As some IP variants switch to reset mode (and thus the register contents is
lost) when setting clocks (due to module standby functionality) to be able
to implement runtime PM and save more power, set the IP's operating mode to
reset at the end of the probe. Along with it, in the ndo_open API the IP
will be switched to configuration, then operation mode. In the ndo_close
API, the IP will be switched back to reset mode. This allows implementing
runtime PM and, along with it, save more power when the IP is not used.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

The initialization sequence for PTP is the same for platforms with ccc_gac
and gptp (according to "Figure 50.71 Flow of gPTP Initialization (Normal,
Common to All Modes)" of the R-Car Series, 3rd generation hardware
manual and "Figure 37A.53 Flow of gPTP Initialization (Normal, Common to
All Modes)" of the RZ/G Series hardware manual).

As some IP variants switch to reset mode (and thus the registers content is
lost) when setting clocks (due to module standby functionality) to be able
to implement runtime PM, move the PTP initialization to the driver's
ndo_open API.

This commit prepares the code for the addition of runtime PM.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

DBAT setup was done in the driver's probe API. As some IP variants switch
to reset mode (and thus registers content is lost) when setting clocks
(due to module standby functionality) to be able to implement runtime PM
move the DBAT configuration in the driver's ndo_open API.

This commit prepares the code for the addition of runtime PM.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Delay parsing and setting were done in the driver's probe API. As some IP
variants switch to reset mode (and thus the register contents is lost) when
setting clocks (due to module standby functionality) to be able to
implement runtime PM keep the delay parsing in the driver's probe function
and move the delay applying function to the driver's ndo_open API.

Along with it, ravb_parse_delay_mode() function was moved close to
ravb_set_delay_mode() function to have the delay specific code in the
same place.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

ravb_set_gti() was computing the value of GTI based on the reference clock
rate and then applied it to register. This was done on the driver's probe
function. In order to implement runtime PM for all IP variants (as some IP
variants switches to reset mode (and thus the registers content is lost)
when module standby is configured through clock APIs) the GTI setup was
split in 2 parts: one computing the value of the GTI register (done in the
driver's probe function) and one applying the computed value to register
(done in the driver's ndo_open API).
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

The runtime PM implementation will disable clocks at the end of
ravb_probe(). As some IP variants switch to reset mode as a result of
setting module standby through clock disable APIs, to implement runtime PM
the resource parsing and requesting are moved in the probe function and IP
settings are moved in the open function. This is done because at the end of
the probe some IP variants will switch anyway to reset mode and the
registers content is lost. Also keeping only register settings operations
in the ravb_open()/ravb_close() functions will make them faster.

Commit moves IRQ requests to ravb_probe() to have all the IRQs ready when
the interface is open. As now getting/requesting IRQs is done in a single
place there is no need to keep intermediary data (like ravb_rx_irqs[] and
ravb_tx_irqs[] arrays or IRQs in struct ravb_private).

In order to avoid accessing the IP registers while the IP is runtime
suspended (e.g. in the timeframe b/w the probe requests shared IRQs and
IP clocks are enabled) in the interrupt handlers were introduced
pm_runtime_active() checks. The device runtime PM usage counter has been
incremented to avoid disabling the device's clocks while the check is in
progress (if any).

This is a preparatory change to add runtime PM support for all IP variants.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Reference clock could be or not be part of the power domain. If it is part
of the power domain, the power domain takes care of properly setting it. In
case it is not part of the power domain and full runtime PM support is
available in driver the clock will not be propertly disabled/enabled at
runtime. For this, keep the prepare/unprepare operations in the driver's
probe()/remove() functions and move the enable/disable in runtime PM
functions.

By doing this, the previous ravb_runtime_nop() function was renamed
ravb_runtime_suspend() and the comment was removed. A proper runtime PM
resume function was added (ravb_runtime_resume()). The current driver
still don't need to make any register settings on runtime suspend/resume
(as expressed in the removed comment) because, currently,
pm_runtime_put_sync() is called on the driver remove function. This will be
changed in the next commits (that extends the runtime PM support) such
that proper register settings (along with runtime resume/suspend) will be
done on ravb_open()/ravb_close().

