openwrt/staging/blogic.git
5 years agonet: dsa: Allow drivers to filter packets they can decode source port from
Vladimir Oltean [Sun, 5 May 2019 10:19:23 +0000 (13:19 +0300)]
net: dsa: Allow drivers to filter packets they can decode source port from

Frames get processed by DSA and redirected to switch port net devices
based on the ETH_P_XDSA multiplexed packet_type handler found by the
network stack when calling eth_type_trans().

The running assumption is that once the DSA .rcv function is called, DSA
is always able to decode the switch tag in order to change the skb->dev
from its master.

However there are tagging protocols (such as the new DSA_TAG_PROTO_SJA1105,
user of DSA_TAG_PROTO_8021Q) where this assumption is not completely
true, since switch tagging piggybacks on the absence of a vlan_filtering
bridge. Moreover, management traffic (BPDU, PTP) for this switch doesn't
rely on switch tagging, but on a different mechanism. So it would make
sense to at least be able to terminate that.

Having DSA receive traffic it can't decode would put it in an impossible
situation: the eth_type_trans() function would invoke the DSA .rcv(),
which could not change skb->dev, then eth_type_trans() would be invoked
again, which again would call the DSA .rcv, and the packet would never
be able to exit the DSA filter and would spiral in a loop until the
whole system dies.

This happens because eth_type_trans() doesn't actually look at the skb
(so as to identify a potential tag) when it deems it as being
ETH_P_XDSA. It just checks whether skb->dev has a DSA private pointer
installed (therefore it's a DSA master) and that there exists a .rcv
callback (everybody except DSA_TAG_PROTO_NONE has that). This is
understandable as there are many switch tags out there, and exhaustively
checking for all of them is far from ideal.

The solution lies in introducing a filtering function for each tagging
protocol. In the absence of a filtering function, all traffic is passed
to the .rcv DSA callback. The tagging protocol should see the filtering
function as a pre-validation that it can decode the incoming skb. The
traffic that doesn't match the filter will bypass the DSA .rcv callback
and be left on the master netdevice, which wasn't previously possible.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Optional VLAN-based port separation for switches without tagging
Vladimir Oltean [Sun, 5 May 2019 10:19:22 +0000 (13:19 +0300)]
net: dsa: Optional VLAN-based port separation for switches without tagging

This patch provides generic DSA code for using VLAN (802.1Q) tags for
the same purpose as a dedicated switch tag for injection/extraction.
It is based on the discussions and interest that has been so far
expressed in https://www.spinics.net/lists/netdev/msg556125.html.

Unlike all other DSA-supported tagging protocols, CONFIG_NET_DSA_TAG_8021Q
does not offer a complete solution for drivers (nor can it). Instead, it
provides generic code that driver can opt into calling:
- dsa_8021q_xmit: Inserts a VLAN header with the specified contents.
  Can be called from another tagging protocol's xmit function.
  Currently the LAN9303 driver is inserting headers that are simply
  802.1Q with custom fields, so this is an opportunity for code reuse.
- dsa_8021q_rcv: Retrieves the TPID and TCI from a VLAN-tagged skb.
  Removing the VLAN header is left as a decision for the caller to make.
- dsa_port_setup_8021q_tagging: For each user port, installs an Rx VID
  and a Tx VID, for proper untagged traffic identification on ingress
  and steering on egress. Also sets up the VLAN trunk on the upstream
  (CPU or DSA) port. Drivers are intentionally left to call this
  function explicitly, depending on the context and hardware support.
  The expected switch behavior and VLAN semantics should not be violated
  under any conditions. That is, after calling
  dsa_port_setup_8021q_tagging, the hardware should still pass all
  ingress traffic, be it tagged or untagged.

For uniformity with the other tagging protocols, a module for the
dsa_8021q_netdev_ops structure is registered, but the typical usage is
to set up another tagging protocol which selects CONFIG_NET_DSA_TAG_8021Q,
and calls the API from tag_8021q.h. Null function definitions are also
provided so that a "depends on" is not forced in the Kconfig.

This tagging protocol only works when switch ports are standalone, or
when they are added to a VLAN-unaware bridge. It will probably remain
this way for the reasons below.

When added to a bridge that has vlan_filtering 1, the bridge core will
install its own VLANs and reset the pvids through switchdev. For the
bridge core, switchdev is a write-only pipe. All VLAN-related state is
kept in the bridge core and nothing is read from DSA/switchdev or from
the driver. So the bridge core will break this port separation because
it will install the vlan_default_pvid into all switchdev ports.

Even if we could teach the bridge driver about switchdev preference of a
certain vlan_default_pvid (task difficult in itself since the current
setting is per-bridge but we would need it per-port), there would still
exist many other challenges.

Firstly, in the DSA rcv callback, a driver would have to perform an
iterative reverse lookup to find the correct switch port. That is
because the port is a bridge slave, so its Rx VID (port PVID) is subject
to user configuration. How would we ensure that the user doesn't reset
the pvid to a different value (which would make an O(1) translation
impossible), or to a non-unique value within this DSA switch tree (which
would make any translation impossible)?

Finally, not all switch ports are equal in DSA, and that makes it
difficult for the bridge to be completely aware of this anyway.
The CPU port needs to transmit tagged packets (VLAN trunk) in order for
the DSA rcv code to be able to decode source information.
But the bridge code has absolutely no idea which switch port is the CPU
port, if nothing else then just because there is no netdevice registered
by DSA for the CPU port.
Also DSA does not currently allow the user to specify that they want the
CPU port to do VLAN trunking anyway. VLANs are added to the CPU port
using the same flags as they were added on the user port.

So the VLANs installed by dsa_port_setup_8021q_tagging per driver
request should remain private from the bridge's and user's perspective,
and should not alter the VLAN semantics observed by the user.

In the current implementation a VLAN range ending at 4095 (VLAN_N_VID)
is reserved for this purpose. Each port receives a unique Rx VLAN and a
unique Tx VLAN. Separate VLANs are needed for Rx and Tx because they
serve different purposes: on Rx the switch must process traffic as
untagged and process it with a port-based VLAN, but with care not to
hinder bridging. On the other hand, the Tx VLAN is where the
reachability restrictions are imposed, since by tagging frames in the
xmit callback we are telling the switch onto which port to steer the
frame.

Some general guidance on how this support might be employed for
real-life hardware (some comments made by Florian Fainelli):

- If the hardware supports VLAN tag stacking, it should somehow back
  up its private VLAN settings when the bridge tries to override them.
  Then the driver could re-apply them as outer tags. Dedicating an outer
  tag per bridge device would allow identical inner tag VID numbers to
  co-exist, yet preserve broadcast domain isolation.

- If the switch cannot handle VLAN tag stacking, it should disable this
  port separation when added as slave to a vlan_filtering bridge, in
  that case having reduced functionality.

- Drivers for old switches that don't support the entire VLAN_N_VID
  range will need to rework the current range selection mechanism.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Export symbols for dsa_port_vid_{add, del}
Vladimir Oltean [Sun, 5 May 2019 10:19:21 +0000 (13:19 +0300)]
net: dsa: Export symbols for dsa_port_vid_{add, del}

This is needed so that the newly introduced tag_8021q may access these
core DSA functions when built as a module.

Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dsa: Call driver's setup callback after setting up its switchdev notifier
Vladimir Oltean [Sun, 5 May 2019 10:19:20 +0000 (13:19 +0300)]
net: dsa: Call driver's setup callback after setting up its switchdev notifier

This allows the driver to perform some manipulations of its own during
setup, using generic switchdev calls. Having the notifiers registered at
setup time is important because otherwise any switchdev transaction
emitted during this time would be ignored (dispatched to an empty call
chain).

One current usage scenario is for the driver to request DSA to set up
802.1Q based switch tagging for its ports.

