linux-2.6-microblaze.git
2 years agowireguard: selftests: support UML
Jason A. Donenfeld [Tue, 2 Aug 2022 12:56:13 +0000 (14:56 +0200)]
wireguard: selftests: support UML

This shoud open up various possibilities like time travel execution, and
is also just another platform to help shake out bugs.

Cc: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agowireguard: allowedips: don't corrupt stack when detecting overflow
Jason A. Donenfeld [Tue, 2 Aug 2022 12:56:12 +0000 (14:56 +0200)]
wireguard: allowedips: don't corrupt stack when detecting overflow

In case push_rcu() and related functions are buggy, there's a
WARN_ON(len >= 128), which the selftest tries to hit by being tricky. In
case it is hit, we shouldn't corrupt the kernel's stack, though;
otherwise it may be hard to even receive the report that it's buggy. So
conditionalize the stack write based on that WARN_ON()'s return value.

Note that this never *actually* happens anyway. The WARN_ON() in the
first place is bounded by IS_ENABLED(DEBUG), and isn't expected to ever
actually hit. This is just a debugging sanity check.

Additionally, hoist the constant 128 into a named enum,
MAX_ALLOWEDIPS_BITS, so that it's clear why this value is chosen.

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Link: https://lore.kernel.org/all/CAHk-=wjJZGA6w_DxA+k7Ejbqsq+uGK==koPai3sqdsfJqemvag@mail.gmail.com/
Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agowireguard: selftests: update config fragments
Lukas Bulwahn [Tue, 2 Aug 2022 12:56:11 +0000 (14:56 +0200)]
wireguard: selftests: update config fragments

The kernel.config and debug.config fragments in wireguard selftests mention
some config symbols that have been reworked:

Commit c5665868183f ("mm: kmemleak: use the memory pool for early
allocations") removes the config DEBUG_KMEMLEAK_EARLY_LOG_SIZE and since
then, the config's feature is available without further configuration.

Commit 4675ff05de2d ("kmemcheck: rip it out") removes kmemcheck and the
corresponding arch config HAVE_ARCH_KMEMCHECK. There is no need for this
config.

Commit 3bf195ae6037 ("netfilter: nat: merge nf_nat_ipv4,6 into nat core")
removes the config NF_NAT_IPV4 and since then, the config's feature is
available without further configuration.

Commit 41a2901e7d22 ("rcu: Remove SPARSE_RCU_POINTER Kconfig option")
removes the config SPARSE_RCU_POINTER and since then, the config's feature
is enabled by default.

Commit dfb4357da6dd ("time: Remove CONFIG_TIMER_STATS") removes the feature
and config CONFIG_TIMER_STATS without any replacement.

Commit 3ca17b1f3628 ("lib/ubsan: remove null-pointer checks") removes the
check and config UBSAN_NULL without any replacement.

Adjust the config fragments to those changes in configs.

Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agowireguard: ratelimiter: use hrtimer in selftest
Jason A. Donenfeld [Tue, 2 Aug 2022 12:56:10 +0000 (14:56 +0200)]
wireguard: ratelimiter: use hrtimer in selftest

Using msleep() is problematic because it's compared against
ratelimiter.c's ktime_get_coarse_boottime_ns(), which means on systems
with slow jiffies (such as UML's forced HZ=100), the result is
inaccurate. So switch to using schedule_hrtimeout().

However, hrtimer gives us access only to the traditional posix timers,
and none of the _COARSE variants. So now, rather than being too
imprecise like jiffies, it's too precise.

One solution would be to give it a large "range" value, but this will
still fire early on a loaded system. A better solution is to align the
timeout to the actual coarse timer, and then round up to the nearest
tick, plus change.

So add the timeout to the current coarse time, and then
schedule_hrtimer() until the absolute computed time.

This should hopefully reduce flakes in CI as well. Note that we keep the
retry loop in case the entire function is running behind, because the
test could still be scheduled out, by either the kernel or by the
hypervisor's kernel, in which case restarting the test and hoping to not
be scheduled out still helps.

Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: net: fix IOAM test skip return code
Kleber Sacilotto de Souza [Mon, 1 Aug 2022 12:46:15 +0000 (14:46 +0200)]
selftests: net: fix IOAM test skip return code

The ioam6.sh test script exits with an error code (1) when tests are
skipped due to lack of support from userspace/kernel or not enough
permissions. It should return the kselftests SKIP code instead.

Reviewed-by: Justin Iurman <justin.iurman@uliege.be>
Signed-off-by: Kleber Sacilotto de Souza <kleber.souza@canonical.com>
Link: https://lore.kernel.org/r/20220801124615.256416-1-kleber.souza@canonical.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: marvell: prestera: remove reduntant code
Sebin Sebastian [Mon, 1 Aug 2022 04:07:31 +0000 (09:37 +0530)]
net: marvell: prestera: remove reduntant code

Fixes the coverity warning 'EVALUATION_ORDER' violation. port is written
twice with the same value.

Signed-off-by: Sebin Sebastian <mailmesebin00@gmail.com>
Link: https://lore.kernel.org/r/20220801040731.34741-1-mailmesebin00@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: devlink: Fix missing mutex_unlock() call
Ammar Faizi [Mon, 1 Aug 2022 11:59:56 +0000 (18:59 +0700)]
net: devlink: Fix missing mutex_unlock() call

Commit 2dec18ad826f forgets to call mutex_unlock() before the function
returns in the error path:

   New smatch warnings:
   net/core/devlink.c:6392 devlink_nl_cmd_region_new() warn: inconsistent \
   returns '&region->snapshot_lock'.

Make sure we call mutex_unlock() in this error path.

Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Fixes: 2dec18ad826f ("net: devlink: remove region snapshots list dependency on devlink->lock")
Signed-off-by: Ammar Faizi <ammarfaizi2@gnuweeb.org>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/20220801115742.1309329-1-ammar.faizi@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/tls: Remove redundant workqueue flush before destroy
Tariq Toukan [Mon, 1 Aug 2022 11:24:44 +0000 (14:24 +0300)]
net/tls: Remove redundant workqueue flush before destroy

destroy_workqueue() safely destroys the workqueue after draining it.
No need for the explicit call to flush_workqueue(). Remove it.

Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/20220801112444.26175-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: txgbe: Fix an error handling path in txgbe_probe()
Christophe JAILLET [Sun, 31 Jul 2022 16:34:15 +0000 (18:34 +0200)]
net: txgbe: Fix an error handling path in txgbe_probe()

A pci_enable_pcie_error_reporting() should be balanced by a corresponding
pci_disable_pcie_error_reporting() call in the error handling path, as
already done in the remove function.

Fixes: 3ce7547e5b71 ("net: txgbe: Add build support for txgbe")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Jiawen Wu <jiawenwu@trustnetic.com>
Link: https://lore.kernel.org/r/082003d00be1f05578c9c6434272ceb314609b8e.1659285240.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: Fix spelling mistakes and cleanup code
Xie Shaowen [Sat, 30 Jul 2022 09:22:54 +0000 (17:22 +0800)]
net: dsa: Fix spelling mistakes and cleanup code

fix follow spelling misktakes:
desconstructed ==> deconstructed
enforcment ==> enforcement

Reported-by: Hacash Robot <hacashRobot@santino.com>
Signed-off-by: Xie Shaowen <studentxswpy@163.com>
Link: https://lore.kernel.org/r/20220730092254.3102875-1-studentxswpy@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoDocumentation: devlink: add add devlink-selftests to the table of contents
Bagas Sanjaya [Sat, 30 Jul 2022 02:20:57 +0000 (09:20 +0700)]
Documentation: devlink: add add devlink-selftests to the table of contents

Commit 08f588fa301bef ("devlink: introduce framework for selftests") adds
documentation for devlink selftests framework, but it is missing from
table of contents.

Add it.

Link: https://lore.kernel.org/linux-doc/202207300406.CUBuyN5i-lkp@intel.com/
Fixes: 08f588fa301bef ("devlink: introduce framework for selftests")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com>
Link: https://lore.kernel.org/r/20220730022058.16813-1-bagasdotme@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonfp: flower: add support for tunnel offload without key ID
Baowen Zheng [Fri, 29 Jul 2022 09:16:41 +0000 (11:16 +0200)]
nfp: flower: add support for tunnel offload without key ID

Currently nfp driver will reject to offload tunnel key action without
tunnel key ID which means tunnel ID is 0. But it is a normal case for tc
flower since user can setup a tunnel with tunnel ID is 0.

So we need to support this case to accept tunnel key action without
tunnel key ID.

Signed-off-by: Baowen Zheng <baowen.zheng@corigine.com>
Reviewed-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20220729091641.354748-1-simon.horman@corigine.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'net-rose-fix-module-unload-issues'
Jakub Kicinski [Mon, 1 Aug 2022 18:59:25 +0000 (11:59 -0700)]
Merge branch 'net-rose-fix-module-unload-issues'

Eric Dumazet says:

====================
net: rose: fix module unload issues

Bernard Pidoux reported that unloading rose module could lead
to infamous "unregistered_netdevice:" issues.

First patch is the fix, stable candidate.
Second patch is adding netdev ref tracker to af_rose.

