platform/kernel/linux-rpi.git
4 years agocxgb4: add TC-MATCHALL IPv6 support
Rahul Lakkireddy [Mon, 3 Aug 2020 18:30:08 +0000 (00:00 +0530)]
cxgb4: add TC-MATCHALL IPv6 support

Matching IPv6 traffic require allocating their own individual slots
in TCAM. So, fetch additional slots to insert IPv6 rules. Also, fetch
the cumulative stats of all the slots occupied by the Matchall rule.

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: sja1105: poll for extts events from a timer
Vladimir Oltean [Mon, 3 Aug 2020 17:51:58 +0000 (20:51 +0300)]
net: dsa: sja1105: poll for extts events from a timer

The current poll interval is enough to ensure that rising and falling
edge events are not lost for a 1 PPS signal with 50% duty cycle.

But when we deliver the events to user space, it will try to infer if
they were corresponding to a rising or to a falling edge (the kernel
driver doesn't know that either). User space will try to make that
inference based on the time at which the PPS master had emitted the
pulse (i.e. if it's a .0 time, it's rising edge, if it's .5 time, it's
falling edge).

But there is no in-kernel API for retrieving the precise timestamp
corresponding to a PPS master (aka perout) pulse. So user space has to
guess even that. It will read the PTP time on the PPS master right after
we've delivered the extts event, and declare that the PPS master time
was just the closest integer second, based on 2 thresholds (lower than
.25, or higher than .75, and ignore anything else).

Except that, if we poll for extts events (and our hardware doesn't
really help us, by not providing an interrupt), then there is a risk
that the poll period (and therefore the time at which the event is
delivered) might confuse user space.

Because we are always scheduling the next extts poll at
SJA1105_EXTTS_INTERVAL "from now" (that's the only thing that the
schedule_delayed_work() API gives us), it means that the start time of
the next delayed workqueue will always be shifted to the right a little
bit (shifted with the SPI access duration of this workqueue run).
In turn, because user space sees extts events that are non-periodic
compared to the PPS master's time, this means that it might start making
wrong guesses about rising/falling edge.

To understand the effect, here is the output of ts2phc currently. Notice
the 'src' timestamps of the 'SKIP extts' events, and how they have a
large wander. They keep increasing until the upper limit for the ignore
threshold (.75 seconds), after which the application starts ignoring the
_other_ edge.

ts2phc[26.624]: /dev/ptp3 SKIP extts index 0 at 21.449898912 src 21.657784518
ts2phc[27.133]: adding tstamp 21.949894240 to clock /dev/ptp3
ts2phc[27.133]: adding tstamp 22.000000000 to clock /dev/ptp1
ts2phc[27.133]: /dev/ptp3 offset        640 s2 freq   +5112
ts2phc[27.636]: /dev/ptp3 SKIP extts index 0 at 22.449889360 src 22.669398022
ts2phc[28.140]: adding tstamp 22.949884376 to clock /dev/ptp3
ts2phc[28.140]: adding tstamp 23.000000000 to clock /dev/ptp1
ts2phc[28.140]: /dev/ptp3 offset         96 s2 freq   +4760
ts2phc[28.644]: /dev/ptp3 SKIP extts index 0 at 23.449879504 src 23.677420422
ts2phc[29.153]: adding tstamp 23.949874704 to clock /dev/ptp3
ts2phc[29.153]: adding tstamp 24.000000000 to clock /dev/ptp1
ts2phc[29.153]: /dev/ptp3 offset       -264 s2 freq   +4429
ts2phc[29.656]: /dev/ptp3 SKIP extts index 0 at 24.449870008 src 24.689407238
ts2phc[30.160]: adding tstamp 24.949865376 to clock /dev/ptp3
ts2phc[30.160]: adding tstamp 25.000000000 to clock /dev/ptp1
ts2phc[30.160]: /dev/ptp3 offset       -280 s2 freq   +4334
ts2phc[30.664]: /dev/ptp3 SKIP extts index 0 at 25.449860760 src 25.697449926
ts2phc[31.168]: adding tstamp 25.949856176 to clock /dev/ptp3
ts2phc[31.168]: adding tstamp 26.000000000 to clock /dev/ptp1
ts2phc[31.168]: /dev/ptp3 offset       -176 s2 freq   +4354
ts2phc[31.672]: /dev/ptp3 SKIP extts index 0 at 26.449851584 src 26.705433606
ts2phc[32.180]: adding tstamp 26.949846992 to clock /dev/ptp3
ts2phc[32.180]: adding tstamp 27.000000000 to clock /dev/ptp1
ts2phc[32.180]: /dev/ptp3 offset        -80 s2 freq   +4397
ts2phc[32.684]: /dev/ptp3 SKIP extts index 0 at 27.449842384 src 27.717415110
ts2phc[33.192]: adding tstamp 27.949837768 to clock /dev/ptp3
ts2phc[33.192]: adding tstamp 28.000000000 to clock /dev/ptp1
ts2phc[33.192]: /dev/ptp3 offset          0 s2 freq   +4453
ts2phc[33.696]: /dev/ptp3 SKIP extts index 0 at 28.449833128 src 28.729412902
ts2phc[34.200]: adding tstamp 28.949828472 to clock /dev/ptp3
ts2phc[34.200]: adding tstamp 29.000000000 to clock /dev/ptp1
ts2phc[34.200]: /dev/ptp3 offset          8 s2 freq   +4461
ts2phc[34.704]: /dev/ptp3 SKIP extts index 0 at 29.449823816 src 29.737416038
ts2phc[35.208]: adding tstamp 29.949819152 to clock /dev/ptp3
ts2phc[35.208]: adding tstamp 30.000000000 to clock /dev/ptp1
ts2phc[35.208]: /dev/ptp3 offset         -8 s2 freq   +4447
ts2phc[35.712]: /dev/ptp3 SKIP extts index 0 at 30.449814496 src 30.745554982
ts2phc[36.216]: adding tstamp 30.949809840 to clock /dev/ptp3
ts2phc[36.216]: adding tstamp 31.000000000 to clock /dev/ptp1
ts2phc[36.216]: /dev/ptp3 offset         -8 s2 freq   +4445
ts2phc[36.468]: /dev/ptp3 SKIP extts index 0 at 31.449805184 src 31.501109446
ts2phc[36.972]: adding tstamp 31.949800536 to clock /dev/ptp3
ts2phc[36.972]: adding tstamp 32.000000000 to clock /dev/ptp1
ts2phc[36.972]: /dev/ptp3 offset         -8 s2 freq   +4442
ts2phc[37.480]: /dev/ptp3 SKIP extts index 0 at 32.449795896 src 32.513320070
ts2phc[37.984]: adding tstamp 32.949791248 to clock /dev/ptp3
ts2phc[37.984]: adding tstamp 33.000000000 to clock /dev/ptp1
ts2phc[37.984]: /dev/ptp3 offset          0 s2 freq   +4448

Fix that by taking the following measures:
- Schedule the poll from a timer. Because we are really scheduling the
  timer periodically, the extts events delivered to user space are
  periodic too, and don't suffer from the "shift-to-the-right" effect.
- Increase the poll period to 6 times a second. This imposes a smaller
  upper bound to the shift that can occur to the delivery time of extts
  events, and makes user space (ts2phc) to always interpret correctly
  which events should be skipped and which shouldn't.
- Move the SPI readout itself to the main PTP kernel thread, instead of
  the generic workqueue. This is because the timer runs in atomic
  context, but is also better than before, because if needed, we can
  chrt & taskset this kernel thread, to ensure it gets enough priority
  under load.

After this patch, one can notice that the wander is greatly reduced, and
that the latencies of one extts poll are not propagated to the next. The
'src' timestamp that is skipped is never larger than .65 seconds (which
means .15 seconds larger than the time at which the real event occurred
at, and .10 seconds smaller than the .75 upper threshold for ignoring
the falling edge):