Along with it, the other clock request operations were moved close to
reference clock request and prepare to have all the clock requests
specific code grouped together.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

RZ/G3S can go to deep sleep states where power to most of the SoC parts is
off. When resuming from such a state, the Ethernet controller needs to be
reinitialized. De-asserting the reset signal for it should also be done.
Thus, add reset assert/de-assert on suspend/resume functions.

On the resume function, the de-assert was not reverted in case of failures
to give the user a chance to restore the interface (e.g., bringing down/up
the interface) in case suspend/resume failed.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Use tabs instead of spaces in the ravb_set_rate_gbeth() function.
This aligns with the coding style requirements.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

SET_SYSTEM_SLEEP_PM_OPS() and SET_RUNTIME_PM_OPS() are deprecated now
and require __maybe_unused protection against unused function warnings.
The usage of pm_ptr() and SYSTEM_SLEEP_PM_OPS()/RUNTIME_PM_OPS() allows
the compiler to see the functions, thus suppressing the warning. Thus
drop the __maybe_unused markings.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

On the RZ/G3S SoC the reset controller is mandatory for the IP to work.
The device tree binding documentation for the ravb driver specifies that
the resets are mandatory. Based on this, make the resets mandatory also in
driver for all ravb devices.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

ravb_rzv2m_hw_info::gptp_ref_clk is enabled only for RZ/V2M. RZ/V2M
is an ARM64-based device which selects power domains by default and
CONFIG_PM. The RZ/V2M Ethernet DT node has proper power-domain binding
available in device tree from the commit that added the Ethernet node.
(4872ca1f ("arm64: dts: renesas: r9a09g011: Add ethernet nodes")).

Power domain support was available in the rzg2l-cpg.c driver when the
Ethernet DT node has been enabled in RZ/V2M device tree.
(ef3c613c ("clk: renesas: Add CPG core wrapper for RZ/G2L SoC")).

Thus, remove the explicit clock enable for gptp_clk (and treat it as the
other clocks are treated) as it is not needed and removing it doesn't
break the ABI according to the above explanations.

By removing the enable/disable operation from the driver we can add
runtime PM support (which operates on clocks) w/o the need to handle
the gptp_clk in the Ethernet driver functions like ravb_runtime_nop().
PM domain does all that is needed.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

ravb_poll() initial code used to interrogate the first descriptor of the
RX queue in case gPTP is false to determine if ravb_rx() should be called.
This is done for non-gPTP IPs. For gPTP IPs the driver PTP-specific
information was used to determine if receive function should be called. As
every IP has its own receive function that interrogates the RX descriptors
list in the same way the ravb_poll() was doing there is no need to double
check this in ravb_poll(). Removing the code from ravb_poll() leads to a
cleaner code.
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

The maple tree register cache is based on a much more modern data structure
than the rbtree cache and makes optimisation choices which are probably
more appropriate for modern systems than those made by the rbtree cache.
Signed-off-by: Bo Liu <liubo03@inspur.com>
Link: https://lore.kernel.org/r/20240202064336.39138-1-liubo03@inspur.comSigned-off-by: Paolo Abeni <pabeni@redhat.com>

This test is time sensitive. It may fail on virtual machines and for
debug builds.

Continue to run in these environments to get code coverage. But
optionally suppress failure for timing errors (only). This is
controlled with environment variable KSFT_MACHINE_SLOW.

The test continues to return 0 (KSFT_PASS), rather than KSFT_XFAIL
as previously discussed. Because making so_txtime.c return that and
then making so_txtime.sh capture runs that pass that vs KSFT_FAIL
and pass it on added a bunch of (fragile bash) boilerplate, while the
result is interpreted the same as KSFT_PASS anyway.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Link: https://lore.kernel.org/r/20240201162130.2278240-1-willemdebruijn.kernel@gmail.comSigned-off-by: Paolo Abeni <pabeni@redhat.com>

Add a Makefile for netdevsim selftests and add selftests path to
MAINTAINERS
Signed-off-by: David Wei <dw@davidwei.uk>
Link: https://lore.kernel.org/r/20240130214620.3722189-5-dw@davidwei.ukSigned-off-by: Jakub Kicinski <kuba@kernel.org>

