In QCA99X0 (QCA99X0, QCA9984, QCA9888 and QCA4019) family chips,
hw adds padding at the begining of the rx payload to make L3
header 4-byte aligned. In the chips doing this type of padding,
the number of bytes padded will be indicated through msdu_end:info1.
Define a hw_rx_desc_ops wrapper to retrieve the number of padded
bytes and use this while doing undecap. This should fix padding
related issues with ethernt decap format with QCA99X0, QCA9984,
QCA9888 and QCA4019 hw.
Signed-off-by: Vasanthakumar Thiagarajan <vthiagar@qti.qualcomm.com>
[Rename operations to hw_ops for other purposes]
Signed-off-by: Benjamin Berg <benjamin@sipsolutions.net>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Commit 3c97f5de1f ("ath10k: implement NAPI support") conflicts with
ath-current. To avoid any merge problems merge ath-current to ath-next already now.
Conflicts:
drivers/net/wireless/ath/ath10k/htt_rx.c
commit 7a0adc83f3 ("ath10k: improve tx scheduling") is causing
severe throughput drop in multi client mode. This issue is originally
reported in veriwave setup with 50 clients with TCP downlink traffic.
While increasing number of clients, the average throughput drops
gradually. With 50 clients, the combined peak throughput is decreased
to 98 Mbps whereas reverting given commit restored it to 550 Mbps.
Processing txqs for every tx completion is causing overhead. Ideally for
management frame tx completion, pending txqs processing can be avoided.
The change partly reverts the commit "ath10k: improve tx scheduling".
Processing pending txqs after all skbs tx completion will yeild enough
room to burst tx frames.
Fixes: 7a0adc83f3 ("ath10k: improve tx scheduling")
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Add NAPI support for rx and tx completion. NAPI poll is scheduled
from interrupt handler. The design is as below
- on interrupt
- schedule napi and mask interrupts
- on poll
- process all pipes (no actual Tx/Rx)
- process Rx within budget
- if quota exceeds budget reschedule napi poll by returning budget
- process Tx completions and update budget if necessary
- process Tx fetch indications (pull-push)
- push any other pending Tx (if possible)
- before resched or napi completion replenish htt rx ring buffer
- if work done < budget, complete napi poll and unmask interrupts
This change also get rid of two tasklets (intr_tq and txrx_compl_task).
Measured peak throughput with NAPI on IPQ4019 platform in controlled
environment. No noticeable reduction in throughput is seen and also
observed improvements in CPU usage. Approx. 15% CPU usage got reduced
in UDP uplink case.
DL: AP DUT Tx
UL: AP DUT Rx
IPQ4019 (avg. cpu usage %)
========
TOT +NAPI
=========== =============
TCP DL 644 Mbps (42%) 645 Mbps (36%)
TCP UL 673 Mbps (30%) 675 Mbps (26%)
UDP DL 682 Mbps (49%) 680 Mbps (49%)
UDP UL 720 Mbps (28%) 717 Mbps (11%)
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
With the %pK format specifier we hide the kernel addresses
with the help of kptr_restrict sysctl.
In this patch, %p is changed to %pK in the driver code.
The sysctl is documented in Documentation/sysctl/kernel.txt.
Signed-off-by: Maharaja Kennadyrajan <c_mkenna@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
It is expected that all pktlog events for 10.4 firmware based solutions
should come through CE8 where as in case of 10.2 firmware based solutions,
it should come through one of the HTT events (HTT_T2H_MSG_TYPE_PKTLOG).
But from experiments with 10.4 based solutions, it is observed that pktlog
event for ATH_PKTLOG_TYPE_TX_MSDU_ID is coming through HTT pktlog event.
Currently, we always parse with 10.2 pktlog header which will lead to
pktlog decoding issues (payload length mismatch exceptions)
For trace points, it is required to provide only the payload size. So
fixing this by simplifying the payload size calculation without the use of
ath10k pktlog headers.
While there, remove the unused ath10k pktlog headers.