There is no danger for the driver setup code to start racing now with
switchdev events emitted from the network stack (such as bridge core)
even if the notifier is registered earlier. This is because the network
stack needs a net_device as a vehicle to perform switchdev operations,
and the slave net_devices are registered later than the core driver
setup anyway (ds->ops->setup in dsa_switch_setup vs dsa_port_setup).

Luckily DSA doesn't need a net_device to carry out switchdev callbacks,
and therefore drivers shouldn't assume either that net_devices are
available at the time their switchdev callbacks get invoked.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-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>
5 years agonet: dsa: mv88e6xxx: refine SMI support
Vivien Didelot [Fri, 3 May 2019 23:28:22 +0000 (19:28 -0400)]
net: dsa: mv88e6xxx: refine SMI support

The Marvell SOHO switches have several ways to access the internal
registers. One of them being the System Management Interface (SMI),
using the MDC and MDIO pins, with direct and indirect variants.

In preparation for adding support for other register accesses, move
the SMI code into its own files. At the same time, refine the code
to make it clear that the indirect variant is implemented using the
direct variant accessing only two registers for command and data.

Signed-off-by: Vivien Didelot <vivien.didelot@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-act_police-offload-support'
David S. Miller [Mon, 6 May 2019 04:49:24 +0000 (21:49 -0700)]
Merge branch 'net-act_police-offload-support'

Jakub Kicinski says:

===================
net: act_police offload support

this set starts by converting cls_matchall to the new flow offload
infrastructure. It so happens that all drivers implementing cls_matchall
offload today also offload cls_flower, so its a little easier for
them to handle the actions in unified flow_rule format, even though
in cls_matchall there is no flow to speak of. If a driver ever appears
which would prefer the old, direct access to TC exts, we can add the
pointer in the offload structure back and support both.

Next the act_police is added to actions supported by flow offload API.

NFP support for act_police offload is added as the final step.  The flower
firmware is configured to perform TX rate limiting in a way which matches
act_police's behaviour.  It does not use DMA.IN back pressure, and
instead drops packets after they had been already DMAed into the NIC.
IOW it uses our standard traffic policing implementation, future patches
will extend it to other ports and traffic directions.
===================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: flower: add qos offload stats request and reply
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:28 +0000 (04:46 -0700)]
nfp: flower: add qos offload stats request and reply

Add stats request function that sends a stats request message to hw for
a specific police-filter. Process stats reply from hw and update the
stored qos structure.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: flower: add qos offload install and remove functionality.
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:27 +0000 (04:46 -0700)]
nfp: flower: add qos offload install and remove functionality.

Add install and remove offload functionality for qos offloads. We
first check that a police filter can be implemented by the VF rate
limiting feature in hw, then we install the filter via the qos
infrastructure. Finally we implement the mechanism for removing
these types of filters.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonfp: flower: add qos offload framework
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:26 +0000 (04:46 -0700)]
nfp: flower: add qos offload framework

Introduce matchall filter offload infrastructure that is needed to
offload qos features like policing. Subsequent patches will make
use of police-filters for ingress rate limiting.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/sched: add block pointer to tc_cls_common_offload structure
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:25 +0000 (04:46 -0700)]
net/sched: add block pointer to tc_cls_common_offload structure

Some actions like the police action are stateful and could share state
between devices. This is incompatible with offloading to multiple devices
and drivers might want to test for shared blocks when offloading.
Store a pointer to the tcf_block structure in the tc_cls_common_offload
structure to allow drivers to determine when offloads apply to a shared
block.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/sched: allow stats updates from offloaded police actions
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:24 +0000 (04:46 -0700)]
net/sched: allow stats updates from offloaded police actions

Implement the stats_update callback for the police action that
will be used by drivers for hardware offload.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/sched: extend matchall offload for hardware statistics
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:23 +0000 (04:46 -0700)]
net/sched: extend matchall offload for hardware statistics

Introduce a new command for matchall classifiers that allows hardware
to update statistics.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/sched: add police action to the hardware intermediate representation
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:22 +0000 (04:46 -0700)]
net/sched: add police action to the hardware intermediate representation

Add police action to the hardware intermediate representation which
would subsequently allow it to be used by drivers for offload.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/sched: move police action structures to header
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:21 +0000 (04:46 -0700)]
net/sched: move police action structures to header

Move tcf_police_params, tcf_police and tc_police_compat structures to a
header. Making them usable to other code for example drivers that would
offload police actions to hardware.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/sched: remove unused functions for matchall offload
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:20 +0000 (04:46 -0700)]
net/sched: remove unused functions for matchall offload

Cleanup unused functions and variables after porting to the newer
intermediate representation.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/dsa: use intermediate representation for matchall offload
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:19 +0000 (04:46 -0700)]
net/dsa: use intermediate representation for matchall offload

Updates dsa hardware switch handling infrastructure to use the newer
intermediate representation for flow actions in matchall offloads.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: use intermediate representation for matchall offload
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:18 +0000 (04:46 -0700)]
mlxsw: use intermediate representation for matchall offload

Updates the Mellanox spectrum driver to use the newer intermediate
representation for flow actions in matchall offloads.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/sched: use the hardware intermediate representation for matchall
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:17 +0000 (04:46 -0700)]
net/sched: use the hardware intermediate representation for matchall

Extends matchall offload to make use of the hardware intermediate
representation. More specifically, this patch moves the native TC
actions in cls_matchall offload to the newer flow_action
representation. This ultimately allows us to avoid a direct
dependency on native TC actions for matchall.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/sched: add sample action to the hardware intermediate representation
Pieter Jansen van Vuuren [Sat, 4 May 2019 11:46:16 +0000 (04:46 -0700)]
net/sched: add sample action to the hardware intermediate representation

Add sample action to the hardware intermediate representation model which
would subsequently allow it to be used by drivers for offload.

Signed-off-by: Pieter Jansen van Vuuren <pieter.jansenvanvuuren@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'of_net-Add-NVMEM-support-to-of_get_mac_address'
David S. Miller [Mon, 6 May 2019 04:47:08 +0000 (21:47 -0700)]
Merge branch 'of_net-Add-NVMEM-support-to-of_get_mac_address'

Petr Štetiar says:

====================
of_net: Add NVMEM support to of_get_mac_address

this patch series is a continuation of my previous attempt[1], where I've
tried to wire MTD layer into of_get_mac_address, so it would be possible to
load MAC addresses from various NVMEMs as EEPROMs etc.

Predecessor of this patch which used directly MTD layer has originated in
OpenWrt some time ago and supports already about 497 use cases in 357
device tree files.

During the review process of my 1st attempt I was told, that I shouldn't be
using MTD directly, but that I should rather use new NVMEM subsystem and
during the review process of v2 I was told, that I should handle
EPROBE_DEFFER error as well, during the review process of v3 I was told,
that returning pointer/NULL/ERR_PTR is considered as wrong API design, so
this v4 patch series tries to accommodate all this previous remarks.

First patch is wiring NVMEM support directly into of_get_mac_address as
it's obvious, that adding support for NVMEM into every other driver would
mean adding a lot of repetitive code. This patch allows us to configure MAC
addresses in various devices like ethernet and wireless adapters directly
from of_get_mac_address, which is used by quite a lot of drivers in the
tree already.

Second patch is simply updating documentation with NVMEM bits, and cleaning
up all current binding documentation referencing any of the MAC address
related properties.

Third and fourth patches are simply removing duplicate NVMEM code which is
no longer needed as the first patch has wired NVMEM support directly into
of_get_mac_address.

Patches 5-10 are converting all current users of of_get_mac_address to the
new ERR_PTR encoded error value, as of_get_mac_address could now return
valid pointer, NULL and ERR_PTR.

