net/mlx5e: Add trace point for neigh update

Allow tracing neigh state during neigh update task that is executed on
workqueue and is scheduled by neigh state change event.

Usage example:
 ># cd /sys/kernel/debug/tracing
 ># echo mlx5:mlx5e_rep_neigh_update >> set_event
 ># cat trace
    ...
    kworker/u48:7-2221  [009] ...1  1475.387435: mlx5e_rep_neigh_update:
netdev: ens1f0 MAC: 24:8a:07:9a:17:9a IPv4: 1.1.1.10 IPv6: ::ffff:1.1.1.10 neigh_connected=1

Added corresponding documentation in
    Documentation/networking/device-driver/mellanox/mlx5.rst

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Dmytro Linkin <dmitrolin@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Add trace point for neigh used value update

Allow tracing result of neigh used value update task that is executed
periodically on workqueue.

Usage example:
 ># cd /sys/kernel/debug/tracing
 ># echo mlx5:mlx5e_tc_update_neigh_used_value >> set_event
 ># cat trace
    ...
    kworker/u48:4-8806  [009] ...1 55117.882428: mlx5e_tc_update_neigh_used_value:
netdev: ens1f0 IPv4: 1.1.1.10 IPv6: ::ffff:1.1.1.10 neigh_used=1

Added corresponding documentation in
    Documentation/networking/device-driver/mellanox/mlx5.rst

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Dmytro Linkin <dmitrolin@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Add tc flower tracepoints

Implemented following tracepoints:
1. Configure flower (mlx5e_configure_flower)
2. Delete flower (mlx5e_delete_flower)
3. Stats flower (mlx5e_stats_flower)

Usage example:
 ># cd /sys/kernel/debug/tracing
 ># echo mlx5:mlx5e_configure_flower >> set_event
 ># cat trace
    ...
    tc-6535  [019] ...1  2672.404466: mlx5e_configure_flower: cookie=0000000067874a55 actions= REDIRECT

Added corresponding documentation in
    Documentation/networking/device-driver/mellanox/mlx5.rst

Signed-off-by: Dmytro Linkin <dmitrolin@mellanox.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Only access fully initialized flows in neigh update

To remove dependency on rtnl lock and prevent neigh update code from
accessing uninitialized flows when executing concurrently with tc, extend
mlx5e_tc_flow with 'init_done' completion. Modify helper
mlx5e_take_all_encap_flows() to wait for flow completion after obtaining
reference to it. Modify mlx5e_tc_encap_flows_del() and
mlx5e_tc_encap_flows_add() to skip flows that don't have OFFLOADED flag
set, which can happen if concurrent flow initialization failed.

This commit finishes neigh update refactoring for concurrent execution
started in previous change in this series.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Refactor neigh update for concurrent execution

In order to remove dependency on rtnl lock and allow neigh update workqueue
task to execute concurrently with tc, refactor mlx5e_rep_neigh_update() for
concurrent execution:

- Lock encap table when accessing encap entry to prevent concurrent
  changes. To do this properly, the initial encap state check is moved from
  mlx5e_rep_neigh_update() into mlx5e_rep_update_flows() to be performed
  under encap_tbl_lock protection.

- Wait for encap to be fully initialized before accessing it by means of
  'res_ready' completion.

- Add mlx5e_take_all_encap_flows() helper which is used to construct a
  temporary list of flows and efi indexes that is used to access current
  encap data in flow which can be attached to multiple encaps
  simultaneously. Release the flows from temporary list after
  encap_tbl_lock critical section. This is necessary because
  mlx5e_flow_put() can't be called while holding encap_tbl_lock.

- Modify mlx5e_tc_encap_flows_add() and mlx5e_tc_encap_flows_del() to work
  with user-provided list of flows built by mlx5e_take_all_encap_flows(),
  instead of traversing encap flow list directly.

This is first step in complex neigh update refactoring, which is finished
by following commit in this series.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Refactor neigh used value update for concurrent execution

In order to remove dependency on rtnl lock and allow neigh used value
update workqueue task to execute concurrently with tc, refactor
mlx5e_tc_update_neigh_used_value() for concurrent execution:

- Lock encap table when accessing encap entry to prevent concurrent
  changes.

- Save offloaded encap flows to temporary list and release them after encap
  entry is updated. Add mlx5e_put_encap_flow_list() helper which is
  intended to be shared with neigh update code in following patch in this
  series. This is necessary because mlx5e_flow_put() can't be called while
  holding encap_tbl_lock.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Protect neigh hash encap list with spinlock and rcu

Rcu-ify mlx5e_neigh_hash_entry->encap_list by changing operations on encap
list to their rcu counterparts and extending encap structure with rcu_head
to free the encap instances after rcu grace period. Use rcu read lock when
traversing encap list. Implement helper mlx5e_get_next_valid_encap()
function that is used by mlx5e_tc_update_neigh_used_value() to safely
iterate over valid entries of nhe->encap_list.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Refactor mlx5e_neigh_update_table->encap_lock

To remove dependency on rtnl lock, always take neigh update encap lock when
modifying neigh update hash table and list. Originally, this lock was only
used to synchronize with netevent handler function, which is called from bh
context and cannot use rtnl lock for synchronization. Take lock in encap
entry attach function to prevent concurrent modifications of neigh update
hash table and list.

Taking the encap lock when creating new nhe introduces a problem that we
need to allocate new entry with sleeping GFP_KERNEL flag while holding a
spinlock. However, since previous patch in this series has already
converted lookup in netevent handler function to user rcu read lock instead
of encap lock, we can safely convert the lock type to mutex.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Extend neigh hash entry with rcu

To remove dependency on rtnl lock and to allow unlocked iteration over list
of neigh hash entries, extend nhe with rcu. Change operations on neigh list
to their rcu counterparts and free neigh hash entry with rcu timeout.

Introduce mlx5e_get_next_nhe() helper that is used to iterate over rcu
neigh list with reference to nhe taken.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Always take reference to neigh entry

Neigh entry has reference counter, however it is only used when scheduling
neigh update event. In all other cases reference to neigh entry is not
taken while working with it. Neigh code relies on synchronization provided
by rtnl lock and uses encap list size as implicit reference counter.

