This allows to re-use the same program.  For example, a nft
ruleset that attaches filter basechains to input, forward, output would
use the same program for all three hook points.

The cache is intentionally netns agnostic, so same config
in different netns will all use same programs.

Signed-off-by: Florian Westphal <fw@strlen.de>

Signed-off-by: Florian Westphal <fw@strlen.de>

We can save a few cycles on netns destruction.
When a hook is removed we can just skip building a new
program with the remaining hooks, those will be removed too
in the immediate future.

Signed-off-by: Florian Westphal <fw@strlen.de>

Add a kernel bpf program generator for netfilter base hooks.

Currently netfilter hooks are invoked by nf_hook_slow:

for i in hooks; do
  verdict = hooks[i]->indirect_func(hooks->[i].hook_arg, skb, state);

  switch (verdict) { ....

The autogenerator unrolls the loop, so we get:

state->priv = hooks->[0].hook_arg;
v = first_hook_function(state);
if (v != ACCEPT) goto done;
state->priv = hooks->[1].hook_arg;
v = second_hook_function(state); ...

Indirections are replaced by direct calls. Invocation of the
autogenerated programs is done via bpf dispatcher from nf_hook().

The autogenerated program has the same return value scheme as
nf_hook_slow(). NF_HOOK() points are converted to call the
autogenerated bpf program instead of nf_hook_slow().

Purpose of this is to eventually add a 'netfilter prog type' to bpf and
permit attachment of (userspace generated) bpf programs to the netfilter
machinery, e.g.  'attach bpf prog id 1234 to ipv6 PREROUTING at prio -300'.

This will require to expose the context structure (program argument,
'__nf_hook_state', with rewriting accesses to match nf_hook_state layout.

TODO:
1. Test !x86_64.
2. Test bridge family.

Future work:
add support for NAT hooks, they still use indirect calls, but those
are less of a problem because these get called only once per
connection.

Could annotate ops struct as to what kind of verdicts the
C function can return.  This would allow to elide retval
check when hook can only return NF_ACCEPT.

Could add extra support for INGRESS hook to move more code from
inline functions to the autogenerated program.

Signed-off-by: Florian Westphal <fw@strlen.de>

1k is huge and will mean we'd need to support tailcalls in the
nf_hook bpf converter.

We need about 5 insns per hook at this time, ignoring prologue/epilogue.

32 should be fine, typically even extreme cases need about 8 hooks per
hook location.

Signed-off-by: Florian Westphal <fw@strlen.de>

BPF conversion requirement: one pointer-to-structure as argument.

Signed-off-by: Florian Westphal <fw@strlen.de>

Previous patch added hook_entry member to nf_hook_state struct, so
use that for passing the index.

Signed-off-by: Florian Westphal <fw@strlen.de>

Makes conversion in followup patch simpler.

Signed-off-by: Florian Westphal <fw@strlen.de>

Rather than passing the index (hook function to call next)
as function argument, store it in the hook state.

This is a prerequesite to allow passing all nf hook arguments in a single
structure.

Signed-off-by: Florian Westphal <fw@strlen.de>

Relax this condition to make add and update commands idempotent for sets
with no timeout. The eval function already checks if the set element
timeout is available and updates it if the update command is used.

Fixes: 22fe54d ("netfilter: nf_tables: add support for dynamic set updates")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

estimation_timer will iterate the est_list to do estimation
for each ipvs stats. When there are lots of services, the
list can be very large.
We found that estimation_timer() run for more then 200ms on a
machine with 104 CPU and 50K services.

yunhong-cgl jiang report the same phenomenon before:
https://www.spinics.net/lists/lvs-devel/msg05426.html

In some cases(for example a large K8S cluster with many ipvs services),
ipvs estimation may not be needed. So adding a sysctl blob to allow
users to disable this completely.

Default is: 1 (enable)

Cc: yunhong-cgl jiang <xintian1976@gmail.com>
Signed-off-by: Dust Li <dust.li@linux.alibaba.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Use 2-factor multiplication argument form devm_kcalloc() instead
of devm_kzalloc().

Link: KSPP#162
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://lore.kernel.org/r/20211006181115.GA913499@embeddedor
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Use 2-factor multiplication argument form kcalloc() instead
of kzalloc().

Link: KSPP#162
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://lore.kernel.org/r/20211006180944.GA913477@embeddedor
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Use 2-factor multiplication argument form kcalloc() instead
of kzalloc().

Link: KSPP#162
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: Dexuan Cui <decui@microsoft.com>
Link: https://lore.kernel.org/r/20211006180927.GA913456@embeddedor
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Use 2-factor multiplication argument form kcalloc() instead
of kzalloc().

