net: switchdev: Replace port attr set SDO with a notification

Drop switchdev_ops.switchdev_port_attr_set. Drop the uses of this field
from all clients, which were migrated to use switchdev notification in
the previous patches.

Add a new function switchdev_port_attr_notify() that sends the switchdev
notifications SWITCHDEV_PORT_ATTR_SET and calls the blocking (process)
notifier chain.

We have one odd case within net/bridge/br_switchdev.c with the
SWITCHDEV_ATTR_ID_PORT_PRE_BRIDGE_FLAGS attribute identifier that
requires executing from atomic context, we deal with that one
specifically.

Drop __switchdev_port_attr_set() and update switchdev_port_attr_set()
likewise.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

staging: fsl-dpaa2: ethsw: Handle SWITCHDEV_PORT_ATTR_SET

Following patches will change the way we communicate setting a port's
attribute and use a blocking notifier to perform those tasks.

Prepare ethsw to support receiving notifier events targeting
SWITCHDEV_PORT_ATTR_SET and simply translate that into the existing
swdev_port_attr_set() call.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: mscc: ocelot: Handle SWITCHDEV_PORT_ATTR_SET

Following patches will change the way we communicate setting a port's
attribute and use notifiers to perform those tasks.

Ocelot does not currently have an atomic notifier registered for
switchdev events, so we need to register one in order to deal with
atomic context SWITCHDEV_PORT_ATTR_SET events.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: spectrum_switchdev: Handle SWITCHDEV_PORT_ATTR_SET

Following patches will change the way we communicate setting a port's
attribute and use a notifier to perform those tasks.

Prepare mlxsw to support receiving notifier events targeting
SWITCHDEV_PORT_ATTR_SET and utilize the switchdev_handle_port_attr_set()
to handle stacking of devices.

Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: Handle SWITCHDEV_PORT_ATTR_SET

Following patches will change the way we communicate setting a port's
attribute and use notifiers towards that goal.

Prepare DSA to support receiving notifier events targeting
SWITCHDEV_PORT_ATTR_SET from both atomic and process context and use a
small helper to translate the event notifier into something that
dsa_slave_port_attr_set() can process.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

rocker: Handle SWITCHDEV_PORT_ATTR_SET

Following patches will change the way we communicate setting a port's
attribute and use notifiers towards that goal.

Prepare rocker to support receiving notifier events targeting
SWITCHDEV_PORT_ATTR_SET from both atomic and process context and use a
small helper to translate the event notifier into something that
rocker_port_attr_set() can process.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

switchdev: Add SWITCHDEV_PORT_ATTR_SET

In preparation for allowing switchdev enabled drivers to veto specific
attribute settings from within the context of the caller, introduce a
new switchdev notifier type for port attributes.

Suggested-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Revert "net: sched: fw: don't set arg->stop in fw_walk() when empty"

This reverts commit 31a998487641 ("net: sched: fw: don't set arg->stop in
fw_walk() when empty")

Cls API function tcf_proto_is_empty() was changed in commit
6676d5e416ee ("net: sched: set dedicated tcf_walker flag when tp is empty")
to no longer depend on arg->stop to determine that classifier instance is
empty. Instead, it adds dedicated arg->nonempty field, which makes the fix
in fw classifier no longer necessary.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ethtool: Use explicit designated initializers for .cmd

Initialize the .cmd member by using a designated struct
initializer. This fixes warning of missing field initializers,
and makes code a little easier to read.

Signed-off-by: Li RongQing <lirongqing@baidu.com>
Reviewed-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: pie: fix 64-bit division

Use div_u64() to resolve build failures on 32-bit platforms.

Fixes: 3f7ae5f3dc52 ("net: sched: pie: add more cases to auto-tune alpha and beta")
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Use RCU_POINTER_INITIALIZER() to init static variable

This pointer is RCU protected, so proper primitives should be used.

Signed-off-by: Zhang Yu <zhangyu31@baidu.com>
Signed-off-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'tcp-cleanups'

Eric Dumazet says:

====================
tcp: cleanups for linux-5.1

This small patch series cleanups few things, and add a small
timewait optimization for hosts not using md5.
====================

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

tcp: remove tcp_queue argument from tso_fragment()

tso_fragment() is only called for packets still in write queue.