To remove dependency on rtnl lock, always take reference to neigh entry
while using it. Remove neigh entry from hash table and delete it only when
reference counter reaches zero. This can result spurious neigh update
events, when there is an event on entry that has zero encaps attached.
However, such events are rare and properly handled by neigh update handler.

Extend encap entry with reference to neigh hash entry in order to be able
to directly release it when encap is detached, instead of lookup nhe by key
through hash table. Extend nhe with reference to device priv structure to
guarantee correctness when nhe is used with stack devices, bond setup, in
which case it is non-trivial to determine correct device when releasing the
nhe.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Extract code that queues neigh update work into function

As a preparation for following refactoring that removes rtnl lock
dependency from neigh hash entry handlers, extract code that enqueues neigh
update work into standalone function. This commit doesn't change
functionality.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jianbo Liu <jianbol@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net: stmmac: dwc-qos: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: dwmac-anarion: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: dwmac-meson: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: dwmac-meson8b: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: systemport: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: bcmgenet: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pxa168_eth: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: mvneta: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fec: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ezchip: nps_enet: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cirrus: cs89x0: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sxgbe: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ks8851-ml: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: socionext: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

via-rhine: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ethernet: ti: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

amd-xgbe: use devm_platform_ioremap_resource() to simplify code

Use devm_platform_ioremap_resource() to simplify the code a bit.
This is detected by coccinelle.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge tag 'mac80211-next-for-davem-2019-08-21' of git://git./linux/kernel/git/jberg/mac80211-next

Johannes Berg:

====================
Here are a few groups of changes:
 * EDMG channel support (60 GHz, just a single patch)
 * initial 6/7 GHz band support (Arend)
 * association timestamp recording (Ben)
 * rate control improvements for better performance with
   the mt76 driver (Felix)
 * various fixes for previous HE support changes (John)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'mlxsw-Add-devlink-trap-support'

Ido Schimmel says:

====================
mlxsw: Add devlink-trap support

This patchset adds devlink-trap support in mlxsw.

Patches #1-#4 add the necessary APIs and defines in mlxsw.

Patch #5 implements devlink-trap support for layer 2 drops. More drops
will be added in the future.

Patches #6-#7 add selftests to make sure that all the new code paths are
exercised and that the feature is working as expected.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: mlxsw: Add a test case for devlink-trap

Test generic devlink-trap functionality over mlxsw. These tests are not
specific to a single trap, but do not check the devlink-trap common
infrastructure either.

Currently, the only test case is device deletion (by reloading the
driver) while packets are being trapped.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: mlxsw: Add test cases for devlink-trap L2 drops

Test that each supported packet trap is triggered under the right
conditions and that packets are indeed dropped and not forwarded.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: spectrum: Add devlink-trap support

Register supported packet traps (layer 2 drops only, currently) and
associated trap group with devlink during driver initialization.

The amount of traffic generated by these packet drop traps is capped at
10Kpps to ensure the CPU is not overwhelmed by incoming packets.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: Add trap group for layer 2 discards

Discard trap groups are defined in a different enum so that they could
all share the same policer ID: MLXSW_REG_HTGT_TRAP_GROUP_MAX + 1.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: Add layer 2 discard trap IDs

Add the trap IDs used to report layer 2 drops.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: reg: Add new trap actions

Subsequent patches will add discard traps support in mlxsw. The driver
cannot configure such traps with a normal trap action, but needs to use
exception trap action, which also increments an error counter.

On the other hand, when these traps are initialized or set to drop
action, they should use the default drop action set by the firmware.
This guarantees that when the feature is disabled we get the exact same
behavior as before the feature was introduced.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: core: Add API to set trap action

Up until now the action of a trap was never changed during its lifetime.
This is going to change by subsequent patches that will allow devlink to
control the action of certain traps.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mac80211: minstrel_ht: improve rate probing for devices with static fallback

On some devices that only support static rate fallback tables sending rate
control probing packets can be really expensive.
Probing lower rates can already hurt throughput quite a bit. What hurts even
more is the fact that on mt76x0/mt76x2, single probing packets can only be
forced by directing packets at a different internal hardware queue, which
causes some heavy reordering and extra latency.
The reordering issue is mainly problematic while pushing lots of packets to
a particular station. If there is little activity, the overhead of probing is
neglegible.

The static fallback behavior is designed to pretty much only handle rate
control algorithms that use only a very limited set of rates on which the
algorithm switches up/down based on packet error rate.

In order to better support that kind of hardware, this patch implements a
different approach to rate probing where it switches to a slightly higher rate,
waits for tx status feedback, then updates the stats and switches back to
the new max throughput rate. This only triggers above a packet rate of 100
per stats interval (~50ms).
For that kind of probing, the code has to reduce the set of probing rates
a lot more compared to single packet probing, so it uses only one packet
per MCS group which is either slightly faster, or as close as possible to
the max throughput rate.
This allows switching between similar rates with different numbers of
streams. The algorithm assumes that the hardware will work its way lower
within an MCS group in case of retransmissions, so that lower rates don't
have to be probed by the high packets per second rate probing code.

To further reduce the search space, it also does not probe rates with lower
channel bandwidth than the max throughput rate.

At the moment, these changes will only affect mt76x0/mt76x2.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
Link: https://lore.kernel.org/r/20190820095449.45255-4-nbd@nbd.name
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: minstrel_ht: fix default max throughput rate indexes

Use the first supported rate instead of 0 (which can be invalid)

Signed-off-by: Felix Fietkau <nbd@nbd.name>
Link: https://lore.kernel.org/r/20190820095449.45255-3-nbd@nbd.name
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: minstrel_ht: reduce unnecessary rate probing attempts

On hardware with static fallback tables (e.g. mt76x2), rate probing attempts
can be very expensive.
On such devices, avoid sampling rates slower than the per-group max throughput
rate, based on the assumption that the fallback table will take care of probing
lower rates within that group if the higher rates fail.

To further reduce unnecessary probing attempts, skip duplicate attempts on
rates slower than the max throughput rate.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
Link: https://lore.kernel.org/r/20190820095449.45255-2-nbd@nbd.name
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: minstrel_ht: fix per-group max throughput rate initialization

The group number needs to be multiplied by the number of rates per group
to get the full rate index

Fixes: 5935839ad735 ("mac80211: improve minstrel_ht rate sorting by throughput & probability")
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Link: https://lore.kernel.org/r/20190820095449.45255-1-nbd@nbd.name
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

nl80211: Add support for EDMG channels

802.11ay specification defines Enhanced Directional Multi-Gigabit
(EDMG) STA and AP which allow channel bonding of 2 channels and more.

