OSDN Git Service

tomoyo/tomoyo-test1.git
22 months agomptcp: add mptcp_for_each_subflow_safe helper
Matthieu Baerts [Tue, 6 Sep 2022 20:55:39 +0000 (22:55 +0200)]
mptcp: add mptcp_for_each_subflow_safe helper

Similar to mptcp_for_each_subflow(): this is clearer now that the _safe
version is used in multiple places.

Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agonet: phy: spi_ks8895: switch to using gpiod API
Dmitry Torokhov [Tue, 6 Sep 2022 20:49:22 +0000 (13:49 -0700)]
net: phy: spi_ks8895: switch to using gpiod API

This patch switches the driver away from legacy gpio/of_gpio API to
gpiod API, and removes use of of_get_named_gpio_flags() which I want to
make private to gpiolib.

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/r/20220906204922.3789922-3-dmitry.torokhov@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agonet: ks8851: switch to using gpiod API
Dmitry Torokhov [Tue, 6 Sep 2022 20:49:21 +0000 (13:49 -0700)]
net: ks8851: switch to using gpiod API

This patch switches the driver away from legacy gpio/of_gpio API to
gpiod API, and removes use of of_get_named_gpio_flags() which I want to
make private to gpiolib.

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/r/20220906204922.3789922-2-dmitry.torokhov@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agonet: davicom: dm9000: switch to using gpiod API
Dmitry Torokhov [Tue, 6 Sep 2022 20:49:20 +0000 (13:49 -0700)]
net: davicom: dm9000: switch to using gpiod API

This patch switches the driver away from legacy gpio/of_gpio API to
gpiod API, and removes use of of_get_named_gpio_flags() which I want to
make private to gpiolib.

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/r/20220906204922.3789922-1-dmitry.torokhov@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agor8169: remove rtl_wol_shutdown_quirk()
Heiner Kallweit [Tue, 6 Sep 2022 18:24:57 +0000 (20:24 +0200)]
r8169: remove rtl_wol_shutdown_quirk()

Since f658b90977d2 ("r8169: fix DMA being used after buffer free if WoL is
enabled") it has been redundant to disable PCI bus mastering in
rtl_wol_shutdown_quirk(). And since 120068481405 ("r8169: fix failing WoL")
CmdRxEnb is still enabled when we get here. So we can remove the function.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/2391ada0-eac5-ac43-f061-a7a44b0e7f33@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agosfc: introduce shutdown entry point in efx pci driver
Pieter Jansen van Vuuren [Tue, 6 Sep 2022 10:56:20 +0000 (11:56 +0100)]
sfc: introduce shutdown entry point in efx pci driver

Make the device inactive when the system shutdown callback has been
invoked. This is achieved by freezing the driver and disabling the
PCI bus mastering.

Co-developed-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@amd.com>
Acked-by: Edward Cree <ecree.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20220906105620.26179-1-pieter.jansen-van-vuuren@amd.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agonet: openvswitch: fix repeated words in comments
Jilin Yuan [Wed, 7 Sep 2022 04:03:46 +0000 (12:03 +0800)]
net: openvswitch: fix repeated words in comments

Delete the redundant word 'is'.

Signed-off-by: Jilin Yuan <yuanjilin@cdjrlc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agoMerge branch 'net-ipa-next'
David S. Miller [Fri, 9 Sep 2022 10:45:25 +0000 (11:45 +0100)]
Merge branch 'net-ipa-next'

Alex Elder says:

====================
net: ipa: don't use lists for transaction state

This is the last series of patches to convert the IPA code so
integer IDs are used rather than lists to track the state of
transactions.

A first series of patches added ID fields to track the state of
transactions:
  https://lore.kernel.org/netdev/20220831224017.377745-1-elder@linaro.org
The second series started transitioning code to use these IDs rather
than lists to manage state:
  https://lore.kernel.org/netdev/20220902210218.745873-1-elder@linaro.org

This final series finishes the transition, to always use IDs instead
of the lists to manage transaction state.  As a result, the list
fields, links, and a spinlock to protect updates are no longer
needed, so they are removed.  This permits a few other improvements
to be implemented.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: ipa: don't have gsi_channel_update() return a value
Alex Elder [Tue, 6 Sep 2022 17:19:42 +0000 (12:19 -0500)]
net: ipa: don't have gsi_channel_update() return a value

If it finds no completed transactions, gsi_channel_trans_complete()
calls gsi_channel_update() to check hardware.  If new transactions
have completed, gsi_channel_update() records that, then calls
gsi_channel_trans_complete() to return the first of those found.
This recursion won't go any further, but can be avoided if we
have gsi_channel_update() only be responsible for updating state
after accessing hardware.

Change gsi_channel_update() so it simply checks for and handles
new completions, without returning a value.  If it needs to call
that function, have gsi_channel_trans_complete() determine whether
there are new transactions available after the update.

Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: ipa: update channel in gsi_channel_trans_complete()
Alex Elder [Tue, 6 Sep 2022 17:19:41 +0000 (12:19 -0500)]
net: ipa: update channel in gsi_channel_trans_complete()

Have gsi_channel_trans_complete() update the known state from
hardware rather than doing so in gsi_channel_poll_one().

Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: ipa: kill all other transaction lists
Alex Elder [Tue, 6 Sep 2022 17:19:40 +0000 (12:19 -0500)]
net: ipa: kill all other transaction lists

None of the transaction lists are actually needed any more, because
transaction IDs (which have been shown to be equivalent) are used
instead.  So we can remove all of them, as well as the spinlock
that protects updates to them.

Not requiring a lock simplifies gsi_trans_free() as well; we only
need to check the reference count once to decide whether we've hit
the last reference.

This makes the links field in the gsi_trans structure unused, so get
rid of that as well.

Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: ipa: kill the allocated transaction list
Alex Elder [Tue, 6 Sep 2022 17:19:39 +0000 (12:19 -0500)]
net: ipa: kill the allocated transaction list

The only place the trans_info->alloc list is used is when
initializing it, when adding a transaction to it when allocation
finishes, and when moving a transaction from that list to the
committed list.

We can just skip putting a transaction on the allocated list, and
add it (rather than move it) to the committed list when it is
committed.

On additional caveat is that an allocated transaction that's
committed without any TREs added will be immediately freed.  Because
we aren't adding allocated transactions to a list any more, the
list links need to be initialized to ensure they're valid at the
time list_del() is called for the transaction.

Then we can safely eliminate the allocated transaction list.

Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: ipa: always use transaction IDs instead of lists
Alex Elder [Tue, 6 Sep 2022 17:19:38 +0000 (12:19 -0500)]
net: ipa: always use transaction IDs instead of lists

In gsi_channel_trans_complete(), use the completed and pending IDs
to determine whether there are any transactions in completed state.

Similarly, in gsi_channel_trans_cancel_pending(), use the pending
and committed IDs to mark pending transactions cancelled.  Rearrange
the logic a bit there for a simpler result.

This removes the only user of list_last_entry_or_null(), so get rid
of that macro.

Remove the temporary warnings added by the previous commit.

Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agoMerge branch 'felix-dsa-ethtool-stats'
David S. Miller [Fri, 9 Sep 2022 09:59:13 +0000 (10:59 +0100)]
Merge branch 'felix-dsa-ethtool-stats'

Vladimir Oltean says:

====================
Standardized ethtool counters for Felix DSA driver

The main purpose of this change set is to add reporting of structured
ethtool statistics counters to the felix DSA driver (see patch 11/14 for
details), as a prerequisite for extending these counters to the
eMAC/pMAC defined by the IEEE MAC Merge layer.

Along the way, the main purpose has diverged into multiple sub-purposes
which are also tackled:

- A bug fix patch submitted to "net" has made ocelot->stats_lock a spin
  lock, which is not an issue currently (all Ocelot switches are MMIO),
  but will be an issue for Colin Foster who is working on a SPI
  controlled Ocelot switch. We restore the hardware access to port stats
  to be sleepable.

- PSFP (tc-gate, tc-police) tc-flower stats on Felix use a non-converged
  procedure to access the hardware counters, although the interface is
  very similar to what is used for the port counters. Benefit from the
  logic used for the port counters, which gains us 64-bit tc-flower
  stats that are resistant to overflows.