Remove the tcp_queue parameter to make this more obvious,
even if the comment clearly states this.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: use tcp_md5_needed for timewait sockets

This might speedup tcp_twsk_destructor() a bit,
avoiding a cache line miss.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: convert tcp_md5_needed to static_branch API

We prefer static_branch_unlikely() over static_key_false() these days.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: get rid of __tcp_add_write_queue_tail()

This helper is only used from tcp_add_write_queue_tail(), and does
not make the code more readable.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: get rid of tcp_check_send_head()

This helper is used only once, and its name is no longer relevant.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tc-testing: gitignore, ignore local tdc config file

Comment in tdc_config.py recommends putting customizations in
tdc_config_local.py file that wasn't included in gitignore. Add the local
config file to gitignore.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: fix typo in walker_check_empty()

Function walker_check_empty() incorrectly verifies that tp pointer is not
NULL, instead of actual filter pointer. Fix conditional to check the right
pointer. Adjust filter pointer naming accordingly to other cls API
functions.

Fixes: 6676d5e416ee ("net: sched: set dedicated tcf_walker flag when tp is empty")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reported-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: pie: fix mistake in reference link

Fix the incorrect reference link to RFC 8033

Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mlxsw: spectrum: remove set but not used variable 'autoneg_status'

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

drivers/net/ethernet/mellanox/mlxsw/spectrum.c: In function 'mlxsw_sp_port_get_link_ksettings':
drivers/net/ethernet/mellanox/mlxsw/spectrum.c:3062:5: warning:
 variable 'autoneg_status' set but not used [-Wunused-but-set-variable]

It's not used since commit 475b33cb66c9 ("mlxsw: spectrum: Remove unsupported
eth_proto_lp_advertise field in PTYS")

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

Merge branch 'vxlan-create-and-changelink-extack-support'

Roopa Prabhu says:

====================
vxlan: create and changelink extack support

This series adds extack support to changelink paths.
In the process re-factors flag sets to a separate helper.
Also adds some changelink testcases to rtnetlink.sh

(This series was initially part of another series that
tried to support changelink for more attributes.
But after some feedback from sabrina, i have dropped the
'support changelink for more attributes' part because some
of them cannot be supported today or may require additional
use-case handling code. These can be done separately
as and when we see the need for it.)
====================

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

tools: selftests: rtnetlink: add testcases for vxlan flag sets

This patch extends rtnetlink.sh to cover some vxlan flag
netlink attribute sets.

Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vxlan: add extack support for create and changelink

This patch adds extack coverage in vxlan link
create and changelink paths. Introduces a new helper
vxlan_nl2flags to consolidate flag attribute validation.

thanks to Johannes Berg for some tips to construct the
generic vxlan flag extack strings.

Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Reviewed-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'devlink-make-ethtool-compat-reliable'

Jakub Kicinski says:

====================
devlink: make ethtool compat reliable

This is a follow up to the series which added device flash
updates via devlink. I went with the approach of adding a
new NDO in the end. It seems to end up looking cleaner.

First patch removes the option to build devlink as a module.
Users can still decide to not build it, but the module option
ends up not being worth the maintenance cost.

Next two patches add a NDO which can be used to ask the driver
to return a devlink instance associated with a given netdev,
instead of iterating over devlink ports. Drivers which implement
this NDO must take into account the potential impact on the
visibility of the devlink instance.

With the new NDO in place we can remove NFP ethtool flash update
code.

Fifth patch makes sure we hold a reference to dev while
callbacks are active.

Last but not least the NULL-check of devlink->ops is moved
to instance allocation time.

Last but not least missing checks for devlink->ops are added.
There is currently no driver registering devlink without ops,
so can just fix this in -next.

v2 (Michal): add netdev_to_devlink() in patch 3.
v3 (Florian):
 - add missing checks for devlink->ops;
 - move locking/holding into devlink_compat_ functions.
v4 (Jiri):
 - hold devlink_mutex around callbacks (patch 2);
 - require non-NULL ops (patch 6).
====================

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

devlink: require non-NULL ops for devlink instances

Commit 76726ccb7f46 ("devlink: add flash update command") and
commit 2d8dc5bbf4e7 ("devlink: Add support for reload")
access devlink ops without NULL-checking. There is, however, no
driver which would pass in NULL ops, so let's just make that
a requirement. Remove the now unnecessary NULL-checking.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

devlink: hold a reference to the netdevice around ethtool compat

When ethtool is calling into devlink compat code make sure we have
a reference on the netdevice on which the operation was invoked.

v3: move the hold/lock logic into devlink_compat_* functions (Florian)

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

nfp: remove ethtool flashing fallback

Now that devlink fallback will be called reliably, we can remove
the ethtool flashing code.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

nfp: add .ndo_get_devlink

Support getting devlink instance from a new NDO.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

devlink: create a special NDO for getting the devlink instance

Instead of iterating over all devlink ports add a NDO which
will return the devlink instance from the driver.

v2: add the netdev_to_devlink() helper (Michal)
v3: check that devlink has ops (Florian)
v4: hold devlink_mutex (Jiri)

Suggested-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: devlink: turn devlink into a built-in

Being able to build devlink as a module causes growing pains.
First all drivers had to add a meta dependency to make sure
they are not built in when devlink is built as a module.  Now
we are struggling to invoke ethtool compat code reliably.