ts2phc[40.076]: adding tstamp 34.949261296 to clock /dev/ptp3
ts2phc[40.076]: adding tstamp 35.000000000 to clock /dev/ptp1
ts2phc[40.076]: /dev/ptp3 offset         48 s2 freq   +4631
ts2phc[40.568]: /dev/ptp3 SKIP extts index 0 at 35.449256496 src 35.595791078
ts2phc[41.064]: adding tstamp 35.949251744 to clock /dev/ptp3
ts2phc[41.064]: adding tstamp 36.000000000 to clock /dev/ptp1
ts2phc[41.064]: /dev/ptp3 offset       -224 s2 freq   +4374
ts2phc[41.552]: /dev/ptp3 SKIP extts index 0 at 36.449247088 src 36.579825574
ts2phc[42.044]: adding tstamp 36.949242456 to clock /dev/ptp3
ts2phc[42.044]: adding tstamp 37.000000000 to clock /dev/ptp1
ts2phc[42.044]: /dev/ptp3 offset       -240 s2 freq   +4290
ts2phc[42.536]: /dev/ptp3 SKIP extts index 0 at 37.449237848 src 37.563828774
ts2phc[43.028]: adding tstamp 37.949233264 to clock /dev/ptp3
ts2phc[43.028]: adding tstamp 38.000000000 to clock /dev/ptp1
ts2phc[43.028]: /dev/ptp3 offset       -144 s2 freq   +4314
ts2phc[43.520]: /dev/ptp3 SKIP extts index 0 at 38.449228656 src 38.547823238
ts2phc[44.012]: adding tstamp 38.949224048 to clock /dev/ptp3
ts2phc[44.012]: adding tstamp 39.000000000 to clock /dev/ptp1
ts2phc[44.012]: /dev/ptp3 offset        -80 s2 freq   +4335
ts2phc[44.508]: /dev/ptp3 SKIP extts index 0 at 39.449219432 src 39.535846118
ts2phc[44.996]: adding tstamp 39.949214816 to clock /dev/ptp3
ts2phc[44.996]: adding tstamp 40.000000000 to clock /dev/ptp1
ts2phc[44.996]: /dev/ptp3 offset        -32 s2 freq   +4359
ts2phc[45.488]: /dev/ptp3 SKIP extts index 0 at 40.449210192 src 40.515824678
ts2phc[45.980]: adding tstamp 40.949205568 to clock /dev/ptp3
ts2phc[45.980]: adding tstamp 41.000000000 to clock /dev/ptp1
ts2phc[45.980]: /dev/ptp3 offset          8 s2 freq   +4390
ts2phc[46.636]: /dev/ptp3 SKIP extts index 0 at 41.449200928 src 41.664176902
ts2phc[47.132]: adding tstamp 41.949196288 to clock /dev/ptp3
ts2phc[47.132]: adding tstamp 42.000000000 to clock /dev/ptp1
ts2phc[47.132]: /dev/ptp3 offset          0 s2 freq   +4384
ts2phc[47.620]: /dev/ptp3 SKIP extts index 0 at 42.449191656 src 42.648117190
ts2phc[48.112]: adding tstamp 42.949187016 to clock /dev/ptp3
ts2phc[48.112]: adding tstamp 43.000000000 to clock /dev/ptp1
ts2phc[48.112]: /dev/ptp3 offset          0 s2 freq   +4384
ts2phc[48.604]: /dev/ptp3 SKIP extts index 0 at 43.449182384 src 43.632112582
ts2phc[49.100]: adding tstamp 43.949177736 to clock /dev/ptp3
ts2phc[49.100]: adding tstamp 44.000000000 to clock /dev/ptp1
ts2phc[49.100]: /dev/ptp3 offset         -8 s2 freq   +4376
ts2phc[49.588]: /dev/ptp3 SKIP extts index 0 at 44.449173096 src 44.616136774
ts2phc[50.080]: adding tstamp 44.949168464 to clock /dev/ptp3
ts2phc[50.080]: adding tstamp 45.000000000 to clock /dev/ptp1
ts2phc[50.080]: /dev/ptp3 offset          8 s2 freq   +4390
ts2phc[50.572]: /dev/ptp3 SKIP extts index 0 at 45.449163816 src 45.600134662
ts2phc[51.064]: adding tstamp 45.949159160 to clock /dev/ptp3
ts2phc[51.064]: adding tstamp 46.000000000 to clock /dev/ptp1
ts2phc[51.064]: /dev/ptp3 offset         -8 s2 freq   +4376
ts2phc[51.556]: /dev/ptp3 SKIP extts index 0 at 46.449154528 src 46.584588550
ts2phc[52.048]: adding tstamp 46.949149896 to clock /dev/ptp3
ts2phc[52.048]: adding tstamp 47.000000000 to clock /dev/ptp1
ts2phc[52.048]: /dev/ptp3 offset          0 s2 freq   +4382
ts2phc[52.540]: /dev/ptp3 SKIP extts index 0 at 47.449145256 src 47.568132198
ts2phc[53.032]: adding tstamp 47.949140616 to clock /dev/ptp3
ts2phc[53.032]: adding tstamp 48.000000000 to clock /dev/ptp1
ts2phc[53.032]: /dev/ptp3 offset          0 s2 freq   +4382
ts2phc[53.524]: /dev/ptp3 SKIP extts index 0 at 48.449135968 src 48.552121446
ts2phc[54.016]: adding tstamp 48.949131320 to clock /dev/ptp3
ts2phc[54.016]: adding tstamp 49.000000000 to clock /dev/ptp1
ts2phc[54.016]: /dev/ptp3 offset          0 s2 freq   +4382
ts2phc[54.512]: /dev/ptp3 SKIP extts index 0 at 49.449126680 src 49.540147014
ts2phc[55.000]: adding tstamp 49.949122040 to clock /dev/ptp3
ts2phc[55.000]: adding tstamp 50.000000000 to clock /dev/ptp1
ts2phc[55.000]: /dev/ptp3 offset          0 s2 freq   +4382
ts2phc[55.492]: /dev/ptp3 SKIP extts index 0 at 50.449117400 src 50.520119078
ts2phc[55.988]: adding tstamp 50.949112768 to clock /dev/ptp3
ts2phc[55.988]: adding tstamp 51.000000000 to clock /dev/ptp1
ts2phc[55.988]: /dev/ptp3 offset          8 s2 freq   +4390
ts2phc[56.476]: /dev/ptp3 SKIP extts index 0 at 51.449108120 src 51.504175910
ts2phc[57.132]: adding tstamp 51.949103480 to clock /dev/ptp3
ts2phc[57.132]: adding tstamp 52.000000000 to clock /dev/ptp1
ts2phc[57.132]: /dev/ptp3 offset          0 s2 freq   +4384
ts2phc[57.624]: /dev/ptp3 SKIP extts index 0 at 52.449098840 src 52.651833574
ts2phc[58.116]: adding tstamp 52.949094200 to clock /dev/ptp3
ts2phc[58.116]: adding tstamp 53.000000000 to clock /dev/ptp1
ts2phc[58.116]: /dev/ptp3 offset          8 s2 freq   +4392
ts2phc[58.612]: /dev/ptp3 SKIP extts index 0 at 53.449089560 src 53.639826918
ts2phc[59.100]: adding tstamp 53.949084920 to clock /dev/ptp3
ts2phc[59.100]: adding tstamp 54.000000000 to clock /dev/ptp1
ts2phc[59.100]: /dev/ptp3 offset          8 s2 freq   +4394
ts2phc[59.592]: /dev/ptp3 SKIP extts index 0 at 54.449080272 src 54.619842278
ts2phc[60.084]: adding tstamp 54.949075624 to clock /dev/ptp3
ts2phc[60.084]: adding tstamp 55.000000000 to clock /dev/ptp1
ts2phc[60.084]: /dev/ptp3 offset          8 s2 freq   +4397
ts2phc[60.576]: /dev/ptp3 SKIP extts index 0 at 55.449070968 src 55.603885542
ts2phc[61.068]: adding tstamp 55.949066312 to clock /dev/ptp3
ts2phc[61.068]: adding tstamp 56.000000000 to clock /dev/ptp1
ts2phc[61.068]: /dev/ptp3 offset          0 s2 freq   +4391
ts2phc[61.560]: /dev/ptp3 SKIP extts index 0 at 56.449061680 src 56.587885798
ts2phc[62.052]: adding tstamp 56.949057032 to clock /dev/ptp3
ts2phc[62.052]: adding tstamp 57.000000000 to clock /dev/ptp1
ts2phc[62.052]: /dev/ptp3 offset         -8 s2 freq   +4383

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomptcp: fix bogus sendmsg() return code under pressure
Paolo Abeni [Mon, 3 Aug 2020 16:40:39 +0000 (18:40 +0200)]
mptcp: fix bogus sendmsg() return code under pressure

In case of memory pressure, mptcp_sendmsg() may call
sk_stream_wait_memory() after succesfully xmitting some
bytes. If the latter fails we currently return to the
user-space the error code, ignoring the succeful xmit.

Address the issue always checking for the xmitted bytes
before mptcp_sendmsg() completes.

Fixes: f296234c98a8 ("mptcp: Add handling of incoming MP_JOIN requests")
Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'mlxsw-Add-support-for-buffer-drop-traps'
David S. Miller [Tue, 4 Aug 2020 01:06:47 +0000 (18:06 -0700)]
Merge branch 'mlxsw-Add-support-for-buffer-drop-traps'

Ido Schimmel says:

====================
mlxsw: Add support for buffer drop traps

Petr says:

A recent patch set added the ability to mirror buffer related drops
(e.g., early drops) through a netdev. This patch set adds the ability to
trap such packets to the local CPU for analysis.

The trapping towards the CPU is configured by using tc-trap action
instead of tc-mirred as was done when the packets were mirrored through
a netdev. A future patch set will also add the ability to sample the
dropped packets using tc-sample action.

The buffer related drop traps are added to devlink, which means that the
dropped packets can be reported to user space via the kernel's
drop_monitor module.

Patch set overview:

Patch #1 adds the early_drop trap to devlink

Patch #2 adds extack to a few devlink operations to facilitate better
error reporting to user space. This is necessary - among other things -
because the action of buffer drop traps cannot be changed in mlxsw

Patch #3 performs a small refactoring in mlxsw, patch #4 fixes a bug that
this patchset would trigger.

Patches #5-#6 add the infrastructure required to support different traps
/ trap groups in mlxsw per-ASIC. This is required because buffer drop
traps are not supported by Spectrum-1

Patch #7 extends mlxsw to register the early_drop trap

Patch #8 adds the offload logic for the "trap" action at a qevent block.

Patch #9 adds a mlxsw-specific selftest.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests: mlxsw: RED: Test offload of trapping on RED qevents
Petr Machata [Mon, 3 Aug 2020 16:11:41 +0000 (19:11 +0300)]
selftests: mlxsw: RED: Test offload of trapping on RED qevents

Add a selftest for RED early_drop and mark qevents when a trap action is
attached at the associated block.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_qdisc: Offload action trap for qevents
Petr Machata [Mon, 3 Aug 2020 16:11:40 +0000 (19:11 +0300)]
mlxsw: spectrum_qdisc: Offload action trap for qevents

When offloading action trap on a qevent, pass to_dev of NULL to the SPAN
module to trigger the mirror to the CPU port. Query the buffer drops
policer and use it for policing of the trapped traffic.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_trap: Add early_drop trap
Ido Schimmel [Mon, 3 Aug 2020 16:11:39 +0000 (19:11 +0300)]
mlxsw: spectrum_trap: Add early_drop trap

As previously explained, packets that are dropped due to buffer related
reasons (e.g., tail drop, early drop) can be mirrored to the CPU port.
These packets are then trapped with one of the "mirror session" traps
and their CQE includes the reason for which the packet was mirrored.

Register with devlink a new trap, early_drop, and initialize the
corresponding Rx listener with the appropriate mirror reason. Return an
error in case user tries to change the traps' action, as this is not
supported.

Since Spectrum-1 does not support these traps, the above is only done
for Spectrum-2 onwards.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_trap: Allow for per-ASIC traps initialization
Ido Schimmel [Mon, 3 Aug 2020 16:11:38 +0000 (19:11 +0300)]
mlxsw: spectrum_trap: Allow for per-ASIC traps initialization

Subsequent patches will need to register different traps for Spectrum-1
and Spectrum-2 onwards.

Enable that by invoking a per-ASIC operation during traps
initialization.

Reviewed-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_trap: Allow for per-ASIC trap groups initialization
Ido Schimmel [Mon, 3 Aug 2020 16:11:37 +0000 (19:11 +0300)]
mlxsw: spectrum_trap: Allow for per-ASIC trap groups initialization

Subsequent patches will need to register different trap groups for
Spectrum-1 and Spectrum-2 onwards.

Enable that by invoking a per-ASIC operation during trap groups
initialization.

Reviewed-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_span: On policer_id_base_ref_count, use dec_and_test
Petr Machata [Mon, 3 Aug 2020 16:11:36 +0000 (19:11 +0300)]
mlxsw: spectrum_span: On policer_id_base_ref_count, use dec_and_test

When unsetting policer base, the SPAN code currently uses refcount_dec().
However that function splats when the counter reaches zero, because
reaching zero without actually testing is in general indicative of a
missing cleanup. There is no cleanup to be done here, but nonetheless, use
refcount_dec_and_test() as required.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_trap: Use 'size_t' for array sizes
Ido Schimmel [Mon, 3 Aug 2020 16:11:35 +0000 (19:11 +0300)]
mlxsw: spectrum_trap: Use 'size_t' for array sizes

Use 'size_t' instead of 'u64' for array sizes, as this this is correct
type to use for expressions involving sizeof().

