For Kiwi, if UMAC force wake is enabled, HIF window register
support is needed as well. currently hal_soc->init_phase just be
true for a very short periord, this lead to hif_force_wake_request
is called frequently when configure register during initialization,
this is not necessary as pld and device is not in power collapse
state during this period.
Enable HIF window register and increase init_phase true period
Change-Id: I0b5394bbc1ca73d20b2fcabbf2a261e6f8335626
CRs-Fixed: 3097991
The current platforms support 4 pci slots, so increasing
WLAN_CFG_MAX_PCIE_GROUPS to 4 to have unique irq name
Change-Id: Ifca25ee075611c75a5fa5e56782835ad89d26f91
Use the tracepoints to trace delayed register write, ce
tasklet scheduling latency, tx, and rx packets.
Change-Id: I63a89276177a9d0466dcb0c831eeb8e938a2bf79
CRs-Fixed: 3081870
Currently EP wake delay timeout is less and there is chance
of timeout to happen, so increase the timeout value.
Change-Id: I59261ade269908cd4d9c17025838807f39913999
CRs-Fixed: 3086547
Add support to affine individual grp irqs to either
perf or non-perf cluster for ipci.
Change-Id: Ifae0f2a5a77068587d99513a1902f850860f2329
CRs-Fixed: 3075398
For specific QCA6174 device, need clear LOW_POWER_INT_MASK register
bit3/4 to disable pcie low power interrupt after suspend, bit3 is for
INTR_L1SS and bit4 is for INTR_CLKPM, this is to fix issue that PMIC
VREG_L30A don't turn off when XO is in shutdown state.
Change-Id: I4833e4143c48a3bf734745a5e772a7a76a94d63c
CRs-Fixed: 3028269
Add HIF events to track NAPI POLL exit time, this gives
information about actual time spent by wlan host in napi softirq.
Change-Id: I7548fd05f2f7db714ac9bcf79171ffb3d478e936
CRs-Fixed: 3052902
Remove logic to enable/disable wake msi for drv systems. Today, during
WoW suspend sequence, between hif pci bus suspend and hif pci bus
suspend no irq, wake msi is getting disable temporarily. In case
fw toggles wake msi during this period, then the wake msi is lost,
which can lead to device not waking up for fw triggered wakeup.
Add the fix to remove enable/disable wake msi from suspend/resume
path and keep wake msi always enabled.
Change-Id: Idad0805f66ead4079d67577fa1151f73567511a3
CRs-Fixed: 3042261
Add check for verifying pld_is_pci_ep_awake support before waiting till
EP vote reset is done after reg completion
Change-Id: I9061c212c115d158aec63464eedcf6ca98e11a56
CRs-Fixed: 3038992
Update WMI CE buffer size to 3520 Bytes to support
large frame size on QCA6490 and WCN7850.
Change-Id: I1f3326f40bf18dda6010def57ea7f2e0961bc725
CRs-Fixed: 3036932
CE 12 and 13 is assigned for CV prefetch. Set attribute flag for these
CE to zero as per current requrement from FW on Emulation.
This may be changed to different value based on future requirement.
Change-Id: I70073aa45b6999a42ed230835a2a7dd7b650edad
CRs-Fixed: 3025570
Even though force wake request timed out there is possibility of
wake request got stuck and handled later at qmi level this leads
to leaving device in wake state for long time unnecessarily.
To avoid this send force wake release when wake request times out.
Change-Id: Ic0e1b3d3fe9d4b0e0f0d4b9be5804ef4f1d3e724
CRs-Fixed: 3019575
hif_ext_group and soc->intr_ctx[i].lro_ctx are allocated
but not free while fail to register parts of ext_group in
hif_register_ext_group, which cause memory leak. Fix is
to detach according resource before return.
Change-Id: I03ecc37437d09bf9275f128b1f2b320124a7df64
CRs-Fixed: 2988891
It is not right that linkstate_vote is 0 when entering
hif_vote_link_down, this indicates there is a mismatch
for vote up/down.
Enable debug to catch such case at first time.
Change-Id: I128bc42917ca5de949ee0d0b657579c14dd96eac
CRs-Fixed: 3009469
Currently there is mismatch in time-stamp source for
hif_hist_skip_event_record with respect to hif_hist_record_event
Updating time-stamp source in hif_hist_record_event to match with
hif_hist_skip_event_record time-stamp source.
Change-Id: I55680c40860eafb5c6ed6117ab7c55bb38cc24bb
CRs-Fixed: 3000319
CE buffer size update to support 3.5k beacon template &
probe response on Pine and Spruce targets.
The support is disabled on low memory profiles including
512M as there is no skb recycler in these profiles.
Change-Id: I451f11d6a20d1246b3e16221f2fb7ebd234b284a
In HMT, pktlog, which uses CE5 traditionally is disabled. Hence donot
configure any entries in target_ce_config.
Change-Id: I31a075fb9c6557599688408ba9452704c7522ec0
CRs-Fixed: 2984362
Current it use __RET_IP to get caller in rtpm resume
function, if it print info by %ps, it will do a
symbol lookup which takes time.
Change as replacing __RET_IP with rtpm_dbgid, so it is
still able to get caller information and avoid time
cost here.