Make devlink code built-in, users can still not build it at
all but the dynamically loadable module option is removed.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: remove unused struct inet_frag_queue.fragments field

Now that all users of struct inet_frag_queue have been converted
to use 'rb_fragments', remove the unused 'fragments' field.

Build with `make allyesconfig` succeeded. ip_defrag selftest passed.

Signed-off-by: Peter Oskolkov <posk@google.com>
Acked-by: Stefan Schmidt <stefan@datenfreihafen.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: wan: z85230: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles

dev_consume_skb_irq() should be called in z8530_tx_done() when skb
xmit done. It makes drop profiles(dropwatch, perf) more friendly.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: wan: cosa: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles

dev_consume_skb_irq() should be called in cosa_net_tx_done() when skb
xmit done. It makes drop profiles(dropwatch, perf) more friendly.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: wan: sbni: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles

dev_consume_skb_irq() should be called in send_complete() when skb
xmit done. It makes drop profiles(dropwatch, perf) more friendly.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: wan: ixp4xx_hss: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles

dev_consume_skb_irq() should be called in hss_hdlc_txdone_irq() when
skb xmit done. It makes drop profiles(dropwatch, perf) more friendly.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: wan: wanxl: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles

dev_consume_skb_irq() should be called in wanxl_tx_intr() when skb
xmit done. It makes drop profiles(dropwatch, perf) more friendly.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: lmc: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles

dev_consume_skb_irq() should be called in lmc_interrupt() when skb
xmit done. It makes drop profiles(dropwatch, perf) more friendly.

Delete a redundant comment line in lmc_interrupt().

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'pie-next'

Leslie Monis says:

====================
net: sched: pie: align PIE implementation with RFC 8033

The current implementation of the PIE queuing discipline is according to the
IETF draft [http://tools.ietf.org/html/draft-pan-aqm-pie-00] and the paper
[PIE: A Lightweight Control Scheme to Address the Bufferbloat Problem].
However, a lot of necessary modifications and enhancements have been proposed
in RFC 8033, which have not yet been incorporated in the source code of Linux.
This patch series helps in achieving the same.

Performance tests carried out using Flent [https://flent.org/]

Changes from v2 to v3:
  - Used div_u64() instead of direct division after explicit type casting as
    recommended by David

Changes from v1 to v2:
  - Excluded the patch setting PIE dynamically active/inactive as the test
    results were unsatisfactory
  - Fixed a scaling issue when adding more auto-tuning cases which caused
    local variables to underflow
  - Changed the long if/else chain to a loop as suggested by Stephen
  - Changed the position of the accu_prob variable in the pie_vars
    structure as recommended by Stephen
====================

Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: pie: update references

RFC 8033 replaces the IETF draft for PIE

Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: pie: add derandomization mechanism

Random dropping of packets to achieve latency control may
introduce outlier situations where packets are dropped too
close to each other or too far from each other. This can
cause the real drop percentage to temporarily deviate from
the intended drop probability. In certain scenarios, such
as a small number of simultaneous TCP flows, these
deviations can cause significant deviations in link
utilization and queuing latency.

RFC 8033 suggests using a derandomization mechanism to avoid
these deviations.

Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: pie: add more cases to auto-tune alpha and beta

The current implementation scales the local alpha and beta
variables in the calculate_probability function by the same
amount for all values of drop probability below 1%.

RFC 8033 suggests using additional cases for auto-tuning
alpha and beta when the drop probability is less than 1%.

In order to add more auto-tuning cases, MAX_PROB must be
scaled by u64 instead of u32 to prevent underflow when
scaling the local alpha and beta variables in the
calculate_probability function.

Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: pie: change initial value of pie_vars->burst_time

RFC 8033 suggests an initial value of 150 milliseconds for
the maximum time allowed for a burst of packets.

Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: pie: change default value of pie_params->tupdate

RFC 8033 suggests a default value of 15 milliseconds for the
update interval.

Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: pie: change default value of pie_params->target

RFC 8033 suggests a default value of 15 milliseconds for the
target queue delay.

Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: pie: change value of QUEUE_THRESHOLD

RFC 8033 recommends a value of 16384 bytes for the queue
threshold.

Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

Merge branch 'aquantia-hwmon'

Heiner Kallweit says:

====================
net: phy: aquantia: add hwmon support

This series adds HWMON support for the temperature sensor and the
related alarms on the 107/108/109 chips.

v2:
- remove struct aqr_priv
- rename header file to aquantia.h
v3:
- add conditional compiling of aquantia_hwmon.c
- improve converting sensor register values to/from long
- add helper aqr_hwmon_test_bit
====================

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

net: phy: aquantia: add hwmon support

This adds HWMON support for the temperature sensor and the related
alarms on the 107/108/109 chips. This patch is based on work from
Nikita and Andrew. I added:
- support for changing alarm thresholds via sysfs
- move HWMON code to a separate source file to improve maintainability
- smaller changes like using IS_REACHABLE instead of ifdef
  (avoids problems if PHY driver is built in and HWMON is a module)

v2:
- remove struct aqr_priv
- rename header file to aquantia.h
v3:
- add conditional compiling of aquantia_hwmon.c
- improve converting sensor register values to/from long
- add helper aqr_hwmon_test_bit

Signed-off-by: Nikita Yushchenko <nikita.yoush@cogentembedded.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: aquantia: rename aquantia.c to aquantia_main.c

Rename aquantia.c to aquantia_main.c to be prepared for adding new
functionality to separate source code files.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
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 2019-02-22

This series contains updates to the ice driver only.

Bruce adds the __always_unused attribute to a parameter to avoid
compiler warnings when using -Wunused-parameter.  Fixed unnecessary
type-casting and the use of sizeof().  Fix the allocation of structs
that have become memory hogs, so allocate them in heaps and fix all the
associated references.  Fixed the "possible" numeric overflow issues
that were caught with static analysis.

Maciej fixes the maximum MTU calculation by taking into account double
VLAN tagging amd ensure that the operations are done in the correct
order.

Victor fixes the supported node calculation, where we were not taking
into account if there is space to add the new VSI or intermediate node
above that layer, then it is not required to continue the calculation.
Added a check for a leaf node presence for a given VSI, which is needed
before removing a VSI.

Jake fixes an issue where the VSI list is shared, so simply removing a
VSI from the list will cause issues for the other users who reference
the list.  Since we also free the memory, this could lead to
segmentation faults.

Brett fixes an issue where driver unload could cause a system reboot
when intel_iommu=on parameter is set.  The issue is that we are not
clearing the CAUSE_ENA bit for the appropriate control queues register
when freeing the miscellaneous interrupt vector.

Mitch is so kind, he prevented spamming the VF with link messages when
the link status really has not changed.  Updates the driver to use the
absolute vector ID and not the per-PF vector ID for the VF MSIx vector
allocation.

Lukasz fixes the ethtool pause parameter for the ice driver, which was
originally based off the link status but is now based off the PHY
configuration.  This is to resolve an issue where pause parameters could
be set while link was down.

Jesse updates the string that reports statistics so the string does not
get modified at runtime and cause reports of string truncation.
====================

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

net: sched: don't release block->lock when dumping chains

Function tc_dump_chain() obtains and releases block->lock on each iteration
of its inner loop that dumps all chains on block. Outputting chain template
info is fast operation so locking/unlocking mutex multiple times is an
overhead when lock is highly contested. Modify tc_dump_chain() to only
obtain block->lock once and dump all chains without releasing it.

Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Suggested-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: sched: set dedicated tcf_walker flag when tp is empty

Using tcf_walker->stop flag to determine when tcf_walker->fn() was called
at least once is unreliable. Some classifiers set 'stop' flag on error
before calling walker callback, other classifiers used to call it with NULL
filter pointer when empty. In order to prevent further regressions, extend
tcf_walker structure with dedicated 'nonempty' flag. Set this flag in
tcf_walker->fn() implementation that is used to check if classifier has
filters configured.

Fixes: 8b64678e0af8 ("net: sched: refactor tp insert/delete for concurrent execution")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Suggested-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: mv88e6xxx: Fix phylink_validate for Topaz family

The Topaz family should have different phylink_validate method from the
Peridot, since on Topaz the port supporting 2500BaseX mode is port 5,
not 9 and 10.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: mv88e6xxx: Default CMODE to 1000BaseX only on 6390X

Commit 787799a9d555 sets the SERDES interfaces of 6390 and 6390X to
1000BaseX, but this is only needed on 6390X, since there are SERDES
interfaces which can be used on lower ports on 6390.

This commit fixes this by returning to previous behaviour on 6390.
(Previous behaviour means that CMODE is not set at all if requested mode
is NA).

This is needed on Turris MOX, where the 88e6190 is connected to CPU in
2500BaseX mode.