Introduce new NL attributes that are needed for enabling and
configuring EDMG support.

Two new attributes are used by kernel to publish driver's EDMG
capabilities to the userspace:
NL80211_BAND_ATTR_EDMG_CHANNELS - bitmap field that indicates the 2.16
GHz channel(s) that are supported by the driver.
When this attribute is not set it means driver does not support EDMG.
NL80211_BAND_ATTR_EDMG_BW_CONFIG - represent the channel bandwidth
configurations supported by the driver.

Additional two new attributes are used by the userspace for connect
command and for AP configuration:
NL80211_ATTR_WIPHY_EDMG_CHANNELS
NL80211_ATTR_WIPHY_EDMG_BW_CONFIG

New rate info flag - RATE_INFO_FLAGS_EDMG, can be reported from driver
and used for bitrate calculation that will take into account EDMG
according to the 802.11ay specification.

Signed-off-by: Alexei Avshalom Lazar <ailizaro@codeaurora.org>
Link: https://lore.kernel.org/r/1566138918-3823-2-git-send-email-ailizaro@codeaurora.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: fix possible NULL pointerderef in obss pd code

he_spr_ie_elem is dereferenced before the NULL check. fix this by moving
the assignment after the check.

fixes commit 697f6c507c74 ("mac80211: propagate HE operation info into
bss_conf")

This was reported by the static code checker.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: John Crispin <john@phrozen.org>
Link: https://lore.kernel.org/r/20190813070712.25509-1-john@phrozen.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: add assoc-at support

Report timestamp for when sta becomes associated.

Signed-off-by: Ben Greear <greearb@candelatech.com>
Link: https://lore.kernel.org/r/20190809180001.26393-2-greearb@candelatech.com
[fix ktime_get_boot_ns() to ktime_get_boottime_ns(), assoc_at type to u64]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: Support assoc-at timer in sta-info

Report timestamp of when sta became associated.

This is the boottime clock, units are nano-seconds.

Signed-off-by: Ben Greear <greearb@candelatech.com>
Link: https://lore.kernel.org/r/20190809180001.26393-1-greearb@candelatech.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: apply same mandatory rate flags for 5GHz and 6GHz

For the new 6GHz band the same rules apply for mandatory rates so
add it to set_mandatory_flags_band() function.

Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
Reviewed-by: Leon Zegers <leon.zegers@broadcom.com>
Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://lore.kernel.org/r/1564745465-21234-9-git-send-email-arend.vanspriel@broadcom.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: ibss: use 11a mandatory rates for 6GHz band operation

The default mandatory rates, ie. when not specified by user-space, is
determined by the band. Select 11a rateset for 6GHz band.

Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
Reviewed-by: Leon Zegers <leon.zegers@broadcom.com>
Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://lore.kernel.org/r/1564745465-21234-8-git-send-email-arend.vanspriel@broadcom.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: use same IR permissive rules for 6GHz band

The function cfg80211_ir_permissive_chan() is applicable for
6GHz band as well so make sure it is handled.

Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
Reviewed-by: Leon Zegers <leon.zegers@broadcom.com>
Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://lore.kernel.org/r/1564745465-21234-7-git-send-email-arend.vanspriel@broadcom.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: add 6GHz in code handling array with NUM_NL80211_BANDS entries

In nl80211.c there is a policy for all bands in NUM_NL80211_BANDS and
in trace.h there is a callback trace for multicast rates which is per
band in NUM_NL80211_BANDS. Both need to be extended for the new
NL80211_BAND_6GHZ.

Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
Reviewed-by: Leon Zegers <leon.zegers@broadcom.com>
Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://lore.kernel.org/r/1564745465-21234-6-git-send-email-arend.vanspriel@broadcom.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: extend ieee80211_operating_class_to_band() for 6GHz

Add 6GHz operating class range as defined in 802.11ax D4.1 Annex E.

Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
Reviewed-by: Leon Zegers <leon.zegers@broadcom.com>
Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://lore.kernel.org/r/1564745465-21234-5-git-send-email-arend.vanspriel@broadcom.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: util: add 6GHz channel to freq conversion and vice versa

Extend the functions ieee80211_channel_to_frequency() and
ieee80211_frequency_to_channel() to support 6GHz band according
specification in 802.11ax D4.1 27.3.22.2.

Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
Reviewed-by: Leon Zegers <leon.zegers@broadcom.com>
Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://lore.kernel.org/r/1564745465-21234-4-git-send-email-arend.vanspriel@broadcom.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: add 6GHz UNII band definitions

For the new 6GHz there are new UNII band definitions as listed
in the FCC notice [1].

[1] https://docs.fcc.gov/public/attachments/FCC-18-147A1_Rcd.pdf

Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
Reviewed-by: Leon Zegers <leon.zegers@broadcom.com>
Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://lore.kernel.org/r/1564745465-21234-3-git-send-email-arend.vanspriel@broadcom.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

nl80211: add 6GHz band definition to enum nl80211_band

In the 802.11ax specification a new band is introduced, which
is also proposed by FCC for unlicensed use. This band is referred
to as 6GHz spanning frequency range from 5925 to 7125 MHz.

Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
Reviewed-by: Leon Zegers <leon.zegers@broadcom.com>
Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Link: https://lore.kernel.org/r/1564745465-21234-2-git-send-email-arend.vanspriel@broadcom.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: add missing length field increment when generating Radiotap header

The code generating the Tx Radiotap header when using tx_status_ext was
missing a field increment after setting the VHT bandwidth.

Fixes: 3d07ffcaf320 ("mac80211: add struct ieee80211_tx_status support to ieee80211_add_tx_radiotap_header")
Signed-off-by: John Crispin <john@phrozen.org>
Link: https://lore.kernel.org/r/20190807075949.32414-4-john@phrozen.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: 80Mhz was not reported properly when using tx_status_ext

When reporting 80MHz, we need to set 4 and not 2 inside the corresponding
field inside the Tx Radiotap header.