Suggested-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Pass extack when setting trap's action and group's parameters
Ido Schimmel [Mon, 3 Aug 2020 16:11:34 +0000 (19:11 +0300)]
devlink: Pass extack when setting trap's action and group's parameters

A later patch will refuse to set the action of certain traps in mlxsw
and also to change the policer binding of certain groups. Pass extack so
that failure could be communicated clearly to user space.

Reviewed-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodevlink: Add early_drop trap
Amit Cohen [Mon, 3 Aug 2020 16:11:33 +0000 (19:11 +0300)]
devlink: Add early_drop trap

Add the packet trap that can report packets that were ECN marked due to RED
AQM.

Signed-off-by: Amit Cohen <amitc@mellanox.com>
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agofib: Fix undef compile warning
YueHaibing [Mon, 3 Aug 2020 13:19:48 +0000 (21:19 +0800)]
fib: Fix undef compile warning

net/core/fib_rules.c:26:7: warning: "CONFIG_IP_MULTIPLE_TABLES" is not defined, evaluates to 0 [-Wundef]
 #elif CONFIG_IP_MULTIPLE_TABLES
       ^~~~~~~~~~~~~~~~~~~~~~~~~

Fixes: 8b66a6fd34f5 ("fib: fix another fib_rules_ops indirect call wrapper problem")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-By: Brian Vazquez <brianvv@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomptcp: use mptcp_for_each_subflow in mptcp_stream_accept
Geliang Tang [Mon, 3 Aug 2020 13:00:44 +0000 (21:00 +0800)]
mptcp: use mptcp_for_each_subflow in mptcp_stream_accept

Use mptcp_for_each_subflow in mptcp_stream_accept instead of
open-coding.

Signed-off-by: Geliang Tang <geliangtang@gmail.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge tag 'mac80211-next-for-davem-2020-08-03' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Tue, 4 Aug 2020 01:00:22 +0000 (18:00 -0700)]
Merge tag 'mac80211-next-for-davem-2020-08-03' of git://git./linux/kernel/git/jberg/mac80211-next

Johannes Berg says:

====================
A few more changes, notably:
 * handle new SAE (WPA3 authentication) status codes in the correct way
 * fix a while that should be an if instead, avoiding infinite loops
 * handle beacon filtering changing better
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: stmmac: fix failed to suspend if phy based WOL is enabled
Jisheng Zhang [Mon, 3 Aug 2020 08:56:47 +0000 (16:56 +0800)]
net: stmmac: fix failed to suspend if phy based WOL is enabled

With the latest net-next tree, if test suspend/resume after enabling
WOL, we get error as below:

[  487.086365] dpm_run_callback(): mdio_bus_suspend+0x0/0x30 returns -16
[  487.086375] PM: Device stmmac-0:00 failed to suspend: error -16

-16 means -EBUSY, this is because I didn't enable wakeup of the correct
device when implementing phy based WOL feature. To be honest, I caught
the issue when implementing phy based WOL and then fix it locally, but
forgot to amend the phy based wol patch. Today, I found the issue by
testing net-next tree.

Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoseg6_iptunnel: Refactor seg6_lwt_headroom out of uapi header
Ioana-Ruxandra Stăncioi [Mon, 3 Aug 2020 07:33:33 +0000 (07:33 +0000)]
seg6_iptunnel: Refactor seg6_lwt_headroom out of uapi header

Refactor the function seg6_lwt_headroom out of the seg6_iptunnel.h uapi
header, because it is only used in seg6_iptunnel.c. Moreover, it is only
used in the kernel code, as indicated by the "#ifdef __KERNEL__".

Suggested-by: David Miller <davem@davemloft.net>
Signed-off-by: Ioana-Ruxandra Stăncioi <stancioi@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agotcp: apply a floor of 1 for RTT samples from TCP timestamps
Jianfeng Wang [Thu, 30 Jul 2020 23:49:16 +0000 (23:49 +0000)]
tcp: apply a floor of 1 for RTT samples from TCP timestamps

For retransmitted packets, TCP needs to resort to using TCP timestamps
for computing RTT samples. In the common case where the data and ACK
fall in the same 1-millisecond interval, TCP senders with millisecond-
granularity TCP timestamps compute a ca_rtt_us of 0. This ca_rtt_us
of 0 propagates to rs->rtt_us.

This value of 0 can cause performance problems for congestion control
modules. For example, in BBR, the zero min_rtt sample can bring the
min_rtt and BDP estimate down to 0, reduce snd_cwnd and result in a
low throughput. It would be hard to mitigate this with filtering in
the congestion control module, because the proper floor to apply would
depend on the method of RTT sampling (using timestamp options or
internally-saved transmission timestamps).

This fix applies a floor of 1 for the RTT sample delta from TCP
timestamps, so that seq_rtt_us, ca_rtt_us, and rs->rtt_us will be at
least 1 * (USEC_PER_SEC / TCP_TS_HZ).

Note that the receiver RTT computation in tcp_rcv_rtt_measure() and
min_rtt computation in tcp_update_rtt_min() both already apply a floor
of 1 timestamp tick, so this commit makes the code more consistent in
avoiding this edge case of a value of 0.

Signed-off-by: Jianfeng Wang <jfwang@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Kevin Yang <yyd@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agotipc: Use is_broadcast_ether_addr() instead of memcmp()
Huang Guobin [Mon, 3 Aug 2020 02:00:55 +0000 (22:00 -0400)]
tipc: Use is_broadcast_ether_addr() instead of memcmp()

Using is_broadcast_ether_addr() instead of directly use
memcmp() to determine if the ethernet address is broadcast
address.

spatch with a semantic match is used to found this problem.
(http://coccinelle.lip6.fr/)

Signed-off-by: Huang Guobin <huangguobin4@huawei.com>
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'DPAA-FMan-driver-fixes'
David S. Miller [Mon, 3 Aug 2020 23:20:15 +0000 (16:20 -0700)]
Merge branch 'DPAA-FMan-driver-fixes'

Florinel Iordache says:

====================
DPAA FMan driver fixes

Here are several fixes for the DPAA FMan driver.

v2 changes:
* corrected patch 4 by removing the line added by mistake
* used longer fixes tags with the first 12 characters of the SHA-1 ID

v3 changes:
* remove the empty line inserted after fixes tag
====================

Acked-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agofsl/fman: fix eth hash table allocation
Florinel Iordache [Mon, 3 Aug 2020 07:07:34 +0000 (10:07 +0300)]
fsl/fman: fix eth hash table allocation

Fix memory allocation for ethernet address hash table.
The code was wrongly allocating an array for eth hash table which
is incorrect because this is the main structure for eth hash table
(struct eth_hash_t) that contains inside a number of elements.

Fixes: 57ba4c9b56d8 ("fsl/fman: Add FMan MAC support")
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agofsl/fman: check dereferencing null pointer
Florinel Iordache [Mon, 3 Aug 2020 07:07:33 +0000 (10:07 +0300)]
fsl/fman: check dereferencing null pointer

Add a safe check to avoid dereferencing null pointer

Fixes: 57ba4c9b56d8 ("fsl/fman: Add FMan MAC support")
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agofsl/fman: fix unreachable code
Florinel Iordache [Mon, 3 Aug 2020 07:07:32 +0000 (10:07 +0300)]
fsl/fman: fix unreachable code

The parameter 'priority' is incorrectly forced to zero which ultimately
induces logically dead code in the subsequent lines.

Fixes: 57ba4c9b56d8 ("fsl/fman: Add FMan MAC support")
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agofsl/fman: fix dereference null return value
Florinel Iordache [Mon, 3 Aug 2020 07:07:31 +0000 (10:07 +0300)]
fsl/fman: fix dereference null return value

Check before using returned value to avoid dereferencing null pointer.

Fixes: 18a6c85fcc78 ("fsl/fman: Add FMan Port Support")
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agofsl/fman: use 32-bit unsigned integer
Florinel Iordache [Mon, 3 Aug 2020 07:07:30 +0000 (10:07 +0300)]
fsl/fman: use 32-bit unsigned integer

Potentially overflowing expression (ts_freq << 16 and intgr << 16)
declared as type u32 (32-bit unsigned) is evaluated using 32-bit
arithmetic and then used in a context that expects an expression of
type u64 (64-bit unsigned) which ultimately is used as 16-bit
unsigned by typecasting to u16. Fixed by using an unsigned 32-bit
integer since the value is truncated anyway in the end.

Fixes: 414fd46e7762 ("fsl/fman: Add FMan support")
Signed-off-by: Florinel Iordache <florinel.iordache@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf-next
David S. Miller [Mon, 3 Aug 2020 23:03:18 +0000 (16:03 -0700)]
Merge git://git./linux/kernel/git/pablo/nf-next

Pablo Neira Ayuso says:

====================
Netfilter updates for net-next

1) UAF in chain binding support from previous batch, from Dan Carpenter.

2) Queue up delayed work to expire connections with no destination,
   from Andrew Sy Kim.

3) Use fallthrough pseudo-keyword, from Gustavo A. R. Silva.

4) Replace HTTP links with HTTPS, from Alexander A. Klimov.

5) Remove superfluous null header checks in ip6tables, from
   Gaurav Singh.

6) Add extended netlink error reporting for expression.

7) Report EEXIST on overlapping chain, set elements and flowtable
   devices.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: spider_net: Remove a useless memset
Christophe JAILLET [Sun, 2 Aug 2020 13:53:48 +0000 (15:53 +0200)]
net: spider_net: Remove a useless memset

Avoid a memset after a call to 'dma_alloc_coherent()'.
This is useless since
commit 518a2f1925c3 ("dma-mapping: zero memory returned from dma_alloc_*")

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: spider_net: Fix the size used in a 'dma_free_coherent()' call
Christophe JAILLET [Sun, 2 Aug 2020 13:53:33 +0000 (15:53 +0200)]
net: spider_net: Fix the size used in a 'dma_free_coherent()' call

Update the size used in 'dma_free_coherent()' in order to match the one
used in the corresponding 'dma_alloc_coherent()', in
'spider_net_init_chain()'.

Fixes: d4ed8f8d1fb7 ("Spidernet DMA coalescing")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: sgi: ioc3-eth: Fix the size used in some 'dma_free_coherent()' calls
Christophe JAILLET [Sun, 2 Aug 2020 13:52:04 +0000 (15:52 +0200)]
net: sgi: ioc3-eth: Fix the size used in some 'dma_free_coherent()' calls

Update the size used in 'dma_free_coherent()' in order to match the one
used in the corresponding 'dma_alloc_coherent()'.