- Also export the ndo_get_stats64 method used by the ocelot switchdev
  driver to Felix, so that ifconfig sees something hardware-based as
  well (but not 100% up to date).

- Create a new ocelot_stats.c file which groups everything stats-related
  together. Along with this, also move some other topic-specific code,
  like FDB and PTP, out of the main ocelot.c.

- Minimize the lines of code for the stats layout definitions. These
  changes alone cause the patch set to have an overall reduction of
  lines of code in the driver, even though we are adding new
  functionality as well.

Tested the port counters with lockdep and friends, with some
garden-variety traffic (ping, iperf3) and the PSFP counters with
tools/testing/selftests/drivers/net/ocelot/psfp.sh.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: mscc: ocelot: share the common stat definitions between all drivers
Vladimir Oltean [Thu, 8 Sep 2022 16:48:16 +0000 (19:48 +0300)]
net: mscc: ocelot: share the common stat definitions between all drivers

All switch families supported by the ocelot lib (ocelot, felix, seville)
export the same registers so far. But for example felix also has TSN
counters, while the others don't.

To reduce the bloat even further, create an OCELOT_COMMON_STATS() macro
which just lists all stats that are common between switches. The array
elements are still replicated among all of vsc9959_stats_layout,
vsc9953_stats_layout and ocelot_stats_layout.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: mscc: ocelot: minimize definitions for stats
Vladimir Oltean [Thu, 8 Sep 2022 16:48:15 +0000 (19:48 +0300)]
net: mscc: ocelot: minimize definitions for stats

The current definition of struct ocelot_stat_layout is long-winded (4
lines per entry, and we have hundreds of entries), so we could make an
effort to use the C preprocessor and reduce the line count.

Create an implicit correspondence between enum ocelot_reg, which tells
us the register address (SYS_COUNT_RX_OCTETS etc) and enum ocelot_stat
which allows us to index the ocelot->stats array (OCELOT_STAT_RX_OCTETS
etc), and don't require us to specify both when we define what stats
each switch family has.

Create an OCELOT_STAT() macro that pairs only an enum ocelot_stat to an
enum ocelot_reg, and an OCELOT_STAT_ETHTOOL() macro which also contains
a name exported to the unstructured ethtool -S stringset API. For now,
we define all counters as having the OCELOT_STAT_ETHTOOL() kind, but we
will add more counters in the future which are not exported to the
unstructured ethtool -S.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: mscc: ocelot: harmonize names of SYS_COUNT_TX_AGING and OCELOT_STAT_TX_AGED
Vladimir Oltean [Thu, 8 Sep 2022 16:48:14 +0000 (19:48 +0300)]
net: mscc: ocelot: harmonize names of SYS_COUNT_TX_AGING and OCELOT_STAT_TX_AGED

The hardware counter is called C_TX_AGED, so rename SYS_COUNT_TX_AGING
to SYS_COUNT_TX_AGED. This will become important since we want to
minimize the way in which we declare struct ocelot_stat_layout elements,
using the C preprocessor.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: mscc: ocelot: add support for all sorts of standardized counters present in DSA
Vladimir Oltean [Thu, 8 Sep 2022 16:48:13 +0000 (19:48 +0300)]
net: mscc: ocelot: add support for all sorts of standardized counters present in DSA

DSA is integrated with the new standardized ethtool -S --groups option,
but the felix driver only exports unstructured statistics.

Reuse the array of 64-bit statistics collected by ocelot_check_stats_work(),
but just export select values from it.

Since ocelot_check_stats_work() runs periodically to avoid 32-bit
overflow, and the ethtool calling context is sleepable, we update the
64-bit stats one more time, to provide up-to-date values. The locking
scheme with a mutex followed by a spinlock is a bit hard to digest, so
we create and use a ocelot_port_stats_run() helper with a callback that
populates the ethool stats group the caller is interested in.

The exported stats are:
ethtool -S swp0 --groups eth-phy
ethtool -S swp0 --groups eth-mac
ethtool -S swp0 --groups eth-ctrl
ethtool -S swp0 --groups rmon
ethtool --include-statistics --show-pause swp0

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: mscc: ocelot: exclude stats from bulk regions based on reg, not name
Vladimir Oltean [Thu, 8 Sep 2022 16:48:12 +0000 (19:48 +0300)]
net: mscc: ocelot: exclude stats from bulk regions based on reg, not name

We want to introduce elements kept in ocelot->stats that aren't exposed
to the unstructured ethtool -S (so they won't have their name populated),
but are otherwise checked for 32-bit wraparounds by
ocelot_port_update_stats().

This isn't possible today because ocelot_prepare_stats_regions() skips
over ocelot_stat_layout elements with no name. Now that we've changed
struct ocelot_stat_layout to keep the absolute register address rather
than the offset relative to SYS_CNT, we can make use of the unpopulated
"reg" value of 0 to mean that the counter isn't present on the current
switch revision, and skip it from the preparation of bulk regions.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: dsa: felix: use ocelot's ndo_get_stats64 method
Vladimir Oltean [Thu, 8 Sep 2022 16:48:11 +0000 (19:48 +0300)]
net: dsa: felix: use ocelot's ndo_get_stats64 method

Move the logic from the ocelot switchdev driver's ocelot_get_stats64()
method to the common switch lib and reuse it for the DSA driver.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: mscc: ocelot: move more PTP code from the lib to ocelot_ptp.c
Vladimir Oltean [Thu, 8 Sep 2022 16:48:10 +0000 (19:48 +0300)]
net: mscc: ocelot: move more PTP code from the lib to ocelot_ptp.c

Decongest ocelot.c a bit more by moving all PTP related logic (including
timestamp processing and PTP packet traps) to ocelot_ptp.c.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: mscc: ocelot: unexport ocelot_port_fdb_do_dump from the common lib
Vladimir Oltean [Thu, 8 Sep 2022 16:48:09 +0000 (19:48 +0300)]
net: mscc: ocelot: unexport ocelot_port_fdb_do_dump from the common lib

ocelot_port_fdb_do_dump() is only used by ocelot_net.c, so move it
there.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: mscc: ocelot: move stats code to ocelot_stats.c
Vladimir Oltean [Thu, 8 Sep 2022 16:48:08 +0000 (19:48 +0300)]
net: mscc: ocelot: move stats code to ocelot_stats.c

The main C file of the ocelot switch lib, ocelot.c, is getting larger
and larger, and there are plans to add more logic related to stats.
So it seems like an appropriate moment to split the statistics code to a
new file.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: mscc: ocelot: sort Makefile files alphabetically
Vladimir Oltean [Thu, 8 Sep 2022 16:48:07 +0000 (19:48 +0300)]
net: mscc: ocelot: sort Makefile files alphabetically

Create a clear ordering of the files used to compile the switch lib and
the switchdev driver.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: mscc: ocelot: report FIFO drop counters through stats->rx_dropped
Vladimir Oltean [Thu, 8 Sep 2022 16:48:06 +0000 (19:48 +0300)]
net: mscc: ocelot: report FIFO drop counters through stats->rx_dropped

if_link.h says:

 * @rx_dropped: Number of packets received but not processed,
 *   e.g. due to lack of resources or unsupported protocol.
 *   For hardware interfaces this counter may include packets discarded
 *   due to L2 address filtering but should not include packets dropped
 *   by the device due to buffer exhaustion which are counted separately in
 *   @rx_missed_errors (since procfs folds those two counters together).

Currently we report "stats->rx_dropped = dev->stats.rx_dropped", the
latter being incremented by various entities in the stack. This is not
wrong, but we'd like to move ocelot_get_stats64() in the common ocelot
switch lib which is independent of struct net_device.

To do that, report the hardware RX drop counters instead. These drops
are due to policer action, or due to no destinations. When we have no
memory in the queue system, report this through rx_missed_errors, as
instructed.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: dsa: felix: check the 32-bit PSFP stats against overflow
Vladimir Oltean [Thu, 8 Sep 2022 16:48:05 +0000 (19:48 +0300)]
net: dsa: felix: check the 32-bit PSFP stats against overflow

The Felix PSFP counters suffer from the same problem as the ocelot
ndo_get_stats64 ones - they are 32-bit, so they can easily overflow and
this can easily go undetected.

Add a custom hook in ocelot_check_stats_work() through which driver
specific actions can be taken, and update the stats for the existing
PSFP filters from that hook.