Just for a better picture, this patch series and one simple patch[2] on top
of it, allows me to configure 8Devices Carambola2 board's MAC addresses
with following DTS (simplified):

 &spi {
  flash@0 {
  partitions {
art: partition@ff0000 {
label = "art";
reg = <0xff0000 0x010000>;
read-only;

nvmem-cells {
compatible = "nvmem-cells";
#address-cells = <1>;
#size-cells = <1>;

eth0_addr: eth-mac-addr@0 {
reg = <0x0 0x6>;
};

eth1_addr: eth-mac-addr@6 {
reg = <0x6 0x6>;
};

wmac_addr: wifi-mac-addr@1002 {
reg = <0x1002 0x6>;
};
};
};
};
};
 };

 &eth0 {
nvmem-cells = <&eth0_addr>;
nvmem-cell-names = "mac-address";
 };

 &eth1 {
nvmem-cells = <&eth1_addr>;
nvmem-cell-names = "mac-address";
 };

 &wmac {
nvmem-cells = <&wmac_addr>;
nvmem-cell-names = "mac-address";
 };

1. https://patchwork.ozlabs.org/patch/1086628/
2. https://patchwork.ozlabs.org/patch/890738/
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agopowerpc: tsi108: support of_get_mac_address new ERR_PTR error
Petr Štetiar [Fri, 3 May 2019 14:27:15 +0000 (16:27 +0200)]
powerpc: tsi108: support of_get_mac_address new ERR_PTR error

There was NVMEM support added to of_get_mac_address, so it could now return
ERR_PTR encoded error values, so we need to adjust all current users of
of_get_mac_address to this new fact.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoARM: Kirkwood: support of_get_mac_address new ERR_PTR error
Petr Štetiar [Fri, 3 May 2019 14:27:14 +0000 (16:27 +0200)]
ARM: Kirkwood: support of_get_mac_address new ERR_PTR error

There was NVMEM support added to of_get_mac_address, so it could now return
ERR_PTR encoded error values, so we need to adjust all current users of
of_get_mac_address to this new fact.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agostaging: octeon-ethernet: support of_get_mac_address new ERR_PTR error
Petr Štetiar [Fri, 3 May 2019 14:27:13 +0000 (16:27 +0200)]
staging: octeon-ethernet: support of_get_mac_address new ERR_PTR error

There was NVMEM support added to of_get_mac_address, so it could now return
ERR_PTR encoded error values, so we need to adjust all current users of
of_get_mac_address to this new fact.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: wireless: support of_get_mac_address new ERR_PTR error
Petr Štetiar [Fri, 3 May 2019 14:27:12 +0000 (16:27 +0200)]
net: wireless: support of_get_mac_address new ERR_PTR error

There was NVMEM support added to of_get_mac_address, so it could now return
ERR_PTR encoded error values, so we need to adjust all current users of
of_get_mac_address to this new fact.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: usb: support of_get_mac_address new ERR_PTR error
Petr Štetiar [Fri, 3 May 2019 14:27:11 +0000 (16:27 +0200)]
net: usb: support of_get_mac_address new ERR_PTR error

There was NVMEM support added to of_get_mac_address, so it could now return
ERR_PTR encoded error values, so we need to adjust all current users of
of_get_mac_address to this new fact.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: davinci: support of_get_mac_address new ERR_PTR error
Petr Štetiar [Fri, 3 May 2019 14:27:09 +0000 (16:27 +0200)]
net: davinci: support of_get_mac_address new ERR_PTR error

There was NVMEM support added directly to of_get_mac_address, and it uses
nvmem_get_mac_address under the hood, so we can remove it. As
of_get_mac_address can now return ERR_PTR encoded error values, adjust to
that as well.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: macb: support of_get_mac_address new ERR_PTR error
Petr Štetiar [Fri, 3 May 2019 14:27:08 +0000 (16:27 +0200)]
net: macb: support of_get_mac_address new ERR_PTR error

There was NVMEM support added directly to of_get_mac_address, and it uses
nvmem_get_mac_address under the hood, so we can remove it. As
of_get_mac_address can now return ERR_PTR encoded error values, adjust to
that as well.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodt-bindings: doc: reflect new NVMEM of_get_mac_address behaviour
Petr Štetiar [Fri, 3 May 2019 14:27:07 +0000 (16:27 +0200)]
dt-bindings: doc: reflect new NVMEM of_get_mac_address behaviour

As of_get_mac_address now supports NVMEM under the hood, we need to update
the bindings documentation with the new nvmem-cell* properties, which would
mean copy&pasting a lot of redundant information to every binding
documentation currently referencing some of the MAC address properties.

So I've just removed all the references to the optional MAC address
properties and replaced them with the small note referencing
net/ethernet.txt file.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoof_net: add NVMEM support to of_get_mac_address
Petr Štetiar [Fri, 3 May 2019 14:27:06 +0000 (16:27 +0200)]
of_net: add NVMEM support to of_get_mac_address

Many embedded devices have information such as MAC addresses stored
inside NVMEMs like EEPROMs and so on. Currently there are only two
drivers in the tree which benefit from NVMEM bindings.

Adding support for NVMEM into every other driver would mean adding a lot
of repetitive code. This patch allows us to configure MAC addresses in
various devices like ethernet and wireless adapters directly from
of_get_mac_address, which is already used by almost every driver in the
tree.

Predecessor of this patch which used directly MTD layer has originated
in OpenWrt some time ago and supports already about 497 use cases in 357
device tree files.

Cc: Alban Bedel <albeu@free.fr>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: John Crispin <john@phrozen.org>
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'bnxt_en-Driver-updates'
David S. Miller [Mon, 6 May 2019 04:42:17 +0000 (21:42 -0700)]
Merge branch 'bnxt_en-Driver-updates'

Michael Chan says:

====================
bnxt_en: Driver updates.

This patch series adds some extended statistics available with the new
firmware interface, package version from firmware, aRFS support on
57500 chips, new PCI IDs, and some miscellaneous fixes and improvements.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Add device IDs 0x1806 and 0x1752 for 57500 devices.
Michael Chan [Sun, 5 May 2019 11:17:08 +0000 (07:17 -0400)]
bnxt_en: Add device IDs 0x1806 and 0x1752 for 57500 devices.

0x1806 and 0x1752 are VF variant and PF variant of the 57500 chip
family.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Add support for aRFS on 57500 chips.
Michael Chan [Sun, 5 May 2019 11:17:07 +0000 (07:17 -0400)]
bnxt_en: Add support for aRFS on 57500 chips.

Set RSS ring table index of the RFS destination ring for the NTUPLE
filters on 57500 chips.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Query firmware capability to support aRFS on 57500 chips.
Michael Chan [Sun, 5 May 2019 11:17:06 +0000 (07:17 -0400)]
bnxt_en: Query firmware capability to support aRFS on 57500 chips.

Query support for the aRFS ring table index in the firmware.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Improve NQ reservations.
Michael Chan [Sun, 5 May 2019 11:17:05 +0000 (07:17 -0400)]
bnxt_en: Improve NQ reservations.

bnxt_need_reserve_rings() determines if any resources have changed and
requires new reservation with firmware.  The NQ checking is currently
just an approximation.  Improve the NQ checking logic to make it
accurate.  NQ reservation is only needed on 57500 PFs.  This fix will
eliminate unnecessary reservations and will reduce NQ reservations
when some NQs have been released on 57500 PFs.

Fixes: c0b8cda05e1d ("bnxt_en: Fix NQ/CP rings accounting on the new 57500 chips.")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Separate RDMA MR/AH context allocation.
Devesh Sharma [Sun, 5 May 2019 11:17:04 +0000 (07:17 -0400)]
bnxt_en: Separate RDMA MR/AH context allocation.

In newer firmware, the context memory for MR (Memory Region)
and AH (Address Handle) to support RDMA are specified separately.
Modify driver to specify and allocate the 2 context memory types
separately when supported by the firmware.

