After flushing software-pending frames to DMA, mt76_has_tx_pending()
only checks DMA ring q->queued. For token-based drivers, q->queued is
decremented at DMA consumption, but firmware may not have transmitted
the frame yet. Waiting for all tokens is not feasible because data
frames may be stuck in firmware powersave/aggregation queues.
Track PSD queue tokens (firmware ALTX) per phy using an atomic counter.
These frames are sent by firmware immediately without PS buffering, so
the counter reliably reaches zero after transmission.
Increment the counter in mt76_token_consume() and decrement it in
mt76_token_release(), only for PSD queue tokens. Include the counter
in mt76_has_tx_pending() so channel switch waits for firmware TX
completion of management and nullfunc frames.
mt7615 (uses mt76_token_get/put) and non-token drivers are unaffected
as they never call mt76_token_consume/release.
Link: https://patch.msgid.link/20260309060730.87840-10-nbd@nbd.name
Signed-off-by: Felix Fietkau <nbd@nbd.name>
ieee80211_is_data() returns true for nullfunc/QoS-nullfunc frames, so
they bypass the PSD queue routing and go through the regular VO data
queue. This means firmware processes them through the normal TID queue
instead of the ALTX queue, which doesn't guarantee immediate
transmission.
Use ieee80211_is_data_present() instead, which returns false for both
management frames and nullfunc/QoS-nullfunc (no payload), routing them
to MT_TXQ_PSD. Firmware maps PSD to the ALTX queue, which transmits
immediately without PS buffering.
This only affects frames from the mt76_tx() pending path. Regular
mac80211 TXQ scheduling is unchanged.
Link: https://patch.msgid.link/20260309060730.87840-9-nbd@nbd.name
Signed-off-by: Felix Fietkau <nbd@nbd.name>
mt76_tx() queues frames on wcid->tx_pending for async processing by
tx_worker. In __mt76_set_channel(), the worker gets disabled before it
may have run, and the subsequent wait only checks DMA ring queues, not
the software pending list. This means frames like nullfunc PS frames
from mt76_offchannel_notify() may never be transmitted on the correct
channel.
Fix this by running mt76_txq_schedule_pending() synchronously after
disabling the tx_worker but before setting MT76_RESET, which would
otherwise cause mt76_txq_schedule_pending_wcid() to bail out.
Link: https://patch.msgid.link/20260309060730.87840-8-nbd@nbd.name
Signed-off-by: Felix Fietkau <nbd@nbd.name>
mt76_txq_send_burst does check if the number of non-AQL frames exceeds
the maximum. In this case the queue is returned to ieee80211_return_txq
when iterating over the scheduled TXQs in mt76_txq_schedule_list.
This has the effect of inserting said TXQ at the head of the list. This
means the loop will get the same TXQ again, which will terminate the
scheduling round. TXQs following in the list thus never get scheduled
for transmission.
This can manifest in high latency low throughput or broken connections
for said STAs.
Check if the non-AQL packet count exceeds the limit and not return the
TXQ in this case.
Schedule all TXQs for the STA in case the non-AQL limit can be satisfied
again.
Signed-off-by: David Bauer <mail@david-bauer.net>
Link: https://patch.msgid.link/20260129232321.276575-1-mail@david-bauer.net
Signed-off-by: Felix Fietkau <nbd@nbd.name>
MediaTek has asked to switch from the ISC license to BSD-3-Clause-Clear,
in order to improve clarity and the legal integrity of the code.
The BSD-3-Clause license includes the "no endorsement" clause, which is
important for protecting the reputation of the original authors and
contributors by preventing unauthorized use of their names for endorsement
purposes.
This clause is absent in the BSD-2-Clause license, which is more permissive
but lacks this specific protection.
This change also cleans up the license of some Kconfig/Makefile files,
which were accidentally marked as GPL.
The GPL 2.0 remains in use on mt76x0, as well as two source files in mt7615
for which the license situation still needs to be clarified.
Link: https://patch.msgid.link/20251008104250.46292-2-nbd@nbd.name
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Only put probe request packets in the offchannel queue if
IEEE80211_TX_CTRL_DONT_USE_RATE_MASK is set and IEEE80211_TX_CTL_TX_OFFCHAN
is unset.
Fixes: 0b3be9d1d3 ("wifi: mt76: add separate tx scheduling queue for off-channel tx")
Reported-by: Chad Monroe <chad.monroe@adtran.com>
Link: https://patch.msgid.link/20250813121106.81559-2-nbd@nbd.name
Signed-off-by: Felix Fietkau <nbd@nbd.name>
When running in AP mode and deauthenticating a client that's in powersave
mode, the disassoc/deauth packet can get stuck in a tx queue along with
other buffered frames. This can fill up hardware queues with frames
that are only released after the WTBL slot is reused for another client.
