Add the rt status of the last assert or 0 if the dump collection was
not initiated by an assert.  Add the number of rx and tx fifos in use.
These fields are added to dump info lst file.
Signed-off-by: NShahar S Matityahu <shahar.s.matityahu@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>
Signed-off-by: NKalle Valo <kvalo@codeaurora.org>

When traffic load is not low or low latency is active, TCM schedules
re-evaluation work so in case traffic stops TCM will detect that
traffic load has become low or that low latency is no longer active.
However, if TCM is paused when the re-evaluation work runs, it does
not re-evaluate and the re-evaluation work is no longer scheduled.
As a result, TCM will not indicate that low latency is no longer
active or that traffic load is low when traffic stops.

Fix this by forcing TCM re-evaluation when TCM is resumed in case
low latency is active or traffic load is not low.
Signed-off-by: NAvraham Stern <avraham.stern@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

FW asserts 0x70, 0x71, and 0x73 all just mean that the real error
happened in another MAC, and to look there for the problem. Add their
descriptions to the assert number lookup table so users get a nicer
error message in the logs.
Also, since the 4 most-significant bits of the assert number are
dynamic, and depend on which MAC the assert occurred on, ignore those
bits when looking up the assert name.
Signed-off-by: NNaftali Goldstein <naftali.goldstein@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Enable low latency for softAP in all modes (standalone, SCM
and DCM).
This is in order to minimize the time the softAP leaves the channel for
other operations
Signed-off-by: NTova Mussai <tova.mussai@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

All the queue management code runs under mvm->mutex, so there are
only very few cases of accessing the data structures without it:
 * TX path, which doesn't take any locks anyway
 * iwl_mvm_wake_sw_queue() and iwl_mvm_stop_sw_queue() where we
   just (atomically) read a bitmap, so the lock isn't needed.

Therefore, we can remove the spinlock. This enables some cleanup
in the ugly locking in iwl_mvm_inactivity_check().
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

There's no need to have a hw refcount if we just mark the
command queue with a (fake) TID; at that point, the refcount
becomes equivalent to the hweight() of the TID bitmap.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

None of these functions really need to be separate, they're all
only used in sta.c, move them there and make them static.

Fix a small typo in related code while at it.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

The triplet of get trigger, is trigger enabled and is trigger stopped
repeats itself.  Group them in a function to avoid code duplication.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Alexei's patch, assumed that all versions of "struct iwl_error_event_table"
are the same, but there are really different versions in different files.

Rather than trying to fix this, or splitting the tracepoint, or anything of
the sort, just remove it entirely - turns out that nobody really uses it.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Support more txq_alloc command types by moving the command declaration
to the gen specific area.  While at it, move some of the code segments
to a common place for re-use.
Signed-off-by: NGolan Ben Ami <golan.ben.ami@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

The Free Software Foundation address is superfluous and causes
checkpatch to issue a warning when present.  Remove all paragraphs
with FSF's address to prevent that.
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

The parameter that indicated whether the LQ command should be sent
as sync or async was removed, causing the LQ command to be sent as
sync from interrupt context (e.g. from the RX path). This resulted
in a kernel warning: "scheduling while atomic" and failing to send
the LQ command, which ultimately leads to a queue hang.

Fix it by adding back the required parameter to send the command as
sync only when it is allowed.

Fixes: d94c5a82 ("iwlwifi: mvm: open BA session only when sta is authorized")
Signed-off-by: NAvraham Stern <avraham.stern@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Different device families may have different flag values
for passing a message to the fw (i.e. SW_RESET).
In order to keep the code readable, and avoid conditioning
upon the family, store a value for each flag, which indicates
the bit that needs to be enabled.
Signed-off-by: NGolan Ben Ami <golan.ben.ami@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Update rs-fw API to match changes in FW.  Specifically, the
TLC_MNG_NOTIF_REQ_CMD command and TLC_MNG_AMSDU_ENABLE_NOTIF
notification are removed, the A-MSDU related info is received from FW
via the TLC_MNG_UPDATE_NOTIF, and the TLC_MNG_CONFIG_CMD uses version
2 of its data structure.