Fixes: 787799a9d555 ("net: dsa: mv88e6xxx: Default ports 9/10 6390X CMODE to 1000BaseX")
Signed-off-by: Marek Behún <marek.behun@nic.cz>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: clean up SOCK_DEBUG()

Per discussion with Daniel[1] and Eric[2], these SOCK_DEBUG() calles in
TCP are not needed now.
We'd better clean up it.

[1] https://patchwork.ozlabs.org/patch/1035573/
[2] https://patchwork.ozlabs.org/patch/1040533/

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: remove unused parameter of tcp_sacktag_bsearch()

parameter state in the tcp_sacktag_bsearch() is not used.
So, it can be removed.

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

ice: fix overlong string, update stats output

A test started warning on a string truncation. This led to an unfortunate
realization that we are likely not accounting for the stats length
correctly before this patch, so fix the issue by putting "port." in front
of all the PF stats, instead of magically prepending it at runtime.

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>

ice: Fix for FC get rx/tx pause params

Ethtool reported pause params based on the currently negotiated
link settings instead of current PHY config. User was not able
to turn off pause params because ethtool was incorrectly reporting
parameters as off when link was down even though PHY was configured
to support pause frames. Now pause params are taken from PHY config
instead of link status.

Signed-off-by: Lukasz Czapnik <lukasz.czapnik@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>

ice: use absolute vector ID for VFs

When the PF driver sets up the VF MSI-X vector allocation, it needs to
use the hardware absolute vector ID, not the per-PF vector ID. Without
this change we see (apparent) TX hangs when using VFs on multiple PFs.

Signed-off-by: Mitch Williams <mitch.a.williams@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>

ice: check for a leaf node presence

Check for a leaf node presence for a given VSI. This check is required
before removing a VSI since VSIs can't be removed with enabled queues
(with leaf nodes) from the FW scheduler tree unless its a reset.

Signed-off-by: Victor Raj <victor.raj@intel.com>
Reviewed-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>

ice: flush Tx pipe on disable queue timeout

Set the flush Tx pipe flag instead of getting an EAGAIN error when FW
times out in processing the disable Tx queue command.

Signed-off-by: Victor Raj <victor.raj@intel.com>
Reviewed-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>

ice: clear VF ARQLEN register on reset

On older devices like X710 and X722, the VF's ARQLEN register is cleared
on reset, so the VF driver uses that register to detect an unannounced
reset. Unfortunately, on devices controlled by ice, this register is NOT
cleared on reset. This causes the VF to miss resets, and even on
properly-announced resets, the VF driver complains that it didn't see
the reset.

To fix this, we'll do it in software. When we handle a VF reset (whether
triggered by software or VFLR), clear this register after the HW reset
is complete.

Signed-off-by: Mitch Williams <mitch.a.williams@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>

ice: don't spam VFs with link messages

Don't send a link message to the VFs unless link actually changes state.
This avoids a small timing hole in some VF drivers that can cause an
apparent TX hang if they receive a link status message at the wrong time.

Although we have fixed the timing hole in the current VF driver, there
are still lots of drivers in the field that have this timing hole. Let's
not fall into it if we can avoid it.

Signed-off-by: Mitch Williams <mitch.a.williams@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>

ice: only use the VF for ICE_VSI_VF in ice_vsi_release

In ice_vsi_release we are always assigning a value to the local VF
variable. Change this to only be assigned if the VSI is a VF VSI.

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>

ice: fix numeric overflow warning

When compiling and analyzing the driver on newer kernels, a static
analyzer warns about the following "numeric overflow" issues:

  "The result of expression: 'budget-1' generates 4-byte type while casting
   to a bigger size of 8-byte".

  "The result of expression: '*words-words_read' generates 4-byte type
   while casting to a bigger size of 8-byte".

Fix them both.

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>

ice: fix issue where host reboots on unload when iommu=on

Currently if the kernel has the intel_iommu=on parameter set, on some
platforms removing the driver causes a system reboot. In initialization
we associate the control queue interrupts with the pf->hw_oicr_idx and
enable the interrupts by setting the CAUSE_ENA bit. The problem comes
on teardown because we are not clearing the CAUSE_ENA bit for the
control queues, but the vector at pf->hw_oicr_idx (miscellaneous
interrupt vector) gets disabled.

Fix this by clearing the CAUSE_ENA bit in the appropriate control queue
registers on when freeing the miscellaneous interrupt vector. Also,
move the call to ice_free_irq_msix_misc() to after ice_deinit_sw() in
ice_remove() because ice_deinit_sw() makes an AQ call, but
ice_free_irq_msix_misc() disables the miscellaneous vector and it's
associated interrupts.

Also, create two small helper functions to enable and disable the
control queue interrupts respectively.

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>

ice: fix ice_remove_rule_internal vsi_list handling

When adding multiple VLANs to the same VSI, the ice_add_vlan code will
share the VSI list, so as not to create multiple unnecessary VSI lists.