Vladimir Oltean says:

====================
Fixups for qca8k ds->user_mii_bus cleanup

The series "ds->user_mii_bus cleanup (part 1)" from the last development
cycle:
https://patchwork.kernel.org/project/netdevbpf/cover/20240104140037.374166-1-vladimir.oltean@nxp.com/

had some review comments I didn't have the time to address at the time.
One from Alvin and one from Luiz. They can reasonably be treated as
improvements for v6.9.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

It was pointed out during the review [1] of commit 68e1010c ("net:
dsa: qca8k: put MDIO bus OF node on qca8k_mdio_register() failure") that
the rest of the qca8k driver uses "int ret" rather than "int err".

Make everything consistent in that regard, not only
qca8k_mdio_register(), but also qca8k_setup_mdio_bus().

[1] https://lore.kernel.org/netdev/qyl2w3ownx5q7363kqxib52j5htar4y6pkn7gen27rj45xr4on@pvy5agi6o2te/Suggested-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

It was pointed out during the review [1] of commit e66bf63a ("net:
dsa: qca8k: skip MDIO bus creation if its OF node has status =
"disabled"") that we now leak a reference to the "mdio" OF node if it is
disabled.

This is only a concern when using dynamic OF as far as I can tell (like
probing on an overlay), since OF nodes are never freed in the regular
case. Additionally, I'm unaware of any actual device trees (in
production or elsewhere) which have status = "disabled" for the MDIO OF
node. So handling this as a simple enhancement.

[1] https://lore.kernel.org/netdev/CAJq09z4--Ug+3FAmp=EimQ8HTQYOWOuVon-PUMGB5a1N=RPv4g@mail.gmail.com/Suggested-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

These got misaligned after commit 6ca80638 ("net: dsa: Use conduit
and user terms").
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

commit 1c870c63 ("net: fill in MODULE_DESCRIPTION()s for ocelot")
got a suggestion from Vladimir Oltean after it had landed in net-next.

Rewrite the module description according to Vladimir's suggestion.

Fixes: 1c870c63 ("net: fill in MODULE_DESCRIPTION()s for ocelot")
Suggested-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

lib/test_blackhole_dev.c sets a variable that is never read, causing
this following building warning:

	lib/test_blackhole_dev.c:32:17: warning: variable 'ethh' set but not used [-Wunused-but-set-variable]

Remove the variable struct ethhdr *ethh, which is unused.

Fixes: 509e56b3 ("blackhole_dev: add a selftest")
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Matthieu Baerts says:

====================
mptcp: annotate lockless access

This is a series of 5 patches from Paolo to annotate lockless access.

The MPTCP locking schema is already quite complex. We need to clarify it
and make the lockless access already there consistent, or later changes
will be even harder to follow and understand.

This series goes through all the msk fields accessed in the RX and TX
path and makes the lockless annotation consistent with the in-use
locking schema.

As a bonus, this should fix data races eventually found by fuzzers --
even if we haven't seen many such reports so far.

Patch 1/5 hints we could remove "local_key" and "remote_key" from the
subflow context, and always use the ones from the msk socket, possibly
reducing the context memory usage. That change is left over as a
possible follow-up.
====================
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

The following MPTCP socket fields:

 - can_ack
 - fully_established
 - rcv_data_fin
 - snd_data_fin_enable
 - rcv_fastclose
 - use_64bit_ack

are accessed without any lock, add the appropriate annotation.

The schema is safe as each field can change its value at most
once in the whole mptcp socket life cycle.
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

The token field is manipulated under the msk socket lock
and accessed lockless in a few spots, add proper ONCE annotation
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

The following fields:

 - ack_seq
 - snd_una
 - wnd_end
 - rmem_fwd_alloc

are protected by the data lock end accessed lockless in a few
spots. Ensure ONCE annotation for write (under such lock) and for
lockless read.
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

The mptcp-level TX path info (write_seq, bytes_sent, snd_nxt) are under
the msk socket lock protection, and are accessed lockless in a few spots.