Signed-off-by: Devesh Sharma <devesh.sharma@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: read the clause type from the PHY ID
Vasundhara Volam [Sun, 5 May 2019 11:17:03 +0000 (07:17 -0400)]
bnxt_en: read the clause type from the PHY ID

Currently driver hard code Clause 45 based on speed supported by the
PHY. Instead read the clause type from the PHY ID provided as input
to the mdio ioctl.

Fixes: 0ca12be99667 ("bnxt_en: Add support for mdio read/write to external PHY")
Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Read package version from firmware.
Vasundhara Volam [Sun, 5 May 2019 11:17:02 +0000 (07:17 -0400)]
bnxt_en: Read package version from firmware.

HWRM_VER_GET firmware command returns package name that is running
actively on the adapter.  Use this version instead of parsing from
the package log in NVRAM.

Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Check new firmware capability to display extended stats.
Vasundhara Volam [Sun, 5 May 2019 11:17:01 +0000 (07:17 -0400)]
bnxt_en: Check new firmware capability to display extended stats.

Newer firmware now advertises the capability for extended stats
support.  Check the new capability in addition to the existing
version check.

Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Add support for PCIe statistics
Vasundhara Volam [Sun, 5 May 2019 11:17:00 +0000 (07:17 -0400)]
bnxt_en: Add support for PCIe statistics

Gather periodic PCIe statistics for ethtool -S.

Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Refactor bnxt_alloc_stats().
Vasundhara Volam [Sun, 5 May 2019 11:16:59 +0000 (07:16 -0400)]
bnxt_en: Refactor bnxt_alloc_stats().

Reverse the condition of the large "if" block and return early.  This
will simplify the follow up patch to add PCIe statistics.

Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Update firmware interface to 1.10.0.69.
Michael Chan [Sun, 5 May 2019 11:16:58 +0000 (07:16 -0400)]
bnxt_en: Update firmware interface to 1.10.0.69.

PTP API updates for 57500 chips, new RX port stats counters and other
miscellaneous updates.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Mon, 6 May 2019 04:40:23 +0000 (21:40 -0700)]
Merge branch '100GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
100GbE Intel Wired LAN Driver Updates 2019-05-04

This series contains updates to the ice driver only.

Jesse updated the driver to make more functions consistent in their use
of a local variable for vsi->back.  Updates the driver to use bit fields
when possible to avoid wasting lots of storage space to store single bit
values.  Optimized the driver to be more memory efficient by moving
structure members around that are not in are hot path.

Michal updates the driver to disable the VF if malicious device driver
(MDD) event is detected by the hardware.  Adds checks to validate the
messages coming from the VF driver.  Tightens up the sniffing of the
driver so that transmit traffic so that VF's cannot see what is on other
VSIs.

Tony fixed the driver so that receive stripping state won't change every
time transmit insertion is changed.  Cleanup the __always_unused
attribute, now that the variable is being used.  Fixed the function
which evaluates setting of features to ensure that can evaluate and set
multiple features in a single function call.

Akeem fixes the driver so that we do not attempt to remove a VLAN filter
that does not exist.  Adds support for adding a ethertype based filter
rule on VSI and describe it in a very long run-on sentence. :-)

Bruce cleans up static analysis warnings by removing a local variable
initialization that is not needed.

Brett makes the allocate/deallocate more consistent in all the driver
flows for VSI q_vectors.  In addition, makes setting/getting coalesce
settings more consistent throughout the driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Mon, 6 May 2019 04:38:25 +0000 (21:38 -0700)]
Merge branch '40GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
40GbE Intel Wired LAN Driver Updates 2019-05-04

This series contains updates to i40e only.

Grzegorz fixes a bug with promiscuous mode not being kept when the VF
switched to a new VLAN.  Fixed a typo in the link mode code, by using
the correct define.  Fixed truncation issue, which changed an function
argument to a smaller value that should have been a larger value.

Aleksandr adds support for new x710 devices and the speeds they support.

Sergey adds a check for the number of vectors against the number of MSIx
vectors to ensure one does not exceed the other.

Martyna adds additional input validation on VF messages handled by the
PF.  Fixed potential memory leaks in the driver where the error paths
were not freeing allocated memory.

Maciej reverts the double ShadowRAM checksum calculation change because
issues were found in the NVM downgrade situation.

Gustavo Silva changes the i40e driver to use struct_size() in kzalloc()
calls, to avoid type mistakes.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf-next
David S. Miller [Mon, 6 May 2019 04:35:08 +0000 (21:35 -0700)]
Merge git://git./linux/kernel/git/pablo/nf-next

Pablo Neira Ayuso says:

===================
Netfilter updates for net-next

The following batch contains Netfilter updates for net-next, they are:

1) Move nft_expr_clone() to nft_dynset, from Paul Gortmaker.

2) Do not include module.h from net/netfilter/nf_tables.h,
   also from Paul.

3) Restrict conntrack sysctl entries to boolean, from Tonghao Zhang.

4) Several patches to add infrastructure to autoload NAT helper
   modules from their respective conntrack helper, this also includes
   the first client of this code in OVS, patches from Flavio Leitner.

5) Add support to match for conntrack ID, from Brett Mastbergen.

6) Spelling fix in connlabel, from Colin Ian King.

7) Use struct_size() from hashlimit, from Gustavo A. R. Silva.

8) Add optimized version of nf_inet_addr_mask(), from Li RongQing.
===================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonetfilter: slightly optimize nf_inet_addr_mask
Li RongQing [Sun, 28 Apr 2019 07:12:19 +0000 (15:12 +0800)]
netfilter: slightly optimize nf_inet_addr_mask

using 64bit computation to slightly optimize nf_inet_addr_mask

Signed-off-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
5 years agonetfilter: xt_hashlimit: use struct_size() helper
Gustavo A. R. Silva [Wed, 1 May 2019 22:01:08 +0000 (17:01 -0500)]
netfilter: xt_hashlimit: use struct_size() helper

Make use of the struct_size() helper instead of an open-coded version
in order to avoid any potential type mistakes, in particular in the
context in which this code is being used.

So, replace code of the following form:

sizeof(struct xt_hashlimit_htable) + sizeof(struct hlist_head) * size

with:

struct_size(hinfo, hash, size)

This code was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
5 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
David S. Miller [Sun, 5 May 2019 20:10:36 +0000 (13:10 -0700)]
Merge branch 'for-upstream' of git://git./linux/kernel/git/bluetooth/bluetooth-next

Johan Hedberg says:

====================
pull request: bluetooth-next 2019-05-05

Here's one more bluetooth-next pull request for 5.2:

 - Fixed Command Complete event handling check for matching opcode
 - Added support for Qualcomm WCN3998 controller, along with DT bindings
 - Added default address for Broadcom BCM2076B1 controllers

Please let me know if there are any issues pulling. Thanks.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agor8169: speed up rtl_loop_wait
Heiner Kallweit [Sat, 4 May 2019 13:20:38 +0000 (15:20 +0200)]
r8169: speed up rtl_loop_wait

When testing I figured out that most operations signal finish even
before we trigger the first delay. Seems like PCI(e) access and
memory barriers typically add enough latency. Therefore move the
first delay after the first check.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agor8169: make use of phy_set_asym_pause
Heiner Kallweit [Sat, 4 May 2019 10:01:03 +0000 (12:01 +0200)]
r8169: make use of phy_set_asym_pause

phy_probe() takes care that all supported modes are advertised,
in addition use phy_support_asym_pause() to advertise pause modes.
This way we don't have to deal with phylib internals directly.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ll_temac: Make some functions static
YueHaibing [Sat, 4 May 2019 10:10:30 +0000 (18:10 +0800)]
net: ll_temac: Make some functions static

Fix sparse warnings:

drivers/net/ethernet/xilinx/ll_temac_main.c:66:5: warning: symbol '_temac_ior_be' was not declared. Should it be static?
drivers/net/ethernet/xilinx/ll_temac_main.c:71:6: warning: symbol '_temac_iow_be' was not declared. Should it be static?
drivers/net/ethernet/xilinx/ll_temac_main.c:76:5: warning: symbol '_temac_ior_le' was not declared. Should it be static?
drivers/net/ethernet/xilinx/ll_temac_main.c:81:6: warning: symbol '_temac_iow_le' was not declared. Should it be static?
drivers/net/ethernet/xilinx/ll_temac_main.c:648:6: warning: symbol 'ptr_to_txbd' was not declared. Should it be static?
drivers/net/ethernet/xilinx/ll_temac_main.c:654:6: warning: symbol 'ptr_from_txbd' was not declared. Should it be static?