Consider the following flow

  ice_add_vlan(hw, <VSI 0 VID 7, VSI 0 VID 8, VSI 0 VID 9>)

Where we add three VLAN filters for VIDs 7, 8, and 9, all for VSI 0.

The ice_add_vlan will create a single vsi_list and share it among all
the filters.

Later, if we try to remove a VLAN,

  ice_remove_vlan(hw, <VSI 0 VID 7>)

Then the removal code will update the vsi_list and remove VSI 0 from it.
But, since the vsi_list is shared, this breaks the list for the other
users who reference it. We actually even free the VSI list memory, and
may result in segmentation faults.

This is due to the way that VLAN rule share VSI lists with reference
counts, and is caused because we call ice_rem_update_vsi_list even when
the ref_cnt is greater than one.

To fix this, handle the case where ref_cnt is greater than one
separately. In this case, we need to remove the associated rule without
modifying the vsi_list, since it is currently being referenced by
another rule. Instead, we just need to decrement the VSI list ref_cnt.

The case for handling sharing of VSI lists with multiple VSIs is not
currently supported by this code. No such rules will be created today,
and this code will require changes if/when such code is added.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-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>

ice: fix stack hogs from struct ice_vsi_ctx structures

struct ice_vsi_ctx has gotten large enough that function local declarations
of it on the stack are causing stack hogs.  Fix that by allocating the
structs on heap.  Cleanup some formatting issues in the code around these
changes and fix incorrect data type uses of returned functions in a couple
places.

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>

ice: sizeof(<type>) should be avoided

With sizeof(), it is preferable to use the variable of type <type> instead
of sizeof(<type>).

There are multiple places where a temporary variable is used to hold a
'size' value which is then used for a subsequent alloc/memset. Get rid
of the temporary variable by calculating size as part of the alloc/memset
statement.

Also remove unnecessary type-cast.

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>

ice: Fix added in VSI supported nodes calc

VSI supported nodes are calculated in order to add the VSI parent or
intermediate nodes to the scheduler tree. If one of the node in below
layers (from VSI layer) has space to add the new VSI or intermediate node
above that layer then it's not required to continue the calculation further
for below layers.

Signed-off-by: Victor Raj <victor.raj@intel.com>
Reviewed-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>

ice: Fix the calculation of ICE_MAX_MTU

Currently ICE_MAX_MTU subtracts only ETH_HLEN from max frame size and
adds ETH_FCS_LEN and VLAN_HLEN, which is not what was intended.
The ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN expression should be surrounded
with parentheses.

Wrap mentioned expression and take into account VLAN double tagging.

Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@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>

ice: Mark extack argument as __always_unused

Commit 87b0984ebfab ("net: Add extack argument to ndo_fdb_add()") in
net-next added an extended parameter to the .ndo_fdb_add op and changed
ice_fdb_add() accordingly. Update the function header and add the
__always_unused attribute to the new parameter to avoid -Wunused-parameter
warnings.

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>

switchdev: Complete removal of switchdev_port_attr_get()

We have no more in tree users of switchdev_port_attr_get() after
d0e698d57a94 ("Merge branch 'net-Get-rid-of-switchdev_port_attr_get'")
so completely remove the function signature and body.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dsa: Remove phydev parameter from disable_port call

No current DSA driver makes use of the phydev parameter passed to the
disable_port call. Remove it.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: fix reading fixed phy status

With the switch to phy_resolve_aneg_linkmode() we don't read from the
chip any longer what is advertised but use phydev->advertising directly.
For a fixed phy however this bitmap is empty so far, what results in
no common mode being found. This breaks DSA. Fix this by advertising
everything that is supported. For a normal phy this done by phy_probe().

Fixes: 5502b218e001 ("net: phy: use phy_resolve_aneg_linkmode in genphy_read_status")
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: improve auto-neg emulation in swphy

Auto-neg emulation currently doesn't set bit BMCR_ANENABLE in BMCR,
add this. Users will ignore speed and duplex settings in BMCR because
we're emulating auto-neg, therefore we can remove related code.
See also following discussion [0].

[0] https://marc.info/?t=155041784900002&r=1&w=2

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'for-upstream' of git://git./linux/kernel/git/bluetooth/bluetooth-next

Johan Hedberg says:

====================
Here's the main bluetooth-next pull request for the 5.1 kernel.