Always mark the write operations with WRITE_ONCE, read operations
outside the lock with READ_ONCE and drop the annotation for read
under such lock.

To simplify the annotations move mptcp_pending_data_fin_ack() from
__mptcp_data_acked() to __mptcp_clean_una(), under the msk socket
lock, where such call would belong.
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Both the local and the remote key follow the same locking
schema, put in place the proper ONCE accessors.
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Similar chunk of code is used in tsnep_rx_poll_zc() and
tsnep_rx_reopen_xsk() to maintain the RX XDP_RING_NEED_WAKEUP flag.
Consolidate the code to common helper function.
Suggested-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

We can change sctp_sk() to propagate its argument const qualifier,
thanks to container_of_const().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Cc: Xin Long <lucien.xin@gmail.com>
Acked-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We can use a global dev_unreg_count counter instead
of a per netns one.

As a bonus we can factorize the changes done on it
for bulk device removals.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Implement the tun .get_channels functionality. This feature is necessary
for some tools, such as libxdp, which need to retrieve the queue count.
Signed-off-by: Yunjian Wang <wangyunjian@huawei.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This fixes the following code style problem:
- WARNING: please, no spaces at the start of a line
- CHECK: Please use a blank line after
         function/struct/union/enum declarations
Signed-off-by: Yunjian Wang <wangyunjian@huawei.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This adds support for the RTL8126A found on Asus z790 Maximus Formula.
It was successfully tested w/o the firmware at 1000Mbps. Firmware file
has been provided by Realtek and submitted to linux-firmware.
2.5G and 5G modes are untested.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Paolo points out that ifconfig is legacy and we should not use it.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

By default lan8841's 1588 clock frequency is 125MHz. But when adjusting
the frequency, it is using the 1PPM format of the lan8814. Which is the
wrong format as lan8814 has a 1588 clock frequency of 250MHz. So then
for each 1PPM adjustment would adjust less than expected.
Therefore fix this by using the correct 1PPM format for lan8841.
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Christian Marangi says:

====================
net: phy: qcom: qca808x: fixup qca808x LED

This is a bit embarassing and totally my fault so sorry for that!

While reworking the patch to phy_modify API, it was done a logic
error and made the brightness_set function broken. It wasn't
notice in last revisions test as the testing method was to verify
if hw control was correctly working.

Noticing this problem also made me notice an additional problem
with the polarity.

The introduced patch made the polarity configurable but I forgot
to add the required code to enable Active High by default.
(the PHY sets active low by default)

This wasn't notice with hw control testing as the LED blink on
traffic and polarity problem are not notice.

It might be worth discussing if needed a change in implementation
where the polarity function is always called but I think it's
better this way where specific PHY apply fixup with the help
of priv struct and on the config_init phase.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

qca808x PHY provide support for the led_polarity_set OP to configure
and apply the active-low property but on PHY reset, the Active High bit
is not set resulting in the LED driven as active-low.

To fix this, check if active-low is not set in DT and enable Active High
polarity by default to restore correct funcionality of the LED.

Fixes: 7196062b ("net: phy: at803x: add LED support for qca808x")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In switching to using phy_modify_mmd and a more short version of the
LED ON/OFF condition in later revision, it was made a logic error where

value ? QCA808X_LED_FORCE_ON : QCA808X_LED_FORCE_OFF is always true as
value is always OR with QCA808X_LED_FORCE_EN due to missing ()
resulting in the testing condition being QCA808X_LED_FORCE_EN | value.

Add the () to apply the correct condition and restore correct
functionality of the brightness ON/OFF.

Fixes: 7196062b ("net: phy: at803x: add LED support for qca808x")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jakub Kicinski says:

====================
tools: ynl: auto-gen for all genetlink families

The code gen has caught up with all features required in genetlink
families in Linux 6.8 already. We have also stopped committing auto-
-generated user space code to the tree. Instead of listing all the
families in the Makefile search the spec directory, and generate
code for everything that's not legacy netlink.
====================

Link: https://lore.kernel.org/r/20240202004926.447803-1-kuba@kernel.orgSigned-off-by: Jakub Kicinski <kuba@kernel.org>