Link: KSPP#162
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://lore.kernel.org/r/20211006180843.GA913399@embeddedor
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

…pub/scm/linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
wireless-drivers-next patches for v5.16

First set of patches for v5.16. ath11k getting most of new features
this time. Other drivers also have few new features, and of course the
usual set of fixes and cleanups all over.

Major changes:

rtw88

* support adaptivity for ETSI/JP DFS region

* 8821c: support RFE type4 wifi NIC

brcmfmac

* DMI nvram filename quirk for Cyberbook T116 tablet

ath9k

* load calibration data and pci init values via nvmem subsystem

ath11k

* include channel rx and tx time in survey dump statistics

* support for setting fixed Wi-Fi 6 rates from user space

* support for 80P80 and 160 MHz bandwidths

* spectral scan support for QCN9074

* support for calibration data files per radio

* support for calibration data via eeprom

* support for rx decapsulation offload (data frames in 802.3 format)

* support channel 2 in 6 GHz band

ath10k

* include frame time stamp in beacon and probe response frames

wcn36xx

* enable Idle Mode Power Save (IMPS) to reduce power consumption during idle
====================

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

Jakub Kicinski says:

====================
net: add a helpers for loading netdev->dev_addr from FW

We're trying to make all writes to netdev->dev_addr go via helpers.
A lot of places pass netdev->dev_addr to of_get_ethdev_address() and
device_get_ethdev_addr() so this set adds new functions which wrap
the functionality.

v2 performs suggested code moves, adds a couple additional clean ups
on the device property side, and an extra patch converting drivers
which can benefit from device_get_ethdev_address().

v3 removes OF_NET and corrects kdoc.
====================

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

Convert a few drivers to device_get_ethdev_address(),
saving a few LoC.

The check if addr is valid in netsec is superfluous,
device_get_ethdev_addr() already checks that (in
fwnode_get_mac_addr()).

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Use the new device_get_ethdev_address() helper for the cases
where dev->dev_addr is passed in directly as the destination.

  @@
  expression dev, np;
  @@
  - device_get_mac_address(np, dev->dev_addr, ETH_ALEN)
  + device_get_ethdev_address(np, dev)

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Commit 406f42f ("net-next: When a bond have a massive amount
of VLANs...") introduced a rbtree for faster Ethernet address look
up. To maintain netdev->dev_addr in this tree we need to make all
the writes to it got through appropriate helpers.

There is a handful of drivers which pass netdev->dev_addr as
the destination buffer to device_get_mac_address(). Add a helper
which takes a dev pointer instead, so it can call an appropriate
helper.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

All callers pass in ETH_ALEN and the function itself
will return -EINVAL for any other address length.
Just assume it's ETH_ALEN like all other mac address
helpers (nvm, of, platform).

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

fwnode_get_mac_address() and device_get_mac_address()
return a pointer to the buffer that was passed to them
on success or NULL on failure. None of the callers
care about the actual value, only if it's NULL or not.

These semantics differ from of_get_mac_address() which
returns an int so to avoid confusion make the device
helpers return an errno.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Move the mac address helpers out, eth.c already contains
a bunch of similar helpers.

Suggested-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Use the new of_get_ethdev_address() helper for the cases
where dev->dev_addr is passed in directly as the destination.

  @@
  expression dev, np;
  @@
  - of_get_mac_address(np, dev->dev_addr)
  + of_get_ethdev_address(np, dev)

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Commit 406f42f ("net-next: When a bond have a massive amount
of VLANs...") introduced a rbtree for faster Ethernet address look
up. To maintain netdev->dev_addr in this tree we need to make all
the writes to it got through appropriate helpers.

There are roughly 40 places where netdev->dev_addr is passed
as the destination to a of_get_mac_address() call. Add a helper
which takes a dev pointer instead, so it can call an appropriate
helper.

Note that of_get_mac_address() already assumes the address is
6 bytes long (ETH_ALEN) so use eth_hw_addr_set().

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Rob suggests to move of_net.c from under drivers/of/ somewhere
to the networking code.

Suggested-by: Rob Herring <robh@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Rikard Falkeborn says:

====================
nfc: pn533: Constify ops-structs

Constify a couple of ops-structs. This allows the compiler to put the
static structs in read-only memory.
====================

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

Neither the driver or the core modifies the pn533_phy_ops struct, so
make them const to allow the compiler to put the static structs in
read-only memory.

Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The only usage of pn532_serdev_ops is to pass its address to
serdev_device_set_client_ops(), which takes a pointer to const
serdev_device_ops as argument. Make it const to allow the compiler to
put it in read-only memory.

Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Russell King says:

====================
Add mdiobus_modify_changed() helper

Sean Anderson's recent patch series is introducing more read-write
operations on the MDIO bus that only need to happen if a change is
being made.