Fixes: 3d07ffcaf320 ("mac80211: add struct ieee80211_tx_status support to ieee80211_add_tx_radiotap_header")
Signed-off-by: John Crispin <john@phrozen.org>
Link: https://lore.kernel.org/r/20190807075949.32414-3-john@phrozen.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: fix bad guard when reporting legacy rates

When reporting legacy rates inside the TX Radiotap header we need to split
the check between "uses tx_statua_ext" and "is legacy rate". Not doing so
would make the code drop into the !tx_status_ext path.

Fixes: 3d07ffcaf320 ("mac80211: add struct ieee80211_tx_status support to ieee80211_add_tx_radiotap_header")
Signed-off-by: John Crispin <john@phrozen.org>
Link: https://lore.kernel.org/r/20190807075949.32414-2-john@phrozen.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: fix TX legacy rate reporting when tx_status_ext is used

The RX Radiotap header length was not calculated properly when reporting
legacy rates using tx_status_ext.

Fixes: 3d07ffcaf320 ("mac80211: add struct ieee80211_tx_status support to ieee80211_add_tx_radiotap_header")
Signed-off-by: John Crispin <john@phrozen.org>
Link: https://lore.kernel.org/r/20190807075949.32414-1-john@phrozen.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Merge tag 'mlx5-updates-2019-08-15' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-updates-2019-08-15

This patchset introduces changes in mlx5 devlink health reporters.
The highlight of these changes is adding a new reporter: RX reporter

mlx5 RX reporter: reports and recovers from timeouts and RX completion
error.

1) Perform TX reporter cleanup. In order to maintain the
code flow as similar as possible between RX and TX reporters, start the
set with cleanup.

2) Prepare for code sharing, generalize and move shared
functionality.

3) Refactor and extend TX reporter diagnostics information
to align the TX reporter diagnostics output with the RX reporter's
diagnostics output.

4) Add helper functions Patch 11: Add RX reporter, initially
supports only the diagnostics call back.

5) Change ICOSQ (Internal Operations Send Queue) open/close flow to
avoid race between interface down and completion error recovery.

6) Introduce recovery flows for RX ring population timeout on ICOSQ,
and for completion errors on ICOSQ and on RQ (Regular receive queues).

7) Include RX reporters in mlx5 documentation.

8) Last two patches of this series, are trivial fixes for previously
submitted patches on this release cycle.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch '100GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
100GbE Intel Wired LAN Driver Updates 2019-08-20

This series contains updates to ice driver only.

Brett fixes the detection of a hung transmit ring by checking the
software based tail (next_to_use) to determine if there is pending work.
Updates the driver to assume that using more than one receive queue per
receive ring container is a rare case, so use unlikely() in the case
were we actually need to divide our budget for multiple queues.  Fixed
an issue where the write back on ITR bit was not being set when
interrupts are disabled, which was causing only write backs when polling
only when a cache line is filled.  Cleans up unnecessary wait times
during VF bring up and reset paths.  Increased the mailbox size for
receive queues that are used to communicate with VFs to accommodate the
large number of VFs that the driver can support.

Akeem restructures the initialization flows for VFs, including how VFs
are configured and resources allocated to improve flows so that when we
clean up resources, we do not try to free resources that were never
allocated.  Organizes code to ensure that VF specific code is located in
the SR-IOV specific file.

Paul fixes an issue when setting the pause parameter which was
incorrectly blocking users from changing receive or transmit pause
settings.  Ensure register access for MSIX vector index is only done in
the PF space and not absolute device space.

Usha fixes a potential kernel hang in the DCB rebuild path when in CEE
mode, where the ETS recommended DCB configuration is not being set or
set correctly.

Mitch updates the driver to process all receive descriptors, regardless
of the size of the associated data.

Tony fixes and issue during the reset/rebuild path of a PF VSI where we
were assuming that the PF VSI was always to be enabled, which can
attempt to bring up a PF VSI on a downed interface which can lead to
various crashes.

Pawel fixes up variable definitions to match the type of data being
stored.

v2: Dropped patch 1 of the series to add ethtool support to query/add
    channels on a VSI, while we re-qork the functionality to match the
    ethtool expected behavior to report combined (Tx and Rx) numbers.
v3: Updated patch 4 to use kzalloc() and kfree() instead devm_kzalloc()
    and devm_kfree().
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

ice: improve print for VF's when adding/deleting MAC filters

When we fail to add/delete MAC filters in the VF, the print doesn't
distinguish between the two. Fix that by printing whether or not we
failed to add/delete the MAC filter respectively.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Change type for queue counts

These queue variables are being assigned values that are type u16.
Change the local variables to match these types. Since these
represent queue counts, they should never be negative.

Signed-off-by: Pawel Kaminski <pawel.kaminski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Move VF resources definition to SR-IOV specific file

In order to use some of the VF resources definition in the SR-IOV specific
virtchnl header file, this patch moves applicable code to
ice_virtchnl_pf.h file accordingly... and they should have been defined in
the destination file originally.

Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Increase size of Mailbox receive queue for many VFs

Currently we use the ICE_MBXQ_LEN for both the Mailbox send and receive
queues that are used to communicate with VFs. This is fine for the send
queue because the PF driver will lock the queue for every single send,
but for the Mailbox receive queue every VF is posting to its Mailbox
send queue and the hardware is then handing the message to the PF on its
Mailbox receive queue. This becomes a problem with many VFs because it
seems to overburden the Mailbox receive queue on the PF. Fix this by
increasing the Mailbox receive queue for the PF to 512 entries.

The number 512 was determined based on the number of VFs supported by
the device. We can have a total of 256 VFs so in the worst case this
allows the VFs to put 2 messages in the PFs Mailbox receive queue at the
same time.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Reduce wait times during VF bringup/reset

Currently there are a couple places where the VF is waiting too long when
checking the status of registers. This is causing the AVF driver to
spin for longer than necessary in the __IAVF_STARTUP state. Sometimes
it causes the AVF to go into the __IAVF_COMM_FAILED, which may retrigger
the __IAVF_STARTUP state. Try to reduce the chance of this happening by
removing unnecessary wait times in VF bringup/resets.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: update GLINT_DYN_CTL and GLINT_VECT2FUNC register access

Register access for GLINT_DYN_CTL and GLINT_VECT2FUNC should be within
the PF space and not the absolute device space.