Fixes: 369a782af0f1 ("net: sgi: ioc3-eth: ensure tx ring is 16k aligned.")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoliquidio: Fix wrong return value in cn23xx_get_pf_num()
Tianjia Zhang [Sun, 2 Aug 2020 11:15:44 +0000 (19:15 +0800)]
liquidio: Fix wrong return value in cn23xx_get_pf_num()

On an error exit path, a negative error code should be returned
instead of a positive return value.

Fixes: 0c45d7fe12c7e ("liquidio: fix use of pf in pass-through mode in a virtual machine")
Cc: Rick Farrington <ricardo.farrington@cavium.com>
Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet/enetc: Fix wrong return value in enetc_psfp_parse_clsflower()
Tianjia Zhang [Sun, 2 Aug 2020 11:15:38 +0000 (19:15 +0800)]
net/enetc: Fix wrong return value in enetc_psfp_parse_clsflower()

In the case of invalid rule, a positive value EINVAL is returned here.
I think this is a typo error. It is necessary to return an error value.

Cc: Po Liu <Po.Liu@nxp.com>
Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ethernet: aquantia: Fix wrong return value
Tianjia Zhang [Sun, 2 Aug 2020 11:15:37 +0000 (19:15 +0800)]
net: ethernet: aquantia: Fix wrong return value

In function hw_atl_a0_hw_multicast_list_set(), when an invalid
request is encountered, a negative error code should be returned.

Fixes: bab6de8fd180b ("net: ethernet: aquantia: Atlantic A0 and B0 specific functions")
Cc: David VomLehn <vomlehn@texas.net>
Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoatm: idt77252: avoid accessing the data mapped to streaming DMA
Jia-Ju Bai [Sun, 2 Aug 2020 09:33:40 +0000 (17:33 +0800)]
atm: idt77252: avoid accessing the data mapped to streaming DMA

In queue_skb(), skb->data is mapped to streaming DMA on line 850:
  dma_map_single(..., skb->data, ...);

Then skb->data is accessed on lines 862 and 863:
  tbd->word_4 = (skb->data[0] << 24) | (skb->data[1] << 16) |
           (skb->data[2] <<  8) | (skb->data[3] <<  0);
and on lines 893 and 894:
  tbd->word_4 = (skb->data[0] << 24) | (skb->data[1] << 16) |
           (skb->data[2] <<  8) | (skb->data[3] <<  0);

These accesses may cause data inconsistency between CPU cache and
hardware.

To fix this problem, the calculation result of skb->data is stored in a
local variable before DMA mapping, and then the driver accesses this
local variable instead of skb->data.

Signed-off-by: Jia-Ju Bai <baijiaju@tsinghua.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoatm: eni: avoid accessing the data mapped to streaming DMA
Jia-Ju Bai [Sun, 2 Aug 2020 09:16:11 +0000 (17:16 +0800)]
atm: eni: avoid accessing the data mapped to streaming DMA

In do_tx(), skb->data is mapped to streaming DMA on line 1111:
  paddr = dma_map_single(...,skb->data,DMA_TO_DEVICE);

Then skb->data is accessed on line 1153:
  (skb->data[3] & 0xf)

This access may cause data inconsistency between CPU cache and hardware.

To fix this problem, skb->data[3] is assigned to a local variable before
DMA mapping, and then the driver accesses this local variable instead of
skb->data[3].

Signed-off-by: Jia-Ju Bai <baijiaju@tsinghua.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: mdio-mvusb: select MDIO_DEVRES in Kconfig
Bartosz Golaszewski [Sun, 2 Aug 2020 07:49:53 +0000 (09:49 +0200)]
net: phy: mdio-mvusb: select MDIO_DEVRES in Kconfig

PHYLIB is not selected by the mvusb driver but it uses mdio devres
helpers. Explicitly select MDIO_DEVRES in this driver's Kconfig entry.

Reported-by: kernel test robot <lkp@intel.com>
Fixes: 1814cff26739 ("net: phy: add a Kconfig option for mdio_devres")
Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoappletalk: Fix atalk_proc_init() return path
Vincent Duvert [Sun, 2 Aug 2020 05:06:51 +0000 (07:06 +0200)]
appletalk: Fix atalk_proc_init() return path

Add a missing return statement to atalk_proc_init so it doesn't return
-ENOMEM when successful.  This allows the appletalk module to load
properly.

Fixes: e2bcd8b0ce6e ("appletalk: use remove_proc_subtree to simplify procfs code")
Link: https://www.downtowndougbrown.com/2020/08/hacking-up-a-fix-for-the-broken-appletalk-kernel-module-in-linux-5-1-and-newer/
Reported-by: Christopher KOBAYASHI <chris@disavowed.jp>
Reported-by: Doug Brown <doug@downtowndougbrown.com>
Signed-off-by: Vincent Duvert <vincent.ldev@duvert.net>
[lukas: add missing tags]
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Cc: stable@vger.kernel.org # v5.1+
Cc: Yue Haibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: qca8k: Add 802.1q VLAN support
Jonathan McDowell [Sat, 1 Aug 2020 17:06:46 +0000 (18:06 +0100)]
net: dsa: qca8k: Add 802.1q VLAN support

This adds full 802.1q VLAN support to the qca8k, allowing the use of
vlan_filtering and more complicated bridging setups than allowed by
basic port VLAN support.

Tested with a number of untagged ports with separate VLANs and then a
trunk port with all the VLANs tagged on it.

v3:
- Pull QCA8K_PORT_VID_DEF changes into separate cleanup patch
- Reverse Christmas tree notation for variable definitions
- Use untagged instead of tagged for consistency
v2:
- Return sensible errnos on failure rather than -1 (rmk)
- Style cleanups based on Florian's feedback
- Silently allow VLAN 0 as device correctly treats this as no tag

Signed-off-by: Jonathan McDowell <noodles@earth.li>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: qca8k: Add define for port VID
Jonathan McDowell [Sat, 1 Aug 2020 17:05:54 +0000 (18:05 +0100)]
net: dsa: qca8k: Add define for port VID

Rather than using a magic value of 1 when configuring the port VIDs add
a QCA8K_PORT_VID_DEF define and use that instead. Also fix up the
bitmask in the process; the top 4 bits are reserved so this wasn't a
problem, but only masking 12 bits is the correct approach.

Signed-off-by: Jonathan McDowell <noodles@earth.li>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Mon, 3 Aug 2020 22:43:59 +0000 (15:43 -0700)]
Merge branch '100GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Tony Nguyen says:

====================
100GbE Intel Wired LAN Driver Updates 2020-08-01

This series contains updates to the ice driver only.

Wei Yongjun marks power management functions with __maybe_unused.

Nick disables VLAN pruning in promiscuous mode and renames grst_delay to
grst_timeout.

Kiran modifies the check for linearization and corrects the vsi_id mask
value.

Vignesh replaces the use of flow profile locks to RSS profile locks for RSS
rule removal. Destroys flow profile lock on clearing XLT table and
clears extraction sequence entries.

Jesse adds some statistics and removes an unreported one.

Brett allows for 2 queue configuration for VFs.

Surabhi adds a check for failed allocation of an extraction sequence
table.

Tony updates the PTYPE lookup table and makes other trivial fixes.

Victor extends profile ID locks to be held until all references are
completed.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: Pass NULL to skb_network_protocol() when we don't care about vlan depth
Miaohe Lin [Sat, 1 Aug 2020 09:36:05 +0000 (17:36 +0800)]
net: Pass NULL to skb_network_protocol() when we don't care about vlan depth

When we don't care about vlan depth, we could pass NULL instead of the
address of a unused local variable to skb_network_protocol() as a param.

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: Use __skb_pagelen() directly in skb_cow_data()
Miaohe Lin [Sat, 1 Aug 2020 09:30:23 +0000 (17:30 +0800)]
net: Use __skb_pagelen() directly in skb_cow_data()

In fact, skb_pagelen() - skb_headlen() is equal to __skb_pagelen(), use it
directly to avoid unnecessary skb_headlen() call.

Also fix the CHECK note of checkpatch.pl:
    Comparison to NULL could be written "!__pskb_pull_tail"

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: qed: use eth_zero_addr() to clear mac address
Miaohe Lin [Sat, 1 Aug 2020 09:14:41 +0000 (17:14 +0800)]
net: qed: use eth_zero_addr() to clear mac address

Use eth_zero_addr() to clear mac address instead of memset().

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: qede: use eth_zero_addr() to clear mac address
Miaohe Lin [Sat, 1 Aug 2020 09:13:54 +0000 (17:13 +0800)]
net: qede: use eth_zero_addr() to clear mac address

Use eth_zero_addr() to clear mac address instead of memset().

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agocxgb4: fix extracting IP addresses in TC-FLOWER rules
Rahul Lakkireddy [Fri, 31 Jul 2020 20:53:40 +0000 (02:23 +0530)]
cxgb4: fix extracting IP addresses in TC-FLOWER rules

commit c8729cac2a11 ("cxgb4: add ethtool n-tuple filter insertion")
has removed checking control key for determining IP address types
for TC-FLOWER rules, which causes all the rules being inserted to
hardware to become IPv6 rule type always. So, add back the check
to select the correct IP address type to extract and hence fix the
correct rule type being inserted to hardware.

Also, ethtool_rx_flow_key doesn't have any control key and instead
directly sets the IPv4/IPv6 address keys. So, explicitly set the
IP address type for ethtool n-tuple filters to reuse the same code.

Fixes: c8729cac2a11 ("cxgb4: add ethtool n-tuple filter insertion")
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agocxgb4: fix check for running offline ethtool selftest
Rahul Lakkireddy [Fri, 31 Jul 2020 20:50:52 +0000 (02:20 +0530)]
cxgb4: fix check for running offline ethtool selftest

The flag indicating the selftest to run is a bitmask. So, fix the
check. Also, the selftests will fail if adapter initialization has
not been completed yet. So, add appropriate check and bail sooner.

