Host time is used to calculate the channel active time on mt7603 and mt7615.
Use the same on mt76x02 and move the lock to core code to get rid of some
duplicated code.
Signed-off-by: Felix Fietkau <[email protected]>
MT_MIB_STAT_PSCCA only counts rx CCA busy time, which does not include
tx time. MT_MIB_STAT_CCA counts full busy time, including Rx, Tx and NAV
Signed-off-by: Felix Fietkau <[email protected]>
Report total rx airtime for valid stations as BSS rx time in survey
mt7615 is left out for now, it will be supported later by reading
hardware counters instead of calculating airtime in software
Signed-off-by: Felix Fietkau <[email protected]>
Move mt76_channel_state() from mt76.h to mac80211.c
Preparation for updating channel state from more places in the drivers/core
Signed-off-by: Felix Fietkau <[email protected]>
tx queues are already disabled by mac80211 during scanning or other
off-channel activity. There is no need to repeat the check in mt76,
since scheduled queues are selected by mac80211 as well.
Signed-off-by: Balakrishna Bandi <[email protected]>
Signed-off-by: Felix Fietkau <[email protected]>
There is no need to send a BAR frame after stopping aggregation, and doing
so could lead to sending class 3 frames after deauthentication from an AP
Signed-off-by: Balakrishna Bandi <[email protected]>
Signed-off-by: Felix Fietkau <[email protected]>
MT7615 chips usually come pre-calibrated, even when used on embedded boards.
In that case, the on-flash EEPROM data needs to be merged with some data
from OTP ROM.
Run this merge if the external EEPROM data is valid and OTP has valid fields.
Signed-off-by: Felix Fietkau <[email protected]>
Bufferable management frames need to be put in the data queue, otherwise
they will not be buffered when the receiver is asleep.
Signed-off-by: Felix Fietkau <[email protected]>
The hardware expects MT_TX_RATE_NSS to be filled with the number of
space/time streams. For non-STBC rates, this is equal to nss.
For 1-stream STBC, this needs to be set to 2.
This is relevant for VHT rates only, on HT, the value is derived from MCS
internally.
Signed-off-by: Felix Fietkau <[email protected]>
- Store the previous and current rate set in the driver + the TSF value
at the time of the switch.
- Use the tx status TSF value to determine which rate set needs to be used
as reference.
- Report only short or long GI rates for a single status event, not a mix.
- The hardware reports the last used rate index. Use it along with the
retry count to figure out what rate was used for the first attempt.
- Use the same retry count value for all rate slots to make this calculation
work.
- Derive the probe rate from the current rateset instead of the skb cb
- Do not wait for a status report for the probe frame before removing the
probe rate from the rate table. Do it immediately after it was referenced
in a tx status report.
- Use the first half of the first rate retry budget for the probe rate
in order to avoid using too many retries on that rate
- Switch from lower rates to higher rates more conservatively
- enable hardware rate up/down selection
Signed-off-by: Felix Fietkau <[email protected]>
These values must be initialized to zero, otherwise the hardware could
reuse previous values, especially the rate index
Signed-off-by: Felix Fietkau <[email protected]>
Since we don't clean that tx queue from the tx tasklet, we need to do it
after the firmware upload is done. This patch also adds a cleanup step during
the upload, to help reclaim memory faster.
Fixes unprocessed queued frames eating up memory long after the firmware
upload has already completed
Signed-off-by: Felix Fietkau <[email protected]>
When beacon length is not a multiple of 4, the beacon could be sent with
the last 1-3 bytes corrupted. The skb data is guaranteed to have enough
room for reading beyond the end, because it is always followed by
skb_shared_info, so rounding up is safe.
All other callers of mt76_wr_copy have multiple-of-4 length already.
Cc: [email protected]
Signed-off-by: Felix Fietkau <[email protected]>
When loading EEPROM data from flash, the RF frontend settings need to be
initialized from flash data. Without this, the chip loads the wrong values
from its internal eFuse ROM.
Signed-off-by: Felix Fietkau <[email protected]>
Now that tx status reporting can figure out the first attempted rate, we can
make switching from lower rates to higher rates more conservative.