Signed-off-by: Paul Greenwalt <paul.greenwalt@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Do not always bring up PF VSI in ice_ena_vsi()

During rebuild ice_ena_vsi() is called to recover the VSI state.
This function assumes the PF VSI is always to be enabled, however,
it's possible that during reset/rebuild the interface can be
brought down.  If this occurs, we can attempt to bring up the PF
VSI on a downed interface which can lead to various crashes. If
the interface is not running, do not bring up the associated VSI.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: allow empty Rx descriptors

In some circumstances, the hardware will hand us a receive descriptor
which has no data attached, but is otherwise valid. The receive code was
improperly ignoring these descriptors, which result in an infinite loop.

To fix this, change the receive code to process all descriptors,
regardless of the size of the associated data. Add checks to the
memory-handling functions to allow for zero size.

Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Fix kernel hang with DCB reset in CEE mode

This patch fixes the set local MIB AQ call failures in the DCB rebuild path
by setting the defaults for the ETS recommended DCB configuration. Also,
willing bits for the DCB configuration needs to be set correctly. Resets
works fine in IEEE mode as the ETS recommended DCB configuration is
populated but not in CEE mode.
Without this patch, PFR causes the kernel hang in CEE mode.

Signed-off-by: Usha Ketineni <usha.k.ketineni@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ice: Set WB_ON_ITR when we don't re-enable interrupts

Currently when busy polling is enabled we aren't setting/enabling
WB_ON_ITR in the driver. This doesn't break the driver, but it does
cause issues. If we don't enable WB_ON_ITR mode we will still get
write-backs from hardware during polling when a cache line has been
filled, but if a cache line is not filled we will not get the
write-back because WB_ON_ITR is not set. Fix this by enabling
WB_ON_ITR in the driver when interrupts are disabled.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

Merge tag 'linux-can-next-for-5.4-20190820' of git://git./linux/kernel/git/mkl/linux-can-next

Marc Kleine-Budde says:

====================
pull-request: can-next 2019-08-20

this is a pull request for net-next/master consisting of 18 patches.

The first patch is by Geert Uytterhoeven, it removes the unused platform
data support from the rcar_can driver.

A patch by Nishka Dasgupta marks the structure peak_pciec_i2c_bit_ops in
the peak_pci driver as constant.

A patch by me removes the custom DMA support from the hi311x driver.

The next 4 patches target the tcan4x5x driver and are also by me, they
first clean up the driver a bit, and then add missing error handling and
fix a bug in the length calculation in the regmap callbacks.

The next 2 patches are by me for the m_can_platform driver, they also
remove unneeded casts and add missing error handling.

The remaining 9 patches all target the mcp251x driver. The first 5 are
clean up patches by me, the next relaxes the timing in the
mcp251x_hw_reset() function. Alexander Shiyan's patch improves the name
which is used while registering the interrupt handler. Phil Elwell's
patch improves the mcp251x_open() function to use the DT-supplied
interrupt flags instead of hard coding them. The final patch is again by
me, it removes the custom DMA support from the hi311x driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

ice: fix set pause param autoneg check

When ETHTOOL_GLINKSETTINGS is defined get pause param pause->autoneg
reports SW configured setting, however when not defined get pause param
pause->autoneg reports the link status. Set pause param needs to compare
pause->autoneg with the same source as get pause param to block the user
from changing autoneg with the set pause param option, or the user
may be incorrectly blocked from changing Rx|Tx pause settings.

Signed-off-by: Paul Greenwalt <paul.greenwalt@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

Merge branch 's390-net-next'

Julian Wiedmann says:

====================
s390/net: updates 2019-08-20

please apply the following patches to net-next. This series brings a mix
of cleanups and small improvements for various parts of qeth's control
path. Also, a minor cleanup for ctcm and lcs.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

s390/lcs: don't use intparm for channel IO

lcs passes an intparm when calling ccw_device_*(), even though lcs_irq()
later makes no use of this.

To reduce the confusion, consistently pass 0 as intparm instead.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Sebastian Ott <sebott@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/ctcm: don't use intparm for channel IO

ctcm passes an intparm when calling ccw_device_*(), even though
ctcm_irq_handler() later makes no use of this.

To reduce the confusion, consistently pass 0 as intparm instead.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Sebastian Ott <sebott@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: streamline control code for promisc mode

We have logic to determine the desired promisc mode in _each_ code path.
Change things around so that there is a clean split between
(a) high-level code that selects the new mode, and (b) implementations
of the various mechanisms to program this mode.

This also keeps qeth_promisc_to_bridge() from polluting the debug logs
on each RX modeset.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Alexandra Winter <wintera@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: get vnicc sub-cmd type from reply data

When processing the reply for a vnicc cmd, there's no need to remember
which specific sub-cmd type we initially sent. The reply itself contains
all the needed information.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: merge qeth_reply struct into qeth_cmd_buffer

Except for card->read_cmd, every cmd we issue now passes through
qeth_send_control_data() and allocates a qeth_reply struct. The way we
use this struct requires additional refcounting, and pointer tracking.

Clean up things by moving most of qeth_reply's content into the main
cmd struct. This keeps things in one place, saves us the additional
refcounting and simplifies the overall code flow.
A nice little benefit is that we can now match incoming replies against
the pending requests themselves, without caching the requests' seqnos.

The qeth_reply struct stays around for a little bit longer in a shrunk
form, to avoid touching every single callback.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: keep cmd alive after IO completion

Current code releases the cmd struct after its initial IO has completed.
Any reply processing is done independently, using a separate qeth_reply
struct.
In preparation for merging the cmd and reply structs together, take an
additional reference on the cmd object so that it stays around all the
way until qeth_send_control_data() returns.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: use correct length field in SNMP cmd callback

qeth_snmp_command_cb() is the only cmd callback that pulls the reply's
data length from a low-level transport header field. This requires
additional complexity (ie. reply->offset) to make the header accessible
to what is supposed to be a pure IPA cmd callback.

Adapter cmds have a length field in their sub-cmd header, get the data
length from there instead.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: propagate length of processed cmd IO data to callback

When an cmd IO completes in qeth_irq(), calculate how much data was
processed by the device and pass this value to the cmd's callback.

This allows cmds that retrieve data from the device to check whether
sufficient data was received, so we do that in qeth_read_conf_data_cb().