Fix this by moving deauth packets to the ALTX queue.
Reported-by: Chad Monroe <chad.monroe@adtran.com>
Link: https://patch.msgid.link/20250707154702.1726-2-nbd@nbd.name
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Several places use rcu_dereference to get a wcid entry without validating
if the index exceeds the array boundary. Fix this by using a helper function,
which handles validation.
Link: https://patch.msgid.link/20250707154702.1726-1-nbd@nbd.name
Signed-off-by: Felix Fietkau <nbd@nbd.name>
The interrupt status polling is unreliable, which can cause status events
to get lost. On all newer chips, txs-timeout is an indication that the
packet was either never sent, or never acked.
Fixes issues with inactivity polling.
Link: https://patch.msgid.link/20250311103646.43346-6-nbd@nbd.name
Signed-off-by: Felix Fietkau <nbd@nbd.name>
When queueing packets, only the MT76_RESET flag of the primary PHY
is checked. If the primary PHY is scanning or changing channels, this can
lead to packet loss for tx on the second PHY.
Fix this by passing the phy to the .tx_queue_skb op and using it to check
the correct flag.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
When drv_tx calls race against local tx scheduling, the queue fill status checks
can potentially race, leading to dma queue entries being overwritten.
Fix this by deferring packets from drv_tx calls to the tx worker, in order to
ensure that all regular queue tx comes from the same context.
Reported-by: Ryder Lee <Ryder.Lee@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
AQL relies on accurate tx rate reporting, so add this patch for
non-binding skb rate when WED is enabled.
Co-developed-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Peter Chiu <chui-hao.chiu@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Concurrent binding/non-binding skbs could be handled anywhere which leads
to mixed byte counting, so switch to use PPDU TxS reporting regardless Tx
paths when WED is active.
Fixes: 43eaa36895 ("wifi: mt76: add PPDU based TxS support for WED device")
Co-developed-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Peter Chiu <chui-hao.chiu@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
According to the documentation, ieee80211_rx_list must not run concurrently
with ieee80211_tx_status (or its variants).
Cc: stable@vger.kernel.org
Fixes: 88046b2c9f ("mt76: add support for reporting tx status with skb")
Reported-by: Brian Coverstone <brian@mainsequence.net>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
The checks of whether or not a frame is bufferable were not
taking into account that some action frames aren't, such as
FTM. Check this, which requires some changes to the function
ieee80211_is_bufferable_mmpdu() since we need the whole skb
for the checks now.
Reviewed-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Take into account possible error conditions of mt76_rx_token_consume
routine in mt7915_mmio_wed_init_rx_buf() and mt76_dma_add_buf()
Fixes: cd372b8c99 ("wifi: mt76: add WED RX support to mt76_dma_{add,get}_buf")
Fixes: 4f831d18d1 ("wifi: mt76: mt7915: enable WED RX support")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit 96f134dc19)
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/20230113105848.34642-2-nbd@nbd.name
This is a preliminary patch to introduce WED RX support for mt7915.
Tested-by: Daniel Golle <daniel@makrotopia.org>
Co-developed-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Sujuan Chen <sujuan.chen@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Report phy_indx in tx_info->hw_queue. This is a preliminary patch to add
newer chipset support
Co-developed-by: Bo Jiao <bo.jiao@mediatek.com>
Signed-off-by: Bo Jiao <bo.jiao@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Introduce phys array in mt76_dev structure to reference mt76_phy
supported by the chipset. This is a preliminary patch to introduce
newer chipset support.
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Previously the code used skb_set_queue_mapping for management or non-bufferable
powersave frames that need to be sent to a different hardware queue.
This can confuse AQL, which expects the value to remain the same until the tx
status report.
The only place that currently uses the altered skb queue mapping is the txwi
write function. Change the code to pass the hardware queue id as a function
parameter instead.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
This patch adds the new struct ieee80211_rate_status and replaces
'struct rate_info *rate' in ieee80211_tx_status with pointer and length
annotation.
The struct ieee80211_rate_status allows to:
(1) receive tx power status feedback for transmit power control (TPC)
per packet or packet retry
(2) dynamic mapping of wifi chip specific multi-rate retry (mrr)
chains with different lengths
(3) increase the limit of annotatable rate indices to support
IEEE802.11ac rate sets and beyond
ieee80211_tx_info, control and status buffer, and ieee80211_tx_rate
cannot be used to achieve these goals due to fixed size limitations.