Signed-off-by: Ashok Raj Nagarajan <arnagara@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
The below warning message seems to hit occasionally with the following
combination (IPQ4019 + ACS scan) where we receive packets as a self peer
when hostapd does ACS when we bring up AP mode . ath10k has the below
fall back mechanism to fetch current operating channel in rx (it will
check for the next channel tracking variable if the current one is NULL)
[scan channel] --> [rx channel] --> [peer channel] -->
[vdev channel] --> [any vdev channel] --> [target oper channel]
'scan channel' and 'target operating channel' are directly fetched from
firmware events. All the others should be updated by mac80211.
During ACS scan we wouldn't have a valid channel context
assigned from mac80211 ('ar->rx_channel'), and also relying on
('ar->scan_channel') is not helpful (it becomes NULL when it goes to
BSS channel and also when the scan event is completed). In short we
cannot always rely on these two channel tracking variables.
'Target Operating Channel' (ar->tgt_oper_chan) seems to keep track of
the current operating even while we are doing ACS scan and etc. Hence
remove this un-necessary warning message and continue with
target_operating channel. At the worst case scenario when the target
operating channel is invalid (NULL) we already have an ath10k warning
message to notify we really don't have a proper channel configured in
rx to update the rx status("no channel configured; ignoring frame(s)!")
WARNING: CPU: 0 PID: 0 at ath/ath10k/htt_rx.c:803
[<c0318838>] (warn_slowpath_null) from [<bf4a0104>]
(ath10k_htt_rx_h_channel+0xe0/0x1b8 [ath10k_core])
[<bf4a0104>] (ath10k_htt_rx_h_channel [ath10k_core]) from
[<bf4a025c>] (ath10k_htt_rx_h_ppdu+0x80/0x288 [ath10k_core])
[<bf4a025c>] (ath10k_htt_rx_h_ppdu [ath10k_core]) from
[<bf4a1a9c>] (ath10k_htt_txrx_compl_task+0x724/0x9d4 [ath10k_core])
[<bf4a1a9c>] (ath10k_htt_txrx_compl_task [ath10k_core])
Fixes:3b0499e9ce42 ("ath10k: reduce warning messages during rx without proper channel context")
Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Several cases of overlapping changes, except the packet scheduler
conflicts which deal with the addition of the free list parameter
to qdisc_enqueue().
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 5c86d97bcc ("ath10k: combine txrx and replenish task")
introduced deadlock while processing rx in order indication message
for qca6174 based devices. While merging replenish and txrx tasklets,
replenish task should be called out of htt rx ring locking since it
is also try to acquire the same lock.
Unfortunately this issue is not exposed by other solutions (qca988x,
qca99x0 & qca4019), as rx_in_ord_ind message is specific to qca6174
based devices. This patch fixes
=============================================
[ INFO: possible recursive locking detected ]
4.7.0-rc2-wt-ath+ #1353 Tainted: G E
---------------------------------------------
swapper/3/0 is trying to acquire lock:
(&(&htt->rx_ring.lock)->rlock){+.-...}, at: [<f8d7ef19>]
ath10k_htt_rx_msdu_buff_replenish+0x29/0x90 [ath10k_core]
but task is already holding lock:
(&(&htt->rx_ring.lock)->rlock){+.-...}, at: [<f8d82cab>]
ath10k_htt_txrx_compl_task+0x21b/0x250 [ath10k_core]
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(&(&htt->rx_ring.lock)->rlock);
lock(&(&htt->rx_ring.lock)->rlock);
*** DEADLOCK ***
May be due to missing lock nesting notation
1 lock held by swapper/3/0:
#0: (&(&htt->rx_ring.lock)->rlock){+.-...}, at: [<f8d82cab>]
ath10k_htt_txrx_compl_task+0x21b/0x250 [ath10k_core]
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=119151
Fixes: 5c86d97bcc ("ath10k: combine txrx and replenish task")
Reported-by: Mike Lothian <mike@fireburn.co.uk>
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Remove unused inline function phy_mode_to_band.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Legacy rate packets may not necessarily be having a rx status
flag of '0' always, for example management frame have flags
like RX_FLAG_ONLY_MONITOR / RX_FLAG_MACTIME_END also set
Just check 'VHT' and 'HT' flags are not set , and simply clasify it as
legacy rate packets
Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
WARN_ON_ONCE when we receive packets for self peer when mac80211
had not assigned a proper channel context. This scenario happens
in QCA4019 when we start the AP via hostapd in background and start
it once again in the background without killing the previous instance!