I chose net-next to not inflict merge conflicts, because
Jakub changed dev_put_track() to netdev_put_track() in net-next.
====================

Link: https://lore.kernel.org/r/20220729091233.1030680-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: rose: add netdev ref tracker to 'struct rose_sock'
Eric Dumazet [Fri, 29 Jul 2022 09:12:33 +0000 (09:12 +0000)]
net: rose: add netdev ref tracker to 'struct rose_sock'

This will help debugging netdevice refcount problems with
CONFIG_NET_DEV_REFCNT_TRACKER=y

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Tested-by: Bernard Pidoux <f6bvp@free.fr>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: rose: fix netdev reference changes
Eric Dumazet [Fri, 29 Jul 2022 09:12:32 +0000 (09:12 +0000)]
net: rose: fix netdev reference changes

Bernard reported that trying to unload rose module would lead
to infamous messages:

unregistered_netdevice: waiting for rose0 to become free. Usage count = xx

This patch solves the issue, by making sure each socket referring to
a netdevice holds a reference count on it, and properly releases it
in rose_release().

rose_dev_first() is also fixed to take a device reference
before leaving the rcu_read_locked section.

Following patch will add ref_tracker annotations to ease
future bug hunting.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: Bernard Pidoux <f6bvp@free.fr>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Bernard Pidoux <f6bvp@free.fr>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue
Jakub Kicinski [Mon, 1 Aug 2022 18:15:39 +0000 (11:15 -0700)]
Merge branch '1GbE' of git://git./linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
1GbE Intel Wired LAN Driver Updates 2022-07-28

Jacob Keller says:

Convert all of the Intel drivers with PTP support to the newer .adjfine
implementation which uses scaled parts per million.

This improves the precision of the frequency adjustments by taking advantage
of the full scaled parts per million input coming from user space.

In addition, all implementations are converted to using the
mul_u64_u64_div_u64 function which better handles the intermediate value.
This function supports architecture specific instructions where possible to
avoid loss of precision if the normal 64-bit multiplication would overflow.

Of note, the i40e implementation is now able to avoid loss of precision on
slower link speeds by taking advantage of this to multiply by the link speed
factor first. This results in a significantly more precise adjustment by
allowing the calculation to impact the lower bits.

This also gets us a step closer to being able to remove the .adjfreq
entirely by removing its use from many drivers.

I plan to follow this up with a series to update the drivers from other
vendors and drop the .adjfreq implementation entirely.

* '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
  igb: convert .adjfreq to .adjfine
  ixgbe: convert .adjfreq to .adjfine
  i40e: convert .adjfreq to .adjfine
  i40e: use mul_u64_u64_div_u64 for PTP frequency calculation
  e1000e: convert .adjfreq to .adjfine
  e1000e: remove unnecessary range check in e1000e_phc_adjfreq
  ice: implement adjfine with mul_u64_u64_div_u64
====================

Link: https://lore.kernel.org/r/20220728181836.3387862-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agodt-bindings: net: fsl,fec: Add i.MX8ULP FEC items
Wei Fang [Tue, 26 Jul 2022 14:38:51 +0000 (00:38 +1000)]
dt-bindings: net: fsl,fec: Add i.MX8ULP FEC items

Add fsl,imx8ulp-fec for i.MX8ULP platform.

Signed-off-by: Wei Fang <wei.fang@nxp.com>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/all/20220726143853.23709-2-wei.fang@nxp.com/
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'funeth-tx-xdp-frags'
David S. Miller [Mon, 1 Aug 2022 11:38:57 +0000 (12:38 +0100)]
Merge branch 'funeth-tx-xdp-frags'

Dimitris Michailidis says:

====================
net/funeth: Tx support for XDP with frags

Support XDP with fragments for XDP_TX and ndo_xdp_xmit.

The first three patches rework existing code used by the skb path to
make it suitable also for XDP. With these all the callees of the main
Tx XDP function, fun_xdp_tx(), are fragment-capable. The last patch
updates fun_xdp_tx() to handle fragments.
====================

Tested-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/funeth: Tx handling of XDP with fragments.
Dimitris Michailidis [Fri, 29 Jul 2022 07:32:57 +0000 (00:32 -0700)]
net/funeth: Tx handling of XDP with fragments.

By now all the functions fun_xdp_tx() calls are shared with the skb path
and thus are fragment-capable. Update fun_xdp_tx(), that up to now has
been passing just one buffer, to check for fragments and call
accordingly.  This makes XDP_TX and ndo_xdp_xmit fragment-capable.

Signed-off-by: Dimitris Michailidis <dmichail@fungible.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/funeth: Unify skb/XDP packet mapping.
Dimitris Michailidis [Fri, 29 Jul 2022 07:32:56 +0000 (00:32 -0700)]
net/funeth: Unify skb/XDP packet mapping.

Instead of passing an skb to the mapping function pass an
skb_shared_info plus an additional address/length pair. This makes it
usable for both skbs and XDP. Call it from the XDP path and adjust the
skb path.

Signed-off-by: Dimitris Michailidis <dmichail@fungible.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/funeth: Unify skb/XDP gather list writing.
Dimitris Michailidis [Fri, 29 Jul 2022 07:32:55 +0000 (00:32 -0700)]
net/funeth: Unify skb/XDP gather list writing.

Extract the Tx gather list writing code that skbs use into a utility
function and use it also for XDP.

Signed-off-by: Dimitris Michailidis <dmichail@fungible.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/funeth: Unify skb/XDP Tx packet unmapping.
Dimitris Michailidis [Fri, 29 Jul 2022 07:32:54 +0000 (00:32 -0700)]
net/funeth: Unify skb/XDP Tx packet unmapping.

Current XDP unmapping is a subset of its skb analog, dealing with
only one buffer. In preparation for multi-frag XDP rename the skb
function and use it also for XDP. The XDP version is removed.

Signed-off-by: Dimitris Michailidis <dmichail@fungible.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'devlink-parallel-commands'
David S. Miller [Mon, 1 Aug 2022 11:14:01 +0000 (12:14 +0100)]
Merge branch 'devlink-parallel-commands'

Jiri Pirko says:

====================
net: devlink: allow parallel commands on multiple devlinks

Aim of this patchset is to remove devlink_mutex and eventually to enable
parallel ops on devlink netlink interface.
====================

Tested-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: devlink: enable parallel ops on netlink interface
Jiri Pirko [Fri, 29 Jul 2022 07:10:38 +0000 (09:10 +0200)]
net: devlink: enable parallel ops on netlink interface

As the devlink_mutex was removed and all devlink instances are protected
individually by devlink->lock mutex, allow the netlink ops to run
in parallel and therefore allow user to execute commands on multiple
devlink instances simultaneously.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: devlink: remove devlink_mutex
Jiri Pirko [Fri, 29 Jul 2022 07:10:37 +0000 (09:10 +0200)]
net: devlink: remove devlink_mutex

All accesses to devlink structure from userspace and drivers are locked
with devlink->lock instance mutex. Also, devlinks xa_array iteration is
taken care of by iteration helpers taking devlink reference.

Therefore, remove devlink_mutex as it is no longer needed.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: devlink: convert reload command to take implicit devlink->lock
Jiri Pirko [Fri, 29 Jul 2022 07:10:36 +0000 (09:10 +0200)]
net: devlink: convert reload command to take implicit devlink->lock

Convert reload command to behave the same way as the rest of the
commands and let if be called with devlink->lock held. Remove the
temporary devl_lock taking from drivers. As the DEVLINK_NL_FLAG_NO_LOCK
flag is no longer used, remove it alongside.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: devlink: introduce "unregistering" mark and use it during devlinks iteration
Jiri Pirko [Fri, 29 Jul 2022 07:10:35 +0000 (09:10 +0200)]
net: devlink: introduce "unregistering" mark and use it during devlinks iteration

Add new mark called "unregistering" to be set at the beginning of
devlink_unregister() function. Check this mark during devlinks
iteration in order to prevent getting a reference of devlink which is
being currently unregistered.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoudp: Remove redundant __udp_sysctl_init() call from udp_init().
Kuniyuki Iwashima [Fri, 29 Jul 2022 03:21:37 +0000 (20:21 -0700)]
udp: Remove redundant __udp_sysctl_init() call from udp_init().

__udp_sysctl_init() is called for init_net via udp_sysctl_ops.

While at it, we can rename __udp_sysctl_init() to udp_sysctl_init().

Fixes: 1e8029515816 ("udp: Move the udp sysctl to namespace.")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge tag 'linux-can-next-for-5.20-20220731' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Mon, 1 Aug 2022 10:42:11 +0000 (11:42 +0100)]
Merge tag 'linux-can-next-for-5.20-20220731' of git://git./linux/kernel/git/mkl/linux-can-next
Marc Kleine-Budde says:

====================
pull-request: can-next 2022-07-31

this is a pull request of 36 patches for net-next/master.

The 1st patch is by me and fixes a typo in the mcp251xfd driver.

Vincent Mailhol contributes a series of 9 patches, which clean up the
drivers to make use of KBUILD_MODNAME instead of hard coded names and
remove DRV_VERSION.

Followed by 3 patches by Vincent Mailhol that directly set the
ethtool_ops in instead of calling a function in the slcan, c_can and
flexcan driver.

Vincent Mailhol contributes a KBUILD_MODNAME and pr_fmt cleanup patch
for the slcan driver. Dario Binacchi contributes 6 patches to clean up
the driver and remove the legacy driver infrastructure.

The next 14 patches are by Vincent Mailhol and target the various
drivers, they add ethtool support and reporting of timestamping
capabilities.

Another patch by Vincent Mailhol for the etas_es58x driver to remove
useless calls to usb_fill_bulk_urb().

The last patch is by Christophe JAILLET and fixes a broken link to
Documentation in the can327 driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agocan: can327: fix a broken link to Documentation
Christophe JAILLET [Sun, 31 Jul 2022 06:32:52 +0000 (08:32 +0200)]
can: can327: fix a broken link to Documentation

Since commit 482a4360c56a ("docs: networking: convert netdevices.txt to
ReST"), Documentation/networking/netdevices.txt has been replaced by
Documentation/networking/netdevices.rst.

