Move frame pad computation out of mt76_skb_adjust_pad routine.
This is a preliminary patch to introduce sdio tx aggregation.
Tested-by: Sean Wang <sean.wang@mediatek.com>
Co-developed-by: Sean Wang <sean.wang@mediatek.com>
Signed-off-by: Sean Wang <sean.wang@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

This is a preliminary patch to introduce sdio tx aggregation
Tested-by: Sean Wang <sean.wang@mediatek.com>
Co-developed-by: Sean Wang <sean.wang@mediatek.com>
Signed-off-by: Sean Wang <sean.wang@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

In order to not update the available quota in case of a tx error, split
mt7663s_tx_update_sched in mt7663s_tx_{pick,update}_quota routines
Tested-by: Sean Wang <sean.wang@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Since the sdio engine does not report quota for altx queue, move
ctl/mgmt traffic to standard data queues
Tested-by: Sean Wang <sean.wang@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Set poll timeout to 3s for mt7622 devices in order to avoid fw hangs.
Swap mt7622_trigger_hif_int and doorbell configuration order in
mt7615_mcu_drv_pmctrl routine.
Introduce mt7615_mcu_lp_drv_pmctrl routine to take care of drv_own
configuration for runtime-pm.

Fixes: 08523a2a ("mt76: mt7615: add mt7615_pm_wake utility routine")
Fixes: 894b7767 ("mt76: mt7615: improve mt7615_driver_own reliability")
Fixes: 757b0e7f ("mt76: mt7615: avoid polling in fw_own for mt7663")
Co-developed-by: Shayne Chen <shayne.chen@mediatek.com>
Signed-off-by: Shayne Chen <shayne.chen@mediatek.com>
Co-developed-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

The MCU field should contain a boolean 0/1, not the flag itself.
Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/net/wireless/mediatek/mt76/mt7615/testmode.c: In function mt7615_tm_set_tx_power
drivers/net/wireless/mediatek/mt76/mt7615/testmode.c:83:7: warning: variable ‘index’ set but not used [-Wunused-but-set-variable]=

commit 4f0bce1c ("mt76: mt7615: implement testmode support")
involved this unused variable, remove it.
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Since the switch to using AQL by default, mtxq->retry_q is never filled anymore
Remove it to get rid of a few more unnecessary cycles in the tx path
Signed-off-by: Felix Fietkau <nbd@nbd.name>

This way we can make some functions static
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Unlike 7615, 7915 uses the same AC index for rx and tx, which matches the
LMAC queue mapping
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Compared to mac80211 ACs, MT7915 queue numbers are in reverse order
There is no need for the defensive WARN_ON_ONCE, so we can simplify
the function to avoid the array lookup
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Accessing them from uncached memory can be expensive, so it's cheaper to
cache them
Signed-off-by: Felix Fietkau <nbd@nbd.name>

he_pe_duration and he_rts_thres have the same unit as the fields in the HE
operation IE
Signed-off-by: Felix Fietkau <nbd@nbd.name>

This improves performance by allowing the scheduler to move the tx scheduling
work to idle CPUs. Since tx scheduling work is very latency sensitive and
kept short via AQL, sched_set_fifo_low is used to keep worker priority above
normal tasks
Signed-off-by: Felix Fietkau <nbd@nbd.name>

In order to avoid keeping work like tx scheduling pinned to the CPU it was
scheduled from, it makes sense to switch from tasklets to kernel threads.

Unlike a workqueue, this one only allows one fixed worker function to be
executed by the worker thread. Because of that, there is less locking
and less code for scheduling involved.
This is important because the tx worker is scheduled often in a hot path
Signed-off-by: Felix Fietkau <nbd@nbd.name>

This avoids running out of available tx tokens
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Do not alter the tx/rx chain settings during channel setup, antennas are
remapped by the testmode specific register writes already
Signed-off-by: Felix Fietkau <nbd@nbd.name>

This is used for testmode tx antenna selection
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Avoid firmware falling into spectrum mode since that will set
unexpected PSE/PLE thresholds which lead to Tx hang.

This mode should be cleaned before firmware download stage.
Signed-off-by: Chih-Min Chen <chih-min.chen@mediatek.com>
Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Increase data queue size to improve performance.
Reduce PS/mgmt queue size
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Stop relying on global tx ring size, reduce size for PS/mgmt queue
Signed-off-by: Felix Fietkau <nbd@nbd.name>

All members except for the struct mt76_queue pointer have been removed
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Now that AQL works well on all mt76 drivers, completely replace the arbitrary
burst sizing and number of bursts logic for tx scheduling.
For the short period of time in which AQL does not work well yet, limit each
stations to 16 in-flight packets that have no estimated tx time.
This should avoid filling the queue if a station connects and queues up a
large number of packets before rate control information is available, especially
with hardware rate control
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Since txq selection was moved to mac80211, it is no longer used
Signed-off-by: Felix Fietkau <nbd@nbd.name>

It is not needed
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Cleanup and preparation for changing tx scheduling behavior
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Their use is reversed compared to DMA. The order for DMA makes more sense,
so let's use that
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Their use is reversed compared to DMA. The order for DMA makes more sense,
so let's use that
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Some devices using MT7628 or MT7603 have only one antenna chain connected.
Detect these using the EEPROM rx/tx path settings
Reported-by: Qin Wei <me@vonger.cn>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Ensure that descriptor memory has been fully written before letting the
hardware read it
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Makes it possible to read/write them via debugfs, similar to mt7615/7603
Signed-off-by: Felix Fietkau <nbd@nbd.name>

On 7615 and newer, DMA completion only triggers unmap, but not free of queued
skbs, since pointers to packets are queued internally.
Because of that, there is no need to process the main data queue immediately
on DMA completion.
To improve performance, mask out the DMA data queue completion interrupt and
process the queue only when we receive a txfree event.
This brings the number of interrupts under load down to a small fraction.
Signed-off-by: Felix Fietkau <nbd@nbd.name>

On 7615 and newer, DMA completion only triggers unmap, but not free of queued
skbs, since pointers to packets are queued internally.
Because of that, there is no need to process the main data queue immediately
on DMA completion.
To improve performance, mask out the DMA data queue completion interrupt and
process the queue only when we receive a txfree event.
This brings the number of interrupts under load down to a small fraction.
Signed-off-by: Felix Fietkau <nbd@nbd.name>

The previous scheduling round may have been limited by AQL.
More frames might be available after the tx free run.
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Tx cleanup and tx enqueuing can run in parallel. In order to avoid queue
starvation issues under load, update q->queued immediately.
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Since DMA completion does not imply tx completion, it makes more sense to
poll for airtime from mt7915_mac_tx_free.
Reduce the runtime of the function by moving all items from dev->sta_poll_list
to a local list once and process any stations that were added afterwards
on the next run
Signed-off-by: Felix Fietkau <nbd@nbd.name>

- change the WTBL LMAC access function to set the mapping window only once
- use ac * 2 as offset, since each AC has separate words for rx and tx
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Fix length field corruption in usb dma header introduced adding sdio
support

Fixes: 75b10f0c ("mt76: mt76u: add mt76_skb_adjust_pad utility routine")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Fix a possible NULL pointer dereference in mt76_testmode_dump() since
nla_nest_start returns NULL in case of error

Fixes: f0efa862 ("mt76: add API for testmode support")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>

Initialize wcid to global_wcid if msta is NULL in mt7615_pm_wake_work
routine since wcid will be dereferenced running mt76_tx()

Fixes: 2b8cdfb2 ("mt76: mt7615: wake device before pushing frames in mt7615_tx")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>