Our new struct contains a struct rate_info to annotate the rate that was
used, retry count of the rate and tx power. It is intended for all
information related to RC and TPC that needs to be passed from driver to
mac80211 and its RC/TPC algorithms like Minstrel_HT. It corresponds to
one stage in an mrr. Multiple subsequent instances of this struct can be
included in struct ieee80211_tx_status via a pointer and a length variable.
Those instances can be allocated on-stack. The former reference to a single
instance of struct rate_info is replaced with our new annotation.
An extension is introduced to struct ieee80211_hw. There are two new
members called 'tx_power_levels' and 'max_txpwr_levels_idx' acting as a
tx power level table. When a wifi device is registered, the driver shall
supply all supported power levels in this list. This allows to support
several quirks like differing power steps in power level ranges or
alike. TPC can use this for algorithm and thus be designed more abstract
instead of handling all possible step widths individually.
Further mandatory changes in status.c, mt76 and ath11k drivers due to the
removal of 'struct rate_info *rate' are also included.
status.c already uses the information in ieee80211_tx_status->rate in
radiotap, this is now changed to use ieee80211_rate_status->rate_idx.
mt76 driver already uses struct rate_info to pass the tx rate to status
path. The new members of the ieee80211_tx_status are set to NULL and 0
because the previously passed rate is not relevant to rate control and
accurate information is passed via tx_info->status.rates.
For ath11k, the txrate can be passed via this struct because ath11k uses
firmware RC and thus the information does not interfere with software RC.
Compile-Tested: current wireless-next tree with all flags on
Tested-on: Xiaomi 4A Gigabit (MediaTek MT7603E, MT7612E) with OpenWrt
Linux 5.10.113
Signed-off-by: Jonas Jelonek <jelonek.jonas@gmail.com>
Link: https://lore.kernel.org/r/20220509173958.1398201-2-jelonek.jonas@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
There is a small race window where ongoing tx activity can lead to a skb
getting added to the status tracking idr after that idr has already been
cleaned up, which will keep the wcid linked in the status poll list.
Fix this by only adding status skbs if the wcid pointer is still assigned
in dev->wcid, which gets cleared early by mt76_sta_pre_rcu_remove
Fixes: bd1e3e7b69 ("mt76: introduce packet_id idr")
Tested-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Reduce MT_TX_STATUS_SKB_TIMEOUT to 250ms
Tested-by: mrkiko.rs@gmail.com
Co-developed-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Substitute sk_buff_head status_list with spinlock_t status_lock since we
just need it for locking
Tested-by: mrkiko.rs@gmail.com
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Remove mt76_wcid pointer from mt76_tx_status_check signature since it is
always set to NULL
Tested-by: mrkiko.rs@gmail.com
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Introduce per-wcid idr to manage packet id for txs. This allow fast idr
lookup and skb queueing at the same time.
Tested-by: mrkiko.rs@gmail.com
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Report tx rate from tx status packets instead of receving periodic mcu
event. This improves flexibility, accuracy and AQL performance, and
simplifies code flow for better readability.
Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
In order to remove duplicated code, move mt76_get_next_pkt_id routine
in mt76.h
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Even if this is not a real issue since mt76_tx is never run with wcid set
to NULL, fix a theoretical NULL pointer dereference in mt76_tx routine
Fixes: db9f11d343 ("mt76: store wcid tx rate info in one u32 reduce locking")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Use ieee80211_tx_status_ext instead of ieee80211_free_skb and
ieee80211_tx_status. This makes it compatible with 802.3 encap offload
and improves performance by removing a redundant sta lookup
Signed-off-by: Felix Fietkau <nbd@nbd.name>
In order to remove duplicated code, move mt76_token_init in
mt76_alloc_device routine
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Move token management in mt76 common module since it is shared between
mt7615, mt7915 and mt7921 drivers
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
since tx/rx path is now relying pm ref counting, get rid of MT76_STATE_PM
check in the tx path
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Introduce mt7615_tx_worker routine as mt76 tx worker callback for
mt7663.
Rely on mt76_connac_pm_ref/mt76_connac_pm_unref to check PM state and
increment/decrement wake counter
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Add support to send larger packet in testmode to meet requirements
of some test cases.
The limit of max packet size is determined based on tx rate mode setting.
Signed-off-by: Shayne Chen <shayne.chen@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Since mt76_txq_send_burst routine can report a negative error code,
check the returned value before incrementing the number of transmitted
frames in mt76_txq_schedule_list routine.
Return -EBUSY directly if the device is in reset or in power management.
Fixes: 90fdc1717b ("mt76: use mac80211 txq scheduling")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