Update the comment accordingly to avoid a 'make htmldocs' warning.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/all/6a54aff884ea4f84b661527d75aabd6632140715.1659249135.git.christophe.jaillet@wanadoo.fr
Fixes: 43da2f07622f ("can: can327: CAN/ldisc driver for ELM327 based OBD-II adapters")
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agoMerge tag 'mlx5-updates-2022-07-28' of git://git.kernel.org/pub/scm/linux/kernel...
Jakub Kicinski [Sat, 30 Jul 2022 04:39:06 +0000 (21:39 -0700)]
Merge tag 'mlx5-updates-2022-07-28' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-updates-2022-07-28

Misc updates to mlx5 driver:

1) Gal corrects to use skb_tcp_all_headers on encapsulated skbs.

2) Roi Adds the support for offloading standalone police actions.

3) lama, did some refactoring to minimize code coupling with
mlx5e_priv "god object" in some of the follows, and converts some of the
objects to pointers to preserve on memory when these objects aren't needed.
This is part one of two parts series.

* tag 'mlx5-updates-2022-07-28' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
  net/mlx5e: Move mlx5e_init_l2_addr to en_main
  net/mlx5e: Split en_fs ndo's and move to en_main
  net/mlx5e: Separate mlx5e_set_rx_mode_work and move caller to en_main
  net/mlx5e: Add mdev to flow_steering struct
  net/mlx5e: Report flow steering errors with mdev err report API
  net/mlx5e: Convert mlx5e_flow_steering member of mlx5e_priv to pointer
  net/mlx5e: Allocate VLAN and TC for featured profiles only
  net/mlx5e: Make mlx5e_tc_table private
  net/mlx5e: Convert mlx5e_tc_table member of mlx5e_flow_steering to pointer
  net/mlx5e: TC, Support tc action api for police
  net/mlx5e: TC, Separate get/update/replace meter functions
  net/mlx5e: Add red and green counters for metering
  net/mlx5e: TC, Allocate post meter ft per rule
  net/mlx5: DR, Add support for flow metering ASO
  net/mlx5e: Fix wrong use of skb_tcp_all_headers() with encapsulation
====================

Link: https://lore.kernel.org/r/20220728205728.143074-1-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next...
Jakub Kicinski [Sat, 30 Jul 2022 04:26:09 +0000 (21:26 -0700)]
Merge branch '100GbE' of git://git./linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
100GbE Intel Wired LAN Driver Updates 2022-07-28

This series contains updates to ice driver only.

Michal allows for VF true promiscuous mode to be set for multiple VFs
and adds clearing of promiscuous filters when VF trust is removed.

Maciej refactors ice_set_features() to track/check changed features
instead of constantly checking against netdev features and adds support for
NETIF_F_LOOPBACK.

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
  ice: allow toggling loopback mode via ndo_set_features callback
  ice: compress branches in ice_set_features()
  ice: Fix promiscuous mode not turning off
  ice: Introduce enabling promiscuous mode on multiple VF's
====================

Link: https://lore.kernel.org/r/20220728195538.3391360-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'sfc-vf-representors-for-ef100-rx-side'
Jakub Kicinski [Sat, 30 Jul 2022 04:22:09 +0000 (21:22 -0700)]
Merge branch 'sfc-vf-representors-for-ef100-rx-side'

Edward Cree says:

====================
sfc: VF representors for EF100 - RX side

This series adds the receive path for EF100 VF representors, plus other
 minor features such as statistics.
====================

Link: https://lore.kernel.org/r/cover.1659034549.git.ecree.xilinx@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosfc: implement ethtool get/set RX ring size for EF100 reps
Edward Cree [Thu, 28 Jul 2022 18:57:52 +0000 (19:57 +0100)]
sfc: implement ethtool get/set RX ring size for EF100 reps

It's not truly a ring, but the maximum length of the list of queued RX
 SKBs is analogous to an RX ring size, so use that API to configure it.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosfc: use a dynamic m-port for representor RX and set it promisc
Edward Cree [Thu, 28 Jul 2022 18:57:51 +0000 (19:57 +0100)]
sfc: use a dynamic m-port for representor RX and set it promisc

Representors do not want to be subject to the PF's Ethernet address
 filters, since traffic from VFs will typically have a destination
 either elsewhere on the link segment or on an overlay network.
So, create a dynamic m-port with promiscuous and all-multicast
 filters, and set it as the egress port of representor default rules.
 Since the m-port is an alias of the calling PF's own m-port, traffic
 will still be delivered to the PF's RXQs, but it will be subject to
 the VNRX filter rules installed on the dynamic m-port (specified by
 the v-port ID field of the filter spec).

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosfc: move table locking into filter_table_{probe,remove} methods
Edward Cree [Thu, 28 Jul 2022 18:57:50 +0000 (19:57 +0100)]
sfc: move table locking into filter_table_{probe,remove} methods

We need to be able to drop the efx->filter_sem in ef100_filter_table_up()
 so that we can call functions that insert filters (and thus take that
 rwsem for read), which means the efx->type->filter_table_probe method
 needs to be responsible for taking the lock in the first place.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosfc: insert default MAE rules to connect VFs to representors
Edward Cree [Thu, 28 Jul 2022 18:57:49 +0000 (19:57 +0100)]
sfc: insert default MAE rules to connect VFs to representors

Default rules are low-priority switching rules which the hardware uses
 in the absence of higher-priority rules.  Each representor requires a
 corresponding rule matching traffic from its representee VF and
 delivering to the PF (where a check on INGRESS_MPORT in
 __ef100_rx_packet() will direct it to the representor).  No rule is
 required in the reverse direction, because representor TX uses a TX
 override descriptor to bypass the MAE and deliver directly to the VF.
Since inserting any rule into the MAE disables the firmware's own
 default rules, also insert a pair of rules to connect the PF to the
 physical network port and vice-versa.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosfc: receive packets from EF100 VFs into representors
Edward Cree [Thu, 28 Jul 2022 18:57:48 +0000 (19:57 +0100)]
sfc: receive packets from EF100 VFs into representors

If the source m-port of a packet in __ef100_rx_packet() is a VF,
 hand off the packet to the corresponding representor with
 efx_ef100_rep_rx_packet().

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosfc: check ef100 RX packets are from the wire
Edward Cree [Thu, 28 Jul 2022 18:57:47 +0000 (19:57 +0100)]
sfc: check ef100 RX packets are from the wire

If not, for now drop them and warn.  A subsequent patch will look up
 the source m-port to try and find a representor to deliver them to.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosfc: determine wire m-port at EF100 PF probe time
Edward Cree [Thu, 28 Jul 2022 18:57:46 +0000 (19:57 +0100)]
sfc: determine wire m-port at EF100 PF probe time

Traffic delivered to the (MAE admin) PF could be from either the wire
 or a VF.  The INGRESS_MPORT field of the RX prefix distinguishes these;
 base_mport is the value this field will have for traffic from the wire
 (which should be delivered to the PF's netdevice, not a representor).

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosfc: ef100 representor RX top half
Edward Cree [Thu, 28 Jul 2022 18:57:45 +0000 (19:57 +0100)]
sfc: ef100 representor RX top half

Representor RX uses a NAPI context driven by a 'fake interrupt': when
 the parent PF receives a packet destined for the representor, it adds
 it to an SKB list (efv->rx_list), and schedules NAPI if the 'fake
 interrupt' is primed.  The NAPI poll then pulls packets off this list
 and feeds them to the stack with netif_receive_skb_list().
This scheme allows us to decouple representor RX from the parent PF's
 RX fast-path.
This patch implements the 'top half', which builds an SKB, copies data
 into it from the RX buffer (which can then be released), adds it to
 the queue and fires the 'fake interrupt' if necessary.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosfc: ef100 representor RX NAPI poll
Edward Cree [Thu, 28 Jul 2022 18:57:44 +0000 (19:57 +0100)]
sfc: ef100 representor RX NAPI poll

This patch adds the 'bottom half' napi->poll routine for representor RX.
See the next patch (with the top half) for an explanation of the 'fake
 interrupt' scheme used to drive this NAPI context.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosfc: plumb ef100 representor stats
Edward Cree [Thu, 28 Jul 2022 18:57:43 +0000 (19:57 +0100)]
sfc: plumb ef100 representor stats

Implement .ndo_get_stats64() method to read values out of struct
 efx_rep_sw_stats.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: marvell: prestera: uninitialized variable bug
Dan Carpenter [Thu, 28 Jul 2022 14:32:36 +0000 (17:32 +0300)]
net: marvell: prestera: uninitialized variable bug

The "ret" variable needs to be initialized at the start.

Fixes: 52323ef75414 ("net: marvell: prestera: add phylink support")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/YuKeBBuGtsmd7QdT@kili
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agodn_route: replace "jiffies-now>0" with "jiffies!=now"
Yu Zhe [Fri, 29 Jul 2022 06:17:12 +0000 (14:17 +0800)]
dn_route: replace "jiffies-now>0" with "jiffies!=now"

Use "jiffies != now" to replace "jiffies - now > 0" to make
code more readable. We want to put a limit on how long the
loop can run for before rescheduling.

Signed-off-by: Yu Zhe <yuzhe@nfschina.com>
Link: https://lore.kernel.org/r/20220729061712.22666-1-yuzhe@nfschina.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'wireless-next-2022-07-29' of git://git.kernel.org/pub/scm/linux/kernel...
Jakub Kicinski [Sat, 30 Jul 2022 02:34:45 +0000 (19:34 -0700)]
Merge tag 'wireless-next-2022-07-29' of git://git./linux/kernel/git/wireless/wireless-next

Kalle Valo says:

====================
wireless-next patches for v5.20

Fourth set of patches for v5.20, last few patches before the merge
window. Only driver changes this time, mostly just fixes and cleanup.