We have similar logic in __mdiobus_modify_changed(), but we didn't
add its correponding locked variant mdiobus_modify_changed() as we
had very few users. Now that we are getting more, let's add the
helper.
====================

Link: https://lore.kernel.org/r/YV2UIa2eU+UjmWaE@shell.armlinux.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Use the mdiobus_modify_changed() helper in the C22 PCS advertisement
helper.

Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Add mdiobus_modify_changed() helper to reflect the phylib and similar
equivalents. This will avoid this functionality being open-coded, as
has already happened in phylink, and it looks like other users will be
appearing soon.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

…r-mode'

Ido Schimmel says:

====================
ethtool: Add ability to control transceiver modules' power mode

This patchset extends the ethtool netlink API to allow user space to
control transceiver modules. Two specific APIs are added, but the plan
is to extend the interface with more APIs in the future (see "Future
plans").

This submission is a complete rework of a previous submission [1] that
tried to achieve the same goal by allowing user space to write to the
EEPROMs of these modules. It was rejected as it could have enabled user
space binary blob drivers.

However, the main issue is that by directly writing to some pages of
these EEPROMs, we are interfering with the entity that is controlling
the modules (kernel / device firmware). In addition, some functionality
cannot be implemented solely by writing to the EEPROM, as it requires
the assertion / de-assertion of hardware signals (e.g., "ResetL" pin in
SFF-8636).

Motivation
==========

The kernel can currently dump the contents of module EEPROMs to user
space via the ethtool legacy ioctl API or the new netlink API. These
dumps can then be parsed by ethtool(8) according to the specification
that defines the memory map of the EEPROM. For example, SFF-8636 [2] for
QSFP and CMIS [3] for QSFP-DD.

In addition to read-only elements, these specifications also define
writeable elements that can be used to control the behavior of the
module. For example, controlling whether the module is put in low or
high power mode to limit its power consumption.

The CMIS specification even defines a message exchange mechanism (CDB,
Command Data Block) on top of the module's memory map. This allows the
host to send various commands to the module. For example, to update its
firmware.

Implementation
==============

The ethtool netlink API is extended with two new messages,
'ETHTOOL_MSG_MODULE_SET' and 'ETHTOOL_MSG_MODULE_GET', that allow user
space to set and get transceiver module parameters. Specifically, the
'ETHTOOL_A_MODULE_POWER_MODE_POLICY' attribute allows user space to
control the power mode policy of the module in order to limit its power
consumption. See detailed description in patch #1.

The user API is designed to be generic enough so that it could be used
for modules with different memory maps (e.g., SFF-8636, CMIS).

The only implementation of the device driver API in this series is for a
MAC driver (mlxsw) where the module is controlled by the device's
firmware, but it is designed to be generic enough so that it could also
be used by implementations where the module is controlled by the kernel.

Testing and introspection
=========================

See detailed description in patches #1 and #5.

Patchset overview
=================

Patch #1 adds the initial infrastructure in ethtool along with the
ability to control transceiver modules' power mode.

Patches #2-#3 add required device registers in mlxsw.

Patch #4 implements in mlxsw the ethtool operations added in patch #1.

Patch #5 adds extended link states in order to allow user space to
troubleshoot link down issues related to transceiver modules.

Patch torvalds#6 adds support for these extended states in mlxsw.

Future plans
============

* Extend 'ETHTOOL_MSG_MODULE_SET' to control Tx output among other
attributes.

* Add new ethtool message(s) to update firmware on transceiver modules.

* Extend ethtool(8) to parse more diagnostic information from CMIS
modules. No kernel changes required.

[1] https://lore.kernel.org/netdev/20210623075925.2610908-1-idosch@idosch.org/
[2] https://members.snia.org/document/dl/26418
[3] http://www.qsfp-dd.com/wp-content/uploads/2021/05/CMIS5p0.pdf

Previous versions:
[4] https://lore.kernel.org/netdev/20211003073219.1631064-1-idosch@idosch.org/
[5] https://lore.kernel.org/netdev/20210824130344.1828076-1-idosch@idosch.org/
[6] https://lore.kernel.org/netdev/20210818155202.1278177-1-idosch@idosch.org/
[7] https://lore.kernel.org/netdev/20210809102152.719961-1-idosch@idosch.org/
====================

Link: https://lore.kernel.org/r/20211006104647.2357115-1-idosch@idosch.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Add support for the transceiver module extended state and sub-state
added in previous patch. The extended state is meant to describe link
issues related to transceiver modules.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Add an extended state and sub-state to describe link issues related to
transceiver modules.

The 'ETHTOOL_LINK_EXT_SUBSTATE_MODULE_CMIS_NOT_READY' extended sub-state
tells user space that port is unable to gain a carrier because the CMIS
Module State Machine did not reach the ModuleReady (Fully Operational)
state. For example, if the module is stuck at ModuleLowPwr or
ModuleFault state. In case of the latter, user space can read the fault
reason from the module's EEPROM and potentially reset it.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>