Additionally, constify some arguments in a couple of functions.
Signed-off-by: NNaftali Goldstein <naftali.goldstein@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

In 22000 devices, aka gen2, the TFS is 256 bytes.
In order to save memory, use shorter TX queue for aux and
mgmt queues, since there isn't much traffic on them.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

As preparation for dynamic queue sizing, add a parameter
of the TX queue size to the dynamic queue allocation op
mode API.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Detect low latency and traffic load per band.  Add support for
deciding on scan type and timings per band.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Try to detect that the AP is not using aggregation even when there's
enough traffic to make it worthwhile; if this is the case and U-APSD
is enabled then assume the AP is broken (like so many) and doesn't
enable aggregation when U-APSD is used. In this case, disconnect from
the AP and blacklist U-APSD for a potential new connection to it.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Traffic condition monitor gathers data about the traffic load and
other conditions and can be used to make decisions regarding latency,
throughput etc.  This patch introduces the code and data structures to
collect this data for future use.
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

fix iwlwifi_dev_ucode_error tracepoint to pass pointer to a table
instead of all 17 arguments by value.
dvm/main.c and mvm/utils.c have 'struct iwl_error_event_table'
defined with very similar yet subtly different fields and offsets.
tracepoint is still common and using definition of 'struct iwl_error_event_table'
from dvm/commands.h while copying fields.
Long term this tracepoint probably should be split into two.
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>
Signed-off-by: NDaniel Borkmann <daniel@iogearbox.net>

Currently we have a boolean variable for each cause.

This costs space, and requires to check each separately
when determining low latency.

Since we have another cause incoming, convert it to an enum.

While at it, move the retrieval of the prev value and the
assignment of the new value to be inside iwl_mvm_update_low_latency
and save the need for each caller to do it separately.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Sometimes iwl_mvm_disable_txq() may be called with mac80211_queue ==
IEEE80211_INVAL_HW_QUEUE, and this would cause us to use BIT(0xFF)
which is way too large for the u16 we used to store it in
hw_queue_to_mac820211.  If this happens the following UBSAN warning
will be generated:

[  167.185167] UBSAN: Undefined behaviour in drivers/net/wireless/intel/iwlwifi/mvm/utils.c:838:5
[  167.185171] shift exponent 255 is too large for 64-bit type 'long unsigned int'

Fix that by checking that it is not IEEE80211_INVAL_HW_QUEUE and,
while at it, add a warning if the queue number is larger than
IEEE80211_MAX_QUEUES.

Fixes: 34e10860 ("iwlwifi: mvm: remove references to queue_info in new TX path")
Reported-by: NPaul Menzel <pmenzel+linux-wireless@molgen.mpg.de>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

This aligns the code with the existing pattern to check
if the firmware has a certain capability.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Recently a new command was added to the firmware
for setting a MAC's low-latency mode.  Use it.
Signed-off-by: NShaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Most of the sw resets in the code are done by one function,
which writes to the relevant CSR.

Use the common function to perform the only reset which was
done separately, redundant to the common code.
Signed-off-by: NGolan Ben Ami <golan.ben.ami@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

When sending LQ command, verify the rate scaling is not in firmware.
Signed-off-by: NGregory Greenman <gregory.greenman@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

This patch adds rate scaling configuration command and
implements a few other handlers.
Signed-off-by: NGregory Greenman <gregory.greenman@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

We might erroneously get to error dumping code when the
device is already stopped.

In that case the driver will detect a defective value and will try to
reset the HW, assuming it is only a bus issue.  The driver than
proceeds with the dumping.

The result has two side effects:

1. The device won't be stopped again, since the transport status is
already stopped, so the device remains powered on while it actually
should be stopped.