This happens intermittently when we start / stop hostapd in a while loop
(incase the hostapd is not properly killed). This results in mac80211
chancontext to be unassigned, while the self peer continuous receive
packets in target operating channel. This results in lot of call traces
in the rx path. Make this as a WARN_ON_ONCE to avoid flooding the console
which result in rebooting low memory systems, while still reporting the
warning once that we are receiving packets in target operating channel and
to indicate that something is happening which is not the expected result.
WARNING: CPU: 0 PID: 0 at ath/ath10k/htt_rx.c:803
[<c0318838>] (warn_slowpath_null) from [<bf4a0104>]
(ath10k_htt_rx_h_channel+0xe0/0x1b8 [ath10k_core])
[<bf4a0104>] (ath10k_htt_rx_h_channel [ath10k_core]) from
[<bf4a025c>] (ath10k_htt_rx_h_ppdu+0x80/0x288 [ath10k_core])
[<bf4a025c>] (ath10k_htt_rx_h_ppdu [ath10k_core]) from
[<bf4a1a9c>] (ath10k_htt_txrx_compl_task+0x724/0x9d4 [ath10k_core])
[<bf4a1a9c>] (ath10k_htt_txrx_compl_task [ath10k_core])
Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Recent changes revolving around implementing
wake_tx_queue support introduced a significant
performance regressions on some (slower, uni-proc)
systems.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
ath.git patches for 4.7. Major changes:
ath10k
* implement set_tsf() for 10.2.4 branch
* remove rare MSI range support
* remove deprecated firmware API 1 support
ath9k
* add module parameter to invert LED polarity
wcn36xx
* fixes to get the driver properly working on Dragonboard 410c
This enum is already perfectly aliased to enum nl80211_band, and
the only reason for it is that we get IEEE80211_NUM_BANDS out of
it. There's no really good reason to not declare the number of
bands in nl80211 though, so do that and remove the cfg80211 one.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Major changes:
iwlwifi
* support for Link Quality measurement
* more work 9000 devices and MSIx
* continuation of the Dynamic Queue Allocation work
* make the paging less memory hungry
* 9000 new Rx path
* removal of IWLWIFI_UAPSD Kconfig option
ath10k
* implement push-pull tx model using mac80211 software queuing support
* enable scan in AP mode (NL80211_FEATURE_AP_SCAN)
wil6210
* add basic PBSS (Personal Basic Service Set) support
* add initial P2P support
* add oob_mode module parameter
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
iQEcBAABAgAGBQJXC5ZnAAoJEG4XJFUm622bcLgIAKxTHWfcirfNpzLg39gfR/jx
vWTX8YifGdZUuQuaEJhc7pPZ6eiTyMlRLZsu9JDT0LErM6uwPKSOXkogzfiSFG00
ntXGOWfQlGdyFyQshOoHWlXSkfE7bBWpLiROgk7dS2oyObQr2ETBT/yVd1bkP6Ka
g13sgzM/a3CbxaZSllaK0S/YPQ6JhXlbeWotqOqE4SotYQD2QfLkYJiXSFRXNcQl
OWDPCTjI3yybOVRLTnLMn2CoWrRqtUElwXaWBKJPhgY3EyxlU6R6UPw8OL8DvXIw
EedpflIekKahwZJbipqdkKbhpxinUAa9kPsSkfCtlPJQWsiQvdwpG7R5KQmJ8Lc=
=A2Lw
-----END PGP SIGNATURE-----
Merge tag 'wireless-drivers-next-for-davem-2016-04-11' of git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/wireless-drivers-next
Kalle Valo says:
====================
wireless-drivers patches for 4.7
Major changes:
iwlwifi
* support for Link Quality measurement
* more work 9000 devices and MSIx
* continuation of the Dynamic Queue Allocation work
* make the paging less memory hungry
* 9000 new Rx path
* removal of IWLWIFI_UAPSD Kconfig option
ath10k
* implement push-pull tx model using mac80211 software queuing support
* enable scan in AP mode (NL80211_FEATURE_AP_SCAN)
wil6210
* add basic PBSS (Personal Basic Service Set) support
* add initial P2P support
* add oob_mode module parameter
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Decrement num_mpdus_ready only when rx amsdu is processed successfully.