Suggested-by: Jens Remus <jremus@linux.ibm.com>
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390/qeth: use node_descriptor struct

Rather than fumbling with hard-coded offsets, use the proper struct to
access the retrieved RCD information.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

netdevsim: Fix build error without CONFIG_INET

If CONFIG_INET is not set, building fails:

drivers/net/netdevsim/dev.o: In function `nsim_dev_trap_report_work':
dev.c:(.text+0x67b): undefined reference to `ip_send_check'

Use ip_fast_csum instead of ip_send_check to avoid
dependencies on CONFIG_INET.

Reported-by: Hulk Robot <hulkci@huawei.com>
Fixes: da58f90f11f5 ("netdevsim: Add devlink-trap support")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Tested-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/mlx5: Fix the order of fc_stats cleanup

Previously, mlx5_cleanup_fc_stats() would cleanup the flow counter
pool beofre releasing all the counters to it, which would result in
flow counter bulks not getting freed. Resolve this by changing the
order in which elements of fc_stats are cleaned up, so that the flow
counter pool is cleaned up after all the counters are released.

Also move cleanup actions for freeing the bulk query memory and
destroying the idr to the end of mlx5_cleanup_fc_stats().

Signed-off-by: Gavi Teitz <gavi@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Fix deallocation of non-fully init encap entries

Recent rtnl lock dependency refactoring changed encap entry attach code to
insert encap entry to hash table before it was fully initialized in order
to allow concurrent tc users to wait on completion for encap entry to
finish initialization. That change required all the users of encap entry to
obtain reference to it first and for caller that creates encap to put
reference to it on error, instead of freeing the entry memory directly.
However, releasing reference to such encap entry that wasn't fully
initialized causes NULL pointer dereference in
mlx5e_rep_encap_entry_detach() which expects e->out_dev to be set and encap
to be attached to nhe:

[ 1092.454517] BUG: unable to handle page fault for address: 00000000000420e8
[ 1092.454571] #PF: supervisor read access in kernel mode
[ 1092.454602] #PF: error_code(0x0000) - not-present page
[ 1092.454632] PGD 800000083032c067 P4D 800000083032c067 PUD 84107d067 PMD 0
[ 1092.454673] Oops: 0000 [#1] SMP PTI
[ 1092.454697] CPU: 20 PID: 22393 Comm: tc Not tainted 5.3.0-rc3+ #589
[ 1092.454733] Hardware name: Supermicro SYS-2028TP-DECR/X10DRT-P, BIOS 2.0b 03/30/2017
[ 1092.454806] RIP: 0010:mlx5e_rep_encap_entry_detach+0x1c/0x630 [mlx5_core]
[ 1092.454845] Code: be f4 ff ff ff e9 11 ff ff ff 0f 1f 40 00 0f 1f 44 00 00 55 48 89 e5 41 57 41 56 41 55 41 54 49 89 fc 53 48 89 f3 48 83 ec 30 <48> 8b 87 28 16 04 00 48 89 f7 48 05 d0 03 00 00 48 89
 45 c8 e8 cb
[ 1092.454942] RSP: 0018:ffffb6f08421f5a0 EFLAGS: 00010286
[ 1092.454974] RAX: 0000000000000000 RBX: ffff8ab668644e00 RCX: ffffb6f08421f56c
[ 1092.455013] RDX: ffff8ab668644e40 RSI: ffff8ab668644e00 RDI: 0000000000000ac0
[ 1092.455053] RBP: ffffb6f08421f5f8 R08: 0000000000000001 R09: 0000000000000000
[ 1092.455092] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000ac0
[ 1092.455131] R13: 00000000ffffff9b R14: ffff8ab63f200ac0 R15: ffff8ab668644e40
[ 1092.455171] FS:  00007fa195bdc480(0000) GS:ffff8ab66fa00000(0000) knlGS:0000000000000000
[ 1092.455216] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 1092.455249] CR2: 00000000000420e8 CR3: 0000000867522001 CR4: 00000000001606e0
[ 1092.455288] Call Trace:
[ 1092.455315]  ? __mutex_unlock_slowpath+0x4d/0x2a0
[ 1092.455365]  mlx5e_encap_dealloc.isra.0+0x31/0x60 [mlx5_core]
[ 1092.455424]  mlx5e_tc_add_fdb_flow+0x596/0x750 [mlx5_core]
[ 1092.455484]  __mlx5e_add_fdb_flow+0x152/0x210 [mlx5_core]
[ 1092.455534]  mlx5e_configure_flower+0x4d5/0xe30 [mlx5_core]
[ 1092.455574]  tc_setup_cb_call+0x67/0xb0
[ 1092.455601]  fl_hw_replace_filter+0x142/0x300 [cls_flower]
[ 1092.455639]  fl_change+0xd24/0x1bdb [cls_flower]
[ 1092.455675]  tc_new_tfilter+0x3e0/0x970
[ 1092.455709]  ? tc_del_tfilter+0x720/0x720
[ 1092.455735]  rtnetlink_rcv_msg+0x389/0x4b0
[ 1092.455763]  ? netlink_deliver_tap+0x95/0x400
[ 1092.455791]  ? rtnl_dellink+0x2d0/0x2d0
[ 1092.455817]  netlink_rcv_skb+0x49/0x110
[ 1092.455844]  netlink_unicast+0x171/0x200
[ 1092.455872]  netlink_sendmsg+0x224/0x3f0
[ 1092.455901]  sock_sendmsg+0x5e/0x60
[ 1092.455924]  ___sys_sendmsg+0x2ae/0x330
[ 1092.455950]  ? task_work_add+0x43/0x50
[ 1092.455976]  ? fput_many+0x45/0x80
[ 1092.456004]  ? __lock_acquire+0x248/0x18e0
[ 1092.456033]  ? find_held_lock+0x2b/0x80
[ 1092.456058]  ? task_work_run+0x7b/0xd0
[ 1092.456085]  __sys_sendmsg+0x59/0xa0
[ 1092.457013]  do_syscall_64+0x5c/0xb0
[ 1092.457924]  entry_SYSCALL_64_after_hwframe+0x49/0xbe
[ 1092.458842] RIP: 0033:0x7fa195da27b8
[ 1092.459918] Code: 89 02 48 c7 c0 ff ff ff ff eb bb 0f 1f 80 00 00 00 00 f3 0f 1e fa 48 8d 05 65 8f 0c 00 8b 00 85 c0 75 17 b8 2e 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 58 c3 0f 1f 80 00 00 00 00 48 83
 ec 28 89 54
[ 1092.462634] RSP: 002b:00007fff94409298 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
[ 1092.464011] RAX: ffffffffffffffda RBX: 000000005d515b0e RCX: 00007fa195da27b8
[ 1092.465391] RDX: 0000000000000000 RSI: 00007fff94409300 RDI: 0000000000000003
[ 1092.466761] RBP: 0000000000000000 R08: 0000000000000001 R09: 0000000000000006
[ 1092.468121] R10: 0000000000404ec2 R11: 0000000000000246 R12: 0000000000000001
[ 1092.469456] R13: 0000000000480640 R14: 0000000000000016 R15: 0000000000000001
[ 1092.470766] Modules linked in: act_mirred act_tunnel_key cls_flower dummy vxlan ip6_udp_tunnel udp_tunnel sch_ingress nfsv3 nfs_acl nfs lockd grace fscache tun bridge stp llc sunrpc rdma_ucm rdma_cm
iw_cm ib_cm mlx5_ib ib_uverbs ib_core intel_rapl_msr intel_rapl_common sb_edac x86_pkg_temp_thermal intel_powerclamp coretemp mlx5_core kvm_intel kvm irqbypass crct10dif_pclmul mei_me crc32_pclmul crc32
c_intel igb iTCO_wdt ghash_clmulni_intel ses mlxfw intel_cstate iTCO_vendor_support ptp intel_uncore lpc_ich pps_core mei i2c_i801 joydev intel_rapl_perf ioatdma enclosure ipmi_ssif pcspkr dca wmi ipmi_
si ipmi_devintf ipmi_msghandler acpi_pad acpi_power_meter ast i2c_algo_bit drm_vram_helper ttm drm_kms_helper drm mpt3sas raid_class scsi_transport_sas
[ 1092.479618] CR2: 00000000000420e8
[ 1092.481214] ---[ end trace ce2e0f4d9a67f604 ]---