Previously, vsc9959_psfp_filter_add() and vsc9959_psfp_filter_del() were
serialized with respect to each other via rtnl_lock(). However, with the
new entry point into &psfp->sfi_list coming from the periodic worker, we
now need an explicit mutex to serialize access to these lists.

We used to keep a struct felix_stream_filter_counters on stack, through
which vsc9959_psfp_stats_get() - a FLOW_CLS_STATS callback - would
retrieve data from vsc9959_psfp_counters_get(). We need to become
smarter about that in 3 ways:

- we need to keep a persistent set of counters for each stream instead
  of keeping them on stack

- we need to promote those counters from u32 to u64, and create a
  procedure that properly keeps 64-bit counters. Since we clear the
  hardware counters anyway, and we poll every 2 seconds, a simple
  increment of a u64 counter with a u32 value will perfectly do the job.

- FLOW_CLS_STATS also expect incremental counters, so we also need to
  zeroize our u64 counters every time sch_flower calls us

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: mscc: ocelot: make access to STAT_VIEW sleepable again
Vladimir Oltean [Thu, 8 Sep 2022 16:48:04 +0000 (19:48 +0300)]
net: mscc: ocelot: make access to STAT_VIEW sleepable again

To support SPI-controlled switches in the future, access to
SYS_STAT_CFG_STAT_VIEW needs to be done outside of any spinlock
protected region, but it still needs to be serialized (by a mutex).

Split the ocelot->stats_lock spinlock into a mutex that serializes
indirect access to hardware registers (ocelot->stat_view_lock) and a
spinlock that serializes access to the u64 ocelot->stats array.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: dsa: felix: add definitions for the stream filter counters
Vladimir Oltean [Thu, 8 Sep 2022 16:48:03 +0000 (19:48 +0300)]
net: dsa: felix: add definitions for the stream filter counters

TSN stream (802.1Qci, 802.1CB) filters are also accessed through
STAT_VIEW, just like the port registers, but these counters are per
stream, rather than per port. So we don't keep them in
ocelot_port_update_stats().

What we can do, however, is we can create register definitions for them
just like we have for the port counters, and delete the last remaining
user of the SYS_CNT register + a group index (read_gix).

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agoMerge branch 'tc_action_ops-refactor'
David S. Miller [Fri, 9 Sep 2022 07:24:43 +0000 (08:24 +0100)]
Merge branch 'tc_action_ops-refactor'

Zhengchao Shao says:

====================
net: refactor the walk and lookup hook functions in tc_action_ops

The implementation logic of the walk/lookup hook function in each action
module is the same. Therefore, the two functions can be reconstructed.
When registering tc_action_ops of each action module, the corresponding
net_id is saved to tc_action_ops. In this way, the net_id of the
corresponding module can be directly obtained in act_api without executing
the specific walk and lookup hook functions. Then, generic functions can
be added to replace the walk and lookup hook functions of each action
module. Last, modify each action module in alphabetical order.

Reserve the walk and lookup interfaces and delete them when they are no
longer used.

This patchset has been tested by using TDC, and I will add selftest in
other patchset.

Last, thanks to Jamal Hadi Salim and Cong Wang for their advice.
---
v3: remove hole from the structure tc_action_ops
v2: save the net_id of each TC action module to the tc_action_ops structure
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_vlan: get rid of tcf_vlan_walker and tcf_vlan_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:54 +0000 (12:14 +0800)]
net: sched: act_vlan: get rid of tcf_vlan_walker and tcf_vlan_search

tcf_vlan_walker() and tcf_vlan_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_tunnel_key: get rid of tunnel_key_walker and tunnel_key_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:53 +0000 (12:14 +0800)]
net: sched: act_tunnel_key: get rid of tunnel_key_walker and tunnel_key_search

tunnel_key_walker() and tunnel_key_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_skbmod: get rid of tcf_skbmod_walker and tcf_skbmod_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:52 +0000 (12:14 +0800)]
net: sched: act_skbmod: get rid of tcf_skbmod_walker and tcf_skbmod_search

tcf_skbmod_walker() and tcf_skbmod_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_skbedit: get rid of tcf_skbedit_walker and tcf_skbedit_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:51 +0000 (12:14 +0800)]
net: sched: act_skbedit: get rid of tcf_skbedit_walker and tcf_skbedit_search

tcf_skbedit_walker() and tcf_skbedit_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_simple: get rid of tcf_simp_walker and tcf_simp_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:50 +0000 (12:14 +0800)]
net: sched: act_simple: get rid of tcf_simp_walker and tcf_simp_search

tcf_simp_walker() and tcf_simp_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_sample: get rid of tcf_sample_walker and tcf_sample_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:49 +0000 (12:14 +0800)]
net: sched: act_sample: get rid of tcf_sample_walker and tcf_sample_search

tcf_sample_walker() and tcf_sample_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_police: get rid of tcf_police_walker and tcf_police_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:48 +0000 (12:14 +0800)]
net: sched: act_police: get rid of tcf_police_walker and tcf_police_search

tcf_police_walker() and tcf_police_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_pedit: get rid of tcf_pedit_walker and tcf_pedit_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:47 +0000 (12:14 +0800)]
net: sched: act_pedit: get rid of tcf_pedit_walker and tcf_pedit_search

tcf_pedit_walker() and tcf_pedit_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_nat: get rid of tcf_nat_walker and tcf_nat_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:46 +0000 (12:14 +0800)]
net: sched: act_nat: get rid of tcf_nat_walker and tcf_nat_search

tcf_nat_walker() and tcf_nat_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_mpls: get rid of tcf_mpls_walker and tcf_mpls_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:45 +0000 (12:14 +0800)]
net: sched: act_mpls: get rid of tcf_mpls_walker and tcf_mpls_search

tcf_mpls_walker() and tcf_mpls_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_mirred: get rid of tcf_mirred_walker and tcf_mirred_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:44 +0000 (12:14 +0800)]
net: sched: act_mirred: get rid of tcf_mirred_walker and tcf_mirred_search

tcf_mirred_walker() and tcf_mirred_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_ipt: get rid of tcf_ipt_walker/tcf_xt_walker and tcf_ipt_search/tcf_x...
Zhengchao Shao [Thu, 8 Sep 2022 04:14:43 +0000 (12:14 +0800)]
net: sched: act_ipt: get rid of tcf_ipt_walker/tcf_xt_walker and tcf_ipt_search/tcf_xt_search

tcf_ipt_walker()/tcf_xt_walker() and tcf_ipt_search()/tcf_xt_search() do
the same thing as generic walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_ife: get rid of tcf_ife_walker and tcf_ife_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:42 +0000 (12:14 +0800)]
net: sched: act_ife: get rid of tcf_ife_walker and tcf_ife_search

tcf_ife_walker() and tcf_ife_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_gate: get rid of tcf_gate_walker and tcf_gate_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:41 +0000 (12:14 +0800)]
net: sched: act_gate: get rid of tcf_gate_walker and tcf_gate_search

tcf_gate_walker() and tcf_gate_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_gact: get rid of tcf_gact_walker and tcf_gact_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:40 +0000 (12:14 +0800)]
net: sched: act_gact: get rid of tcf_gact_walker and tcf_gact_search

tcf_gact_walker() and tcf_gact_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_ctinfo: get rid of tcf_ctinfo_walker and tcf_ctinfo_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:39 +0000 (12:14 +0800)]
net: sched: act_ctinfo: get rid of tcf_ctinfo_walker and tcf_ctinfo_search

tcf_ctinfo_walker() and tcf_ctinfo_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_ct: get rid of tcf_ct_walker and tcf_ct_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:38 +0000 (12:14 +0800)]
net: sched: act_ct: get rid of tcf_ct_walker and tcf_ct_search

tcf_ct_walker() and tcf_ct_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_csum: get rid of tcf_csum_walker and tcf_csum_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:37 +0000 (12:14 +0800)]
net: sched: act_csum: get rid of tcf_csum_walker and tcf_csum_search

tcf_csum_walker() and tcf_csum_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_connmark: get rid of tcf_connmark_walker and tcf_connmark_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:36 +0000 (12:14 +0800)]
net: sched: act_connmark: get rid of tcf_connmark_walker and tcf_connmark_search

tcf_connmark_walker() and tcf_connmark_search() do the same thing as
generic walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_bpf: get rid of tcf_bpf_walker and tcf_bpf_search
Zhengchao Shao [Thu, 8 Sep 2022 04:14:35 +0000 (12:14 +0800)]
net: sched: act_bpf: get rid of tcf_bpf_walker and tcf_bpf_search

tcf_bpf_walker() and tcf_bpf_search() do the same thing as generic
walk/search function, so remove them.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act_api: implement generic walker and search for tc action
Zhengchao Shao [Thu, 8 Sep 2022 04:14:34 +0000 (12:14 +0800)]
net: sched: act_api: implement generic walker and search for tc action