This reduces retries under bad link conditions and ensures that fallback
rates get more test coverage
Signed-off-by: Felix Fietkau <[email protected]>
Tx status reporting on mt7603 has a number of issues:
- the hardware can alter the first rate index, but it is not reported to
the driver
- probing is very imprecise, because it alters the per-client rate set,
but only considers info->status.rates for rate selection of a single probe
packet
- short/long GI selection has limitations, which are not accurately reported
to mac80211
- if rates are changed while packets are in flight, tx status reports for
the old rate set might be processed based on the new selection
This led to very suboptimal rate selection with minstrel_ht.
This patch completely reworks tx status reporting to get rid of these
limitations:
- Store the previous and current rate set in the driver + the TSF value
at the time of the switch.
- Use the tx status TSF value to determine which rate set needs to be used
as reference.
- Report only short or long GI rates for a single status event, not a mix.
- The hardware reports the last used rate index. Use it along with the
retry count to figure out what rate was used for the first attempt.
- Use the same retry count value for all rate slots to make this calculation
work.
- Derive the probe rate from the current rateset instead of the skb cb
- Do not wait for a status report for the probe frame before removing the
probe rate from the rate table. Do it immediately after it was referenced
in a tx status report.
- Use the first half of the first rate retry budget for the probe rate
in order to avoid using too many retries on that rate
With this patch, throughput under bad link conditions is improved
significantly, and there is a lot less rate fluctuation going on.
Signed-off-by: Felix Fietkau <[email protected]>
This didn't bring the performance benefit that I had hoped for, so the mac80211
patch for this will not make it upstream
Signed-off-by: Felix Fietkau <[email protected]>
To avoid aggregating rate control probing packets with other traffic, and to
ensure that the probing rate gets used, probing packets get assigned a different
internal queueing priority.
This causes packets to be transmitted in a different order, which is compensated
by the receiver side reordering.
However, if A-MPDU is disabled, this reordering can become visible to upper
layers on the receiver side. Disable the priority change if A-MPDU is disabled.
Signed-off-by: Felix Fietkau <[email protected]>
When the hardware falls back to lower rates for a transmit attempt, only the
first status report will show the number of retries correctly. The frames
that follow will report the correct final rate, but number of retries set to 0.
This can cause the rate control module to vastly underestimate the number of
retransmissions per rate.
To fix this, we need to keep track of the initial requested tx rate per packet
and pass it to the status information.
For frames with tx status requested, this is simple: use the rate configured
in info->control.rates[0] as reference.
For no-skb tx status information, we have to encode the requested tx rate in
the packet id (and make it possible to distinguish it from real packet ids).
To do that, reduce the packet id field size by one bit, and use that bit to
indicate packet id vs rate.
This change also improves reporting by filling the status rate array with
rates from first rate to final rate, taking the same steps as the hardware
fallback table. This matters in corner cases like MCS8 on HT, where the
fallback target is MCS0, not MCS7.
Signed-off-by: Felix Fietkau <[email protected]>
When setting chan->max_power after registering the wiphy, chan->max_reg_power
needs to be used as a limit
Signed-off-by: Felix Fietkau <[email protected]>
For the external PA (TSSI OFF) case, the target power needs to be read
from a different location in EEPROM
Signed-off-by: Felix Fietkau <[email protected]>
- tx power is stored in the channels after ieee80211_register_hw, so
chan->orig_mpwr needs to be updated as well
- for non-TSSI devices, mt76x2e needs to use a different target power value
from the EEPROM
- fix a rounding error in a few places (need to round up, not down)
Signed-off-by: Felix Fietkau <[email protected]>
Now that the sequence number allocation is fixed, we can finally send a BAR
at powersave wakeup time to refresh the receiver side reorder window
Signed-off-by: Felix Fietkau <[email protected]>
If the MT_TXD3_SN_VALID flag is not set in the tx descriptor, the hardware
assigns the sequence number. However, the rest of the code assumes that the
sequence number specified in the 802.11 header gets transmitted.
This was causing issues with the aggregation setup, which worked for the
initial one (where the sequence numbers were still close), but not for
further teardown/re-establishing of sessions.