To fix the issue, set e->compl_result to positive value after encap was
initialized successfully. Check e->compl_result value in
mlx5e_encap_dealloc() and only detach and dealloc encap if the value is
positive.

Fixes: d589e785baf5 ("net/mlx5e: Allow concurrent creation of encap entries")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

Documentation: net: mlx5: Devlink health documentation updates

Add documentation for devlink health rx reporter supported by mlx5.
Update tx reporter documentation.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Report and recover from CQE with error on RQ

Add support for report and recovery from error on completion on RQ by
setting the queue back to ready state. Handle only errors with a
syndrome indicating the RQ might enter error state and could be
recovered.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: RX, Handle CQE with error at the earliest stage

Just to be aligned with the MPWQE handlers, handle RX WQE with error
for legacy RQs in the top RX handlers, just before calling skb_from_cqe().

CQE error handling will now be called at the same stage regardless of
the RQ type or netdev mode NIC, Representor, IPoIB, etc ..

This will be useful for down stream patch to improve error CQE
handling.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Report and recover from rx timeout

Add support for report and recovery from rx timeout. On driver open we
post NOP work request on the rx channels to trigger napi in order to
fillup the rx rings. In case napi wasn't scheduled due to a lost
interrupt, perform EQ recovery.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Report and recover from CQE error on ICOSQ

Add support for report and recovery from error on completion on ICOSQ.
Deactivate RQ and flush, then deactivate ICOSQ. Set the queue back to
ready state (firmware) and reset the ICOSQ and the RQ (software
resources). Finally, activate the ICOSQ and the RQ.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Split open/close ICOSQ into stages

Align ICOSQ open/close behaviour with RQ and SQ. Split open flow into
open and activate where open handles creation and activate enables the
queue. Do a symmetric thing in close flow: split into close and
deactivate.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Add support to rx reporter diagnose

Add rx reporter, which supports diagnose call-back. Diagnostics output
include: information common to all RQs: RQ type, RQ size, RQ stride
size, CQ size and CQ stride size. In addition advertise information per
RQ and its related icosq and attached CQ.

$ devlink health diagnose pci/0000:00:0b.0 reporter rx
 Common config:
   RQ:
     type: 2 stride size: 2048 size: 8
   CQ:
     stride size: 64 size: 1024
 RQs:
   channel ix: 0 rqn: 4308 HW state: 1 SW state: 3 posted WQEs: 7 cc: 7 ICOSQ HW state: 1
   CQ:
     cqn: 1032 HW status: 0
   channel ix: 1 rqn: 4313 HW state: 1 SW state: 3 posted WQEs: 7 cc: 7 ICOSQ HW state: 1
   CQ:
     cqn: 1036 HW status: 0
   channel ix: 2 rqn: 4318 HW state: 1 SW state: 3 posted WQEs: 7 cc: 7 ICOSQ HW state: 1
   CQ:
     cqn: 1040 HW status: 0
   channel ix: 3 rqn: 4323 HW state: 1 SW state: 3 posted WQEs: 7 cc: 7 ICOSQ HW state: 1
   CQ:
     cqn: 1044 HW status: 0

$ devlink health diagnose pci/0000:00:0b.0 reporter rx -jp
{
    "Common config": {
        "RQ": {
            "type": 2,
            "stride size": 2048,
            "size": 8
        },
        "CQ": {
            "stride size": 64,
            "size": 1024
        }
    },
    "RQs": [ {
            "channel ix": 0,
            "rqn": 4308,
            "HW state": 1,
            "SW state": 3,
            "posted WQEs": 7,
            "cc": 7,
            "ICOSQ HW state": 1,
            "CQ": {
                "cqn": 1032,
                "HW status": 0
            }
        },{
            "channel ix": 1,
            "rqn": 4313,
            "HW state": 1,
            "SW state": 3,
            "posted WQEs": 7,
            "cc": 7,
            "ICOSQ HW state": 1,
            "CQ": {
                "cqn": 1036,
                "HW status": 0
            }
        },{
            "channel ix": 2,
            "rqn": 4318,
            "HW state": 1,
            "SW state": 3,
            "posted WQEs": 7,
            "cc": 7,
            "ICOSQ HW state": 1,
            "CQ": {
                "cqn": 1040,
                "HW status": 0
            }
        },{
            "channel ix": 3,
            "rqn": 4323,
            "HW state": 1,
            "SW state": 3,
            "posted WQEs": 7,
            "cc": 7,
            "ICOSQ HW state": 1,
            "CQ": {
                "cqn": 1044,
                "HW status": 0
            }
        } ]
}

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Add helper functions for reporter's basics

Introduce helper functions for create and destroy reporters and update
channels. In the following patch, rx reporter is added and it will use
these helpers too.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Add cq info to tx reporter diagnose

Add cq information to general diagnose output: CQ size and stride size.
Per SQ add information about the related CQ: cqn and CQ's HW status.

