nfc: s3fwrn5: Remove wrong vendor prefix from GPIOs

The device tree property prefix describes the vendor, which in case of
S3FWRN5 chip is Samsung.  Therefore the "s3fwrn5" prefix for "en-gpios"
and "fw-gpios" is not correct and should be deprecated.  Introduce
properly named properties for these GPIOs and rename the fw-gpios" to
"wake-gpios" to better describe its purpose.

Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agodt-bindings: net: nfc: s3fwrn5: Convert to dtschema
Krzysztof Kozlowski [Thu, 10 Sep 2020 16:12:12 +0000 (18:12 +0200)]
dt-bindings: net: nfc: s3fwrn5: Convert to dtschema

Convert the Samsung S3FWRN5 NCI NFC controller bindings to dtschema.
This is conversion only so it includes properties with invalid prefixes
(s3fwrn5,en-gpios) which should be addressed later.

Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'hns-kdoc'
David S. Miller [Thu, 10 Sep 2020 22:18:19 +0000 (15:18 -0700)]
Merge branch 'hns-kdoc'

Wang Hai says:

====================
Fix some kernel-doc warnings for hns.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns: Fix a kernel-doc warning in hinic_hw_eqs.c
Wang Hai [Thu, 10 Sep 2020 14:56:20 +0000 (22:56 +0800)]
net: hns: Fix a kernel-doc warning in hinic_hw_eqs.c

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

drivers/net/ethernet/huawei/hinic/hinic_hw_eqs.c:115: warning: Excess function parameter 'hw_handler' description in 'hinic_aeq_register_hw_cb'

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns: Fix a kernel-doc warning in hinic_hw_api_cmd.c
Wang Hai [Thu, 10 Sep 2020 14:56:19 +0000 (22:56 +0800)]
net: hns: Fix a kernel-doc warning in hinic_hw_api_cmd.c

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

drivers/net/ethernet/huawei/hinic/hinic_hw_api_cmd.c:382: warning: Excess function parameter 'size' description in 'api_cmd'

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns: Fix some kernel-doc warnings in hns_enet.c
Wang Hai [Thu, 10 Sep 2020 14:56:18 +0000 (22:56 +0800)]
net: hns: Fix some kernel-doc warnings in hns_enet.c

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

drivers/net/ethernet/hisilicon/hns/hns_enet.c:1841: warning: Excess function parameter 'netdev' description in 'hns_set_multicast_list'
drivers/net/ethernet/hisilicon/hns/hns_enet.c:1841: warning: Excess function parameter 'p' description in 'hns_set_multicast_list'

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns: Fix some kernel-doc warnings in hns_dsaf_xgmac.c
Wang Hai [Thu, 10 Sep 2020 14:56:17 +0000 (22:56 +0800)]
net: hns: Fix some kernel-doc warnings in hns_dsaf_xgmac.c

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

drivers/net/ethernet/hisilicon/hns/hns_dsaf_xgmac.c:137: warning: Excess function parameter 'drv' description in 'hns_xgmac_enable'
drivers/net/ethernet/hisilicon/hns/hns_dsaf_xgmac.c:497: warning: Excess function parameter 'cmd' description in 'hns_xgmac_get_regs'

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns: fix 'cdev' kernel-doc warning in hnae_ae_unregister()
Wang Hai [Thu, 10 Sep 2020 14:56:16 +0000 (22:56 +0800)]
net: hns: fix 'cdev' kernel-doc warning in hnae_ae_unregister()

Rename cdev to hdev.