Not doing so, will result in leak and impact stabilty under low memory
cases. Also commit 3128b3d8a2 ("ath10k: speedup htt rx descriptor
processing for rx_ind") missed to removed unused skb list rx_q.
Fixes: 3128b3d8a2 ("ath10k: speedup htt rx descriptor processing for rx_ind")
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
When HW crypto is used, there's no need for the CCMP/GCMP MIC to
be available to mac80211, and the hardware might have removed it
already after checking. The MIC is also useless to have when the
frame is already decrypted, so allow indicating that it's not
present.
Since we are running out of bits in mac80211_rx_flags, make
the flags field a u64.
Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Since tx completion and rx indication processing are moved out
of txrx tasklet and rx ring lock contention also removed from txrx
for rx_ind messages, it would be efficient to combine both replenish
and txrx tasks. Refill threshold is adjusted for both AP135 and AP148
(low and high end systems). With this adjustment in AP135, TCP DL is
improved from 603 Mbps to 620 Mbps and UDP DL is improved from 758 Mbps
to 803 Mbps. Also no watchdog are observed on UDP BiDi.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Except qca61x4 family chips (qca6164, qca6174), copy engine 5 is used
for receiving target to host htt messages. In follow up patch, CE5
descriptors will be reused. In such case, same API can not be used as
htc layer callback where the response messages will be freed at the end.
Hence register new API for HTC layer that free up received message and
keep the message handler common for both HTC and HIF layers.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
In follow up patch, htt rx descriptors will be reused instead of
dealloc and refill. To achieve that htt rx indication messages
should not be deferred and should be processed in pci tasklet itself.
Also from rx indication message, mpdu_count alone is used. So it is
maintained as atomic variable and all rx amsdu handlers are done
processed from txrx tasklet. This change get rid of rx_compl_q usage.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Make amsdu handlers (i.e amsdu_pop and rx_h_handler) common to both
rx_ind and frag_ind htt events. It is sufficient to hold rx_ring lock
for amsdu_pop alone and no need to hold it until the packets are
delivered to mac80211. This helps to reduce rx_lock contention as well.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
The fw descriptor was never used and probably never will be. It makes
little sense to maintain support for it. Remove it and simplify rx
processing. This will make it easier to optimize rx processing later
as well.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
To optmize CPU usage htt rx descriptors will be reused instead of
refilling it for htt rx copy engine (CE5). To support that all htt rx
indications should be proecssed at same context. Instead of queueing
actual indication message, queue copied message for txrx processing.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
To optimize CPU usage htt rx descriptors will be reused instead of
refilling it for htt rx copy engine (CE5). To support that all htt rx
indications should be processed at same context. FIFO queue is used
to maintain tx completion status for each msdu. This helps to retain
the order of tx completion.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Frames that are transmitted via MGMT_TX are using reserved descriptor
slots in firmware. This limitation is for the htt_mgmt_tx path itself,
not for mgmt frames per se. In 16 MBSSID scenario, these reserved slots
will be easy exhausted due to frequent probe responses. So for 10.4
based solutions, probe responses are limited by a threshold (24).
management tx path is separate for all except tlv based solutions. Since
tlv solutions (qca6174 & qca9377) do not support 16 AP interfaces, it is
safe to move management descriptor limitation check under mgmt_tx
function. Though CPU improvement is negligible, unlikely conditions or
never hit conditions in hot path can be avoided on data transmission.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Whenever firmware is configuring operating channel during scan or
home channel, channel change event will be indicated to host. In some
cases (device probe/ last vdev down), target will be configured to
default channel whereas host is unaware of target's operating channel.
This leads to packet drop due to unknown channel and kernel log will be
filled up with "no channel configured; ignoring frame(s)!". Fix that
by handling HTT_T2H_MSG_TYPE_CHAN_CHANGE event.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Until now only WMI originating mgmt frames were
reported to mac80211. Management frames on HTT
were basically dropped (except frames which looked
like management but had FCS error).