Being able to get tc_action_net by using net_id stored in tc_action_ops
and execute the generic walk/search function, add __tcf_generic_walker()
and __tcf_idr_search() helpers.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: sched: act: move global static variable net_id to tc_action_ops
Zhengchao Shao [Thu, 8 Sep 2022 04:14:33 +0000 (12:14 +0800)]
net: sched: act: move global static variable net_id to tc_action_ops

Each tc action module has a corresponding net_id, so put net_id directly
into the structure tc_action_ops.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agoMerge branch 'lan743x-next'
David S. Miller [Fri, 9 Sep 2022 07:20:58 +0000 (08:20 +0100)]
Merge branch 'lan743x-next'

Raju Lakkaraju says:

====================
net: lan743x: Fix to use multiqueue start/stop APIs

This patch series address the fix to use multiqueue start/stop APIs and Add Rx IP and TCP checksum offload

Changes:
========
V1 -> V2:
 - Fix the sparse warnings

V0 -> V1:
 - Remove chip SKU check conditionals
 - Update the changes description
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: lan743x: Add support for Rx IP & TCP checksum offload
Raju Lakkaraju [Thu, 8 Sep 2022 08:28:34 +0000 (13:58 +0530)]
net: lan743x: Add support for Rx IP & TCP checksum offload

Add Rx IP and TCP checksum offload

Signed-off-by: Raju Lakkaraju <Raju.Lakkaraju@microchip.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: lan743x: Fix to use multiqueue start/stop APIs
Raju Lakkaraju [Thu, 8 Sep 2022 08:28:33 +0000 (13:58 +0530)]
net: lan743x: Fix to use multiqueue start/stop APIs

- Fix to use multiqueue start/stop APIs
 - Change to return NETDEV_TX_BUSY instead of holding the TX skb when busy
 - Increase Tx ring size to 128 to address performance issues in some platforms
 - Use NAPI_POLL_WEIGHT for Tx Napi handler instead of ring dependent value
 - Use multiqueue to register 4 Rx channels

Signed-off-by: Raju Lakkaraju <Raju.Lakkaraju@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf-next
David S. Miller [Fri, 9 Sep 2022 07:08:51 +0000 (08:08 +0100)]
Merge git://git./linux/kernel/git/netfilter/nf-next

Florian Westphal says:

====================
The following set contains changes for your *net-next* tree:

- make conntrack ignore packets that are delayed (containing
  data already acked).  The current behaviour to flag them as INVALID
  causes more harm than good, let them pass so peer can send an
  immediate ACK for the most recent sequence number.
- make conntrack recognize when both peers have sent 'invalid' FINs:
  This helps cleaning out stale connections faster for those cases where
  conntrack is no longer in sync with the actual connection state.
- Now that DECNET is gone, we don't need to reserve space for DECNET
  related information.
- compact common 'find a free port number for the new inbound
  connection' code and move it to a helper, then cap number of tries
  the new helper will make until it gives up.
- replace various instances of strlcpy with strscpy, from Wolfram Sang.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Paolo Abeni [Thu, 8 Sep 2022 16:34:54 +0000 (18:34 +0200)]
Merge git://git./linux/kernel/git/netdev/net

drivers/net/ethernet/freescale/fec.h
  7d650df99d52 ("net: fec: add pm_qos support on imx6q platform")
  40c79ce13b03 ("net: fec: add stop mode support for imx8 platform")

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agonet: sparx5: fix function return type to match actual type
Casper Andersson [Tue, 6 Sep 2022 06:58:15 +0000 (08:58 +0200)]
net: sparx5: fix function return type to match actual type

Function returns error integer, not bool.

Does not have any impact on functionality.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Casper Andersson <casper.casan@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20220906065815.3856323-1-casper.casan@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agoMerge tag 'net-6.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 8 Sep 2022 12:15:01 +0000 (08:15 -0400)]
Merge tag 'net-6.0-rc5' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Paolo Abeni:
 "Including fixes from rxrpc, netfilter, wireless and bluetooth
  subtrees.

  Current release - regressions:

   - skb: export skb drop reaons to user by TRACE_DEFINE_ENUM

   - bluetooth: fix regression preventing ACL packet transmission

  Current release - new code bugs:

   - dsa: microchip: fix kernel oops on ksz8 switches

   - dsa: qca8k: fix NULL pointer dereference for
     of_device_get_match_data

  Previous releases - regressions:

   - netfilter: clean up hook list when offload flags check fails

   - wifi: mt76: fix crash in chip reset fail

   - rxrpc: fix ICMP/ICMP6 error handling

   - ice: fix DMA mappings leak

   - i40e: fix kernel crash during module removal

  Previous releases - always broken:

   - ipv6: sr: fix out-of-bounds read when setting HMAC data.

   - tcp: TX zerocopy should not sense pfmemalloc status

   - sch_sfb: don't assume the skb is still around after
     enqueueing to child

   - netfilter: drop dst references before setting

   - wifi: wilc1000: fix DMA on stack objects

   - rxrpc: fix an insufficiently large sglist in
     rxkad_verify_packet_2()

   - fec: use a spinlock to guard `fep->ptp_clk_on`

  Misc:

   - usb: qmi_wwan: add Quectel RM520N"

* tag 'net-6.0-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (50 commits)
  sch_sfb: Also store skb len before calling child enqueue
  net: phy: lan87xx: change interrupt src of link_up to comm_ready
  net/smc: Fix possible access to freed memory in link clear
  net: ethernet: mtk_eth_soc: check max allowed hash in mtk_ppe_check_skb
  net: skb: export skb drop reaons to user by TRACE_DEFINE_ENUM
  net: ethernet: mtk_eth_soc: fix typo in __mtk_foe_entry_clear
  net: dsa: felix: access QSYS_TAG_CONFIG under tas_lock in vsc9959_sched_speed_set
  net: dsa: felix: disable cut-through forwarding for frames oversized for tc-taprio
  net: dsa: felix: tc-taprio intervals smaller than MTU should send at least one packet
  net: usb: qmi_wwan: add Quectel RM520N
  net: dsa: qca8k: fix NULL pointer dereference for of_device_get_match_data
  tcp: fix early ETIMEDOUT after spurious non-SACK RTO
  stmmac: intel: Simplify intel_eth_pci_remove()
  net: mvpp2: debugfs: fix memory leak when using debugfs_lookup()
  ipv6: sr: fix out-of-bounds read when setting HMAC data.
  bonding: accept unsolicited NA message
  bonding: add all node mcast address when slave up
  bonding: use unspecified address if no available link local address
  wifi: use struct_group to copy addresses
  wifi: mac80211_hwsim: check length for virtio packets
  ...

22 months agofs: only do a memory barrier for the first set_buffer_uptodate()
Linus Torvalds [Wed, 31 Aug 2022 16:46:12 +0000 (09:46 -0700)]
fs: only do a memory barrier for the first set_buffer_uptodate()

Commit d4252071b97d ("add barriers to buffer_uptodate and
set_buffer_uptodate") added proper memory barriers to the buffer head
BH_Uptodate bit, so that anybody who tests a buffer for being up-to-date
will be guaranteed to actually see initialized state.

However, that commit didn't _just_ add the memory barrier, it also ended
up dropping the "was it already set" logic that the BUFFER_FNS() macro
had.

That's conceptually the right thing for a generic "this is a memory
barrier" operation, but in the case of the buffer contents, we really
only care about the memory barrier for the _first_ time we set the bit,
in that the only memory ordering protection we need is to avoid anybody
seeing uninitialized memory contents.

Any other access ordering wouldn't be about the BH_Uptodate bit anyway,
and would require some other proper lock (typically BH_Lock or the folio
lock).  A reader that races with somebody invalidating the buffer head
isn't an issue wrt the memory ordering, it's a serialization issue.