Fixes: 7235ffae3d2c ("cxgb4: add loopback ethtool self-test")
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'ionic-txrx-updates'
David S. Miller [Mon, 3 Aug 2020 22:32:02 +0000 (15:32 -0700)]
Merge branch 'ionic-txrx-updates'

Shannon Nelson says:

====================
ionic txrx updates

These are a few patches to do some cleanup in the packet
handling and give us more flexibility in tuning performance
by allowing us to put Tx handling on separate interrupts
when it makes sense for particular traffic loads.

v3: simplified queue count change logging, removed unnecessary
    check for no count change
v2: dropped the original patch 2 for ringsize change
    changed the separated tx/rx interrupts to use ethtool -L
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoionic: separate interrupt for Tx and Rx
Shannon Nelson [Fri, 31 Jul 2020 20:15:36 +0000 (13:15 -0700)]
ionic: separate interrupt for Tx and Rx

Add the capability to split the Tx queues onto their own
interrupts with their own napi contexts.  This gives the
opportunity for more direct control of Tx interrupt
handling, such as CPU affinity and interrupt coalescing,
useful for some traffic loads.

v2: use ethtool -L, not a vendor specific priv-flag
v3: simplify logging, drop unnecessary "no-change" tests

Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoionic: tx separate servicing
Shannon Nelson [Fri, 31 Jul 2020 20:15:35 +0000 (13:15 -0700)]
ionic: tx separate servicing

We give the tx clean path its own budget and service routine in
order to give a little more leeway to be more aggressive, and
in preparation for coming changes.  We've found this gives us
a little better performance in some packet processing scenarios
without hurting other scenarios.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoionic: use fewer firmware doorbells on rx fill
Shannon Nelson [Fri, 31 Jul 2020 20:15:34 +0000 (13:15 -0700)]
ionic: use fewer firmware doorbells on rx fill

We really don't need to hit the Rx queue doorbell so many times,
we can wait to the end and cause a little less thrash.

Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: gre: recompute gre csum for sctp over gre tunnels
Lorenzo Bianconi [Fri, 31 Jul 2020 18:12:05 +0000 (20:12 +0200)]
net: gre: recompute gre csum for sctp over gre tunnels

The GRE tunnel can be used to transport traffic that does not rely on a
Internet checksum (e.g. SCTP). The issue can be triggered creating a GRE
or GRETAP tunnel and transmitting SCTP traffic ontop of it where CRC
offload has been disabled. In order to fix the issue we need to
recompute the GRE csum in gre_gso_segment() not relying on the inner
checksum.
The issue is still present when we have the CRC offload enabled.
In this case we need to disable the CRC offload if we require GRE
checksum since otherwise skb_checksum() will report a wrong value.

Fixes: 90017accff61 ("sctp: Add GSO support")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Reviewed-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: bridge: clear bridge's private skb space on xmit
Nikolay Aleksandrov [Fri, 31 Jul 2020 16:26:16 +0000 (19:26 +0300)]
net: bridge: clear bridge's private skb space on xmit

We need to clear all of the bridge private skb variables as they can be
stale due to the packet being recirculated through the stack and then
transmitted through the bridge device. Similar memset is already done on
bridge's input. We've seen cases where proxyarp_replied was 1 on routed
multicast packets transmitted through the bridge to ports with neigh
suppress which were getting dropped. Same thing can in theory happen with
the port isolation bit as well.

Fixes: 821f1b21cabb ("bridge: add new BR_NEIGH_SUPPRESS port flag to suppress arp and nd flood")
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoipv6/addrconf: use a boolean to choose between UNREGISTER/DOWN
Florent Fourcot [Fri, 31 Jul 2020 13:32:07 +0000 (15:32 +0200)]
ipv6/addrconf: use a boolean to choose between UNREGISTER/DOWN

"how" was used as a boolean. Change the type to bool, and improve
variable name

Signed-off-by: Florent Fourcot <florent.fourcot@wifirst.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoipv6/addrconf: call addrconf_ifdown with consistent values
Florent Fourcot [Fri, 31 Jul 2020 13:32:06 +0000 (15:32 +0200)]
ipv6/addrconf: call addrconf_ifdown with consistent values

Second parameter of addrconf_ifdown "how" is used as a boolean
internally. It does not make sense to call it with something different
of 0 or 1.

This value is set to 2 in all git history.

Signed-off-by: Florent Fourcot <florent.fourcot@wifirst.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'net-openvswitch-masks-cache-enhancements'
David S. Miller [Mon, 3 Aug 2020 22:17:48 +0000 (15:17 -0700)]
Merge branch 'net-openvswitch-masks-cache-enhancements'

Eelco Chaudron says:

====================
net: openvswitch: masks cache enhancements

This patchset adds two enhancements to the Open vSwitch masks cache.

Changes in v4 [patch 2/2 only]:
 - Remove null check before calling free_percpu()
 - Make ovs_dp_change() return appropriate error codes

Changes in v3 [patch 2/2 only]:
 - Use is_power_of_2() function
 - Use array_size() function
 - Fix remaining sparse errors

Changes in v2 [patch 2/2 only]:
 - Fix sparse warnings
 - Fix netlink policy items reported by Florian Westphal
====================

Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: openvswitch: make masks cache size configurable
Eelco Chaudron [Fri, 31 Jul 2020 12:21:34 +0000 (14:21 +0200)]
net: openvswitch: make masks cache size configurable

This patch makes the masks cache size configurable, or with
a size of 0, disable it.

Reviewed-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: openvswitch: add masks cache hit counter
Eelco Chaudron [Fri, 31 Jul 2020 12:20:56 +0000 (14:20 +0200)]
net: openvswitch: add masks cache hit counter

Add a counter that counts the number of masks cache hits, and
export it through the megaflow netlink statistics.

Reviewed-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: mvpp2: fix memory leak in mvpp2_rx
Lorenzo Bianconi [Fri, 31 Jul 2020 08:38:32 +0000 (10:38 +0200)]
net: mvpp2: fix memory leak in mvpp2_rx

Release skb memory in mvpp2_rx() if mvpp2_rx_refill routine fails

Fixes: b5015854674b ("net: mvpp2: fix refilling BM pools in RX path")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Acked-by: Matteo Croce <mcroce@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoethtool: ethnl_set_linkmodes: remove redundant null check
Gaurav Singh [Fri, 31 Jul 2020 04:58:44 +0000 (00:58 -0400)]
ethtool: ethnl_set_linkmodes: remove redundant null check

info cannot be NULL here since its being accessed earlier
in the function: nlmsg_parse(info->nlhdr...). Remove this
redundant NULL check.

Signed-off-by: Gaurav Singh <gaurav1086@gmail.com>
Reviewed-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoopenvswitch: Prevent kernel-infoleak in ovs_ct_put_key()
Peilin Ye [Fri, 31 Jul 2020 04:48:38 +0000 (00:48 -0400)]
openvswitch: Prevent kernel-infoleak in ovs_ct_put_key()

ovs_ct_put_key() is potentially copying uninitialized kernel stack memory
into socket buffers, since the compiler may leave a 3-byte hole at the end
of `struct ovs_key_ct_tuple_ipv4` and `struct ovs_key_ct_tuple_ipv6`. Fix
it by initializing `orig` with memset().

Fixes: 9dd7f8907c37 ("openvswitch: Add original direction conntrack tuple to sw_flow_key.")
Suggested-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Peilin Ye <yepeilin.cs@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet/sched: act_ct: fix miss set mru for ovs after defrag in act_ct
wenxu [Fri, 31 Jul 2020 02:45:01 +0000 (10:45 +0800)]
net/sched: act_ct: fix miss set mru for ovs after defrag in act_ct

When openvswitch conntrack offload with act_ct action. Fragment packets
defrag in the ingress tc act_ct action and miss the next chain. Then the
packet pass to the openvswitch datapath without the mru. The over
mtu packet will be dropped in output action in openvswitch for over mtu.

"kernel: net2: dropped over-mtu packet: 1528 > 1500"

This patch add mru in the tc_skb_ext for adefrag and miss next chain
situation. And also add mru in the qdisc_skb_cb. The act_ct set the mru
to the qdisc_skb_cb when the packet defrag. And When the chain miss,
The mru is set to tc_skb_ext which can be got by ovs datapath.

Fixes: b57dc7c13ea9 ("net/sched: Introduce action ct")
Signed-off-by: wenxu <wenxu@ucloud.cn>
Reviewed-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'Improve-MDIO-Ethernet-PHY-reset'
David S. Miller [Mon, 3 Aug 2020 22:01:02 +0000 (15:01 -0700)]
Merge branch 'Improve-MDIO-Ethernet-PHY-reset'

Bruno Thomsen says:

====================
Improve MDIO Ethernet PHY reset

This patch series is a result of trying to upstream a new device
tree for a TQMa7D based board[1][2]. Initial this DTS used some
deprecated PHY reset properties on the FEC device; NXP Ethernet
MAC also known as Freescale Fast Ethernet Controller.

When switching from FEC properties[3]:
"phy-reset-gpios"
"phy-reset-duration"
"phy-reset-post-delay"

To MDIO PHY properties[4]:
"reset-gpios"
"reset-assert-us"
"reset-deassert-us"

The result was that no Ethernet PHY device was detected on boot.

This issue could be worked around by disabling PHY type ID auto-
detection by using "ethernet-phy-id0022.1560" as compatible
string and not "ethernet-phy-ieee802.3-c22".

Upstreaming a DTS with this workaround was not accepted, so I
digged into the MDIO reset flow and found that it had a few
missing parts compared to the deprecated FEC reset function.
After some more testing and logic analyzer traces it was
revealed that the failed PHY communication was due to missing
initial device reset.

I was suggested[5] in a earlier mail thread to use MDIO bus
reset as that was performed before auto-detection, but current
device tree binding was limited to reset assert in usec.
Microchip/Micrel Ethernet PHYs recommended reset circuit[8],
figure 7-12, is a little "slow" after reset deassert as that
is left to a RC circuit with a tau of ~100ms; using a 10k PU
resistor together with a 10uF decoupling capacitor. The diode
in serie of the reset signal converts the GPIO push-pull output
into a open-drain output. So a post reset delay in the range
of 500-1000ms is needed, depending on component tolerances
and general hardware design margins.

In the first version of this patch series[6] I reused the
"reset-delay-us" property for reset deassert in usec as that
would cause 50/50% duty-cycle, but that would always apply.
The solution in this patch series is to add a new MDIO bus
property, so post reset delay is optional and configured
separately.

MDIO bus properties[7]:
"reset-delay-us"
"reset-post-delay-us" (new)

I have not marked this with "Fixes:" as no single commit is the
cause and historically this code has only supported MDIO devices
that need reset after auto-detection. The patch series also uses
a new flexible sleep helper function that was introduced in
5.8-rc1, so the driver uses the optimal sleep function depending
on value loaded from device tree.

Future work in this area could add new properties on the MDIO
device, so reset points are configurable, e.g. no reset,
before/after auto-detection or both.