To allow sniffing all frames (including offloaded
frames) without interfering with mac80211
operation and states a new rx_flag was introduced
and is not being used to distinguish frames and
classify them for mac80211.
Signed-off-by: Grzegorz Bajorski <grzegorz.bajorski@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
The current/old tx path design was that host, at
its own leisure, pushed tx frames to the device.
For HTT there was ~1000-1400 msdu queue depth.
After reaching that limit the driver would request
mac80211 to stop queues. There was little control
over what packets got in there as far as
DA/RA was considered so it was rather easy to
starve per-station traffic flows.
With MU-MIMO this became a significant problem
because the queue depth was insufficient to buffer
frames from multiple clients (which could have
different signal quality and capabilities) in an
efficient fashion.
Hence the new tx path in 10.4 was introduced: a
pull-push mode.
Firmware and host can share tx queue state via
DMA. The state is logically a 2 dimensional array
addressed via peer_id+tid pair. Each entry is a
counter (either number of bytes or packets. Host
keeps it updated and firmware uses it for
scheduling Tx pull requests to host.
This allows MU-MIMO to become a lot more effective
with 10+ clients.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
This implements very basic support for software
queueing. It also contains some knobs that will be
patched later.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
This merely adds some parsing, generation and
sanity checks with placeholders for real
code/functionality to be added later.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
On multicore systems, it is possible that txrx tasket can run
in parallel with pci tasklet (i.e smp affinity of ath10k irq is
assigned to multiple CPUs). Feeding and consuming from the same
rx completion list leads to txrx tasklet runs for longer period.
Prevent this by processing a snapshot of rx queue by moving list
into temporary list. Consecutive received frames will be processed
in next batch.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Received frame indications are queued into a skb list and latest
processed by txrx tasklet. This skb queue is protected by htt rx lock.
Since the entire rx processing till delivering frame to mac80211 and
replenish tasks are processed under rx_lock protection, there might be
some delay in queuing newly received rx frame into that list on
multicore systems. Optimize this by using skb list lock while accessing
rx completion queue instead of htt rx lock.
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Currently, we are providing wrong payload data of pktlog to trace points.
Data we receive from FW through copy engine 8 contains pktlog data alone.
We don't need to parse anything in driver before handing it to trace
points.
Fixes: afb0bf7f53 ("ath10k: add support for pktlog in QCA99X0")
Signed-off-by: Ashok Raj Nagarajan <arnagara@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
This allows to use the new firmware which
implements the new tx data path. Without this
patch firmware supporting new tx path stops
responding shortly after booting.
This patch doesn't implement the entire pull-push
logic available in the new firmware. This will be
done in subsequent patches.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
On multicore systems it was possible for the txrx
tasklet to keep on running for long periods of
time on a single CPU due to tx completion
processing. Another CPU could feed the running
tasklet for an indefinite amount of time.
The tasklet is now guaranteed to run a finite
amount of time and is limited by HTT CE Rx ring
depth.
This improves behavior when RPS is used on target
system and might improve TCP handling as well.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Allocations from the DMA zone were originally added for legacy ISA
stuff, or PCI devices that have specific limitations in their DMA
addressing capabilities. It has no place in ath10k, which can do
full 32-bit DMA.
Fixes memory allocation errors on some platforms.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
The QCA99X0 4 addresses RX packets pad 2 bytes at the beginning of
MSDU instead the end of ieee80211 header to keep alignment. The currently RX
data path can't parse the header correctly in this case. This patch fixes it
for QCA99X0.
Signed-off-by: Yanbo Li <yanbol@qca.qualcomm.com>
[kvalo@qca.qualcomm.com: checkpatch fixes and naming changes]
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
The RX rate idx is not correct for 11G mode OFDM packet.
Because the bitrate table start with CCK index instead of OFDM.
Signed-off-by: Yanbo Li <yanbol@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
This patch adds pktlog support for 10.4 fw versions.