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

drivers/net/ethernet/hisilicon/hns/hnae.c:444: warning: Excess function parameter 'cdev' description in 'hnae_ae_unregister'

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agohinic: Fix some kernel-doc warnings in hinic_hw_io.c
Wang Hai [Thu, 10 Sep 2020 14:56:15 +0000 (22:56 +0800)]
hinic: Fix some kernel-doc warnings in hinic_hw_io.c

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

drivers/net/ethernet/huawei/hinic/hinic_hw_io.c:373: warning: Excess function parameter 'sq_msix_entry' description in 'hinic_io_create_qps'
drivers/net/ethernet/huawei/hinic/hinic_hw_io.c:373: warning: Excess function parameter 'rq_msix_entry' description in 'hinic_io_create_qps'

Rename these wrong names.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: mvpp2: ptp: Fix unused variables
Alex Dewar [Thu, 10 Sep 2020 13:49:10 +0000 (14:49 +0100)]
net: mvpp2: ptp: Fix unused variables

In the functions mvpp2_isr_handle_xlg() and
mvpp2_isr_handle_gmac_internal(), the bool variable link is assigned a
true value in the case that a given bit of val is set. However, if the
bit is unset, no value is assigned to link and it is then passed to
mvpp2_isr_handle_link() without being initialised. Fix by assigning to
link the value of the bit test.

Build-tested on x86.

Fixes: 36cfd3a6e52b ("net: mvpp2: restructure "link status" interrupt handling")
Signed-off-by: Alex Dewar <alex.dewar90@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: cxgb3: Fix some kernel-doc warnings
Wang Hai [Thu, 10 Sep 2020 13:36:16 +0000 (21:36 +0800)]
net: cxgb3: Fix some kernel-doc warnings

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

drivers/net/ethernet/chelsio/cxgb3/t3_hw.c:2209: warning: Excess function parameter 'adapter' description in 'clear_sge_ctxt'
drivers/net/ethernet/chelsio/cxgb3/t3_hw.c:2975: warning: Excess function parameter 'adapter' description in 't3_set_proto_sram'

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'Enhance-current-features-in-ena-driver'
David S. Miller [Thu, 10 Sep 2020 22:12:27 +0000 (15:12 -0700)]
Merge branch 'Enhance-current-features-in-ena-driver'

Sameeh Jubran says:

====================
Enhance current features in ena driver

This series adds the following:
* Exposes new device stats using ethtool.
* Adds and exposes the stats of xdp TX queues through ethtool.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ena: xdp: add queue counters for xdp actions
Sameeh Jubran [Thu, 10 Sep 2020 13:07:13 +0000 (13:07 +0000)]
net: ena: xdp: add queue counters for xdp actions

When using XDP every ingress packet is passed to an eBPF (xdp) program
which returns an action for this packet.

This patch adds counters for the number of times each such action was
received. It also counts all the invalid actions received from the eBPF
program.

Signed-off-by: Shay Agroskin <shayagr@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ena: ethtool: add stats printing to XDP queues
Sameeh Jubran [Thu, 10 Sep 2020 13:07:12 +0000 (13:07 +0000)]
net: ena: ethtool: add stats printing to XDP queues

Added statistics for TX queues that are used for XDP TX. The statistics
are the same as the ones printed for regular non-XDP TX queues.

The XDP queue statistics can be queried using
`ethtool -S <ifname>`

Signed-off-by: Shay Agroskin <shayagr@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ena: ethtool: Add new device statistics
Sameeh Jubran [Thu, 10 Sep 2020 13:07:11 +0000 (13:07 +0000)]
net: ena: ethtool: Add new device statistics

The new metrics provide granular visibility along multiple network
dimensions and enable troubleshooting and remediation of issues caused
by instances exceeding network performance allowances.

The new statistics can be queried using ethtool command.

Signed-off-by: Guy Tzalik <gtzalik@amazon.com>
Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ena: ethtool: convert stat_offset to 64 bit resolution
Sameeh Jubran [Thu, 10 Sep 2020 13:07:10 +0000 (13:07 +0000)]
net: ena: ethtool: convert stat_offset to 64 bit resolution

The type of all stat fields is u64, therefore when iterating over stat
fields in a stats struct, it makes sense to use an offset in 64 bit
resolution. Doing so allows us to drop some of the casting that is
currently used when referencing stats.