[1] https://lore.kernel.org/linux-devicetree/20200629114927.17379-2-bruno.thomsen@gmail.com/
[2] https://lore.kernel.org/linux-devicetree/20200716172611.5349-2-bruno.thomsen@gmail.com/
[3] https://elixir.bootlin.com/linux/v5.7.8/source/Documentation/devicetree/bindings/net/fsl-fec.txt#L44
[4] https://elixir.bootlin.com/linux/v5.8-rc4/source/Documentation/devicetree/bindings/net/mdio.yaml#L78
[5] https://lore.kernel.org/netdev/CAOMZO5DtYDomD8FDCZDwYCSr2AwNT81Ay4==aDxXyBxtyvPiJA@mail.gmail.com/
[6] https://lore.kernel.org/netdev/20200728090203.17313-1-bruno.thomsen@gmail.com/
[7] https://elixir.bootlin.com/linux/v5.8-rc4/source/Documentation/devicetree/bindings/net/mdio.yaml#L36
[8] http://ww1.microchip.com/downloads/en/DeviceDoc/00002202C.pdf
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: mdio device: use flexible sleeping in reset function
Bruno Thomsen [Thu, 30 Jul 2020 19:57:49 +0000 (21:57 +0200)]
net: mdio device: use flexible sleeping in reset function

MDIO device reset assert and deassert length was created by
usleep_range() but that does not ensure optimal handling of
all the different values from device tree properties.
By switching to the new flexible sleeping helper function,
fsleep(), the correct delay function is called depending on
delay length, e.g. udelay(), usleep_range() or msleep().

Signed-off-by: Bruno Thomsen <bruno.thomsen@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: mdiobus: add reset-post-delay-us handling
Bruno Thomsen [Thu, 30 Jul 2020 19:57:48 +0000 (21:57 +0200)]
net: mdiobus: add reset-post-delay-us handling

Load new "reset-post-delay-us" value from MDIO properties,
and if configured to a greater then zero delay do a
flexible sleeping delay after MDIO bus reset deassert.
This allows devices to exit reset state before start
bus communication.

Signed-off-by: Bruno Thomsen <bruno.thomsen@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: mdiobus: use flexible sleeping for reset-delay-us
Bruno Thomsen [Thu, 30 Jul 2020 19:57:47 +0000 (21:57 +0200)]
net: mdiobus: use flexible sleeping for reset-delay-us

MDIO bus reset pulse width is created by using udelay()
and that function might not be optimal depending on
device tree value. By switching to the new fsleep() helper
the correct delay function is called depending on
delay length, e.g. udelay(), usleep_range() or msleep().

Signed-off-by: Bruno Thomsen <bruno.thomsen@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agodt-bindings: net: mdio: add reset-post-delay-us property
Bruno Thomsen [Thu, 30 Jul 2020 19:57:46 +0000 (21:57 +0200)]
dt-bindings: net: mdio: add reset-post-delay-us property

Add "reset-post-delay-us" parameter to MDIO bus properties,
so it's possible to add a delay after reset deassert.
This is optional in case external hardware slows down
release of the reset signal.

Signed-off-by: Bruno Thomsen <bruno.thomsen@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomac80211: Do not report beacon loss if beacon filtering enabled
Loic Poulain [Thu, 18 Jun 2020 09:17:42 +0000 (11:17 +0200)]
mac80211: Do not report beacon loss if beacon filtering enabled

mac80211.h says: Beacon filter support is advertised with the
IEEE80211_VIF_BEACON_FILTER interface capability. The driver needs to
enable beacon filter support whenever power save is enabled, that is
IEEE80211_CONF_PS is set. When power save is enabled, the stack will
not check for beacon loss and the driver needs to notify about loss
of beacons with ieee80211_beacon_loss().

Some controllers may want to dynamically enable the beacon filter
capabilities on power save entry (CONF_PS) and disable it on exit.
This is the case for the wcn36xx driver which only supports beacon
filtering in PS mode (no CONNECTION_MONITOR support).

When the mac80211 beacon monitor timer expires, the beacon filter
flag must be checked again in case it as been changed in between
(e.g. vif moved to PS mode).

Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Link: https://lore.kernel.org/r/1592471863-31402-1-git-send-email-loic.poulain@linaro.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac80211: fix misplaced while instead of if
Johannes Berg [Mon, 3 Aug 2020 09:02:10 +0000 (11:02 +0200)]
mac80211: fix misplaced while instead of if

This never was intended to be a 'while' loop, it should've
just been an 'if' instead of 'while'. Fix this.

I noticed this while applying another patch from Ben that
intended to fix a busy loop at this spot.

Cc: stable@vger.kernel.org
Fixes: b16798f5b907 ("mac80211: mark station unauthorized before key removal")
Reported-by: Ben Greear <greearb@candelatech.com>
Link: https://lore.kernel.org/r/20200803110209.253009ae41ff.I3522aad099392b31d5cf2dcca34cbac7e5832dde@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agonl80211: use eth_zero_addr() to clear mac address
Miaohe Lin [Sat, 1 Aug 2020 09:15:49 +0000 (17:15 +0800)]
nl80211: use eth_zero_addr() to clear mac address

Use eth_zero_addr() to clear mac address instead of memset().

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Link: https://lore.kernel.org/r/1596273349-24333-1-git-send-email-linmiaohe@huawei.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac80211: use eth_zero_addr() to clear mac address
Miaohe Lin [Sat, 1 Aug 2020 09:12:38 +0000 (17:12 +0800)]
mac80211: use eth_zero_addr() to clear mac address

Use eth_zero_addr() to clear mac address instead of memset().

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Link: https://lore.kernel.org/r/1596273158-24183-1-git-send-email-linmiaohe@huawei.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac8211: fix struct initialisation
John Crispin [Mon, 3 Aug 2020 08:45:40 +0000 (10:45 +0200)]
mac8211: fix struct initialisation

Sparse showed up with the following error.
net/mac80211/agg-rx.c:480:43: warning: Using plain integer as NULL pointer

Fixes: 2ab45876756f (mac80211: add support for the ADDBA extension element)
Signed-off-by: John Crispin <john@phrozen.org>
Link: https://lore.kernel.org/r/20200803084540.179908-1-john@phrozen.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac80211: Handle special status codes in SAE commit
Jouni Malinen [Fri, 31 Jul 2020 18:38:30 +0000 (21:38 +0300)]
mac80211: Handle special status codes in SAE commit

SAE authentication has been extended with H2E (IEEE 802.11 REVmd) and PK
(WFA) options. Those extensions use special status code values in the
SAE commit messages (Authentication frame with transaction sequence
number 1) to identify which extension is in use. mac80211 was
interpreting those new values as the AP denying authentication and that
resulted in failure to complete SAE authentication in some cases.

Fix this by adding exceptions for the new status code values 126 and
127.

Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Link: https://lore.kernel.org/r/20200731183830.18735-1-jouni@codeaurora.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agonetfilter: nf_tables: report EEXIST on overlaps
Pablo Neira Ayuso [Sun, 2 Aug 2020 01:05:25 +0000 (03:05 +0200)]
netfilter: nf_tables: report EEXIST on overlaps

Replace EBUSY by EEXIST in the following cases:

- If the user adds a chain with a different configuration such as different
  type, hook and priority.

- If the user adds a non-base chain that clashes with an existing basechain.

- If the user adds a { key : value } mapping element and the key exists
  but the value differs.

- If the device already belongs to an existing flowtable.

User describe that this error reporting is confusing:

- https://bugzilla.netfilter.org/show_bug.cgi?id=1176
- https://bugzilla.netfilter.org/show_bug.cgi?id=1413

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
4 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
David S. Miller [Sun, 2 Aug 2020 08:02:12 +0000 (01:02 -0700)]
Merge git://git./linux/kernel/git/netdev/net

Resolved kernel/bpf/btf.c using instructions from merge commit
69138b34a7248d2396ab85c8652e20c0c39beaba

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonetfilter: nf_tables: extended netlink error reporting for expressions
Pablo Neira Ayuso [Sun, 2 Aug 2020 00:56:37 +0000 (02:56 +0200)]
netfilter: nf_tables: extended netlink error reporting for expressions

This patch extends 36dd1bcc07e5 ("netfilter: nf_tables: initial support
for extended ACK reporting") to include netlink extended error reporting
for expressions. This allows userspace to identify what rule expression
is triggering the error.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
4 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Sat, 1 Aug 2020 23:47:24 +0000 (16:47 -0700)]
Merge git://git./linux/kernel/git/netdev/net

Pull networking fixes from David Miller:

 1) Encap offset calculation is incorrect in esp6, from Sabrina Dubroca.

 2) Better parameter validation in pfkey_dump(), from Mark Salyzyn.

 3) Fix several clang issues on powerpc in selftests, from Tanner Love.

 4) cmsghdr_from_user_compat_to_kern() uses the wrong length, from Al
    Viro.

 5) Out of bounds access in mlx5e driver, from Raed Salem.

 6) Fix transfer buffer memleak in lan78xx, from Johan Havold.

 7) RCU fixups in rhashtable, from Herbert Xu.

 8) Fix ipv6 nexthop refcnt leak, from Xiyu Yang.

 9) vxlan FDB dump must be done under RCU, from Ido Schimmel.

10) Fix use after free in mlxsw, from Ido Schimmel.

11) Fix map leak in HASH_OF_MAPS bpf code, from Andrii Nakryiko.

12) Fix bug in mac80211 Tx ack status reporting, from Vasanthakumar
    Thiagarajan.

13) Fix memory leaks in IPV6_ADDRFORM code, from Cong Wang.

14) Fix bpf program reference count leaks in mlx5 during
    mlx5e_alloc_rq(), from Xin Xiong.

* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (86 commits)
  vxlan: fix memleak of fdb
  rds: Prevent kernel-infoleak in rds_notify_queue_get()
  net/sched: The error lable position is corrected in ct_init_module
  net/mlx5e: fix bpf_prog reference count leaks in mlx5e_alloc_rq
  net/mlx5e: E-Switch, Specify flow_source for rule with no in_port
  net/mlx5e: E-Switch, Add misc bit when misc fields changed for mirroring
  net/mlx5e: CT: Support restore ipv6 tunnel
  net: gemini: Fix missing clk_disable_unprepare() in error path of gemini_ethernet_port_probe()
  ionic: unlock queue mutex in error path
  atm: fix atm_dev refcnt leaks in atmtcp_remove_persistent
  net: ethernet: mtk_eth_soc: fix MTU warnings
  net: nixge: fix potential memory leak in nixge_probe()
  devlink: ignore -EOPNOTSUPP errors on dumpit
  rxrpc: Fix race between recvmsg and sendmsg on immediate call failure
  MAINTAINERS: Replace Thor Thayer as Altera Triple Speed Ethernet maintainer
  selftests/bpf: fix netdevsim trap_flow_action_cookie read
  ipv6: fix memory leaks on IPV6_ADDRFORM path
  net/bpfilter: Initialize pos in __bpfilter_process_sockopt
  igb: reinit_locked() should be called with rtnl_lock
  e1000e: continue to init PHY even when failed to disable ULP
  ...