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: aquantia: Make aq_ndev_driver_name static
YueHaibing [Sat, 4 May 2019 09:57:55 +0000 (17:57 +0800)]
net: aquantia: Make aq_ndev_driver_name static

Fix sparse warning:

drivers/net/ethernet/aquantia/atlantic/aq_main.c:26:12:
 warning: symbol 'aq_ndev_driver_name' was not declared. Should it be static?

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonetdevsim: Make nsim_num_vf static
YueHaibing [Sat, 4 May 2019 08:12:07 +0000 (16:12 +0800)]
netdevsim: Make nsim_num_vf static

Fix sparse warning:

drivers/net/netdevsim/bus.c:253:5: warning:
 symbol 'nsim_num_vf' was not declared. Should it be static?

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: mvpp2: cls: Remove set but not used variable 'act'
YueHaibing [Sat, 4 May 2019 04:04:05 +0000 (04:04 +0000)]
net: mvpp2: cls: Remove set but not used variable 'act'

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.c: In function 'mvpp2_cls_c2_build_match':
drivers/net/ethernet/marvell/mvpp2/mvpp2_cls.c:1159:28: warning:
 variable 'act' set but not used [-Wunused-but-set-variable]

It is never used since introduction in
commit 90b509b39ac9 ("net: mvpp2: cls: Add Classification offload support")

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoum: vector netdev: adjust to xmit_more API change
Johannes Berg [Fri, 3 May 2019 18:21:33 +0000 (20:21 +0200)]
um: vector netdev: adjust to xmit_more API change

Replace skb->xmit_more usage by netdev_xmit_more().

Fixes: 4f296edeb9d4 ("drivers: net: aurora: use netdev_xmit_more helper")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'net-extend-indirect-calls-helper-usage'
David S. Miller [Sun, 5 May 2019 17:38:04 +0000 (10:38 -0700)]
Merge branch 'net-extend-indirect-calls-helper-usage'

Paolo Abeni says:

====================
net: extend indirect calls helper usage

This series applies the indirect calls helper introduced with commit
283c16a2dfd3 ("indirect call wrappers: helpers to speed-up indirect
calls of builtin") to more hooks inside the network stack.

Overall this avoids up to 4 indirect calls for each RX packets,
giving small but measurable gain TCP_RR workloads and 5% under UDP
flood.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: use indirect calls helpers at the socket layer
Paolo Abeni [Fri, 3 May 2019 15:01:39 +0000 (17:01 +0200)]
net: use indirect calls helpers at the socket layer

This avoids an indirect call per {send,recv}msg syscall in
the common (IPv6 or IPv4 socket) case.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: use indirect calls helpers at early demux stage
Paolo Abeni [Fri, 3 May 2019 15:01:38 +0000 (17:01 +0200)]
net: use indirect calls helpers at early demux stage

So that we avoid another indirect call per RX packet, if
early demux is enabled.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: use indirect calls helpers for L3 handler hooks
Paolo Abeni [Fri, 3 May 2019 15:01:37 +0000 (17:01 +0200)]
net: use indirect calls helpers for L3 handler hooks

So that we avoid another indirect call per RX packet in the common
case.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: use indirect calls helpers for ptype hook
Paolo Abeni [Fri, 3 May 2019 15:01:36 +0000 (17:01 +0200)]
net: use indirect calls helpers for ptype hook

This avoids an indirect call per RX IPv6/IPv4 packet.
Note that we don't want to use the indirect calls helper for taps.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoBluetooth: hci_qca: Rename STATE_<flags> to QCA_<flags>
Matthias Kaehlcke [Mon, 29 Apr 2019 23:21:30 +0000 (16:21 -0700)]
Bluetooth: hci_qca: Rename STATE_<flags> to QCA_<flags>

Rename STATE_IN_BAND_SLEEP_ENABLED to QCA_IBS_ENABLED. The constant
represents a flag (multiple flags can be set at once), not a unique
state of the controller or driver.

Also make the flag an enum value instead of a pre-processor constant
(more flags will be added to the enum group by another patch).

Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
Reviewed-by: Balakrishna Godavarthi <bgodavar@codeaurora.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agonet: ll_temac: remove an unnecessary condition
Dan Carpenter [Fri, 3 May 2019 12:50:51 +0000 (15:50 +0300)]
net: ll_temac: remove an unnecessary condition

The "pdata->mdio_bus_id" is unsigned so this condition is always true.
This patch just removes it.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: ll_temac: Fix an NULL vs IS_ERR() check in temac_open()
Dan Carpenter [Fri, 3 May 2019 12:50:24 +0000 (15:50 +0300)]
net: ll_temac: Fix an NULL vs IS_ERR() check in temac_open()

The phy_connect() function doesn't return NULL pointers.  It returns
error pointers on error, so I have updated the check.

Fixes: 8425c41d1ef7 ("net: ll_temac: Extend support to non-device-tree platforms")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoBluetooth: Ignore CC events not matching the last HCI command
João Paulo Rechi Vita [Thu, 2 May 2019 02:01:52 +0000 (10:01 +0800)]
Bluetooth: Ignore CC events not matching the last HCI command

This commit makes the kernel not send the next queued HCI command until
a command complete arrives for the last HCI command sent to the
controller. This change avoids a problem with some buggy controllers
(seen on two SKUs of QCA9377) that send an extra command complete event
for the previous command after the kernel had already sent a new HCI
command to the controller.

The problem was reproduced when starting an active scanning procedure,
where an extra command complete event arrives for the LE_SET_RANDOM_ADDR
command. When this happends the kernel ends up not processing the
command complete for the following commmand, LE_SET_SCAN_PARAM, and
ultimately behaving as if a passive scanning procedure was being
performed, when in fact controller is performing an active scanning
procedure. This makes it impossible to discover BLE devices as no device
found events are sent to userspace.

This problem is reproducible on 100% of the attempts on the affected
controllers. The extra command complete event can be seen at timestamp
27.420131 on the btmon logs bellow.