Major changes:

brcmfmac
 - support brcm,ccode-map-trivial DT property

wcn36xx
 - add debugfs file to show firmware feature strings

* tag 'wireless-next-2022-07-29' of git://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next: (36 commits)
  wifi: rtw88: check the return value of alloc_workqueue()
  wifi: rtw89: 8852a: adjust IMR for SER L1
  wifi: rtw89: 8852a: update RF radio A/B R56
  wifi: wcn36xx: Add debugfs entry to read firmware feature strings
  wifi: wcn36xx: Move capability bitmap to string translation function to firmware.c
  wifi: wcn36xx: Move firmware feature bit storage to dedicated firmware.c file
  wifi: wcn36xx: Rename clunky firmware feature bit enum
  wifi: brcmfmac: prevent double-free on hardware-reset
  wifi: brcmfmac: support brcm,ccode-map-trivial DT property
  dt-bindings: bcm4329-fmac: add optional brcm,ccode-map-trivial
  wifi: brcmfmac: Replace default (not configured) MAC with a random MAC
  wifi: brcmfmac: Add brcmf_c_set_cur_etheraddr() helper
  wifi: brcmfmac: Remove #ifdef guards for PM related functions
  wifi: brcmfmac: use strreplace() in brcmf_of_probe()
  wifi: plfxlc: Use eth_zero_addr() to assign zero address
  wifi: wilc1000: use existing iftype variable to store the interface type
  wifi: wilc1000: add 'isinit' flag for SDIO bus similar to SPI
  wifi: wilc1000: cancel the connect operation during interface down
  wifi: wilc1000: get correct length of string WID from received config packet
  wifi: wilc1000: set station_info flag only when signal value is valid
  ...
====================

Link: https://lore.kernel.org/r/20220729192832.A5011C433D6@smtp.kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next
Jakub Kicinski [Sat, 30 Jul 2022 02:04:28 +0000 (19:04 -0700)]
Merge https://git./linux/kernel/git/bpf/bpf-next

Andrii Nakryiko says:

====================
 bpf-next 2022-07-29

We've added 22 non-merge commits during the last 4 day(s) which contain
a total of 27 files changed, 763 insertions(+), 120 deletions(-).

The main changes are:

1) Fixes to allow setting any source IP with bpf_skb_set_tunnel_key() helper,
   from Paul Chaignon.

2) Fix for bpf_xdp_pointer() helper when doing sanity checking, from Joanne Koong.

3) Fix for XDP frame length calculation, from Lorenzo Bianconi.

4) Libbpf BPF_KSYSCALL docs improvements and fixes to selftests to accommodate
   s390x quirks with socketcall(), from Ilya Leoshkevich.

5) Allow/denylist and CI configs additions to selftests/bpf to improve BPF CI,
   from Daniel Müller.

6) BPF trampoline + ftrace follow up fixes, from Song Liu and Xu Kuohai.

7) Fix allocation warnings in netdevsim, from Jakub Kicinski.

8) bpf_obj_get_opts() libbpf API allowing to provide file flags, from Joe Burton.

9) vsnprintf usage fix in bpf_snprintf_btf(), from Fedor Tokarev.

10) Various small fixes and clean ups, from Daniel Müller, Rongguang Wei,
    Jörn-Thorben Hinz, Yang Li.

* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (22 commits)
  bpf: Remove unneeded semicolon
  libbpf: Add bpf_obj_get_opts()
  netdevsim: Avoid allocation warnings triggered from user space
  bpf: Fix NULL pointer dereference when registering bpf trampoline
  bpf: Fix test_progs -j error with fentry/fexit tests
  selftests/bpf: Bump internal send_signal/send_signal_tracepoint timeout
  bpftool: Don't try to return value from void function in skeleton
  bpftool: Replace sizeof(arr)/sizeof(arr[0]) with ARRAY_SIZE macro
  bpf: btf: Fix vsnprintf return value check
  libbpf: Support PPC in arch_specific_syscall_pfx
  selftests/bpf: Adjust vmtest.sh to use local kernel configuration
  selftests/bpf: Copy over libbpf configs
  selftests/bpf: Sort configuration
  selftests/bpf: Attach to socketcall() in test_probe_user
  libbpf: Extend BPF_KSYSCALL documentation
  bpf, devmap: Compute proper xdp_frame len redirecting frames
  bpf: Fix bpf_xdp_pointer return pointer
  selftests/bpf: Don't assign outer source IP to host
  bpf: Set flow flag to allow any source IP in bpf_tunnel_key
  geneve: Use ip_tunnel_key flow flags in route lookups
  ...
====================

Link: https://lore.kernel.org/r/20220729230948.1313527-1-andrii@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agobpf: Remove unneeded semicolon
Yang Li [Mon, 25 Jul 2022 22:27:33 +0000 (06:27 +0800)]
bpf: Remove unneeded semicolon

Eliminate the following coccicheck warning:
/kernel/bpf/trampoline.c:101:2-3: Unneeded semicolon

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220725222733.55613-1-yang.lee@linux.alibaba.com
2 years agolibbpf: Add bpf_obj_get_opts()
Joe Burton [Fri, 29 Jul 2022 20:27:27 +0000 (20:27 +0000)]
libbpf: Add bpf_obj_get_opts()

Add an extensible variant of bpf_obj_get() capable of setting the
`file_flags` parameter.

This parameter is needed to enable unprivileged access to BPF maps.
Without a method like this, users must manually make the syscall.

Signed-off-by: Joe Burton <jevburton@google.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220729202727.3311806-1-jevburton.kernel@gmail.com
2 years agonetdevsim: Avoid allocation warnings triggered from user space
Jakub Kicinski [Tue, 26 Jul 2022 21:36:05 +0000 (14:36 -0700)]
netdevsim: Avoid allocation warnings triggered from user space

We need to suppress warnings from sily map sizes. Also switch
from GFP_USER to GFP_KERNEL_ACCOUNT, I'm pretty sure I misunderstood
the flags when writing this code.

Fixes: 395cacb5f1a0 ("netdevsim: bpf: support fake map offload")
Reported-by: syzbot+ad24705d3fd6463b18c6@syzkaller.appspotmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220726213605.154204-1-kuba@kernel.org
2 years agobpf: Fix NULL pointer dereference when registering bpf trampoline
Xu Kuohai [Thu, 28 Jul 2022 11:40:48 +0000 (07:40 -0400)]
bpf: Fix NULL pointer dereference when registering bpf trampoline

A panic was reported on arm64:

[   44.517109] audit: type=1334 audit(1658859870.268:59): prog-id=19 op=LOAD
[   44.622031] Unable to handle kernel NULL pointer dereference at
virtual address 0000000000000010
[   44.624321] Mem abort info:
[   44.625049]   ESR = 0x0000000096000004
[   44.625935]   EC = 0x25: DABT (current EL), IL = 32 bits
[   44.627182]   SET = 0, FnV = 0
[   44.627930]   EA = 0, S1PTW = 0
[   44.628684]   FSC = 0x04: level 0 translation fault
[   44.629788] Data abort info:
[   44.630474]   ISV = 0, ISS = 0x00000004
[   44.631362]   CM = 0, WnR = 0
[   44.632041] user pgtable: 4k pages, 48-bit VAs, pgdp=0000000100ab5000
[   44.633494] [0000000000000010] pgd=0000000000000000, p4d=0000000000000000
[   44.635202] Internal error: Oops: 96000004 [#1] SMP
[   44.636452] Modules linked in: xfs crct10dif_ce ghash_ce virtio_blk
virtio_console virtio_mmio qemu_fw_cfg
[   44.638713] CPU: 2 PID: 1 Comm: systemd Not tainted 5.19.0-rc7 #1
[   44.640164] Hardware name: QEMU KVM Virtual Machine, BIOS 0.0.0 02/06/2015
[   44.641799] pstate: 00400005 (nzcv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   44.643404] pc : ftrace_set_filter_ip+0x24/0xa0
[   44.644659] lr : bpf_trampoline_update.constprop.0+0x428/0x4a0
[   44.646118] sp : ffff80000803b9f0
[   44.646950] x29: ffff80000803b9f0 x28: ffff0b5d80364400 x27: ffff80000803bb48
[   44.648721] x26: ffff8000085ad000 x25: ffff0b5d809d2400 x24: 0000000000000000
[   44.650493] x23: 00000000ffffffed x22: ffff0b5dd7ea0900 x21: 0000000000000000
[   44.652279] x20: 0000000000000000 x19: 0000000000000000 x18: ffffffffffffffff
[   44.654067] x17: 0000000000000000 x16: 0000000000000000 x15: ffffffffffffffff
[   44.655787] x14: ffff0b5d809d2498 x13: ffff0b5d809d2432 x12: 0000000005f5e100
[   44.657535] x11: abcc77118461cefd x10: 000000000000005f x9 : ffffa7219cb5b190
[   44.659254] x8 : ffffa7219c8e0000 x7 : 0000000000000000 x6 : ffffa7219db075e0
[   44.661066] x5 : ffffa7219d3130e0 x4 : ffffa7219cab9da0 x3 : 0000000000000000
[   44.662837] x2 : 0000000000000000 x1 : ffffa7219cb7a5c0 x0 : 0000000000000000
[   44.664675] Call trace:
[   44.665274]  ftrace_set_filter_ip+0x24/0xa0
[   44.666327]  bpf_trampoline_update.constprop.0+0x428/0x4a0
[   44.667696]  __bpf_trampoline_link_prog+0xcc/0x1c0
[   44.668834]  bpf_trampoline_link_prog+0x40/0x64
[   44.669919]  bpf_tracing_prog_attach+0x120/0x490
[   44.671011]  link_create+0xe0/0x2b0
[   44.671869]  __sys_bpf+0x484/0xd30
[   44.672706]  __arm64_sys_bpf+0x30/0x40
[   44.673678]  invoke_syscall+0x78/0x100
[   44.674623]  el0_svc_common.constprop.0+0x4c/0xf4
[   44.675783]  do_el0_svc+0x38/0x4c
[   44.676624]  el0_svc+0x34/0x100
[   44.677429]  el0t_64_sync_handler+0x11c/0x150
[   44.678532]  el0t_64_sync+0x190/0x194
[   44.679439] Code: 2a0203f4 f90013f5 2a0303f5 f9001fe1 (f9400800)
[   44.680959] ---[ end trace 0000000000000000 ]---
[   44.682111] Kernel panic - not syncing: Oops: Fatal exception
[   44.683488] SMP: stopping secondary CPUs
[   44.684551] Kernel Offset: 0x2721948e0000 from 0xffff800008000000
[   44.686095] PHYS_OFFSET: 0xfffff4a380000000
[   44.687144] CPU features: 0x010,00022811,19001080
[   44.688308] Memory Limit: none
[   44.689082] ---[ end Kernel panic - not syncing: Oops: Fatal exception ]---