4 years agoMerge tag 'for-linus-2020-08-01' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 1 Aug 2020 23:40:59 +0000 (16:40 -0700)]
Merge tag 'for-linus-2020-08-01' of git://git./linux/kernel/git/brauner/linux

Pull thread fix from Christian Brauner:
 "A simple spelling fix for dequeue_synchronous_signal()"

* tag 'for-linus-2020-08-01' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux:
  signal: fix typo in dequeue_synchronous_signal()

4 years agoMerge tag 'perf-tools-fixes-2020-08-01' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 1 Aug 2020 20:08:50 +0000 (13:08 -0700)]
Merge tag 'perf-tools-fixes-2020-08-01' of git://git./linux/kernel/git/acme/linux

Pull perf tooling fixes from Arnaldo Carvalho de Melo:

 - Fix libtraceevent build with binutils 2.35

 - Fix memory leak in process_dynamic_array_len in libtraceevent

 - Fix 'perf test 68' zstd compression for s390

 - Fix record failure when mixed with ARM SPE event

* tag 'perf-tools-fixes-2020-08-01' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux:
  libtraceevent: Fix build with binutils 2.35
  perf tools: Fix record failure when mixed with ARM SPE event
  perf tests: Fix test 68 zstd compression for s390
  tools lib traceevent: Fix memory leak in process_dynamic_array_len

4 years agomptcp: fix syncookie build error on UP
Florian Westphal [Sat, 1 Aug 2020 14:39:59 +0000 (16:39 +0200)]
mptcp: fix syncookie build error on UP

kernel test robot says:
net/mptcp/syncookies.c: In function 'mptcp_join_cookie_init':
include/linux/kernel.h:47:38: warning: division by zero [-Wdiv-by-zero]
 #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))

I forgot that spinock_t size is 0 on UP, so ARRAY_SIZE cannot be used.

Fixes: 9466a1ccebbe54 ("mptcp: enable JOIN requests even if cookies are in use")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agovxlan: fix memleak of fdb
Taehee Yoo [Sat, 1 Aug 2020 07:07:50 +0000 (07:07 +0000)]
vxlan: fix memleak of fdb

When vxlan interface is deleted, all fdbs are deleted by vxlan_flush().
vxlan_flush() flushes fdbs but it doesn't delete fdb, which contains
all-zeros-mac because it is deleted by vxlan_uninit().
But vxlan_uninit() deletes only the fdb, which contains both all-zeros-mac
and default vni.
So, the fdb, which contains both all-zeros-mac and non-default vni
will not be deleted.

Test commands:
    ip link add vxlan0 type vxlan dstport 4789 external
    ip link set vxlan0 up
    bridge fdb add to 00:00:00:00:00:00 dst 172.0.0.1 dev vxlan0 via lo \
    src_vni 10000 self permanent
    ip link del vxlan0

kmemleak reports as follows:
unreferenced object 0xffff9486b25ced88 (size 96):
  comm "bridge", pid 2151, jiffies 4294701712 (age 35506.901s)
  hex dump (first 32 bytes):
    02 00 00 00 ac 00 00 01 40 00 09 b1 86 94 ff ff  ........@.......
    46 02 00 00 00 00 00 00 a7 03 00 00 12 b5 6a 6b  F.............jk
  backtrace:
    [<00000000c10cf651>] vxlan_fdb_append.part.51+0x3c/0xf0 [vxlan]
    [<000000006b31a8d9>] vxlan_fdb_create+0x184/0x1a0 [vxlan]
    [<0000000049399045>] vxlan_fdb_update+0x12f/0x220 [vxlan]
    [<0000000090b1ef00>] vxlan_fdb_add+0x12a/0x1b0 [vxlan]
    [<0000000056633c2c>] rtnl_fdb_add+0x187/0x270
    [<00000000dd5dfb6b>] rtnetlink_rcv_msg+0x264/0x490
    [<00000000fc44dd54>] netlink_rcv_skb+0x4a/0x110
    [<00000000dff433e7>] netlink_unicast+0x18e/0x250
    [<00000000b87fb421>] netlink_sendmsg+0x2e9/0x400
    [<000000002ed55153>] ____sys_sendmsg+0x237/0x260
    [<00000000faa51c66>] ___sys_sendmsg+0x88/0xd0
    [<000000006c3982f1>] __sys_sendmsg+0x4e/0x80
    [<00000000a8f875d2>] do_syscall_64+0x56/0xe0
    [<000000003610eefa>] entry_SYSCALL_64_after_hwframe+0x44/0xa9
unreferenced object 0xffff9486b1c40080 (size 128):
  comm "bridge", pid 2157, jiffies 4294701754 (age 35506.866s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 f8 dc 42 b2 86 94 ff ff  ..........B.....
    6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b  kkkkkkkkkkkkkkkk
  backtrace:
    [<00000000a2981b60>] vxlan_fdb_create+0x67/0x1a0 [vxlan]
    [<0000000049399045>] vxlan_fdb_update+0x12f/0x220 [vxlan]
    [<0000000090b1ef00>] vxlan_fdb_add+0x12a/0x1b0 [vxlan]
    [<0000000056633c2c>] rtnl_fdb_add+0x187/0x270
    [<00000000dd5dfb6b>] rtnetlink_rcv_msg+0x264/0x490
    [<00000000fc44dd54>] netlink_rcv_skb+0x4a/0x110
    [<00000000dff433e7>] netlink_unicast+0x18e/0x250
    [<00000000b87fb421>] netlink_sendmsg+0x2e9/0x400
    [<000000002ed55153>] ____sys_sendmsg+0x237/0x260
    [<00000000faa51c66>] ___sys_sendmsg+0x88/0xd0
    [<000000006c3982f1>] __sys_sendmsg+0x4e/0x80
    [<00000000a8f875d2>] do_syscall_64+0x56/0xe0
    [<000000003610eefa>] entry_SYSCALL_64_after_hwframe+0x44/0xa9

Fixes: 3ad7a4b141eb ("vxlan: support fdb and learning in COLLECT_METADATA mode")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Acked-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agofib: fix another fib_rules_ops indirect call wrapper problem
Brian Vazquez [Sat, 1 Aug 2020 03:01:10 +0000 (20:01 -0700)]
fib: fix another fib_rules_ops indirect call wrapper problem

It turns out that on commit 41d707b7332f ("fib: fix fib_rules_ops
indirect calls wrappers") I forgot to include the case when
CONFIG_IP_MULTIPLE_TABLES is not set.

Fixes: 41d707b7332f ("fib: fix fib_rules_ops indirect calls wrappers")
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Brian Vazquez <brianvv@google.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agotcp: fix build fong CONFIG_MPTCP=n
Eric Dumazet [Sat, 1 Aug 2020 02:09:29 +0000 (19:09 -0700)]
tcp: fix build fong CONFIG_MPTCP=n

Fixes these errors:

net/ipv4/syncookies.c: In function 'tcp_get_cookie_sock':
net/ipv4/syncookies.c:216:19: error: 'struct tcp_request_sock' has no
member named 'drop_req'
  216 |   if (tcp_rsk(req)->drop_req) {
      |                   ^~
net/ipv4/syncookies.c: In function 'cookie_tcp_reqsk_alloc':
net/ipv4/syncookies.c:289:27: warning: unused variable 'treq'
[-Wunused-variable]
  289 |  struct tcp_request_sock *treq;
      |                           ^~~~
make[3]: *** [scripts/Makefile.build:280: net/ipv4/syncookies.o] Error 1
make[3]: *** Waiting for unfinished jobs....

Fixes: 9466a1ccebbe ("mptcp: enable JOIN requests even if cookies are in use")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Acked-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge tag 'pinctrl-v5.8-4' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Sat, 1 Aug 2020 17:11:42 +0000 (10:11 -0700)]
Merge tag 'pinctrl-v5.8-4' of git://git./linux/kernel/git/linusw/linux-pinctrl

Pull pin control fix from Linus Walleij:
 "A single last minute pin control fix to the Qualcomm driver fixing
  missing dual edge PCH interrupts"

* tag 'pinctrl-v5.8-4' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
  pinctrl: qcom: Handle broken/missing PDC dual edge IRQs on sc7180

4 years agoice: Misc minor fixes
Tony Nguyen [Thu, 30 Jul 2020 00:19:22 +0000 (17:19 -0700)]
ice: Misc minor fixes

This is a collection of minor fixes including typos, white space, and
style. No functional changes.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
4 years agoice: adjust profile ID map locks
Victor Raj [Thu, 30 Jul 2020 00:19:21 +0000 (17:19 -0700)]
ice: adjust profile ID map locks

The profile ID map lock should be held till the caller completes
all references of that profile entries.

The current code releases the lock right after the match search.
This caused a driver issue when the profile map entries were
referenced after it was freed in other thread after the lock was
released earlier.

Signed-off-by: Victor Raj <victor.raj@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoice: update PTYPE lookup table
Tony Nguyen [Thu, 30 Jul 2020 00:19:20 +0000 (17:19 -0700)]
ice: update PTYPE lookup table

Update the PTYPE lookup table to reflect values that can be set by the
hardware.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
4 years agoice: Disable VLAN pruning in promiscuous mode
Nick Nunley [Thu, 30 Jul 2020 00:19:19 +0000 (17:19 -0700)]
ice: Disable VLAN pruning in promiscuous mode

Disable VLAN pruning when entering promiscuous mode, and re-enable it
when exiting.

Without this VLAN-over-bridge topologies created on the device won't be
functional unless rx-vlan-filter is explicitly disabled with ethtool.

Signed-off-by: Nick Nunley <nicholas.d.nunley@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoice: Graceful error handling in HW table calloc failure
Surabhi Boob [Thu, 30 Jul 2020 00:19:18 +0000 (17:19 -0700)]
ice: Graceful error handling in HW table calloc failure

In the ice_init_hw_tbls, if the devm_kcalloc for es->written fails, catch
that error and bail out gracefully, instead of continuing with a NULL
pointer.

Fixes: 32d63fa1e9f3 ("ice: Initialize DDP package structures")
Signed-off-by: Surabhi Boob <surabhi.boob@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoice: port fix for chk_linearlize
Kiran Patil [Thu, 30 Jul 2020 00:19:17 +0000 (17:19 -0700)]
ice: port fix for chk_linearlize

This is a port of commit 248de22e638f ("i40e/i40evf: Account for frags
split over multiple descriptors in check linearize")

As part of testing workloads (read/write) using larger IO size (128K)
tx_timeout is observed and whenever it happens, it was due to
tx_linearize.