Signed-off-by: Vivek Natarajan <nataraja@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Some special copy engines delivers messages directly to HTT by
bypassing HTC layer. Hence exporting tx_completion and rx_handler
for delivering the data to HTT layer.
Reviewed-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Sometimes hardware reports invalid mcs index in rx descriptor
when operating in VHT80 mode and all packets with invalid mcs
will be eventually dropped in mac80211. This issue is observerd during
testing on QCA99X0 chipsets.
This patch adds a warn message for dumping the rx desc info which helps
in analysing the issue when invalid mcs is received.
Signed-off-by: Manikanta Pubbisetty <c_mpubbi@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
MU-MIMO Rx involves different interpretation of
the VHT-SIG-A compared to SU-MIMO.
The incorrect interpretation led ath10k to report
VHT MCS values greater than 9 which subsequently
prompted mac80211 to drop such frames. This
effectively broke Rx with MU-MIMO in many cases
and manifested with a kernel warning in the log
which looked like this:
[ 14.552520] WARNING: CPU: 2 PID: 0 at net/mac80211/rx.c:3578 ieee80211_rx+0x26c/0x940 [mac80211]()
[ 14.552522] Rate marked as a VHT rate but data is invalid: MCS: 10, NSS: 2
... call trace follows ...
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
This patch enables raw Rx/Tx encap mode to support software based
crypto engine. This patch introduces a new module param 'cryptmode'.
cryptmode:
0: Use hardware crypto engine globally with native Wi-Fi mode TX/RX
encapsulation to the firmware. This is the default mode.
1: Use sofware crypto engine globally with raw mode TX/RX
encapsulation to the firmware.
Known limitation:
A-MSDU must be disabled for RAW Tx encap mode to perform well when
heavy traffic is applied.
Testing: (by Michal Kazior <michal.kazior@tieto.com>)
a) Performance Testing
cryptmode=1
ap=qca988x sta=killer1525
killer1525 -> qca988x 194.496 mbps [tcp1 ip4]
killer1525 -> qca988x 238.309 mbps [tcp5 ip4]
killer1525 -> qca988x 266.958 mbps [udp1 ip4]
killer1525 -> qca988x 477.468 mbps [udp5 ip4]
qca988x -> killer1525 301.378 mbps [tcp1 ip4]
qca988x -> killer1525 297.949 mbps [tcp5 ip4]
qca988x -> killer1525 331.351 mbps [udp1 ip4]
qca988x -> killer1525 371.528 mbps [udp5 ip4]
ap=killer1525 sta=qca988x
qca988x -> killer1525 331.447 mbps [tcp1 ip4]
qca988x -> killer1525 328.783 mbps [tcp5 ip4]
qca988x -> killer1525 375.309 mbps [udp1 ip4]
qca988x -> killer1525 403.379 mbps [udp5 ip4]
killer1525 -> qca988x 203.689 mbps [tcp1 ip4]
killer1525 -> qca988x 222.339 mbps [tcp5 ip4]
killer1525 -> qca988x 264.199 mbps [udp1 ip4]
killer1525 -> qca988x 479.371 mbps [udp5 ip4]
Note:
- only open network tested for RAW vs nwifi performance comparison
- killer1525 (qca6174 hw2.2) is 2x2 device (hence max 866mbps)
- used iperf
- OTA, devices a few cm apart from each other, no shielding
- tcpX/udpX, X - means number of threads used
Overview:
- relative Tx performance drop is seen but is within reasonable and
expected threshold (A-MSDU must be disabled with RAW Tx)
b) Connectivity Testing
cryptmode=1
ap=iwl6205 sta1=qca988x crypto=open topology-1ap1sta OK
ap=iwl6205 sta1=qca988x crypto=wep1 topology-1ap1sta OK
ap=iwl6205 sta1=qca988x crypto=wpa topology-1ap1sta OK