$ devlink health diagnose pci/0000:00:0b.0 reporter tx
 Common Config:
   SQ:
     stride size: 64 size: 1024
   CQ:
     stride size: 64 size: 1024
 SQs:
   channel ix: 0 tc: 0 txq ix: 0 sqn: 4307 HW state: 1 stopped: false cc: 0 pc: 0
   CQ:
     cqn: 1030 HW status: 0
   channel ix: 1 tc: 0 txq ix: 1 sqn: 4312 HW state: 1 stopped: false cc: 0 pc: 0
   CQ:
     cqn: 1034 HW status: 0
   channel ix: 2 tc: 0 txq ix: 2 sqn: 4317 HW state: 1 stopped: false cc: 0 pc: 0
   CQ:
     cqn: 1038 HW status: 0
   channel ix: 3 tc: 0 txq ix: 3 sqn: 4322 HW state: 1 stopped: false cc: 0 pc: 0
   CQ:
     cqn: 1042 HW status: 0

$ devlink health diagnose pci/0000:00:0b.0 reporter tx -jp
{
    "Common Config": {
        "SQ": {
            "stride size": 64,
            "size": 1024
        },
        "CQ": {
            "stride size": 64,
            "size": 1024
        }
    },
    "SQs": [ {
            "channel ix": 0,
            "tc": 0,
            "txq ix": 0,
            "sqn": 4307,
            "HW state": 1,
            "stopped": false,
            "cc": 0,
            "pc": 0,
            "CQ": {
                "cqn": 1030,
                "HW status": 0
            }
        },{
            "channel ix": 1,
            "tc": 0,
            "txq ix": 1,
            "sqn": 4312,
            "HW state": 1,
            "stopped": false,
            "cc": 0,
            "pc": 0,
            "CQ": {
                "cqn": 1034,
                "HW status": 0
            }
        },{
            "channel ix": 2,
            "tc": 0,
            "txq ix": 2,
            "sqn": 4317,
            "HW state": 1,
            "stopped": false,
            "cc": 0,
            "pc": 0,
            "CQ": {
                "cqn": 1038,
                "HW status": 0
            }
        },{
            "channel ix": 3,
            "tc": 0,
            "txq ix": 3,
            "sqn": 4322,
            "HW state": 1,
            "stopped": false,
            "cc": 0,
            "pc": 0,
            "CQ": {
                "cqn": 1042,
                "HW status": 0
        } ]
}

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Extend tx reporter diagnostics output

Enhance tx reporter's diagnostics output to include: information common
to all SQs: SQ size, SQ stride size.
In addition add channel ix, tc, txq ix, cc and pc.

$ devlink health diagnose pci/0000:00:0b.0 reporter tx
 Common config:
   SQ:
     stride size: 64 size: 1024
 SQs:
   channel ix: 0 tc: 0 txq ix: 0 sqn: 4307 HW state: 1 stopped: false cc: 0 pc: 0
   channel ix: 1 tc: 0 txq ix: 1 sqn: 4312 HW state: 1 stopped: false cc: 0 pc: 0
   channel ix: 2 tc: 0 txq ix: 2 sqn: 4317 HW state: 1 stopped: false cc: 0 pc: 0
   channel ix: 3 tc: 0 txq ix: 3 sqn: 4322 HW state: 1 stopped: false cc: 0 pc: 0

$ devlink health diagnose pci/0000:00:0b.0 reporter tx -jp
{
    "Common config": {
        "SQ": {
            "stride size": 64,
            "size": 1024
        }
    },
    "SQs": [ {
            "channel ix": 0,
            "tc": 0,
            "txq ix": 0,
            "sqn": 4307,
            "HW state": 1,
            "stopped": false,
            "cc": 0,
            "pc": 0
        },{
            "channel ix": 1,
            "tc": 0,
            "txq ix": 1,
            "sqn": 4312,
            "HW state": 1,
            "stopped": false,
            "cc": 0,
            "pc": 0
        },{
            "channel ix": 2,
            "tc": 0,
            "txq ix": 2,
            "sqn": 4317,
            "HW state": 1,
            "stopped": false,
            "cc": 0,
            "pc": 0
        },{
            "channel ix": 3,
            "tc": 0,
            "txq ix": 3,
            "sqn": 4322,
            "HW state": 1,
            "stopped": false,
            "cc": 0,
            "pc": 0
         } ]
}

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Extend tx diagnose function

The following patches in the set enhance the diagnostics info of tx
reporter. Therefore, it is better to pass a pointer to the SQ for
further data extraction.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Generalize tx reporter's functionality

Prepare for code sharing with rx reporter, which is added in the
following patches in the set. Introduce a generic error_ctx for
agnostic recovery despatch.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Change naming convention for reporter's functions

Change from mlx5e_tx_reporter_* to mlx5e_reporter_tx_*. In the following
patches in the set rx reporter is added, the new naming convention is
more uniformed.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

net/mlx5e: Rename reporter header file

Rename reporter.h -> health.h so patches in the set can use it for
health related functionality.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

Merge branch 'net-dsa-enable-and-disable-all-ports'

Vivien Didelot says:

====================
net: dsa: enable and disable all ports

The DSA stack currently calls the .port_enable and .port_disable switch
callbacks for slave ports only. However, it is useful to call them for all
port types. For example this allows some drivers to delay the optimization
of power consumption after the switch is setup. This can also help reducing
the setup code of drivers a bit.

The first DSA core patches enable and disable all ports of a switch, regardless
their type. The last mv88e6xxx patches remove redundant code from the driver
setup and the said callbacks, now that they handle SERDES power for all ports.

Changes in v2: do not guard .port_disable for broadcom switches.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: mv88e6xxx: wrap SERDES IRQ in power function

Now that mv88e6xxx_serdes_power is only called after driver setup,
we can wrap the SERDES IRQ code directly within it for clarity.

Signed-off-by: Vivien Didelot <vivien.didelot@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>