 - Fixes & improvements to mediatek, hci_qca, btrtl, and btmrvl HCI drivers
 - Fixes to parsing invalid L2CAP config option sizes
 - Locking fix to bt_accept_enqueue()
 - Add support for new Marvel sd8977 chipset
 - Various other smaller fixes & cleanups
====================

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

net: fix double-free in bpf_lwt_xmit_reroute

dst_output() frees skb when it fails (see, for example,
ip_finish_output2), so it must not be freed in this case.

Fixes: 3bd0b15281af ("bpf: add handling of BPF_LWT_REROUTE to lwt_bpf.c")
Signed-off-by: Peter Oskolkov <posk@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ip_tunnel: Add ip tunnel tun_info type dst_cache in ip_tunnel_xmit

ip l add dev tun type gretap key 1000

Non-tunnel-dst ip tunnel device can send packet through lwtunnel
This patch provide the tun_inf dst cache support for this mode.

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

Merge branch 'dsa-mv88e6xxx-lockdep'

Andrew Lunn says:

====================
mv88e6xxx: Avoid false positive Lockdep splats

When acquiring the GPIO interrupt line for the switch, it is possible
to trigger lockdep splats. These are false positives, the mutex is in
a different IRQ descriptor. But fix it anyway, since it could mask
real locking issues.
====================

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

net: dsa: mv88e6xxx: Release lock while requesting IRQ

There is no need to hold the register lock while requesting the GPIO
interrupt. By not holding it we can also avoid a false positive
lockdep splat.

Reported-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: mv88e6xxx: Add lockdep classes to fix false positive splat

The following false positive lockdep splat has been observed.

======================================================
WARNING: possible circular locking dependency detected
4.20.0+ #302 Not tainted
------------------------------------------------------
systemd-udevd/160 is trying to acquire lock:
edea6080 (&chip->reg_lock){+.+.}, at: __setup_irq+0x640/0x704

but task is already holding lock:
edff0340 (&desc->request_mutex){+.+.}, at: __setup_irq+0xa0/0x704

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (&desc->request_mutex){+.+.}:
       mutex_lock_nested+0x1c/0x24
       __setup_irq+0xa0/0x704
       request_threaded_irq+0xd0/0x150
       mv88e6xxx_probe+0x41c/0x694 [mv88e6xxx]
       mdio_probe+0x2c/0x54
       really_probe+0x200/0x2c4
       driver_probe_device+0x5c/0x174
       __driver_attach+0xd8/0xdc
       bus_for_each_dev+0x58/0x7c
       bus_add_driver+0xe4/0x1f0
       driver_register+0x7c/0x110
       mdio_driver_register+0x24/0x58
       do_one_initcall+0x74/0x2e8
       do_init_module+0x60/0x1d0
       load_module+0x1968/0x1ff4
       sys_finit_module+0x8c/0x98
       ret_fast_syscall+0x0/0x28
       0xbedf2ae8

-> #0 (&chip->reg_lock){+.+.}:
       __mutex_lock+0x50/0x8b8
       mutex_lock_nested+0x1c/0x24
       __setup_irq+0x640/0x704
       request_threaded_irq+0xd0/0x150
       mv88e6xxx_g2_irq_setup+0xcc/0x1b4 [mv88e6xxx]
       mv88e6xxx_probe+0x44c/0x694 [mv88e6xxx]
       mdio_probe+0x2c/0x54
       really_probe+0x200/0x2c4
       driver_probe_device+0x5c/0x174
       __driver_attach+0xd8/0xdc
       bus_for_each_dev+0x58/0x7c
       bus_add_driver+0xe4/0x1f0
       driver_register+0x7c/0x110
       mdio_driver_register+0x24/0x58
       do_one_initcall+0x74/0x2e8
       do_init_module+0x60/0x1d0
       load_module+0x1968/0x1ff4
       sys_finit_module+0x8c/0x98
       ret_fast_syscall+0x0/0x28
       0xbedf2ae8

other info that might help us debug this:

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&desc->request_mutex);
                               lock(&chip->reg_lock);
                               lock(&desc->request_mutex);
  lock(&chip->reg_lock);

&desc->request_mutex refer to two different mutex. #1 is the GPIO for
the chip interrupt. #2 is the chained interrupt between global 1 and
global 2.

Add lockdep classes to the GPIO interrupt to avoid this.

Reported-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

ip_tunnel: Add dst_cache support in lwtunnel_state of ip tunnel

The lwtunnel_state is not init the dst_cache Which make the
ip_md_tunnel_xmit can't use the dst_cache. It will lookup
route table every packets.

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

tls: Return type of non-data records retrieved using MSG_PEEK in recvmsg

The patch enables returning 'type' in msghdr for records that are
retrieved with MSG_PEEK in recvmsg. Further it prevents records peeked
from socket from getting clubbed with any other record of different
type when records are subsequently dequeued from strparser.