Bluetooth monitor ver 5.50
= Note: Linux version 5.0.0+ (x86_64)                                  0.352340
= Note: Bluetooth subsystem version 2.22                               0.352343
= New Index: 80:C5:F2:8F:87:84 (Primary,USB,hci0)               [hci0] 0.352344
= Open Index: 80:C5:F2:8F:87:84                                 [hci0] 0.352345
= Index Info: 80:C5:F2:8F:87:84 (Qualcomm)                      [hci0] 0.352346
@ MGMT Open: bluetoothd (privileged) version 1.14             {0x0001} 0.352347
@ MGMT Open: btmon (privileged) version 1.14                  {0x0002} 0.352366
@ MGMT Open: btmgmt (privileged) version 1.14                {0x0003} 27.302164
@ MGMT Command: Start Discovery (0x0023) plen 1       {0x0003} [hci0] 27.302310
        Address type: 0x06
          LE Public
          LE Random
< HCI Command: LE Set Random Address (0x08|0x0005) plen 6   #1 [hci0] 27.302496
        Address: 15:60:F2:91:B2:24 (Non-Resolvable)
> HCI Event: Command Complete (0x0e) plen 4                 #2 [hci0] 27.419117
      LE Set Random Address (0x08|0x0005) ncmd 1
        Status: Success (0x00)
< HCI Command: LE Set Scan Parameters (0x08|0x000b) plen 7  #3 [hci0] 27.419244
        Type: Active (0x01)
        Interval: 11.250 msec (0x0012)
        Window: 11.250 msec (0x0012)
        Own address type: Random (0x01)
        Filter policy: Accept all advertisement (0x00)
> HCI Event: Command Complete (0x0e) plen 4                 #4 [hci0] 27.420131
      LE Set Random Address (0x08|0x0005) ncmd 1
        Status: Success (0x00)
< HCI Command: LE Set Scan Enable (0x08|0x000c) plen 2      #5 [hci0] 27.420259
        Scanning: Enabled (0x01)
        Filter duplicates: Enabled (0x01)
> HCI Event: Command Complete (0x0e) plen 4                 #6 [hci0] 27.420969
      LE Set Scan Parameters (0x08|0x000b) ncmd 1
        Status: Success (0x00)
> HCI Event: Command Complete (0x0e) plen 4                 #7 [hci0] 27.421983
      LE Set Scan Enable (0x08|0x000c) ncmd 1
        Status: Success (0x00)
@ MGMT Event: Command Complete (0x0001) plen 4        {0x0003} [hci0] 27.422059
      Start Discovery (0x0023) plen 1
        Status: Success (0x00)
        Address type: 0x06
          LE Public
          LE Random
@ MGMT Event: Discovering (0x0013) plen 2             {0x0003} [hci0] 27.422067
        Address type: 0x06
          LE Public
          LE Random
        Discovery: Enabled (0x01)
@ MGMT Event: Discovering (0x0013) plen 2             {0x0002} [hci0] 27.422067
        Address type: 0x06
          LE Public
          LE Random
        Discovery: Enabled (0x01)
@ MGMT Event: Discovering (0x0013) plen 2             {0x0001} [hci0] 27.422067
        Address type: 0x06
          LE Public
          LE Random
        Discovery: Enabled (0x01)

Signed-off-by: João Paulo Rechi Vita <jprvita@endlessm.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: btbcm: Add default address for BCM2076B1
Stephan Gerhold [Wed, 1 May 2019 07:18:23 +0000 (09:18 +0200)]
Bluetooth: btbcm: Add default address for BCM2076B1

BCM2076B1 appears to use 20:76:A0:00:56:79 as default address.
This address is used by at least 5 devices with the AMPAK AP6476
module and is also suspicious because it starts with the chip name
2076 (followed by a different revision A0 for some reason).

Add it to the list of default addresses and leave it up to the
user to configure a valid one.

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agonet: atm: clean up a range check
Dan Carpenter [Fri, 3 May 2019 12:39:48 +0000 (15:39 +0300)]
net: atm: clean up a range check

The code works fine but the problem is that check for negatives is a
no-op:

if (arg < 0)
i = 0;

The "i" value isn't used.  We immediately overwrite it with:

i = array_index_nospec(arg, MAX_LEC_ITF);

The array_index_nospec() macro returns zero if "arg" is out of bounds so
this works, but the dead code is confusing and it doesn't look very
intentional.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'wireless-drivers-next-for-davem-2019-05-03' of git://git.kernel.org/pub...
David S. Miller [Sun, 5 May 2019 17:20:53 +0000 (10:20 -0700)]
Merge tag 'wireless-drivers-next-for-davem-2019-05-03' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
wireless-drivers-next patches for 5.2

Most likely the last patchset of new feature for 5.2, and this time we
have quite a lot of new features. Most obvious being rtw88 from
Realtek which supports RTL8822BE and RTL8822CE 802.11ac devices. We
have also new hardware support for existing drivers and improvements.

There's one conflict in iwlwifi, my example conflict resolution below.

Major changes:

iwlwifi

* bump the 20000-series FW API version

* work on new hardware continues

* RTT confidence indication support for Fine Timing Measurement (FTM)

* an improvement in HE (802.11ax) rate-scaling

* add command version parsing from the fimware TLVs

* add support for a new WoWLAN patterns firmware API

rsi

* add support for rs9116

mwifiex

* add support for SD8987

brcmfmac

* add quirk for ACEPC T8 and T11 mini PCs

rt2x00

* add RT3883 support

qtnfmac

* fix debugfs interface to support multiple cards

rtw88

* new driver

mt76

* share more code across drivers

* add support for MT7615 chipset

* rework DMA API

* tx/rx performance optimizations

* use NAPI for tx cleanup on mt76x02

* AP mode support for USB devices

* USB stability fixes

* tx power handling fixes for 76x2

* endian fixes

Conflicts:

There's a trivial conflict in
drivers/net/wireless/intel/iwlwifi/fw/file.h, just leave
IWL_UCODE_TLV_FW_FSEQ_VERSION to the file. 'git diff' output should be
just empty:

diff --cc drivers/net/wireless/intel/iwlwifi/fw/file.h
index cd622af90077,b0671e16e1ce..000000000000
--- a/drivers/net/wireless/intel/iwlwifi/fw/file.h
+++ b/drivers/net/wireless/intel/iwlwifi/fw/file.h
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodrivers: net: davinci_mdio: fix return value check in davinci_mdio_probe()
Wei Yongjun [Fri, 3 May 2019 11:18:59 +0000 (11:18 +0000)]
drivers: net: davinci_mdio: fix return value check in davinci_mdio_probe()

In case of error, the function devm_ioremap() returns NULL pointer not
ERR_PTR(). The IS_ERR() test in the return value check should be
replaced with NULL test.

Fixes: 03f66f067560 ("net: ethernet: ti: davinci_mdio: use devm_ioremap()")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoopenvswitch: check for null pointer return from nla_nest_start_noflag
Colin Ian King [Wed, 1 May 2019 13:41:58 +0000 (14:41 +0100)]
openvswitch: check for null pointer return from nla_nest_start_noflag

The call to nla_nest_start_noflag can return null in the unlikely
event that nla_put returns -EMSGSIZE.  Check for this condition to
avoid a null pointer dereference on pointer nla_reply.

Addresses-Coverity: ("Dereference null return value")
Fixes: 11efd5cb04a1 ("openvswitch: Support conntrack zone limit")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Yi-Hung Wei <yihung.wei@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'ipv4-Move-location-of-pcpu-route-cache-and-exceptions'
David S. Miller [Sun, 5 May 2019 07:47:16 +0000 (00:47 -0700)]
Merge branch 'ipv4-Move-location-of-pcpu-route-cache-and-exceptions'

David Ahern says:

====================
ipv4: Move location of pcpu route cache and exceptions

This series moves IPv4 pcpu cached routes from fib_nh to fib_nh_common
to make the caches available for IPv6 nexthops (fib6_nh) with IPv4
routes. This allows a fib6_nh struct to be used with both IPv4 and
and IPv6 routes.

v4
- fixed memleak if encap_type is not set as noticed by Ido

v3
- dropped ipv6 patches for now. Will resubmit those once the existing
  refcnt problem is fixed

v2
- reverted patch 2 to use ifdef CONFIG_IP_ROUTE_CLASSID instead
  of IS_ENABLED(CONFIG_IP_ROUTE_CLASSID) to fix compile issues
  reported by kbuild test robot
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoipv4: Move exception bucket to nh_common
David Ahern [Tue, 30 Apr 2019 14:45:50 +0000 (07:45 -0700)]
ipv4: Move exception bucket to nh_common

Similar to the cached routes, make IPv4 exceptions accessible when
using an IPv6 nexthop struct with IPv4 routes. Simplify the exception
functions by passing in fib_nh_common since that is all it needs,
and then cleanup the call sites that have extraneous fib_nh conversions.