It's caused by a NULL tr->fops passed to ftrace_set_filter_ip(). tr->fops
is initialized to NULL and is assigned to an allocated memory address if
CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS is enabled. Since there is no
direct call on arm64 yet, the config can't be enabled.

To fix it, call ftrace_set_filter_ip() only if tr->fops is not NULL.

Fixes: 00963a2e75a8 ("bpf: Support bpf_trampoline on functions with IPMODIFY (e.g. livepatch)")
Reported-by: Bruno Goncalves <bgoncalv@redhat.com>
Signed-off-by: Xu Kuohai <xukuohai@huawei.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Tested-by: Bruno Goncalves <bgoncalv@redhat.com>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Link: https://lore.kernel.org/bpf/20220728114048.3540461-1-xukuohai@huaweicloud.com
2 years agobpf: Fix test_progs -j error with fentry/fexit tests
Song Liu [Fri, 29 Jul 2022 19:41:06 +0000 (12:41 -0700)]
bpf: Fix test_progs -j error with fentry/fexit tests

When multiple threads are attaching/detaching fentry/fexit programs to
the same trampoline, we may call register_fentry on the same trampoline
twice: register_fentry(), unregister_fentry(), then register_fentry again.
This causes ftrace_set_filter_ip() for the same ip on tr->fops twice,
which leaves duplicated ip in tr->fops. The extra ip is not cleaned up
properly on unregister and thus causes failures with further register in
register_ftrace_direct_multi():

register_ftrace_direct_multi()
{
        ...
        for (i = 0; i < size; i++) {
                hlist_for_each_entry(entry, &hash->buckets[i], hlist) {
                        if (ftrace_find_rec_direct(entry->ip))
                                goto out_unlock;
                }
        }
        ...
}

This can be triggered with parallel fentry/fexit tests with test_progs:

  ./test_progs -t fentry,fexit -j

Fix this by resetting tr->fops in ftrace_set_filter_ip(), so that there
will never be duplicated entries in tr->fops.

Fixes: 00963a2e75a8 ("bpf: Support bpf_trampoline on functions with IPMODIFY (e.g. livepatch)")
Reported-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Song Liu <song@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220729194106.1207472-1-song@kernel.org
2 years agoselftests/bpf: Bump internal send_signal/send_signal_tracepoint timeout
Daniel Müller [Wed, 27 Jul 2022 18:29:55 +0000 (18:29 +0000)]
selftests/bpf: Bump internal send_signal/send_signal_tracepoint timeout

The send_signal/send_signal_tracepoint is pretty flaky, with at least
one failure in every ten runs on a few attempts I've tried it:
  > test_send_signal_common:PASS:pipe_c2p 0 nsec
  > test_send_signal_common:PASS:pipe_p2c 0 nsec
  > test_send_signal_common:PASS:fork 0 nsec
  > test_send_signal_common:PASS:skel_open_and_load 0 nsec
  > test_send_signal_common:PASS:skel_attach 0 nsec
  > test_send_signal_common:PASS:pipe_read 0 nsec
  > test_send_signal_common:PASS:pipe_write 0 nsec
  > test_send_signal_common:PASS:reading pipe 0 nsec
  > test_send_signal_common:PASS:reading pipe error: size 0 0 nsec
  > test_send_signal_common:FAIL:incorrect result unexpected incorrect result: actual 48 != expected 50
  > test_send_signal_common:PASS:pipe_write 0 nsec
  > #139/1   send_signal/send_signal_tracepoint:FAIL

The reason does not appear to be a correctness issue in the strict
sense. Rather, we merely do not receive the signal we are waiting for
within the provided timeout.
Let's bump the timeout by a factor of ten. With that change I have not
been able to reproduce the failure in 150+ iterations. I am also sneaking
in a small simplification to the test_progs test selection logic.

Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20220727182955.4044988-1-deso@posteo.net
2 years agobpftool: Don't try to return value from void function in skeleton
Jörn-Thorben Hinz [Tue, 26 Jul 2022 13:32:03 +0000 (15:32 +0200)]
bpftool: Don't try to return value from void function in skeleton

A skeleton generated by bpftool previously contained a return followed
by an expression in OBJ_NAME__detach(), which has return type void. This
did not hurt, the bpf_object__detach_skeleton() called there returns
void itself anyway, but led to a warning when compiling with e.g.
-pedantic.

Signed-off-by: Jörn-Thorben Hinz <jthinz@mailbox.tu-berlin.de>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20220726133203.514087-1-jthinz@mailbox.tu-berlin.de
2 years agobpftool: Replace sizeof(arr)/sizeof(arr[0]) with ARRAY_SIZE macro
Rongguang Wei [Tue, 26 Jul 2022 09:30:45 +0000 (17:30 +0800)]
bpftool: Replace sizeof(arr)/sizeof(arr[0]) with ARRAY_SIZE macro

Use the ARRAY_SIZE macro and make the code more compact.

Signed-off-by: Rongguang Wei <weirongguang@kylinos.cn>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20220726093045.3374026-1-clementwei90@163.com
2 years agobpf: btf: Fix vsnprintf return value check
Fedor Tokarev [Mon, 11 Jul 2022 21:13:17 +0000 (23:13 +0200)]
bpf: btf: Fix vsnprintf return value check

vsnprintf returns the number of characters which would have been written if
enough space had been available, excluding the terminating null byte. Thus,
the return value of 'len_left' means that the last character has been
dropped.

Signed-off-by: Fedor Tokarev <ftokarev@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Alan Maguire <alan.maguire@oracle.com>
Link: https://lore.kernel.org/bpf/20220711211317.GA1143610@laptop
2 years agoMerge ath-next from git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git
Kalle Valo [Fri, 29 Jul 2022 13:38:03 +0000 (16:38 +0300)]
Merge ath-next from git://git./linux/kernel/git/kvalo/ath.git

ath.git patches for v5.20. Major changes:

ath11k:

* fix WCN9074 to work again

* revert rfkill support as it was causing problems

2 years agowifi: rtw88: check the return value of alloc_workqueue()
William Dean [Sat, 23 Jul 2022 06:37:56 +0000 (14:37 +0800)]
wifi: rtw88: check the return value of alloc_workqueue()

The function alloc_workqueue() in rtw_core_init() can fail, but
there is no check of its return value. To fix this bug, its return value
should be checked with new error handling code.

Fixes: fe101716c7c9d ("rtw88: replace tx tasklet with work queue")
Reported-by: Hacash Robot <hacashRobot@santino.com>
Signed-off-by: William Dean <williamsukatube@gmail.com>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/20220723063756.2956189-1-williamsukatube@163.com
2 years agowifi: rtw89: 8852a: adjust IMR for SER L1
Zong-Zhe Yang [Thu, 21 Jul 2022 07:49:52 +0000 (15:49 +0800)]
wifi: rtw89: 8852a: adjust IMR for SER L1

SER (system error recovery) L1 (level 1) has a step-by-step handshake
process with FW. These handshakes still rely on B_AX_HS0ISR_IND_INT_EN.
So, even already during recovery, we enable this bit in IMR.