Now, you'd think that the buffer head operations don't matter in this
day and age (and I certainly thought so), but apparently some loads
still end up being heavy users of buffer heads.  In particular, the
kernel test robot reported that not having this bit access optimization
in place caused a noticeable direct IO performance regression on ext4:

  fxmark.ssd_ext4_no_jnl_DWTL_54_directio.works/sec -26.5% regression

although you presumably need a fast disk and a lot of cores to actually
notice.

Link: https://lore.kernel.org/all/Yw8L7HTZ%2FdE2%2Fo9C@xsang-OptiPlex-9020/
Reported-by: kernel test robot <oliver.sang@intel.com>
Tested-by: Fengwei Yin <fengwei.yin@intel.com>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
22 months agoMerge tag 'efi-urgent-for-v6.0-1' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 8 Sep 2022 11:37:38 +0000 (07:37 -0400)]
Merge tag 'efi-urgent-for-v6.0-1' of git://git./linux/kernel/git/efi/efi

Pull EFI fixes from Ard Biesheuvel:
 "A couple of low-priority EFI fixes:

   - prevent the randstruct plugin from re-ordering EFI protocol
     definitions

   - fix a use-after-free in the capsule loader

   - drop unused variable"

* tag 'efi-urgent-for-v6.0-1' of git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi:
  efi: capsule-loader: Fix use-after-free in efi_capsule_write
  efi/x86: libstub: remove unused variable
  efi: libstub: Disable struct randomization

22 months agor8169: merge support for chip versions 10, 13, 16
Heiner Kallweit [Mon, 5 Sep 2022 19:23:12 +0000 (21:23 +0200)]
r8169: merge support for chip versions 10, 13, 16

These chip versions are closely related and all of them have no
chip-specific MAC/PHY initialization. Therefore merge support
for the three chip versions.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/469d27e0-1d06-9b15-6c96-6098b3a52e35@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agosch_sfb: Also store skb len before calling child enqueue
Toke Høiland-Jørgensen [Mon, 5 Sep 2022 19:21:36 +0000 (21:21 +0200)]
sch_sfb: Also store skb len before calling child enqueue

Cong Wang noticed that the previous fix for sch_sfb accessing the queued
skb after enqueueing it to a child qdisc was incomplete: the SFB enqueue
function was also calling qdisc_qstats_backlog_inc() after enqueue, which
reads the pkt len from the skb cb field. Fix this by also storing the skb
len, and using the stored value to increment the backlog after enqueueing.

Fixes: 9efd23297cca ("sch_sfb: Don't assume the skb is still around after enqueueing to child")
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
Acked-by: Cong Wang <cong.wang@bytedance.com>
Link: https://lore.kernel.org/r/20220905192137.965549-1-toke@toke.dk
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agonet: phy: lan87xx: change interrupt src of link_up to comm_ready
Arun Ramadoss [Mon, 5 Sep 2022 15:27:50 +0000 (20:57 +0530)]
net: phy: lan87xx: change interrupt src of link_up to comm_ready

Currently phy link up/down interrupt is enabled using the
LAN87xx_INTERRUPT_MASK register. In the lan87xx_read_status function,
phy link is determined using the T1_MODE_STAT_REG register comm_ready bit.
comm_ready bit is set using the loc_rcvr_status & rem_rcvr_status.
Whenever the phy link is up, LAN87xx_INTERRUPT_SOURCE link_up bit is set
first but comm_ready bit takes some time to set based on local and
remote receiver status.
As per the current implementation, interrupt is triggered using link_up
but the comm_ready bit is still cleared in the read_status function. So,
link is always down.  Initially tested with the shared interrupt
mechanism with switch and internal phy which is working, but after
implementing interrupt controller it is not working.
It can fixed either by updating the read_status function to read from
LAN87XX_INTERRUPT_SOURCE register or enable the interrupt mask for
comm_ready bit. But the validation team recommends the use of comm_ready
for link detection.
This patch fixes by enabling the comm_ready bit for link_up in the
LAN87XX_INTERRUPT_MASK_2 register (MISC Bank) and link_down in
LAN87xx_INTERRUPT_MASK register.

Fixes: 8a1b415d70b7 ("net: phy: added ethtool master-slave configuration support")
Signed-off-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20220905152750.5079-1-arun.ramadoss@microchip.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agonet: stmmac: Disable automatic FCS/Pad stripping
Kurt Kanzenbach [Mon, 5 Sep 2022 13:01:55 +0000 (15:01 +0200)]
net: stmmac: Disable automatic FCS/Pad stripping

The stmmac has the possibility to automatically strip the padding/FCS for IEEE
802.3 type frames. This feature is enabled conditionally. Therefore, the stmmac
receive path has to have a determination logic whether the FCS has to be
stripped in software or not.

In fact, for DSA this ACS feature is disabled and the determination logic
doesn't check for it properly. For instance, when using DSA in combination with
an older stmmac (pre version 4), the FCS is not stripped by hardware or software
which is problematic.

So either add another check for DSA to the fast path or simply disable ACS
feature completely. The latter approach has been chosen, because most of the
time the FCS is stripped in software anyway and it removes conditionals from the
receive fast path.

Signed-off-by: Kurt Kanzenbach <kurt@linutronix.de>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Link: https://lore.kernel.org/r/87v8q8jjgh.fsf@kurt/
Link: https://lore.kernel.org/r/20220905130155.193640-1-kurt@linutronix.de
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
22 months agoefi: capsule-loader: Fix use-after-free in efi_capsule_write
Hyunwoo Kim [Wed, 7 Sep 2022 16:07:14 +0000 (09:07 -0700)]
efi: capsule-loader: Fix use-after-free in efi_capsule_write

A race condition may occur if the user calls close() on another thread
during a write() operation on the device node of the efi capsule.

This is a race condition that occurs between the efi_capsule_write() and
efi_capsule_flush() functions of efi_capsule_fops, which ultimately
results in UAF.

So, the page freeing process is modified to be done in
efi_capsule_release() instead of efi_capsule_flush().

Cc: <stable@vger.kernel.org> # v4.9+
Signed-off-by: Hyunwoo Kim <imv4bel@gmail.com>
Link: https://lore.kernel.org/all/20220907102920.GA88602@ubuntu/
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
22 months agoMerge branch 'hns3-new-features'
David S. Miller [Wed, 7 Sep 2022 15:20:15 +0000 (16:20 +0100)]
Merge branch 'hns3-new-features'

Guangbin Huang says:

====================
hns3: add some new features

This series adds some new features for the HNS3 ethernet driver.

Patches #1~#3 support configuring dscp map to tc.

Patch 4# supports querying FEC statistics by command "ethtool -I --show-fec eth0".

Patch 5# supports querying and setting Serdes lane number.

Change logs:
V1 -> V2:
 - fix build error of patch 1# reported by robot lkp@intel.com.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: hns3: add support to query and set lane number by ethtool
Hao Chen [Tue, 6 Sep 2022 09:12:23 +0000 (17:12 +0800)]
net: hns3: add support to query and set lane number by ethtool

When serdes lane support setting 25Gb/s or 50Gb/s speed and user wants to
set port speed as 50Gb/s, it can be setted as one 50Gb/s serdes lane or
two 25Gb/s serdes lanes.

So, this patch adds support to query and set lane number by ethtool
to satisfy this scenario.

Signed-off-by: Hao Chen <chenhao418@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: hns3: add querying fec statistics
Hao Lan [Tue, 6 Sep 2022 09:12:22 +0000 (17:12 +0800)]
net: hns3: add querying fec statistics

FEC statistics can be used to check the transmission quality of links.
This patch implements the get_fec_stats callback of ethtool_ops to support
querying FEC statistics by command "ethtool -I --show-fec eth0".

Signed-off-by: Hao Lan <lanhao@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: hns3: debugfs add dump dscp map info
Guangbin Huang [Tue, 6 Sep 2022 09:12:21 +0000 (17:12 +0800)]
net: hns3: debugfs add dump dscp map info

This patch add dump the map relation for dscp, priority and TC, and
the current tc map mode.

Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: hns3: support ndo_select_queue()
Guangbin Huang [Tue, 6 Sep 2022 09:12:20 +0000 (17:12 +0800)]
net: hns3: support ndo_select_queue()

To support tx packets to select queue according to its dscp field after
setting dscp and tc map relationship, this patch implements
ndo_select_queue() to set skb->priority according to the user's setting
dscp and priority map relationship.

Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: hns3: add support config dscp map to tc
Guangbin Huang [Tue, 6 Sep 2022 09:12:19 +0000 (17:12 +0800)]
net: hns3: add support config dscp map to tc

This patch add support config dscp map to tc by implementing ieee_setapp
and ieee_delapp of struct dcbnl_rtnl_ops. Driver will convert mapping
relationship from dscp-prio to dscp-tc.

Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/smc: Fix possible access to freed memory in link clear
Yacan Liu [Tue, 6 Sep 2022 13:01:39 +0000 (21:01 +0800)]
net/smc: Fix possible access to freed memory in link clear

After modifying the QP to the Error state, all RX WR would be completed
with WC in IB_WC_WR_FLUSH_ERR status. Current implementation does not
wait for it is done, but destroy the QP and free the link group directly.
So there is a risk that accessing the freed memory in tasklet context.

Here is a crash example:

 BUG: unable to handle page fault for address: ffffffff8f220860
 #PF: supervisor write access in kernel mode
 #PF: error_code(0x0002) - not-present page
 PGD f7300e067 P4D f7300e067 PUD f7300f063 PMD 8c4e45063 PTE 800ffff08c9df060
 Oops: 0002 [#1] SMP PTI
 CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: G S         OE     5.10.0-0607+ #23
 Hardware name: Inspur NF5280M4/YZMB-00689-101, BIOS 4.1.20 07/09/2018
 RIP: 0010:native_queued_spin_lock_slowpath+0x176/0x1b0
 Code: f3 90 48 8b 32 48 85 f6 74 f6 eb d5 c1 ee 12 83 e0 03 83 ee 01 48 c1 e0 05 48 63 f6 48 05 00 c8 02 00 48 03 04 f5 00 09 98 8e <48> 89 10 8b 42 08 85 c0 75 09 f3 90 8b 42 08 85 c0 74 f7 48 8b 32
 RSP: 0018:ffffb3b6c001ebd8 EFLAGS: 00010086
 RAX: ffffffff8f220860 RBX: 0000000000000246 RCX: 0000000000080000
 RDX: ffff91db1f86c800 RSI: 000000000000173c RDI: ffff91db62bace00
 RBP: ffff91db62bacc00 R08: 0000000000000000 R09: c00000010000028b
 R10: 0000000000055198 R11: ffffb3b6c001ea58 R12: ffff91db80e05010
 R13: 000000000000000a R14: 0000000000000006 R15: 0000000000000040
 FS:  0000000000000000(0000) GS:ffff91db1f840000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: ffffffff8f220860 CR3: 00000001f9580004 CR4: 00000000003706e0
 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
 Call Trace:
  <IRQ>
  _raw_spin_lock_irqsave+0x30/0x40
  mlx5_ib_poll_cq+0x4c/0xc50 [mlx5_ib]
  smc_wr_rx_tasklet_fn+0x56/0xa0 [smc]
  tasklet_action_common.isra.21+0x66/0x100
  __do_softirq+0xd5/0x29c
  asm_call_irq_on_stack+0x12/0x20
  </IRQ>
  do_softirq_own_stack+0x37/0x40
  irq_exit_rcu+0x9d/0xa0
  sysvec_call_function_single+0x34/0x80
  asm_sysvec_call_function_single+0x12/0x20

Fixes: bd4ad57718cc ("smc: initialize IB transport incl. PD, MR, QP, CQ, event, WR")
Signed-off-by: Yacan Liu <liuyacan@corp.netease.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agoMerge branch '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next...
David S. Miller [Wed, 7 Sep 2022 14:57:48 +0000 (15:57 +0100)]
Merge branch '40GbE' of git://git./linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2022-09-06 (i40e, iavf)

This series contains updates to i40e and iavf drivers.

Stanislaw adds support for new device id for i40e.

Jaroslaw tidies up some code around MSI-X configuration by adding/
reworking comments and introducing a couple of macros for i40e.

Michal resolves some races around reset and close by deferring and deleting
some pending AdminQ operations and reworking filter additions and deletions
during these operations for iavf.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next...
David S. Miller [Wed, 7 Sep 2022 14:56:21 +0000 (15:56 +0100)]
Merge branch '100GbE' of git://git./linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2022-09-06 (ice)

This series contains updates to ice driver only.

Tony reduces device MSI-X request/usage when entire request can't be fulfilled.

Michal adds check for reset when waiting for PTP offsets.

Paul refactors firmware version checks to use a common helper.

Christophe Jaillet changes a couple of local memory allocation to not
use the devm variant.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonetfilter: nat: avoid long-running port range loop
Florian Westphal [Tue, 6 Sep 2022 15:20:36 +0000 (17:20 +0200)]
netfilter: nat: avoid long-running port range loop

Looping a large port range takes too long. Instead select a random
offset within [ntohs(exp->saved_proto.tcp.port), 65535] and try 128
ports.

This is a rehash of an erlier patch to do the same, but generalized
to handle other helpers as well.

Link: https://patchwork.ozlabs.org/project/netfilter-devel/patch/20210920204439.13179-2-Cole.Dishington@alliedtelesis.co.nz/
Signed-off-by: Florian Westphal <fw@strlen.de>
22 months agonetfilter: nat: move repetitive nat port reserve loop to a helper
Florian Westphal [Tue, 6 Sep 2022 15:20:35 +0000 (17:20 +0200)]
netfilter: nat: move repetitive nat port reserve loop to a helper

Almost all nat helpers reserve an expecation port the same way:
Try the port inidcated by the peer, then move to next port if that
port is already in use.

We can squash this into a helper.

Suggested-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
22 months agonetfilter: move from strlcpy with unused retval to strscpy
Wolfram Sang [Thu, 18 Aug 2022 21:02:24 +0000 (23:02 +0200)]
netfilter: move from strlcpy with unused retval to strscpy

Follow the advice of the below link and prefer 'strscpy' in this
subsystem. Conversion is 1:1 because the return value is not used.
Generated by a coccinelle script.

Link: https://lore.kernel.org/r/CAHk-=wgfRnXz0W3D37d01q3JFkr_i_uTL=V6A6G1oUZcprmknw@mail.gmail.com/
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Reviewed-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Florian Westphal <fw@strlen.de>
22 months agonetfilter: remove NFPROTO_DECNET
Florian Westphal [Mon, 22 Aug 2022 14:41:21 +0000 (16:41 +0200)]
netfilter: remove NFPROTO_DECNET

Decnet has been removed. so no need to reserve space in arrays for it.

Signed-off-by: Florian Westphal <fw@strlen.de>
22 months agonetfilter: conntrack: reduce timeout when receiving out-of-window fin or rst
Florian Westphal [Fri, 26 Aug 2022 13:32:27 +0000 (15:32 +0200)]
netfilter: conntrack: reduce timeout when receiving out-of-window fin or rst

In case the endpoints and conntrack go out-of-sync, i.e. there is
disagreement wrt. validy of sequence/ack numbers between conntracks
internal state and those of the endpoints, connections can hang for a
long time (until ESTABLISHED timeout).

This adds a check to detect a fin/fin exchange even if those are
invalid.  The timeout is then lowered to UNACKED (default 300s).

Signed-off-by: Florian Westphal <fw@strlen.de>
22 months agonetfilter: conntrack: remove unneeded indent level
Florian Westphal [Fri, 26 Aug 2022 13:32:26 +0000 (15:32 +0200)]
netfilter: conntrack: remove unneeded indent level

After previous patch, the conditional branch is obsolete, reformat it.
gcc generates same code as before this change.

Signed-off-by: Florian Westphal <fw@strlen.de>
22 months agonet: sysctl: remove unused variable long_max
Liu Shixin [Mon, 5 Sep 2022 12:50:42 +0000 (20:50 +0800)]
net: sysctl: remove unused variable long_max

The variable long_max is replaced by bpf_jit_limit_max and no longer be
used. So remove it.

No functional change.