As with the cached routes this is a change in location only, from fib_nh
up to fib_nh_common; no functional change intended.

Signed-off-by: David Ahern <dsahern@gmail.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoipv4: Pass fib_nh_common to rt_cache_route
David Ahern [Tue, 30 Apr 2019 14:45:49 +0000 (07:45 -0700)]
ipv4: Pass fib_nh_common to rt_cache_route

Now that the cached routes are in fib_nh_common, pass it to
rt_cache_route and simplify its callers. For rt_set_nexthop,
the tclassid becomes the last user of fib_nh so move the
container_of under the #ifdef CONFIG_IP_ROUTE_CLASSID.

Signed-off-by: David Ahern <dsahern@gmail.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoipv4: Move cached routes to fib_nh_common
David Ahern [Tue, 30 Apr 2019 14:45:48 +0000 (07:45 -0700)]
ipv4: Move cached routes to fib_nh_common

While the cached routes, nh_pcpu_rth_output and nh_rth_input, are IPv4
specific, a later patch wants to make them accessible for IPv6 nexthops
with IPv4 routes using a fib6_nh. Move the cached routes from fib_nh to
fib_nh_common and update references.

Initialization of the cached entries is moved to fib_nh_common_init,
and free is moved to fib_nh_common_release.

Change in location only, from fib_nh up to fib_nh_common; no functional
change intended.

Signed-off-by: David Ahern <dsahern@gmail.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoi40e: Memory leak in i40e_config_iwarp_qvlist
Martyna Szapar [Thu, 18 Apr 2019 20:31:53 +0000 (13:31 -0700)]
i40e: Memory leak in i40e_config_iwarp_qvlist

Added freeing the old allocation of vf->qvlist_info in function
i40e_config_iwarp_qvlist before overwriting it with
the new allocation.

Signed-off-by: Martyna Szapar <martyna.szapar@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: Fix of memory leak and integer truncation in i40e_virtchnl.c
Martyna Szapar [Mon, 15 Apr 2019 21:43:07 +0000 (14:43 -0700)]
i40e: Fix of memory leak and integer truncation in i40e_virtchnl.c

Fixed possible memory leak in i40e_vc_add_cloud_filter function:
cfilter is being allocated and in some error conditions
the function returns without freeing the memory.

Fix of integer truncation from u16 (type of queue_id value) to u8
when calling i40e_vc_isvalid_queue_id function.

Signed-off-by: Martyna Szapar <martyna.szapar@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: Use struct_size() in kzalloc()
Gustavo A. R. Silva [Fri, 29 Mar 2019 23:38:49 +0000 (16:38 -0700)]
i40e: Use struct_size() in kzalloc()

One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:

struct foo {
    int stuff;
    struct boo entry[];
};

size = sizeof(struct foo) + count * sizeof(struct boo);
instance = kzalloc(size, GFP_KERNEL)

Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:

instance = kzalloc(struct_size(instance, entry, count), GFP_KERNEL)

Notice that, in this case, variable size is not necessary, hence it
is removed.

This code was detected with the help of Coccinelle.

Signed-off-by: "Gustavo A. R. Silva" <gustavo@embeddedor.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: Revert ShadowRAM checksum calculation change
Maciej Paczkowski [Fri, 29 Mar 2019 22:08:40 +0000 (15:08 -0700)]
i40e: Revert ShadowRAM checksum calculation change

The reason of this revert is unexpected issue found in NVM Update tool
during NVM image downgrade. The implementation is no longer needed
since the QV tools are already aware of new FW double ShadowRAM dump
mechanism.

This patch reverts ShadowRAM checksum calculation change introduced in
commit 9d12f0c4e436 ("i40e: Revert ShadowRAM checksum calculation change")

Signed-off-by: Maciej Paczkowski <maciej.paczkowski@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: missing input validation on VF message handling by the PF
Martyna Szapar [Fri, 29 Mar 2019 22:08:39 +0000 (15:08 -0700)]
i40e: missing input validation on VF message handling by the PF

Patch is adding missing input validation on VF message handling
by the PF to the functions with opcodes:
VIRTCHNL_OP_CONFIG_VSI_QUEUES = 6
VIRTCHNL_OP_CONFIG_IRQ_MAP = 7,
VIRTCHNL_OP_DISABLE_QUEUES = 9,
VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE = 14,

Signed-off-by: Martyna Szapar <martyna.szapar@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: Add support for X710 B/P & SFP+ cards
Aleksandr Loktionov [Fri, 29 Mar 2019 22:08:38 +0000 (15:08 -0700)]
i40e: Add support for X710 B/P & SFP+ cards

New device ids are created to support X710 backplane and SFP+ cards.

This patch adds in i40e driver support for 2.5GbaseT and 5GbaseT speed.
It's implemented by checking I40E_CAP_PHY_TYPE_2_5GBASE_T,
I40E_CAP_PHY_TYPE_5GBASE_T bits from f/w and setting corresponding bits
in ethtool link ksettings supported and advertising masks.

Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Signed-off-by: Alice Michael <alice.michael@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: Wrong truncation from u16 to u8
Grzegorz Siwik [Fri, 29 Mar 2019 22:08:37 +0000 (15:08 -0700)]
i40e: Wrong truncation from u16 to u8

In this patch fixed wrong truncation method from u16 to u8 during
validation.

It was changed by changing u8 to u32 parameter in method declaration
and arguments were changed to u32.

Signed-off-by: Grzegorz Siwik <grzegorz.siwik@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: add num_vectors checker in iwarp handler
Sergey Nemov [Fri, 29 Mar 2019 22:08:36 +0000 (15:08 -0700)]
i40e: add num_vectors checker in iwarp handler

Field num_vectors from struct virtchnl_iwarp_qvlist_info should not be
larger than num_msix_vectors_vf in the hw struct.  The iwarp uses the
same set of vectors as the LAN VF driver.

Signed-off-by: Sergey Nemov <sergey.nemov@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: Fix the typo in adding 40GE KR4 mode
Grzegorz Siwik [Fri, 29 Mar 2019 22:08:35 +0000 (15:08 -0700)]
i40e: Fix the typo in adding 40GE KR4 mode

This patch fixes the typo in I40E_CAP_PHY_TYPE mode link code.
It was fixed by changing 40000baseLR4_Full to 40000baseKR4_Full

Signed-off-by: Grzegorz Siwik <grzegorz.siwik@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: Setting VF to VLAN 0 requires restart
Grzegorz Siwik [Fri, 29 Mar 2019 22:08:32 +0000 (15:08 -0700)]
i40e: Setting VF to VLAN 0 requires restart

This patch fixes a bug where changing VLAN to 0 was not set until VF
restart.

Now we are setting pvid info to 0 when we have to change VLAN to 0.
Without this change when VF VLAN was changed to 0 nothing happened until
VF restart. For changing to VLAN different than 0 it worked correctly.

Signed-off-by: Grzegorz Siwik <grzegorz.siwik@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: add new pci id for X710/XXV710 N3000 cards
Aleksandr Loktionov [Fri, 29 Mar 2019 22:08:31 +0000 (15:08 -0700)]
i40e: add new pci id for X710/XXV710 N3000 cards

New device ids are created to support X710/XXV710 N3000 cards.

Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoi40e: VF's promiscuous attribute is not kept
Grzegorz Siwik [Fri, 29 Mar 2019 22:08:30 +0000 (15:08 -0700)]
i40e: VF's promiscuous attribute is not kept

This patch fixes a bug where the promiscuous mode was not being
kept when the VF switched to a new VLAN.
Now we are config two times a promiscuous mode when we switch VLAN.
Without this change when we change VF VLAN we still receive
all the packets from previous VLAN and only unicast from new VLAN.