Signed-off-by: Zong-Zhe Yang <kevin_yang@realtek.com>
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/20220721074952.19676-1-pkshih@realtek.com
2 years agowifi: rtw89: 8852a: update RF radio A/B R56
Ping-Ke Shih [Wed, 22 Jun 2022 09:19:37 +0000 (17:19 +0800)]
wifi: rtw89: 8852a: update RF radio A/B R56

Update to internal tag HALRF_027_00_060.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/20220622091937.11325-1-pkshih@realtek.com
2 years agowifi: wcn36xx: Add debugfs entry to read firmware feature strings
Bryan O'Donoghue [Wed, 27 Jul 2022 16:16:55 +0000 (17:16 +0100)]
wifi: wcn36xx: Add debugfs entry to read firmware feature strings

Add in the ability to easily find the firmware feature bits reported in the
get feature exchange without having to compile-in debug prints.

root@linaro-alip:~# cat /sys/kernel/debug/ieee80211/phy0/wcn36xx/firmware_feat_caps
MCC
P2P
DOT11AC
SLM_SESSIONIZATION
DOT11AC_OPMODE
SAP32STA
TDLS
P2P_GO_NOA_DECOUPLE_INIT_SCAN
WLANACTIVE_OFFLOAD
BEACON_OFFLOAD
SCAN_OFFLOAD
BCN_MISS_OFFLOAD
STA_POWERSAVE
STA_ADVANCED_PWRSAVE
BCN_FILTER
RTT
RATECTRL
WOW
WLAN_ROAM_SCAN_OFFLOAD
SPECULATIVE_PS_POLL
IBSS_HEARTBEAT_OFFLOAD
WLAN_SCAN_OFFLOAD
WLAN_PERIODIC_TX_PTRN
ADVANCE_TDLS
BATCH_SCAN
FW_IN_TX_PATH
EXTENDED_NSOFFLOAD_SLOT
CH_SWITCH_V1
HT40_OBSS_SCAN
UPDATE_CHANNEL_LIST
WLAN_MCADDR_FLT
WLAN_CH144
TDLS_SCAN_COEXISTENCE
LINK_LAYER_STATS_MEAS
MU_MIMO
EXTENDED_SCAN
DYNAMIC_WMM_PS
MAC_SPOOFED_SCAN
FW_STATS
WPS_PRBRSP_TMPL
BCN_IE_FLT_DELTA

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20220727161655.2286867-5-bryan.odonoghue@linaro.org
2 years agowifi: wcn36xx: Move capability bitmap to string translation function to firmware.c
Bryan O'Donoghue [Wed, 27 Jul 2022 16:16:54 +0000 (17:16 +0100)]
wifi: wcn36xx: Move capability bitmap to string translation function to firmware.c

Move wcn36xx_get_cap_name() function in main.c into firmware.c as
wcn36xx_firmware_get_cap_name().

Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20220727161655.2286867-4-bryan.odonoghue@linaro.org
2 years agowifi: wcn36xx: Move firmware feature bit storage to dedicated firmware.c file
Bryan O'Donoghue [Wed, 27 Jul 2022 16:16:53 +0000 (17:16 +0100)]
wifi: wcn36xx: Move firmware feature bit storage to dedicated firmware.c file

The naming of the get/set/clear firmware feature capability bits doesn't
really follow the established namespace pattern of
wcn36xx_logicalblock_do_something();

The feature bits are accessed by smd.c and main.c. It would be nice to
display the found feature bits in debugfs. To do so though we should tidy
up the namespace a bit.

Move the firmware feature exchange API to its own file - firmware.c giving
us the opportunity to functionally decompose other firmware related
accessors as appropriate in future.

Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20220727161655.2286867-3-bryan.odonoghue@linaro.org
2 years agowifi: wcn36xx: Rename clunky firmware feature bit enum
Bryan O'Donoghue [Wed, 27 Jul 2022 16:16:52 +0000 (17:16 +0100)]
wifi: wcn36xx: Rename clunky firmware feature bit enum

The enum name "place_holder_in_cap_bitmap" is self descriptively asking to
be changed to something else.

Rename place_holder_in_cap_bitmap to wcn36xx_firmware_feat_caps so that the
contents and intent of the enum is obvious.

Reviewed-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20220727161655.2286867-2-bryan.odonoghue@linaro.org
2 years agoMerge branch 'mtk_eth_soc-xdp-multi-frame'
David S. Miller [Fri, 29 Jul 2022 11:18:10 +0000 (12:18 +0100)]
Merge branch 'mtk_eth_soc-xdp-multi-frame'

Lorenzo Bianconi says:

====================
mtk_eth_soc: introduce xdp multi-frag support

Convert mtk_eth_soc driver to xdp_return_frame_bulk APIs.
===================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: ethernet: mtk_eth_soc: add xdp tx return bulking support
Lorenzo Bianconi [Wed, 27 Jul 2022 21:20:52 +0000 (23:20 +0200)]
net: ethernet: mtk_eth_soc: add xdp tx return bulking support

Convert mtk_eth_soc driver to xdp_return_frame_bulk APIs.

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: ethernet: mtk_eth_soc: introduce xdp multi-frag support
Lorenzo Bianconi [Wed, 27 Jul 2022 21:20:51 +0000 (23:20 +0200)]
net: ethernet: mtk_eth_soc: introduce xdp multi-frag support

Add the capability to map non-linear xdp frames in XDP_TX and
ndo_xdp_xmit callback.

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: ethernet: mtk_eth_soc: introduce mtk_xdp_frame_map utility routine
Lorenzo Bianconi [Wed, 27 Jul 2022 21:20:50 +0000 (23:20 +0200)]
net: ethernet: mtk_eth_soc: introduce mtk_xdp_frame_map utility routine

This is a preliminary patch to add xdp multi-frag support to mtk_eth_soc
driver

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'seg6-headend-reduced'
David S. Miller [Fri, 29 Jul 2022 11:14:03 +0000 (12:14 +0100)]
Merge branch 'seg6-headend-reduced'

Andrea Mayer says:

====================
seg6: add support for SRv6 Headend Reduced

This patchset adds support for SRv6 Headend behavior with Reduced
Encapsulation. It introduces the H.Encaps.Red and H.L2Encaps.Red versions
of the SRv6 H.Encaps and H.L2Encaps behaviors, according to RFC 8986 [1].

In details, the patchset is made of:
 - patch 1/4: add support for SRv6 H.Encaps.Red behavior;
 - Patch 2/4: add support for SRv6 H.L2Encaps.Red behavior;
 - patch 2/4: add selftest for SRv6 H.Encaps.Red behavior;
 - patch 3/4: add selftest for SRv6 H.L2Encaps.Red behavior.

The corresponding iproute2 patch for supporting SRv6 H.Encaps.Red and
H.L2Encaps.Red behaviors is provided in a separated patchset.

[1] - https://datatracker.ietf.org/doc/html/rfc8986

V4 -> v5:
 - Fix skb checksum for SRH Reduced encapsulation/insertion;

 - Improve selftests by:
      i) adding a random suffix to network namespaces;
     ii) creating net devices directly into network namespaces;
    iii) using trap EXIT command to properly clean up selftest networks.

 Thanks to Paolo Abeni.

v3 -> v4:
 - Add selftests to the Makefile, thanks to Jakub Kicinski.

v2 -> v3:
 - Keep SRH when HMAC TLV is present;

 - Split the support for H.Encaps.Red and H.L2Encaps.Red behaviors in two
   patches (respectively, patch 1/4 and patch 2/4);

 - Add selftests for SRv6 H.Encaps.Red and H.L2Encaps.Red.

v1 -> v2:
 - Fixed sparse warnings;

 - memset now uses sizeof() instead of hardcoded value;

 - Removed EXPORT_SYMBOL_GPL.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: seg6: add selftest for SRv6 H.L2Encaps.Red behavior
Andrea Mayer [Wed, 27 Jul 2022 18:54:08 +0000 (20:54 +0200)]
selftests: seg6: add selftest for SRv6 H.L2Encaps.Red behavior

This selftest is designed for testing the H.L2Encaps.Red behavior. It
instantiates a virtual network composed of several nodes: hosts and SRv6
routers. Each node is realized using a network namespace that is
properly interconnected to others through veth pairs.
The test considers SRv6 routers implementing a L2 VPN leveraged by hosts
for communicating with each other. Such routers make use of the SRv6
H.L2Encaps.Red behavior for applying SRv6 policies to L2 traffic coming
from hosts.

The correct execution of the behavior is verified through reachability
tests carried out between hosts belonging to the same VPN.

Signed-off-by: Andrea Mayer <andrea.mayer@uniroma2.it>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: seg6: add selftest for SRv6 H.Encaps.Red behavior
Andrea Mayer [Wed, 27 Jul 2022 18:54:07 +0000 (20:54 +0200)]
selftests: seg6: add selftest for SRv6 H.Encaps.Red behavior

This selftest is designed for testing the H.Encaps.Red behavior. It
instantiates a virtual network composed of several nodes: hosts and SRv6
routers. Each node is realized using a network namespace that is
properly interconnected to others through veth pairs.
The test considers SRv6 routers implementing L3 VPNs leveraged by hosts
for communicating with each other. Such routers make use of the SRv6
H.Encaps.Red behavior for applying SRv6 policies to L3 traffic coming
from hosts.

The correct execution of the behavior is verified through reachability
tests carried out between hosts belonging to the same VPN.

Signed-off-by: Andrea Mayer <andrea.mayer@uniroma2.it>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoseg6: add support for SRv6 H.L2Encaps.Red behavior
Andrea Mayer [Wed, 27 Jul 2022 18:54:06 +0000 (20:54 +0200)]
seg6: add support for SRv6 H.L2Encaps.Red behavior

The SRv6 H.L2Encaps.Red behavior described in [1] is an optimization of
the SRv6 H.L2Encaps behavior [2].

H.L2Encaps.Red reduces the length of the SRH by excluding the first
segment (SID) in the SRH of the pushed IPv6 header. The first SID is
only placed in the IPv6 Destination Address field of the pushed IPv6
header.
When the SRv6 Policy only contains one SID the SRH is omitted, unless
there is an HMAC TLV to be carried.

[1] - https://datatracker.ietf.org/doc/html/rfc8986#section-5.4
[2] - https://datatracker.ietf.org/doc/html/rfc8986#section-5.3

Signed-off-by: Andrea Mayer <andrea.mayer@uniroma2.it>
Signed-off-by: Anton Makarov <anton.makarov11235@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoseg6: add support for SRv6 H.Encaps.Red behavior
Andrea Mayer [Wed, 27 Jul 2022 18:54:05 +0000 (20:54 +0200)]
seg6: add support for SRv6 H.Encaps.Red behavior

The SRv6 H.Encaps.Red behavior described in [1] is an optimization of
the SRv6 H.Encaps behavior [2].