Signed-off-by: Liu Shixin <liushixin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: ethernet: mtk_eth_soc: remove mtk_foe_entry_timestamp
Lorenzo Bianconi [Mon, 5 Sep 2022 12:46:01 +0000 (14:46 +0200)]
net: ethernet: mtk_eth_soc: remove mtk_foe_entry_timestamp

Get rid of mtk_foe_entry_timestamp routine since it is no longer used.

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: ethernet: mtk_eth_soc: check max allowed hash in mtk_ppe_check_skb
Lorenzo Bianconi [Mon, 5 Sep 2022 12:41:28 +0000 (14:41 +0200)]
net: ethernet: mtk_eth_soc: check max allowed hash in mtk_ppe_check_skb

Even if max hash configured in hw in mtk_ppe_hash_entry is
MTK_PPE_ENTRIES - 1, check theoretical OOB accesses in
mtk_ppe_check_skb routine

Fixes: c4f033d9e03e9 ("net: ethernet: mtk_eth_soc: rework hardware flow table management")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: skb: export skb drop reaons to user by TRACE_DEFINE_ENUM
Menglong Dong [Mon, 5 Sep 2022 03:50:15 +0000 (11:50 +0800)]
net: skb: export skb drop reaons to user by TRACE_DEFINE_ENUM

As Eric reported, the 'reason' field is not presented when trace the
kfree_skb event by perf:

$ perf record -e skb:kfree_skb -a sleep 10
$ perf script
  ip_defrag 14605 [021]   221.614303:   skb:kfree_skb:
  skbaddr=0xffff9d2851242700 protocol=34525 location=0xffffffffa39346b1
  reason:

The cause seems to be passing kernel address directly to TP_printk(),
which is not right. As the enum 'skb_drop_reason' is not exported to
user space through TRACE_DEFINE_ENUM(), perf can't get the drop reason
string from the 'reason' field, which is a number.

Therefore, we introduce the macro DEFINE_DROP_REASON(), which is used
to define the trace enum by TRACE_DEFINE_ENUM(). With the help of
DEFINE_DROP_REASON(), now we can remove the auto-generate that we
introduced in the commit ec43908dd556
("net: skb: use auto-generation to convert skb drop reason to string"),
and define the string array 'drop_reasons'.

Hmmmm...now we come back to the situation that have to maintain drop
reasons in both enum skb_drop_reason and DEFINE_DROP_REASON. But they
are both in dropreason.h, which makes it easier.

After this commit, now the format of kfree_skb is like this:

$ cat /tracing/events/skb/kfree_skb/format
name: kfree_skb
ID: 1524
format:
        field:unsigned short common_type;       offset:0;       size:2; signed:0;
        field:unsigned char common_flags;       offset:2;       size:1; signed:0;
        field:unsigned char common_preempt_count;       offset:3;       size:1; signed:0;
        field:int common_pid;   offset:4;       size:4; signed:1;

        field:void * skbaddr;   offset:8;       size:8; signed:0;
        field:void * location;  offset:16;      size:8; signed:0;
        field:unsigned short protocol;  offset:24;      size:2; signed:0;
        field:enum skb_drop_reason reason;      offset:28;      size:4; signed:0;

print fmt: "skbaddr=%p protocol=%u location=%p reason: %s", REC->skbaddr, REC->protocol, REC->location, __print_symbolic(REC->reason, { 1, "NOT_SPECIFIED" }, { 2, "NO_SOCKET" } ......

Fixes: ec43908dd556 ("net: skb: use auto-generation to convert skb drop reason to string")
Link: https://lore.kernel.org/netdev/CANn89i+bx0ybvE55iMYf5GJM48WwV1HNpdm9Q6t-HaEstqpCSA@mail.gmail.com/
Reported-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Menglong Dong <imagedong@tencent.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet: ethernet: mtk_eth_soc: fix typo in __mtk_foe_entry_clear
Lorenzo Bianconi [Tue, 6 Sep 2022 14:36:32 +0000 (16:36 +0200)]
net: ethernet: mtk_eth_soc: fix typo in __mtk_foe_entry_clear

Set ib1 state to MTK_FOE_STATE_UNBIND in __mtk_foe_entry_clear routine.

Fixes: 33fc42de33278 ("net: ethernet: mtk_eth_soc: support creating mac address based offload entries")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonetfilter: conntrack: ignore overly delayed tcp packets
Florian Westphal [Fri, 26 Aug 2022 13:32:25 +0000 (15:32 +0200)]
netfilter: conntrack: ignore overly delayed tcp packets

If 'nf_conntrack_tcp_loose' is off (the default), tcp packets that are
outside of the current window are marked as INVALID.

nf/iptables rulesets often drop such packets via 'ct state invalid' or
similar checks.

For overly delayed acks, this can be a nuisance if such 'invalid' packets
are also logged.

Since they are not invalid in a strict sense, just ignore them, i.e.
conntrack won't extend timeout or change state so that they do not match
invalid state rules anymore.

This also avoids unwantend connection stalls in case conntrack considers
retransmission (of data that did not reach the peer) as too old.

The else branch of the conditional becomes obsolete.
Next patch will reformant the now always-true if condition.

The existing workaround for data that exceeds the calculated receive
window is adjusted to use the 'ignore' state so that these packets do
not refresh the timeout or change state other than updating ->td_end.

Signed-off-by: Florian Westphal <fw@strlen.de>
22 months agonetfilter: conntrack: prepare tcp_in_window for ternary return value
Florian Westphal [Fri, 26 Aug 2022 13:32:24 +0000 (15:32 +0200)]
netfilter: conntrack: prepare tcp_in_window for ternary return value

tcp_in_window returns true if the packet is in window and false if it is
not.

If its outside of window, packet will be treated as INVALID.

There are corner cases where the packet should still be tracked, because
rulesets may drop or log such packets, even though they can occur during
normal operation, such as overly delayed acks.

In extreme cases, connection may hang forever because conntrack state
differs from real state.

There is no retransmission for ACKs.

In case of ACK loss after conntrack processing, its possible that a
connection can be stuck because the actual retransmits are considered
stale ("SEQ is under the lower bound (already ACKed data
retransmitted)".

The problem is made worse by carrier-grade-nat which can also result
in stale packets from old connections to get treated as 'recent' packets
in conntrack (it doesn't support tcp timestamps at this time).

Prepare tcp_in_window() to return an enum that tells the desired
action (in-window/accept, bogus/drop).

A third action (accept the packet as in-window, but do not change
state) is added in a followup patch.

Signed-off-by: Florian Westphal <fw@strlen.de>
22 months agoMerge branch 'macsec-offload-mlx5'
David S. Miller [Wed, 7 Sep 2022 13:02:09 +0000 (14:02 +0100)]
Merge branch 'macsec-offload-mlx5'

Saeed Mahameed says:

====================
Introduce MACsec skb_metadata_dst and mlx5 macsec offload

v1->v2:
   - attach mlx5 implementation patches.

This patchset introduces MACsec skb_metadata_dst to lay the ground
for MACsec HW offload.

MACsec is an IEEE standard (IEEE 802.1AE) for MAC security.
It defines a way to establish a protocol independent connection
between two hosts with data confidentiality, authenticity and/or
integrity, using GCM-AES. MACsec operates on the Ethernet layer and
as such is a layer 2 protocol, which means it’s designed to secure
traffic within a layer 2 network, including DHCP or ARP requests.

Linux has a software implementation of the MACsec standard and
HW offloading support.
The offloading is re-using the logic, netlink API and data
structures of the existing MACsec software implementation.

For Tx:
In the current MACsec offload implementation, MACsec interfaces shares
the same MAC address by default.
Therefore, HW can't distinguish from which MACsec interface the traffic
originated from.

MACsec stack will use skb_metadata_dst to store the SCI value, which is
unique per MACsec interface, skb_metadat_dst will be used later by the
offloading device driver to associate the SKB with the corresponding
offloaded interface (SCI) to facilitate HW MACsec offload.

For Rx:
Like in the Tx changes, if there are more than one MACsec device with
the same MAC address as in the packet's destination MAC, the packet will
be forward only to one of the devices and not neccessarly to the desired one.

Offloading device driver sets the MACsec skb_metadata_dst sci
field with the appropriaate Rx SCI for each SKB so the MACsec rx handler
will know to which port to divert those skbs, instead of wrongly solely
relaying on dst MAC address comparison.