For each record, we now retain its type in sk_buff's control buffer
cb[]. Inside control buffer, record's full length and offset are already
stored by strparser in 'struct strp_msg'. We store record type after
'struct strp_msg' inside 'struct tls_msg'. For tls1.2, the type is
stored just after record dequeue. For tls1.3, the type is stored after
record has been decrypted.

Inside process_rx_list(), before processing a non-data record, we check
that we must be able to return back the record type to the user
application. If not, the decrypted records in tls context's rx_list is
left there without consuming any data.

Fixes: 692d7b5d1f912 ("tls: Fix recvmsg() to be able to peek across multiple records")
Signed-off-by: Vakul Garg <vakul.garg@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'ipv4-v6-icmp-small-cleanup-and-update'

Kefeng Wang says:

====================
ipv4/v6: icmp: small cleanup and update

v2:
- Add cover letter and user proper patch subject-prefix suggested-by Eric Dumazet

This patch series contains some small cleanup and update,
1) use icmp/v6_sk_exit when icmp_sk_init fails instead of open-code
2) use new percpu allocation interface for the ipv6.icmp_sk
====================

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

ipv6: icmp: use percpu allocation

Use percpu allocation for the ipv6.icmp_sk.

Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: icmp: use icmpv6_sk_exit()

Simply use icmpv6_sk_exit() when inet_ctl_sock_create() fail
in icmpv6_sk_init().

Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: icmp: use icmp_sk_exit()

Simply use icmp_sk_exit() when inet_ctl_sock_create() fail in icmp_sk_init().

Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ila: Fix uninitialised return value in ila_xlat_nl_cmd_flush

This patch fixes an uninitialised return value error in
ila_xlat_nl_cmd_flush.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Fixes: 6c4128f65857 ("rhashtable: Remove obsolete...")
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/sched: act_tunnel_key: Add dst_cache support

The metadata_dst is not init the dst_cache which make the
ip_md_tunnel_xmit can't use the dst_cache. It will lookup
route table every packets.

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

Merge branch 'code-optimizations-and-bugfixes-for-HNS3-driver'

Huazhong Tan says:

====================
code optimizations & bugfixes for HNS3 driver

This patchset includes bugfixes and code optimizations for
the HNS3 ethernet controller driver.
====================

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

net: hns3: fix improper error handling for hns3_client_start

If hns3_client_start() failed in the hns3_client_init(),
register_dev() should be undo in its error handling.

Fixes: a6d818e31d08 ("net: hns3: Add vport alive state checking support")
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hns3: fix setting of the hns reset_type for rdma hw errors

Presently the hns reset_type for the roce errors is set
in the hclge_log_and_clear_rocee_ras_error function.
This function is also called to detect and clear roce errors
while enabling the rdma error interrupts. However there is no hns
reset requested for this case. This can cause issue of wrong
reset_type used with subsequent hns reset as the
reset_type set in the above case was not cleared.

This patch moves setting of hns reset_type for the roce errors from
hclge_log_and_clear_rocee_ras_error function
to hclge_handle_rocee_ras_error.

Fixes: 630ba007f475 ("net: hns3: add handling of RDMA RAS errors")
Reported-by: Huazhong Tan <tanhuazhong@huawei.com>
Reported-by: Xiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: Shiju Jose <shiju.jose@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hns3: fix get VF RSS issue

For revision 0x20, VF shares the same RSS config with PF.
In original codes, it always return 0 when query RSS hash
key for VF. This patch fixes it by return the hash key
got from PF.

Fixes: 374ad291762a ("net: hns3: net: hns3: Add RSS general configuration support for VF")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hns3: enable VF VLAN filter for each VF when initializing

For revision 0x21, the switch of VF VLAN filter is per function.
It's necessary to enable VF VLAN filter for each VF when initializing.
Otherwise, VF will be able to receive broadcast packets with unknown
VLAN when PF enters promisc mode.

Fixes: 64d114f0a750 ("net: hns3: Add egress/ingress vlan filter for revision 0x21")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hns3: add support to config depth for tx|rx ring separately

This patch adds support to config depth for tx|rx ring separately
by ethtool command "-G".

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hns3: remove hnae3_get_bit in data path

The hnae3_get_bit uses hnae3_get_field, and hnae3_get_field
masks the data, which is unnecessary in data path.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hns3: replace hnae3_set_bit and hnae3_set_field in data path

hnae3_set_bit and hnae3_set_field masks the data before setting
the field or bit, which is unnecessary because the data is already
zero initialized.

Suggested-by: John Garry <john.garry@huawei.com>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hns3: add unlikely for error handling in data path

This patch adds unlikely hint for error handling in critical data
path.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>