H.Encaps.Red reduces the length of the SRH by excluding the first
segment (SID) in the SRH of the pushed IPv6 header. The first SID is
only placed in the IPv6 Destination Address field of the pushed IPv6
header.
When the SRv6 Policy only contains one SID the SRH is omitted, unless
there is an HMAC TLV to be carried.

[1] - https://datatracker.ietf.org/doc/html/rfc8986#section-5.2
[2] - https://datatracker.ietf.org/doc/html/rfc8986#section-5.1

Signed-off-by: Andrea Mayer <andrea.mayer@uniroma2.it>
Signed-off-by: Anton Makarov <anton.makarov11235@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agovmxnet3: do not reschedule napi for rx processing
Ronak Doshi [Wed, 27 Jul 2022 17:30:37 +0000 (10:30 -0700)]
vmxnet3: do not reschedule napi for rx processing

Commit '2c5a5748105a ("vmxnet3: add support for out of order rx
completion")' added support for out of order rx completion. Within
that patch, an enhancement was done to reschedule napi for processing
rx completions.

However, it can lead to missing an interrupt. So, this patch reverts
that part of the code.

Fixes: 2c5a5748105a ("vmxnet3: add support for out of order rx completion")
Signed-off-by: Ronak Doshi <doshir@vmware.com>
Acked-by: Guolin Yang <gyang@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoDocumentation: Describe net.ipv4.tcp_reflect_tos.
Guillaume Nault [Wed, 27 Jul 2022 11:18:21 +0000 (13:18 +0200)]
Documentation: Describe net.ipv4.tcp_reflect_tos.

The tcp_reflect_tos option was introduced in Linux 5.10 but was still
undocumented.

Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/af_packet: check len when min_header_len equals to 0
Zhengchao Shao [Wed, 27 Jul 2022 09:33:12 +0000 (17:33 +0800)]
net/af_packet: check len when min_header_len equals to 0

User can use AF_PACKET socket to send packets with the length of 0.
When min_header_len equals to 0, packet_snd will call __dev_queue_xmit
to send packets, and sock->type can be any type.

Reported-by: syzbot+5ea725c25d06fb9114c4@syzkaller.appspotmail.com
Fixes: fd1894224407 ("bpf: Don't redirect packets with invalid pkt_len")
Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'mlxsw-ptp-spectrum-2'
David S. Miller [Fri, 29 Jul 2022 10:02:24 +0000 (11:02 +0100)]
Merge branch 'mlxsw-ptp-spectrum-2'

Ido Schimmel says:

====================
mlxsw: Add PTP support for Spectrum-2 and newer ASICs

This patchset adds PTP support for Spectrum-{2,3,4} switch ASICs. They
all act largely the same with respect to PTP except for a workaround
implemented for Spectrum-{2,3} in patch #6.

Spectrum-2 and newer ASICs essentially implement a transparent clock
between all the switch ports, including the CPU port. The hardware will
generate the UTC time stamp for transmitted / received packets at the
CPU port, but will compensate for forwarding delays in the ASIC by
adjusting the correction field in the PTP header (for PTP events) at the
ingress and egress ports.

Specifically, the hardware will subtract the current time stamp from the
correction field at the ingress port and will add the current time stamp
to the correction field at the egress port. For the purpose of an
ordinary or boundary clock (this patchset), the correction field will
always be adjusted between the CPU port and one of the front panel
ports, but never between two front panel ports.

Patchset overview:

Patch #1 extracts a helper to configure traps for PTP packets (event and
general messages). The helper is shared between all Spectrum
generations.

Patch #2 transitions Spectrum-2 and newer ASICs to use a different
format of Tx completions that includes the UTC time stamp of transmitted
packets.

Patch #3 adds basic initialization required for Spectrum-2 PTP support.
It mainly invokes the helper from patch #1.

Patch #4 adds helpers to read the UTC time (seconds and nanoseconds)
from the device over memory-mapped I/O instead of going through firmware
which is slower and therefore inaccurate. The helpers will be used to
implement various PHC operations (e.g., gettimex64) and to construct the
full UTC time stamp from the truncated one reported over Tx / Rx
completions.

Patch #5 implements the various PHC operations.

Patch #6 implements the previously described workaround for
Spectrum-{2,3}.

Patch #7 adds the ability to report a hardware time stamp for a received
/ transmitted packet based off the associated Rx / Tx completion that
includes a truncated UTC time stamp.

Patches #8 and #9 implement support for the SIOCGHWTSTAMP /
SIOCSHWTSTAMP ioctls and the get_ts_info ethtool callback, respectively.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum: Support ethtool 'get_ts_info' callback in Spectrum-2
Danielle Ratson [Wed, 27 Jul 2022 06:23:28 +0000 (09:23 +0300)]
mlxsw: spectrum: Support ethtool 'get_ts_info' callback in Spectrum-2

The 'get_ts_info' callback is used for obtaining information about
time stamping and PTP hardware clock capabilities of a network device.

The existing function of Spectrum-1 is used to advertise the PHC
capabilities and the supported RX and TX filters. Implement a similar
function for Spectrum-2, expose that the supported 'rx_filters' are all
PTP event packets, as for these packets the driver fills the time stamp
from the CQE in the SKB.

In the future, mlxsw driver will be extended to support one-step PTP in
Spectrum-2 and newer ASICs. Then additional 'tx_types' will be supported.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_ptp: Support SIOCGHWTSTAMP, SIOCSHWTSTAMP ioctls
Danielle Ratson [Wed, 27 Jul 2022 06:23:27 +0000 (09:23 +0300)]
mlxsw: spectrum_ptp: Support SIOCGHWTSTAMP, SIOCSHWTSTAMP ioctls

The SIOCSHWTSTAMP ioctl configures HW timestamping on a given port. In
Spectrum-2 and above, each packet gets time stamp by default, but in
order to provide an accurate time stamp, software should configure to
update the correction field. In addition, the PTP traps are not enabled
by default, software should enable it per port or for all ports.

The switch behaves like a transparent clock between CPU port and each
front panel port. If ingress correction is set on a port for a given packet
type, then when such a packet is received via the port, the current time
stamp is subtracted from the correction field. If egress correction is set
on a port for a given packet type, then when such a packet is transmitted
via the port, the current time stamp is added to the correction field.

The result is that as the packet ingresses through a port with ingress
correction enabled, and egresses through a port with egress correction
enabled, the PTP correction field is updated to reflect the time that the
packet spent in the ASIC.

This can be used to update the correction field of trapped packets by
enabling ingress correction on a port where time stamping was enabled,
and egress correction on the CPU port. Similarly, for packets transmitted
from the host, ingress correction should be enabled on the CPU port, and
egress correction on a front-panel port.

However, since the correction fields will be updated for all PTP packets
crossing the CPU port, in order not to mangle the correction field, the
front panel port involved in the packet transfer must have the
corresponding correction enabled as well.

Therefore, when HW timestamping is enabled on at least one port, we have
to configure hardware to update the correction field and trap PTP event
packets on all ports.

Add reference count as part of 'struct mlxsw_sp_ptp_state', to maintain
how many ports use HW timestamping. Handle the correction field
configuration only when the first port enables time stamping and when the
last port disables time stamping. Store the configuration as part of
'struct mlxsw_sp_ptp_state', as it is global for all ports.

The SIOCGHWTSTAMP ioctl is a getter for the current configuration,
implement it and use the global configuration.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum: Support time stamping on Spectrum-2
Danielle Ratson [Wed, 27 Jul 2022 06:23:26 +0000 (09:23 +0300)]
mlxsw: spectrum: Support time stamping on Spectrum-2

As opposed to Spectrum-1, in which time stamps arrive through a pair of
dedicated events into a queue and later are being matched to the
corresponding packets, in Spectrum-2 we are reading the time stamps
directly from the CQE. Software can get the time stamp in UTC format
using CQEv2.

Add a time stamp field to 'struct mlxsw_skb_cb'. In
mlxsw_pci_cqe_{rdq,sdq}_handle() extract the time stamp from the CQE into
the new time stamp field. Note that the time stamp in the CQE is
represented by 38 bits, which is a short representation of UTC time.
Software should create the full time stamp using the global UTC clock.
Read UTC clock from hardware only for PTP packets which were trapped to CPU
with PTP0 trap ID (event packets).

Use the time stamp from the SKB when packet is received or transmitted.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: Send PTP packets as data packets to overcome a limitation
Danielle Ratson [Wed, 27 Jul 2022 06:23:25 +0000 (09:23 +0300)]
mlxsw: Send PTP packets as data packets to overcome a limitation

In Spectrum-2 and Spectrum-3, the correction field of PTP packets which are
sent as control packets is not updated at egress port. To overcome this
limitation, PTP packets which require time stamp, should be sent as data
packets with the following details:
1. FID valid = 1
2. FID value above the maximum FID
3. rx_router_port = 1

>From Spectrum-4 and on, this limitation will be solved.

Extend the function which handles TX header, in case that the packet is
a PTP packet, add TX header with type=data and all the above mentioned
requirements. Add operation as part of 'struct mlxsw_sp_ptp_ops', to be
able to separate the handling of PTP packets between different ASICs. Use
the data packet solution only for Spectrum-2 and Spectrum-3. Therefore, add
a dedicated operation structure for Spectrum-4, as it will be same to
Spectrum-2 in PTP implementation, just will not have the limitation of
control packets.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_ptp: Add implementation for physical hardware clock operations
Danielle Ratson [Wed, 27 Jul 2022 06:23:24 +0000 (09:23 +0300)]
mlxsw: spectrum_ptp: Add implementation for physical hardware clock operations

Implement physical hardware clock operations. The main difference between
the existing operations of Spectrum-1 and the new operations of Spectrum-2
is the usage of UTC hardware clock instead of FRC.