2. The dump in that case is completely garbaged and useless.

Detect and avoid this.  It will also make debugging such issues
easier.
Signed-off-by: NSara Sharon <sara.sharon@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

The MONITOR type is missing in the interface type switch.
Add it.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

When we act as an AP, new firmware versions handle
internally the power saving clients and the driver doesn't
know that the peers went to sleep. It is, hence, possible
that a peer goes to sleep for a long time and stop pulling
frames. This will cause its transmit queue to hang which is
a condition that triggers the recovery flow in the driver.

While this client is certainly buggy (it should have pulled
the frame based on the TIM IE in the beacon), we can't blow
up because of a buggy client.

Change the current implementation to not enable the
transmit queue hang detection on queues that serve peers
when we act as an AP / GO.

We can still enable this mechanism using the debug
configuration which can come in handy when we want to
debug why the client doesn't wake up.

Cc: stable@vger.kernel.org # v4.13
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

There's no point in checking the validity of the
umac_error_event_table pointer every time we generate a dump.  It's
cleaner to do so when we read the value, namely when we receive the
alive data.
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Currently, UMAC error data reading is restricted to DCCM.
A000 NICs use SMEM for this data.
Signed-off-by: NBeni Lev <beni.lev@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

The trigger that collects data when a frame is released
because of the timer of the reordering buffer was not
implemented for 9000 devices.
Fix this.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

This was never used by any product. Remove it.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Apart from DVM, all firmware uses the same base API, and there's
code outside iwlmvm that needs to interact with it. Reflect this
in the source better and reorganize the firmware API to a new
fw/api/ directory.

While at it, split the already pretty large fw-api.h file into a
number of smaller files, going from almost 3k lines in there to
a maximum number of lines less than 1k.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Split out the firmware debug code to be more general, so that it
can be used by different subdrivers.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

Track the current firmware image in the common code instead
of in the opmode so that later patches can access it there
in a common way.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

The newest devices need a longer time to reset because of
their more complex hardware. Wait 5ms after device reset.
Consolidate all the places that reset the device in the
PCIe transport to avoid future bugs.

While at it, unify the flow to use set_bit instead of full
write as requested by the hardware designers.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

In the driver, we track which hardware queue is associated with
which mac80211 "hw_queue", in order to be able to stop and wake
it. When moving these bitmaps out of the queue_info structures,
the type of the bitmap was erroneously changed from u32 to u8,
presumably in order to save memory.

Turns out that u32 isn't needed, because the highest queue we
can ever tell mac80211 is always < 16, but a u16 definitely is
needed, queues >=8 do happen.

While at it, throw a BUILD_BUG_ON() into the place where we set
the limit (mvm->first_agg_queue) and a warning when it actually
gets put into the bitmap.

The consequence of this bug is that full HW queues associated
with such a too-high mac80211 number never stop higher layer
queues when full, and thus would simply drop all packets that
couldn't be enqueued to the hardware queue.

Fixes: 34e10860 ("iwlwifi: mvm: remove references to queue_info in new TX path")
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>

In case we need to move the scheduler write pointer by
steps of 0x40, 0x80 or 0xc0, the scheduler gets stuck.
This leads to hardware error interrupts with status:
0x5A5A5A5A or alike.

In order to work around this, detect in the transport
layer that we are going to hit this case and tell iwlmvm
to increment the sequence number of the packets. This
allows to keep the requirement that the WiFi sequence
number is in sync with the index in the scheduler Tx queue
and it also allows to avoid the problematic sequence.
This means that from time to time, we will start a queue
from ssn + 1, but that shouldn't be a problem since we
don't switch to new queues for AMPDU now that we have
DQA which allows to keep the same queue while toggling
the AMPDU state.

This bug has been fixed on 9000 devices and up.
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NLuca Coelho <luciano.coelho@intel.com>