1) patch 1,2, Add support to skb_metadata_dst in MACsec code:
net/macsec: Add MACsec skb_metadata_dst Tx Data path support
net/macsec: Add MACsec skb_metadata_dst Rx Data path support

2) patch 3, Move some MACsec driver code for sharing with various
drivers that implements offload:
net/macsec: Move some code for sharing with various drivers that
implements offload

3) The rest of the patches introduce mlx5 implementation for macsec
offloads TX and RX via steering tables.
  a) TX, intercept skbs with macsec offlad mark in skb_metadata_dst and mark
the descriptor for offload.
  b) RX, intercept offloaded frames and prepare the proper
skb_metadata_dst to mark offloaded rx frames.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5e: Add support to configure more than one macsec offload device
Lior Nahmanson [Tue, 6 Sep 2022 05:21:29 +0000 (22:21 -0700)]
net/mlx5e: Add support to configure more than one macsec offload device

Add the ability to add up to 16 MACsec offload interfaces
over the same physical interface

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5e: Add MACsec stats support for Rx/Tx flows
Lior Nahmanson [Tue, 6 Sep 2022 05:21:28 +0000 (22:21 -0700)]
net/mlx5e: Add MACsec stats support for Rx/Tx flows

Add the following statistics:
RX successfully decrypted MACsec packets:
macsec_rx_pkts : Number of packets decrypted successfully
macsec_rx_bytes : Number of bytes decrypted successfully

Rx dropped MACsec packets:
macsec_rx_pkts_drop : Number of MACsec packets dropped
macsec_rx_bytes_drop : Number of MACsec bytes dropped

TX successfully encrypted MACsec packets:
macsec_tx_pkts : Number of packets encrypted/authenticated successfully
macsec_tx_bytes : Number of bytes encrypted/authenticated successfully

Tx dropped MACsec packets:
macsec_tx_pkts_drop : Number of MACsec packets dropped
macsec_tx_bytes_drop : Number of MACsec bytes dropped

The above can be seen using:
ethtool -S <ifc> |grep macsec

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5e: Add MACsec offload SecY support
Lior Nahmanson [Tue, 6 Sep 2022 05:21:27 +0000 (22:21 -0700)]
net/mlx5e: Add MACsec offload SecY support

Add offload support for MACsec SecY callbacks - add/update/delete.
add_secy is called when need to create a new MACsec interface.
upd_secy is called when source MAC address or tx SC was changed.
del_secy is called when need to destroy the MACsec interface.

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5e: Implement MACsec Rx data path using MACsec skb_metadata_dst
Lior Nahmanson [Tue, 6 Sep 2022 05:21:26 +0000 (22:21 -0700)]
net/mlx5e: Implement MACsec Rx data path using MACsec skb_metadata_dst

MACsec driver need to distinguish to which offload device the MACsec
is target to, in order to handle them correctly.
This can be done by attaching a metadata_dst to a SKB with a SCI,
when there is a match on MACsec rule.
To achieve that, there is a map between fs_id to SCI, so for each RX SC,
there is a unique fs_id allocated when creating RX SC.
fs_id passed to device driver as metadata for packets that passed Rx
MACsec offload to aid the driver to retrieve the matching SCI.

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5e: Add MACsec RX steering rules
Lior Nahmanson [Tue, 6 Sep 2022 05:21:25 +0000 (22:21 -0700)]
net/mlx5e: Add MACsec RX steering rules

Rx flow steering consists of two flow tables (FTs).

The first FT (crypto table) have one default miss rule so non MACsec
offloaded packets bypass the MACSec tables.
All others flow table entries (FTEs) are divided to two equal groups
size, both of them are for MACsec packets:
The first group is for MACsec packets which contains SCI field in the
SecTAG header.
The second group is for MACsec packets which doesn't contain SCI,
where need to match on the source MAC address (only if the SCI
is built from default MACsec port).
Destination MAC address, ethertype and some of SecTAG fields
are also matched for both groups.
In case of match, invoke decrypt action on the packet.
For each MACsec Rx offloaded SA two rules are created: one with SCI
and one without SCI.

The second FT (check table) has two fixed rules:
One rule is for verifying that the previous offload actions were
finished successfully.
In this case, need to decap the SecTAG header and forward the packet
for further processing.
Another default rule for dropping packets that failed in the previous
decrypt actions.

The MACsec FTs are created on demand when the first MACsec rule is
added and destroyed when the last MACsec rule is deleted.

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5: Add MACsec Rx tables support to fs_core
Lior Nahmanson [Tue, 6 Sep 2022 05:21:24 +0000 (22:21 -0700)]
net/mlx5: Add MACsec Rx tables support to fs_core

Add new namespace for MACsec RX flows.
Encrypted MACsec packets should be first decrypted and stripped
from MACsec header and then continues with the kernel's steering
pipeline.

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5e: Add MACsec offload Rx command support
Lior Nahmanson [Tue, 6 Sep 2022 05:21:23 +0000 (22:21 -0700)]
net/mlx5e: Add MACsec offload Rx command support

Add a support for Connect-X MACsec offload Rx SA & SC commands:
add, update and delete.

SCs are created on demend and aren't limited by number and unique by SCI.
Each Rx SA must be associated with Rx SC according to SCI.

Follow-up patches will implement the Rx steering.

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5e: Implement MACsec Tx data path using MACsec skb_metadata_dst
Lior Nahmanson [Tue, 6 Sep 2022 05:21:22 +0000 (22:21 -0700)]
net/mlx5e: Implement MACsec Tx data path using MACsec skb_metadata_dst

MACsec driver marks Tx packets for device offload using a dedicated
skb_metadata_dst which holds a 64 bits SCI number.
A previously set rule will match on this number so the correct SA is used
for the MACsec operation.
As device driver can only provide 32 bits of metadata to flow tables,
need to used a mapping from 64 bit to 32 bits marker or id,
which is can be achieved by provide a 32 bit unique flow id in the
control path, and used a hash table to map 64 bit to the unique id in the
datapath.

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5e: Add MACsec TX steering rules
Lior Nahmanson [Tue, 6 Sep 2022 05:21:21 +0000 (22:21 -0700)]
net/mlx5e: Add MACsec TX steering rules

Tx flow steering consists of two flow tables (FTs).

The first FT (crypto table) has two fixed rules:
One default miss rule so non MACsec offloaded packets bypass the MACSec
tables, another rule to make sure that MACsec key exchange (MKE) traffic
passes unencrypted as expected (matched of ethertype).
On each new MACsec offload flow, a new MACsec rule is added.
This rule is matched on metadata_reg_a (which contains the id of the
flow) and invokes the MACsec offload action on match.

The second FT (check table) has two fixed rules:
One rule for verifying that the previous offload actions were
finished successfully and packet need to be transmitted.
Another default rule for dropping packets that were failed in the
offload actions.

The MACsec FTs should be created on demand when the first MACsec rule is
added and destroyed when the last MACsec rule is deleted.

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5: Add MACsec Tx tables support to fs_core
Lior Nahmanson [Tue, 6 Sep 2022 05:21:20 +0000 (22:21 -0700)]
net/mlx5: Add MACsec Tx tables support to fs_core

Changed EGRESS_KERNEL namespace to EGRESS_IPSEC and add new
namespace for MACsec TX.
This namespace should be the last namespace for transmitted packets.

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5: Add MACsec offload Tx command support
Lior Nahmanson [Tue, 6 Sep 2022 05:21:19 +0000 (22:21 -0700)]
net/mlx5: Add MACsec offload Tx command support

This patch adds support for Connect-X MACsec offload Tx SA commands:
add, update and delete.

In Connect-X MACsec, a Security Association (SA) is added or deleted
via allocating a HW context of an encryption/decryption key and
a HW context of a matching SA (MACsec object).

When new SA is added:
- Use a separate crypto key HW context.
- Create a separate MACsec context in HW to include the SA properties.

Introduce a new compilation flag MLX5_EN_MACSEC for it.

Follow-up patches will implement the Tx steering.

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
22 months agonet/mlx5: Introduce MACsec Connect-X offload hardware bits and structures
Lior Nahmanson [Tue, 6 Sep 2022 05:21:18 +0000 (22:21 -0700)]
net/mlx5: Introduce MACsec Connect-X offload hardware bits and structures

Add MACsec offload related IFC structs, layouts and enumerations.

Signed-off-by: Lior Nahmanson <liorna@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>