Signed-off-by: Grzegorz Siwik <grzegorz.siwik@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Disable sniffing VF traffic on PF
Michal Swiatkowski [Tue, 16 Apr 2019 17:21:27 +0000 (10:21 -0700)]
ice: Disable sniffing VF traffic on PF

Delete code that add default Tx rule on PF. With this rule PF can see
Tx VF traffic that should go outside. For traffic from VF to another
VF default Tx rule on PF doesn't apply because of lower priority than
VF mac rule.

With this change on PF in promisc mode we can see only Rx traffic that
doesn't match any other rule (mac etc.). We can't see Tx traffic from
other VSI.

Signed-off-by: Michal Swiatkowski <michal.swiatkowski@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Use more efficient structures
Jesse Brandeburg [Tue, 16 Apr 2019 17:21:26 +0000 (10:21 -0700)]
ice: Use more efficient structures

Move a bunch of members around to make more efficient use of
memory, eliminating holes where possible. None of these members
are hot path so cache line alignment is not very important here.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Use bitfields where possible
Jesse Brandeburg [Tue, 16 Apr 2019 17:21:25 +0000 (10:21 -0700)]
ice: Use bitfields where possible

The driver was converted to not use bool, but it was
neglected that the bools should have been converted to bit fields
as bit fields in software structures are ok, as long as they
use the correct kinds of unsigned types. This avoids
wasting lots of storage space to store single bit values.

One of the change hunks moves a variable lport out of
a group of "combinable" bit fields because all bits of
the u8 lport are valid and the variable can be packed in the
struct in struct holes.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Add function to program ethertype based filter rule on VSIs
Akeem G Abodunrin [Tue, 16 Apr 2019 17:21:24 +0000 (10:21 -0700)]
ice: Add function to program ethertype based filter rule on VSIs

This patch adds function to program VSI with ethertype based filter rule,
so that all flow control frames would be disallowed from being transmitted
to the client, in order to prevent malicious VSI, especially VF from
sending out PAUSE or PFC frames, and then control other VSIs traffic.

Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Separate if conditions for ice_set_features()
Tony Nguyen [Tue, 16 Apr 2019 17:21:23 +0000 (10:21 -0700)]
ice: Separate if conditions for ice_set_features()

Set features can have multiple features turned on|off in a single
call.  Grouping these all in an if/else means after one condition
is met, other conditions/features will not be evaluated.  Break
the if/else statements by feature to ensure all features will be
handled properly.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Remove __always_unused attribute
Tony Nguyen [Tue, 16 Apr 2019 17:21:22 +0000 (10:21 -0700)]
ice: Remove __always_unused attribute

The variable netdev is being used in this function; remove the
__always_unused attribute from it.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Suppress false-positive style issues reported by static analyzer
Bruce Allan [Tue, 16 Apr 2019 17:21:21 +0000 (10:21 -0700)]
ice: Suppress false-positive style issues reported by static analyzer

A recent version of cppcheck falsely reports-
    Variable ip.hdr is assigned a value that is never used.

ip is a union so the pointer ip.hdr is actually used when referenced as
ip.v4 and ip.v6.  Silence these false reports when using cppcheck with the
--inline-suppr command-line option.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Refactor getting/setting coalesce
Brett Creeley [Tue, 16 Apr 2019 17:21:20 +0000 (10:21 -0700)]
ice: Refactor getting/setting coalesce

Currently if the driver has an uneven amount of Rx/Tx queues
setting the coalesce settings through ethtool will result in
an error. This is happening because in the setting coalesce
flow we are reporting an error if either Rx or Tx fails.

Also, the flow for setting/getting per_q_coalesce and
setting/getting coalesce settings for the entire device
is different.

Fix these issues by adding one function, ice_set_q_coalesce(),
and another, ice_get_q_coalesce(), that both getting/setting
per_q and entire device coalesce can use. This makes handling
the error cases generic between the two flows and simplifies
__ice_set_coalesce() and __ice_get_coalesce().

Also, add a header comment to __ice_set_coalesce().

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Always free/allocate q_vectors
Brett Creeley [Tue, 16 Apr 2019 17:21:19 +0000 (10:21 -0700)]
ice: Always free/allocate q_vectors

Currently when probing/removing the driver we allocate/deallocate
each vsi->q_vectors array in ice_vsi_alloc_arrays() and
ice_vsi_free_arrays() respectively. However, we don't do this
during the reset and VSI rebuild flow. This is inconsistent
and unnecessary to have a difference between the two flows.

This patch makes the change to always allocate/deallocate the
vsi->q_vectors array regardless of the driver flow we are in.

Also, update the comment for ice_vsi_free_arrays() to be more
descriptive.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Do not unnecessarily initialize local variable
Bruce Allan [Tue, 16 Apr 2019 17:21:18 +0000 (10:21 -0700)]
ice: Do not unnecessarily initialize local variable

The local variable speed does not need to be initialized and can cause some
static analysis tools to complain the initial assigned value is never used.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Add more validation in ice_vc_cfg_irq_map_msg
Michal Swiatkowski [Tue, 16 Apr 2019 17:21:17 +0000 (10:21 -0700)]
ice: Add more validation in ice_vc_cfg_irq_map_msg

Add few checks to validate msg from iavf driver.

Test if we have got enough q_vectors allocated in VSI connected with VF.
Add masks for itr_indx and msix_indx to avoid writing to reserved fieldi
of QINT. Clear q_vector->num_ring_rx/tx, without it we can increment this
value every time we send irq map msg from VF. So after second call this
value will be incorrect.

Decrement num_vectors from msg, because last vector in iavf msg is misc
vector (we don't set map for it).

Signed-off-by: Michal Swiatkowski <michal.swiatkowski@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Don't remove VLAN filters that were never programmed
Akeem G Abodunrin [Tue, 16 Apr 2019 17:21:16 +0000 (10:21 -0700)]
ice: Don't remove VLAN filters that were never programmed

In case of non-trusted VFs, it is possible to program VLAN filter far
less than what is requested by the VF originally, thereby makes number of
VLAN elements being tracked by VF different from actual VLAN tags. This
patch makes sure that we are not attempting to remove VLAN filter that
does not exist.

Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Preserve VLAN Rx stripping settings
Tony Nguyen [Tue, 16 Apr 2019 17:21:15 +0000 (10:21 -0700)]
ice: Preserve VLAN Rx stripping settings

When Tx insertion is set, we are not accounting for the state of Rx
stripping.  This causes Rx stripping to be enabled any time Tx
insertion is changed, even when it's supposed to be disabled.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Fix for allowing too many MDD events on VF
Michal Swiatkowski [Tue, 16 Apr 2019 17:21:14 +0000 (10:21 -0700)]
ice: Fix for allowing too many MDD events on VF

Disable VF if any malicious device driver (MDD) event is detected by
hardware. Track vf->num_mdd_events for information about VF MDD events.

Signed-off-by: Michal Swiatkowski <michal.swiatkowski@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoice: Use pf instead of vsi-back
Jesse Brandeburg [Thu, 28 Feb 2019 23:26:03 +0000 (15:26 -0800)]
ice: Use pf instead of vsi-back

Many times in our functions we have a local variable pf, which is
equivalent to vsi->back. Just use pf consistently instead of vsi->back
where available.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agonet: openvswitch: return an error instead of doing BUG_ON()
Eelco Chaudron [Thu, 2 May 2019 20:12:38 +0000 (16:12 -0400)]
net: openvswitch: return an error instead of doing BUG_ON()

For all other error cases in queue_userspace_packet() the error is
returned, so it makes sense to do the same for these two error cases.

Reported-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>
Signed-off-by: David S. Miller <davem@davemloft.net>