Signed-off-by: Sameeh Jubran <sameehj@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests/mptcp: Better delay & reordering configuration
Christoph Paasch [Thu, 10 Sep 2020 05:07:36 +0000 (22:07 -0700)]
selftests/mptcp: Better delay & reordering configuration

The delay was intended to be configured to "simulate" a high(er) BDP
link. As such, it needs to be set as part of the loss-configuration and
not as part of the netem reordering configuration.

The reordering-config also requires a delay but that delay is the
reordering-extend. So, a good approach is to set the reordering-extend
as a function of the configured latency. E.g., 25% of the overall
latency.

To speed up the selftests, we limit the delay to 50ms maximum to avoid
having the selftests run for too long.

Finally, the intention of tc_reorder was that when it is unset, the test
picks a random configuration. However, currently it is always initialized
and thus the random config won't be picked up.

Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/6
Reported-and-reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Christoph Paasch <cpaasch@apple.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'tcp-add-tos-reflection-feature'
David S. Miller [Thu, 10 Sep 2020 20:15:40 +0000 (13:15 -0700)]
Merge branch 'tcp-add-tos-reflection-feature'

Wei Wang says:

====================
tcp: add tos reflection feature

This patch series adds a new tcp feature to reflect TOS value received in
SYN, and send it out in SYN-ACK, and eventually set the TOS value of the
established socket with this reflected TOS value. This provides a way to
set the traffic class/QoS level for all traffic in the same connection
to be the same as the incoming SYN. It could be useful for datacenters
to provide equivalent QoS according to the incoming request.
This feature is guarded by /proc/sys/net/ipv4/tcp_reflect_tos, and is by
default turned off.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agotcp: reflect tos value received in SYN to the socket
Wei Wang [Thu, 10 Sep 2020 00:50:48 +0000 (17:50 -0700)]
tcp: reflect tos value received in SYN to the socket

This commit adds a new TCP feature to reflect the tos value received in
SYN, and send it out on the SYN-ACK, and eventually set the tos value of
the established socket with this reflected tos value. This provides a
way to set the traffic class/QoS level for all traffic in the same
connection to be the same as the incoming SYN request. It could be
useful in data centers to provide equivalent QoS according to the
incoming request.
This feature is guarded by /proc/sys/net/ipv4/tcp_reflect_tos, and is by
default turned off.

Signed-off-by: Wei Wang <weiwan@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoip: pass tos into ip_build_and_send_pkt()
Wei Wang [Thu, 10 Sep 2020 00:50:47 +0000 (17:50 -0700)]
ip: pass tos into ip_build_and_send_pkt()

This commit adds tos as a new passed in parameter to
ip_build_and_send_pkt() which will be used in the later commit.
This is a pure restructure and does not have any functional change.

Signed-off-by: Wei Wang <weiwan@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agotcp: record received TOS value in the request socket
Wei Wang [Thu, 10 Sep 2020 00:50:46 +0000 (17:50 -0700)]
tcp: record received TOS value in the request socket

A new field is added to the request sock to record the TOS value
received on the listening socket during 3WHS:
When not under syn flood, it is recording the TOS value sent in SYN.
When under syn flood, it is recording the TOS value sent in the ACK.
This is a preparation patch in order to do TOS reflection in the later
commit.

Signed-off-by: Wei Wang <weiwan@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: mventa: drop mvneta_stats from mvneta_swbm_rx_frame signature
Lorenzo Bianconi [Wed, 9 Sep 2020 21:05:23 +0000 (23:05 +0200)]
net: mventa: drop mvneta_stats from mvneta_swbm_rx_frame signature

Remove mvneta_stats from mvneta_swbm_rx_frame signature since now stats
are accounted in mvneta_run_xdp routine

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'netpoll-make-sure-napi_list-is-safe-for-RCU-traversal'
David S. Miller [Thu, 10 Sep 2020 20:08:47 +0000 (13:08 -0700)]
Merge branch 'netpoll-make-sure-napi_list-is-safe-for-RCU-traversal'

Jakub Kicinski says:

====================
netpoll: make sure napi_list is safe for RCU traversal

This series is a follow-up to the fix in commit 96e97bc07e90 ("net:
disable netpoll on fresh napis"). To avoid any latent race conditions
convert dev->napi_list to a proper RCU list. We need minor restructuring
because it looks like netif_napi_del() used to be idempotent, and
it may be quite hard to track down everyone who depends on that.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: make sure napi_list is safe for RCU traversal
Jakub Kicinski [Wed, 9 Sep 2020 17:37:53 +0000 (10:37 -0700)]
net: make sure napi_list is safe for RCU traversal

netpoll needs to traverse dev->napi_list under RCU, make
sure it uses the right iterator and that removal from this
list is handled safely.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: manage napi add/del idempotence explicitly
Jakub Kicinski [Wed, 9 Sep 2020 17:37:52 +0000 (10:37 -0700)]
net: manage napi add/del idempotence explicitly

To RCUify napi->dev_list we need to replace list_del_init()
with list_del_rcu(). There is no _init() version for RCU for
obvious reasons. Up until now netif_napi_del() was idempotent
so to make sure it remains such add a bit which is set when
NAPI is listed, and cleared when it removed. Since we don't
expect multiple calls to netif_napi_add() to be correct,
add a warning on that side.

Now that napi_hash_add / napi_hash_del are only called by
napi_add / del we can actually steal its bit. We just need
to make sure hash node is initialized correctly.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: remove napi_hash_del() from driver-facing API
Jakub Kicinski [Wed, 9 Sep 2020 17:37:51 +0000 (10:37 -0700)]
net: remove napi_hash_del() from driver-facing API

We allow drivers to call napi_hash_del() before calling
netif_napi_del() to batch RCU grace periods. This makes
the API asymmetric and leaks internal implementation details.
Soon we will want the grace period to protect more than just
the NAPI hash table.

Restructure the API and have drivers call a new function -
__netif_napi_del() if they want to take care of RCU waits.

Note that only core was checking the return status from
napi_hash_del() so the new helper does not report if the
NAPI was actually deleted.

Some notes on driver oddness:
 - veth observed the grace period before calling netif_napi_del()
   but that should not matter
 - myri10ge observed normal RCU flavor
 - bnx2x and enic did not actually observe the grace period
   (unless they did so implicitly)
 - virtio_net and enic only unhashed Rx NAPIs

The last two points seem to indicate that the calls to
napi_hash_del() were a left over rather than an optimization.
Regardless, it's easy enough to correct them.

This patch may introduce extra synchronize_net() calls for
interfaces which set NAPI_STATE_NO_BUSY_POLL and depend on
free_netdev() to call netif_napi_del(). This seems inevitable
since we want to use RCU for netpoll dev->napi_list traversal,
and almost no drivers set IFF_DISABLE_NETPOLL.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'mlx4-avoid-devlink-port-type-not-set-warnings'
David S. Miller [Thu, 10 Sep 2020 19:49:00 +0000 (12:49 -0700)]
Merge branch 'mlx4-avoid-devlink-port-type-not-set-warnings'

Jakub Kicinski says:

====================
mlx4: avoid devlink port type not set warnings

This small set addresses the issue of mlx4 potentially not setting
devlink port type when Ethernet or IB driver is not built, but
port has that type.

v2:
 - add patch 1
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agomlx4: make sure to always set the port type
Jakub Kicinski [Tue, 8 Sep 2020 22:21:14 +0000 (15:21 -0700)]
mlx4: make sure to always set the port type

Even tho mlx4_core registers the devlink ports, it's mlx4_en
and mlx4_ib which set their type. In situations where one of
the two is not built yet the machine has ports of given type
we see the devlink warning from devlink_port_type_warn() trigger.

Having ports of a type not supported by the kernel may seem
surprising, but it does occur in practice - when the unsupported
port is not plugged in to a switch anyway users are more than happy
not to see it (and potentially allocate any resources to it).

Set the type in mlx4_core if type-specific driver is not built.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>