Add support for init() and fini() functions for PTP clock in Spectrum-2.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: Query UTC sec and nsec PCI offsets and values
Danielle Ratson [Wed, 27 Jul 2022 06:23:23 +0000 (09:23 +0300)]
mlxsw: Query UTC sec and nsec PCI offsets and values

Query UTC sec and nsec PCI offsets during the pci_init(), to be able to
read UTC time later.

Implement functions to read UTC seconds and nanoseconds from the offset
which was read as part of initialization.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_ptp: Add PTP initialization / finalization for Spectrum-2
Danielle Ratson [Wed, 27 Jul 2022 06:23:22 +0000 (09:23 +0300)]
mlxsw: spectrum_ptp: Add PTP initialization / finalization for Spectrum-2

Lay the groundwork for Spectrum-2 support. On Spectrum-2, the packets get
the time stamps from the CQE, which means that the time stamp is attached
to its packet.

Configure MTPTPT to set which message types should arrive under which
PTP trap. PTP0 will be used for event message types, which means that
the packets require time stamp. PTP1 will be used for other packets.

Note that in Spectrum-2, all packets contain time stamp by default. The two
types of traps (PTP0, PTP1) will be used to separate between PTP_EVENT
traps and PTP_GENERAL traps, so then the driver will fill the time stamp as
part of the SKB only for event message types.

Later the driver will enable the traps using 'MTPCPC.ptp_trap_en' bit.
Then, PTP packets start arriving through the PTP traps.

Currently, the structure 'mlxsw_sp2_ptp_state' contains only the common
structure, the next patches will extend it.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: Support CQEv2 for SDQ in Spectrum-2 and newer ASICs
Danielle Ratson [Wed, 27 Jul 2022 06:23:21 +0000 (09:23 +0300)]
mlxsw: Support CQEv2 for SDQ in Spectrum-2 and newer ASICs

Currently, Tx completions are reported using Completion Queue Element
version 1 (CQEv1). These elements do not contain the Tx time stamp,
which is fine as Spectrum-1 reads Tx time stamps via a dedicated FIFO
and Spectrum-2 does not currently support PTP.

In preparation for Spectrum-2 PTP support, use CQEv2 for Spectrum-2 and
newer ASICs, as this CQE format encodes the Tx time stamp.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomlxsw: spectrum_ptp: Add helper functions to configure PTP traps
Amit Cohen [Wed, 27 Jul 2022 06:23:20 +0000 (09:23 +0300)]
mlxsw: spectrum_ptp: Add helper functions to configure PTP traps

MTPTPT register is used to set which message types should arrive under
which PTP trap. Currently, PTP0 is used for event message types, which
means that the packets require time stamp. PTP1 is used for other packets.

This configuration will be same for Spectrum-2 and newer ASICs. In
preparation for Spectrum-2 PTP support, add helper functions to
configure PTP traps and use them for Spectrum-1. These functions will be
used later also for Spectrum-2.

Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agocan: etas_es58x: remove useless calls to usb_fill_bulk_urb()
Vincent Mailhol [Fri, 29 Jul 2022 08:09:02 +0000 (17:09 +0900)]
can: etas_es58x: remove useless calls to usb_fill_bulk_urb()

Aside of urb->transfer_buffer_length and urb->context which might
change in the TX path, all the other URB parameters remains constant
during runtime. So, there is no reasons to call usb_fill_bulk_urb()
each time before submitting an URB.

Make sure to initialize all the fields of the URB at allocation
time. For the TX branch, replace the call usb_fill_bulk_urb() by an
assignment of urb->context. urb->urb->transfer_buffer_length is
already set by the caller functions, no need to set it again. For the
RX branch, because all parameters are unchanged, simply remove the
call to usb_fill_bulk_urb().

Signed-off-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr>
Link: https://lore.kernel.org/all/20220729080902.25839-1-mailhol.vincent@wanadoo.fr
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
2 years agodt-bindings: nfc: use spi-peripheral-props.yaml
Krzysztof Kozlowski [Wed, 27 Jul 2022 16:41:29 +0000 (18:41 +0200)]
dt-bindings: nfc: use spi-peripheral-props.yaml

Instead of listing directly properties typical for SPI peripherals,
reference the spi-peripheral-props.yaml schema.  This allows using all
properties typical for SPI-connected devices, even these which device
bindings author did not tried yet.

Remove the spi-* properties which now come via spi-peripheral-props.yaml
schema, except for the cases when device schema adds some constraints
like maximum frequency.

While changing additionalProperties->unevaluatedProperties, put it in
typical place, just before example DTS.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Acked-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20220727164130.385411-1-krzysztof.kozlowski@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'net-dsa-qca8k-code-split-for-qca8k'
Jakub Kicinski [Fri, 29 Jul 2022 05:24:41 +0000 (22:24 -0700)]
Merge branch 'net-dsa-qca8k-code-split-for-qca8k'

Christian Marangi says:

====================
net: dsa: qca8k: code split for qca8k

This is needed ad ipq4019 SoC have an internal switch that is
based on qca8k with very minor changes. The general function is equal.

Because of this we split the driver to common and specific code.

As the common function needs to be moved to a different file to be
reused, we had to convert every remaining user of qca8k_read/write/rmw
to regmap variant.
We had also to generilized the special handling for the ethtool_stats
function that makes use of the autocast mib. (ipq4019 will have a
different tagger and use mmio so it could be quicker to use mmio instead
of automib feature)
And we had to convert the regmap read/write to bulk implementation to
drop the special function that makes use of it. This will be compatible
with ipq4019 and at the same time permits normal switch to use the eth
mgmt way to send the entire ATU table read/write in one go.
====================

Link: https://lore.kernel.org/r/20220727113523.19742-1-ansuelsmth@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: move read_switch_id function to common code
Christian Marangi [Wed, 27 Jul 2022 11:35:23 +0000 (13:35 +0200)]
net: dsa: qca8k: move read_switch_id function to common code

The same function to read the switch id is used by drivers based on
qca8k family switch. Move them to common code to make them accessible
also by other drivers.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: move port LAG functions to common code
Christian Marangi [Wed, 27 Jul 2022 11:35:22 +0000 (13:35 +0200)]
net: dsa: qca8k: move port LAG functions to common code

The same port LAG functions are used by drivers based on qca8k family
switch. Move them to common code to make them accessible also by other
drivers.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: move port VLAN functions to common code
Christian Marangi [Wed, 27 Jul 2022 11:35:21 +0000 (13:35 +0200)]
net: dsa: qca8k: move port VLAN functions to common code

The same port VLAN functions are used by drivers based on qca8k family
switch. Move them to common code to make them accessible also by other
drivers.
Also drop exposing busy_wait and make it static.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: move port mirror functions to common code
Christian Marangi [Wed, 27 Jul 2022 11:35:20 +0000 (13:35 +0200)]
net: dsa: qca8k: move port mirror functions to common code

The same port mirror functions are used by drivers based on qca8k family
switch. Move them to common code to make them accessible also by other
drivers.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: move port FDB/MDB function to common code
Christian Marangi [Wed, 27 Jul 2022 11:35:19 +0000 (13:35 +0200)]
net: dsa: qca8k: move port FDB/MDB function to common code

The same port FDB/MDB function are used by drivers based on qca8k family
switch. Move them to common code to make them accessible also by other
drivers.
Also drop bulk read/write functions and make them static

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: move set age/MTU/port enable/disable functions to common code
Christian Marangi [Wed, 27 Jul 2022 11:35:18 +0000 (13:35 +0200)]
net: dsa: qca8k: move set age/MTU/port enable/disable functions to common code

The same set age, MTU and port enable/disable function are used by
driver based on qca8k family switch.
Move them to common code to make them accessible also by other drivers.
While at it also drop unnecessary qca8k_priv cast for void pointers.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: move bridge functions to common code
Christian Marangi [Wed, 27 Jul 2022 11:35:17 +0000 (13:35 +0200)]
net: dsa: qca8k: move bridge functions to common code

The same bridge functions are used by drivers based on qca8k family
switch. Move them to common code to make them accessible also by other
drivers.
While at it also drop unnecessary qca8k_priv cast for void pointers.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: move port set status/eee/ethtool stats function to common code
Christian Marangi [Wed, 27 Jul 2022 11:35:16 +0000 (13:35 +0200)]
net: dsa: qca8k: move port set status/eee/ethtool stats function to common code

The same logic to disable/enable port, set eee and get ethtool stats is
used by drivers based on qca8k family switch.
Move it to common code to make it accessible also by other drivers.
While at it also drop unnecessary qca8k_priv cast for void pointers.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: move mib init function to common code
Christian Marangi [Wed, 27 Jul 2022 11:35:15 +0000 (13:35 +0200)]
net: dsa: qca8k: move mib init function to common code

The same mib function is used by drivers based on qca8k family switch.
Move it to common code to make it accessible also by other drivers.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: move qca8k bulk read/write helper to common code
Christian Marangi [Wed, 27 Jul 2022 11:35:14 +0000 (13:35 +0200)]
net: dsa: qca8k: move qca8k bulk read/write helper to common code

The same ATU function are used by drivers based on qca8k family switch.
Move the bulk read/write helper to common code to declare these shared
ATU functions in common code.
These helper will be dropped when regmap correctly support bulk
read/write.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: qca8k: move qca8k read/write/rmw and reg table to common code
Christian Marangi [Wed, 27 Jul 2022 11:35:13 +0000 (13:35 +0200)]
net: dsa: qca8k: move qca8k read/write/rmw and reg table to common code

The same reg table and read/write/rmw function are used by drivers
based on qca8k family switch.
Move them to common code to make it accessible also by other drivers.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>