ap=iwl6205 sta1=qca988x crypto=wpa-ccmp topology-1ap1sta OK
ap=qca988x sta1=iwl6205 crypto=open topology-1ap1sta OK
ap=qca988x sta1=iwl6205 crypto=wep1 topology-1ap1sta OK
ap=qca988x sta1=iwl6205 crypto=wpa topology-1ap1sta OK
ap=qca988x sta1=iwl6205 crypto=wpa-ccmp topology-1ap1sta OK
ap=iwl6205 sta1=qca988x crypto=open topology-1ap1sta2br OK
ap=iwl6205 sta1=qca988x crypto=wep1 topology-1ap1sta2br OK
ap=iwl6205 sta1=qca988x crypto=wpa topology-1ap1sta2br OK
ap=iwl6205 sta1=qca988x crypto=wpa-ccmp topology-1ap1sta2br OK
ap=qca988x sta1=iwl6205 crypto=open topology-1ap1sta2br OK
ap=qca988x sta1=iwl6205 crypto=wep1 topology-1ap1sta2br OK
ap=qca988x sta1=iwl6205 crypto=wpa topology-1ap1sta2br OK
ap=qca988x sta1=iwl6205 crypto=wpa-ccmp topology-1ap1sta2br OK
ap=iwl6205 sta1=qca988x crypto=open topology-1ap1sta2br1vlan OK
ap=iwl6205 sta1=qca988x crypto=wep1 topology-1ap1sta2br1vlan OK
ap=iwl6205 sta1=qca988x crypto=wpa topology-1ap1sta2br1vlan OK
ap=iwl6205 sta1=qca988x crypto=wpa-ccmp topology-1ap1sta2br1vlan OK
ap=qca988x sta1=iwl6205 crypto=open topology-1ap1sta2br1vlan OK
ap=qca988x sta1=iwl6205 crypto=wep1 topology-1ap1sta2br1vlan OK
ap=qca988x sta1=iwl6205 crypto=wpa topology-1ap1sta2br1vlan OK
ap=qca988x sta1=iwl6205 crypto=wpa-ccmp topology-1ap1sta2br1vlan OK
Note:
- each test takes all possible endpoint pairs and pings
- each pair-ping flushes arp table
- ip6 is used
c) Testbed Topology:
1ap1sta:
[ap] ---- [sta]
endpoints: ap, sta
1ap1sta2br:
[veth0] [ap] ---- [sta] [veth2]
| | | |
[veth1] | \ [veth3]
\ / \ /
[br0] [br1]
endpoints: veth0, veth2, br0, br1
note: STA works in 4addr mode, AP has wds_sta=1
1ap1sta2br1vlan:
[veth0] [ap] ---- [sta] [veth2]
| | | |
[veth1] | \ [veth3]
\ / \ /
[br0] [br1]
| |
[vlan0_id2] [vlan1_id2]
endpoints: vlan0_id2, vlan1_id2
note: STA works in 4addr mode, AP has wds_sta=1
Credits:
Thanks to Michal Kazior <michal.kazior@tieto.com> who helped find the
amsdu issue, contributed a workaround (already squashed into this
patch), and contributed the throughput and connectivity tests results.
Signed-off-by: David Liu <cfliu.tw@gmail.com>
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Tested-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
During tx completion, tx_lock is held for longer than required, preventing
efficient refill of htt->pending_tx. Refactor the code so that only MSDU
related operations are protected by the lock.
Improves downstream performance on a dual-core ARM Freescale LS1024A
(f.k.a. Mindspeed Comcerto 2000) AP with a 3x3 client from 495 to 580 Mbps.
Other CPU bound multicore systems may also benefit.
Signed-off-by: Denton Gentry <dgentry@google.com>
Signed-off-by: Avery Pennarun <apenwarr@google.com>
[mfaltesek@google.com: removed conflicting code for tracking msdu_ids.]
Signed-off-by: Marty Faltesek <mfaltesek@google.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
QCA99X0 chip has an extra 4 bytes in rx_msdu_start,
20 bytes in rx_msdu_end and 20 bytes in rx_ppdu_end structure
which are used in htt_rx_desc and HTT Rx ring offset setup.
This is necessary for correct Rx for QCA99X0 or Rx descriptors
will be overwritten and corrupted.
With this patch QCA988X and QCA6174 will have extra 44 bytes
padding in Rx descriptor layout which is harmless.
Signed-off-by: Peter Oh <poh@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
Vasanthakumar Thiagarajan