Additionally, the overwrite of the TID sequence number in WTBL2 was resetting
the hardware assigned sequence numbers, causing them to drift further apart.
Fix this by using the software assigned sequence numbers
Signed-off-by: Felix Fietkau <[email protected]>
mutex->lock must not be held while cancelling workqueues, because it is held
by the workqueue items themselves
Signed-off-by: Felix Fietkau <[email protected]>
q->tail can be safely updated without locking, because there is no
concurrent access. If called from outside of the tasklet (for flushing),
the tasklet is always disabled.
q->queued can be safely read without locking, as long as the decrement
happens within the locked section.
This patch allows cleaning up tx packets outside of the section that holds
the queue lock for improved performance
Signed-off-by: Felix Fietkau <[email protected]>
Use a separate lock and spin_trylock to avoid disabling interrupts.
Should improve performance and latency
Signed-off-by: Felix Fietkau <[email protected]>
__sw_hweight8() is only defined if CONFIG_GENERIC_HWEIGHT is enabled.
The function that works on all architectures is hweight8().
Signed-off-by: Felix Fietkau <[email protected]>
The previous version only accidentally disabled A-MSDU deaggregation by
using the wrong mask for rx length configuration, which left previous
length value in place.
Fix the length and initialize the register completely to keep A-MSDU
de-aggregation remaining disabled
Signed-off-by: Felix Fietkau <[email protected]>
This feature has been reported to cause stability issues on several systems.
Disable it until it has been fixed and verified. It can still be enabled
through debugfs
Signed-off-by: Felix Fietkau <[email protected]>
These packets have no txwi entry in the ring, so tracking via tx status does
not work. To prevent PS poll requests from being unanswered, end the service
period right away
Signed-off-by: Felix Fietkau <[email protected]>
AGC register 35, 37 override for the low gain setting should only be done
on 5 GHz. Also, 2.4 GHz needs a different value for register 35
Signed-off-by: Felix Fietkau <[email protected]>
Use the correct variable in the check. Fixes an uninitialized variable warning
Reported-by: Gustavo A. R. Silva <[email protected]>
Fixes: c8846e1015022 ("mt76: add driver for MT7603E and MT7628/7688")
Signed-off-by: Felix Fietkau <[email protected]>
Full tx blocking (as opposed to CCA blocking) should only happen if there
is a continuous non-802.11 signal above the energy detect threshold.
Unfortunately the ED/CCA counter can't detect that, as it also counts 802.11
signals as busy.
Similar to the vendor code, implement a learning mode that waits until the AGC
gain has already been adjusted to the lowest value (due to false CCA events),
and the number of false CCA events still remains high, and the blocking
threshold is exceeded for more than 5 seconds.
Signed-off-by: Felix Fietkau <[email protected]>
Also update the mask first before calculating the vif index.
Fixes an issue where adding back the same interfaces in a different order
fails because of duplicate vif index use
Fixes: 06662264ce2ad ("mt76x02: use mask for vifs")
Signed-off-by: Felix Fietkau <[email protected]>
The vendor driver also does not enable it, and it has been shown to cause
some issues. It can be re-enabled for testing through debugfs
Signed-off-by: Felix Fietkau <[email protected]>
It needs to be always enabled for 76x2 and conditionally enabled/disabled
for 76x0. Since mt76x2_mac_stop clears this bit, ensure that it is enabled
in mt76x02_edcca_init
Signed-off-by: Felix Fietkau <[email protected]>
ED/CCA needs to be disable before stopping the MAC to avoid hangs from tx
being blocked. It must only be enabled again after the MAC has been started
again.
In many places this wasn't done properly, so fix this by always clearing
the relevant ED/CCA bits in mt76x2_mac_stop and set it up again after
channel change or calibration is done
Signed-off-by: Felix Fietkau <[email protected]>
MCU request timeout usually indicates that the device is no longer responsive,
and it usually does not recover without a reset
Signed-off-by: Felix Fietkau <[email protected]>
When the timer is already running, there is no need to reset the counter,
because the drift will remain the same.
Signed-off-by: Felix Fietkau <[email protected]>
Felix Fietkau