Signed-off-by: Kiran Patil <kiran.patil@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoice: Allow 2 queue pairs per VF on SR-IOV initialization
Brett Creeley [Thu, 30 Jul 2020 00:19:16 +0000 (17:19 -0700)]
ice: Allow 2 queue pairs per VF on SR-IOV initialization

Currently VFs are only allowed to get 16, 4, and 1 queue pair by
default, which require 17, 5, and 2 MSI-X vectors respectively. This
is because each VF needs a MSI-X per data queue and a MSI-X for its
other interrupt. The calculation is based on the number of VFs created,
MSI-X available, and queue pairs available at the time of VF creation.

Unfortunately the values above exclude 2 queue pairs when only 3 MSI-X
are available to each VF based on resource constraints. The current
calculation would default to 2 MSI-X and 1 queue pair. This is a waste
of resources, so fix this by allowing 2 queue pairs per VF when there
are between 2 and 5 MSI-X available per VF.

Signed-off-by: Brett Creeley <brett.creeley@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoice: Clear and free XLT entries on reset
Vignesh Sridhar [Thu, 30 Jul 2020 00:19:15 +0000 (17:19 -0700)]
ice: Clear and free XLT entries on reset

This fix has been added to address memory leak issues resulting from
triggering a sudden driver reset which does not allow us to follow our
normal removal flows for SW XLT entries for advanced features.

- Adding call to destroy flow profile locks when clearing SW XLT tables.

- Extraction sequence entries were not correctly cleared previously
which could cause ownership conflicts for repeated reset-replay calls.

Fixes: 31ad4e4ee1e4 ("ice: Allocate flow profile")
Signed-off-by: Vignesh Sridhar <vignesh.sridhar@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoice: add useful statistics
Jesse Brandeburg [Thu, 30 Jul 2020 00:19:14 +0000 (17:19 -0700)]
ice: add useful statistics

Display and count some useful hot-path statistics. The usefulness is as
follows:

- tx_restart: use to determine if the transmit ring size is too small or
  if the transmit interrupt rate is too low.
- rx_gro_dropped: use to count drops from GRO layer, which previously were
  completely uncounted when occurring.
- tx_busy: use to determine when the driver is miscounting number of
  descriptors needed for an skb.
- tx_timeout: as our other drivers, count the number of times we've reset
  due to timeout because the kernel only prints a warning once per netdev.

Several of these were already counted but not displayed.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoice: remove page_reuse statistic
Jesse Brandeburg [Thu, 30 Jul 2020 00:19:13 +0000 (17:19 -0700)]
ice: remove page_reuse statistic

The page reuse statistic wasn't even being displayed to the user, even
though the driver counted it. Don't waste the struct space and hot-path
cycles since the driver doesn't display it.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoice: Fix RSS profile locks
Vignesh Sridhar [Thu, 30 Jul 2020 00:19:12 +0000 (17:19 -0700)]
ice: Fix RSS profile locks

Replacing flow profile locks with RSS profile locks in the function to
remove all RSS rules for a given VSI. This is to align the locks used
for RSS rule addition to VSI and removal during VSI teardown to avoid
a race condition owing to several iterations of the above operations.
In function to get RSS rules for given VSI and protocol header replacing
the pointer reference of the RSS entry with a copy of hash value to
ensure thread safety.

Signed-off-by: Vignesh Sridhar <vignesh.sridhar@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoice: fix the vsi_id mask to be 10 bit for set_rss_lut
Kiran Patil [Thu, 30 Jul 2020 00:19:11 +0000 (17:19 -0700)]
ice: fix the vsi_id mask to be 10 bit for set_rss_lut

set_rss_lut can fail due to incorrect vsi_id mask. vsi_id is 10 bit
but mask was 0x1FF whereas it should be 0x3FF.

For vsi_num >= 512, FW set_rss_lut can fail with return code
EACCESS (VSI ownership issue) because software was providing
incorrect vsi_num (dropping 10th bit due to incorrect mask) for
set_rss_lut admin command

Signed-off-by: Kiran Patil <kiran.patil@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoice: rename misleading grst_delay variable
Nick Nunley [Thu, 30 Jul 2020 00:19:10 +0000 (17:19 -0700)]
ice: rename misleading grst_delay variable

The grst_delay variable in ice_check_reset contains the maximum time
(in 100 msec units) that the driver will wait for a reset event to
transition to the Device Active state. The value is the sum of three
separate components:
1) The maximum time it may take for the firmware to process its
outstanding command before handling the reset request.
2) The value in RSTCTL.GRSTDEL (the delay firmware inserts between first
seeing the driver reset request and the actual hardware assertion).
3) The maximum expected reset processing time in hardware.

Referring to this total time as "grst_delay" is misleading and
potentially confusing to someone checking the code and cross-referencing
the hardware specification.

Fix this by renaming the variable to "grst_timeout", which is more
descriptive of its actual use.

Signed-off-by: Nick Nunley <nicholas.d.nunley@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoice: mark PM functions as __maybe_unused
Wei Yongjun [Tue, 28 Jul 2020 01:41:53 +0000 (09:41 +0800)]
ice: mark PM functions as __maybe_unused

In certain configurations without power management support, the
following warnings happen:

drivers/net/ethernet/intel/ice/ice_main.c:4214:12: warning:
 'ice_resume' defined but not used [-Wunused-function]
 4214 | static int ice_resume(struct device *dev)
      |            ^~~~~~~~~~
drivers/net/ethernet/intel/ice/ice_main.c:4150:12: warning:
 'ice_suspend' defined but not used [-Wunused-function]
 4150 | static int ice_suspend(struct device *dev)
      |            ^~~~~~~~~~~

Mark these functions as __maybe_unused to make it clear to the
compiler that this is going to happen based on the configuration,
which is the standard for these types of functions.

Fixes: 769c500dcc1e ("ice: Add advanced power mgmt for WoL")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
4 years agoMerge tag 'mac80211-next-for-davem-2020-07-31' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Sat, 1 Aug 2020 01:51:40 +0000 (18:51 -0700)]
Merge tag 'mac80211-next-for-davem-2020-07-31' of git://git./linux/kernel/git/jberg/mac80211-next

Johannes Berg says:

====================
We have a number of changes
 * code cleanups and fixups as usual
 * AQL & internal TXQ improvements from Felix
 * some mesh 802.1X support bits
 * some injection improvements from Mathy of KRACK
   fame, so we'll see what this results in ;-)
 * some more initial S1G supports bits, this time
   (some of?) the userspace APIs
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agortnetlink: add support for protodown reason
Roopa Prabhu [Sat, 1 Aug 2020 00:34:01 +0000 (17:34 -0700)]
rtnetlink: add support for protodown reason

netdev protodown is a mechanism that allows protocols to
hold an interface down. It was initially introduced in
the kernel to hold links down by a multihoming protocol.
There was also an attempt to introduce protodown
reason at the time but was rejected. protodown and protodown reason
is supported by almost every switching and routing platform.
It was ok for a while to live without a protodown reason.
But, its become more critical now given more than
one protocol may need to keep a link down on a system
at the same time. eg: vrrp peer node, port security,
multihoming protocol. Its common for Network operators and
protocol developers to look for such a reason on a networking
box (Its also known as errDisable by most networking operators)

This patch adds support for link protodown reason
attribute. There are two ways to maintain protodown
reasons.
(a) enumerate every possible reason code in kernel
    - A protocol developer has to make a request and
      have that appear in a certain kernel version
(b) provide the bits in the kernel, and allow user-space
(sysadmin or NOS distributions) to manage the bit-to-reasonname
map.
- This makes extending reason codes easier (kind of like
      the iproute2 table to vrf-name map /etc/iproute2/rt_tables.d/)

This patch takes approach (b).

a few things about the patch:
- It treats the protodown reason bits as counter to indicate
active protodown users
- Since protodown attribute is already an exposed UAPI,
the reason is not enforced on a protodown set. Its a no-op
if not used.
the patch follows the below algorithm:
  - presence of reason bits set indicates protodown
    is in use
  - user can set protodown and protodown reason in a
    single or multiple setlink operations
  - setlink operation to clear protodown, will return -EBUSY
    if there are active protodown reason bits
  - reason is not included in link dumps if not used

example with patched iproute2:
$cat /etc/iproute2/protodown_reasons.d/r.conf
0 mlag
1 evpn
2 vrrp
3 psecurity

$ip link set dev vxlan0 protodown on protodown_reason vrrp on
$ip link set dev vxlan0 protodown_reason mlag on
$ip link show
14: vxlan0: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode
DEFAULT group default qlen 1000
    link/ether f6:06:be:17:91:e7 brd ff:ff:ff:ff:ff:ff protodown on <mlag,vrrp>

$ip link set dev vxlan0 protodown_reason mlag off
$ip link set dev vxlan0 protodown off protodown_reason vrrp off

Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
David S. Miller [Sat, 1 Aug 2020 00:19:47 +0000 (17:19 -0700)]
Merge git://git./pub/scm/linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2020-07-31

The following pull-request contains BPF updates for your *net* tree.

We've added 5 non-merge commits during the last 21 day(s) which contain
a total of 5 files changed, 126 insertions(+), 18 deletions(-).

The main changes are:

1) Fix a map element leak in HASH_OF_MAPS map type, from Andrii Nakryiko.

2) Fix a NULL pointer dereference in __btf_resolve_helper_id() when no
   btf_vmlinux is available, from Peilin Ye.

3) Init pos variable in __bpfilter_process_sockopt(), from Christoph Hellwig.

4) Fix a cgroup sockopt verifier test by specifying expected attach type,
   from Jean-Philippe Brucker.

Note that when net gets merged into net-next later on, there is a small
merge conflict in kernel/bpf/btf.c between commit 5b801dfb7feb ("bpf: Fix
NULL pointer dereference in __btf_resolve_helper_id()") from the bpf tree
and commit 138b9a0511c7 ("bpf: Remove btf_id helpers resolving") from the
net-next tree.

Resolve as follows: remove the old hunk with the __btf_resolve_helper_id()
function. Change the btf_resolve_helper_id() so it actually tests for a
NULL btf_vmlinux and bails out:

int btf_resolve_helper_id(struct bpf_verifier_log *log,
                          const struct bpf_func_proto *fn, int arg)
{
        int id;

        if (fn->arg_type[arg] != ARG_PTR_TO_BTF_ID || !btf_vmlinux)
                return -EINVAL;
        id = fn->btf_id[arg];
        if (!id || id > btf_vmlinux->nr_types)
                return -EINVAL;
        return id;
}

Let me know if you run into any others issues (CC'ing Jiri Olsa so he's in
the loop with regards to merge conflict resolution).
====================

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