Change-Id: Ifc70c118d621bb9e6d12de87582de09316ae9cad
CRs-Fixed: 3001690
%ps does symbol table lookup in kernel and if used in irq context,
it can lead to irq_handler running for longer time, and jank issues
as a side effect. So, remove it from wakeup_irq_handler.
Change-Id: Ifa42d9a46c838c2458d3220c6d3cb008232e3839
CRs-Fixed: 2999176
Based on the bitmap whether the issue is because of the
HIF_DETECT_TASKLET or HIF_DETECT_CREDIT, the hif_latency detection
code needs to be refactor to different functions so that it will
help parsing tools to find the reason of crash.
Change-Id: I2d8a83493354dfbd1c484cb7727018ddeb14c158
CRs-Fixed: 2929736
Currently sleep based wait is used for checking EP request/release
vote is taken effect or not. But once worker thread goes to sleep
it is taking in order of milli seconds to schedule back worker thread
this scheduling delay is causing srng HP/TP writes to be updated in
delayed manner.
To avoid this first try checking EP vote is taken affect by busy wait
then go for sleep based method.
Change-Id: I72e81941a23752a41c9b5307e82c2e971aee1ec8
CRs-Fixed: 2999541
IPA and rx_err ring are processed in same CPU resulting
in low Tput.
To fix this, not allow processing of rx_err ring in
CPU 0
Change-Id: Id53a03c9290607beb1a595c84bfb0fd8d9f5d105
CRs-Fixed: 2949569
Currently timeout for checking pld_is_pci_ep_awake in
hif_prevent_link_low_power_states might be higher than
the defined timeout depends on how long qdf_sleep takes
to get scheduled again, so avoid this uncertain timeout
adding reference to the system time.
Change-Id: I67a28d46c996ccc56680cf8e20c09a70729a33c5
CRs-Fixed: 2994746
After reg work is complete there is possiblity EP vote reset
request still not processed by F.W. So wait till EP vote reset
is done after reg work completion.
Change-Id: I1f4e318ac96ba3a15c613c5faf5095d4be7c8e99
CRs-Fixed: 2994233
Add api to set intermediate EP vote access state.
Make sure reg work handler is complete before setting
intermediate vote access.
Change-Id: Ib229d9af8b1e58505a9ae5fbf39aa912ca21a3c3
CRs-Fixed: 2994040
Currently, excess logging is causing a crash when there is no
data to reap.
Rate limiting the log messages to avoid the crash due to excessive
logging in case of no data to reap.
Change-Id: I175f3f1b3c0b0173f171e971494f6ffda694b4a5
CRs-Fixed: 2988147
Wake-IRQ can be received just before Kernel issuing
suspend_noirq callback, such wake-IRQ would not be
handled by Kernel. So check if any tasklet most likely
CE2 is pending and abort suspend_noirq if true.
Change-Id: Ie3d6de97179aa0d8f383a086f4f8df27d3fc5254
CRs-Fixed: 2981882
During deinit process we free the IRQs and call kill tasklet.
There is the possibility that Tasklet might be running when we
call kill tasklet and as part of tasklet exit it will try to
enable the corresponding CE IRQ which has already been freed.
Adding check for IRQ free condition to avoid any IRQ enable
calls after IRQ free is done.
Change-Id: I9d58ffbb3738251e3d8114631e2daacd07e33d7d
CRs-Fixed: 2973309
CE irqs affinity is set to gold cores, but affinity
hint is not removed while freeing irqs.
So remove CE irq affinity before freeing them.
Change-Id: If3822422f823cb3caa6e291387a2beadc62049b7
CRs-Fixed: 2975529
WCN7850 has support for near full indication for
the consumer srngs. This interrupt is used to take
preventive actions to avoid ring full watchdog irq
trigger.
Register for the near full irq and add the necessary
ext groups for these near-full irqs.
Change-Id: Ic16381fceabc54e6c52b34dd13abea74cad4d38c
CRs-Fixed: 2965081
In API hif_batch_send(), the variable ce_tx_hdl is dereferenced in
if_batch_send() without null check. Check for null before use to
avoid dereference a NULL pointer.
Change-Id: I05c38e180a448522d13d68369b77b45314eac227
CRs-Fixed: 2968584
Runtime put should be done in hif_pm_runtime_get when
return value of __hif_pm_runtime_get is negative. In
the issue scenario, the ret value will be set to 0 when
the return value of __hif_pm_runtime_get is positive.
The store operation for ret variable somehow did not take
effect and the if check for non-zero ret value to do
runtime put did go through. The return value to
dp_tx_hw_enqueue from hif_pm_runtime_get is the updated
ret value of 0. This will result in double runtime put
for a single runtime get.
Fix is to modify the ret variable check to negative
instead of non-zero to ensure runtime put happens
correctly.
Change-Id: Idc380a11c82b6d1acf7c750e7b93776ac9d6b4f2
CRs-Fixed: 2969879
Some latency stats need to do book keeping before interrupts are
enabled to avoid the counters being overwirtten by ISRs.
Also include latency stats in the case when rx is still pending.
Change-Id: Icff6f7876cd5db103693c6d38a396f1571b1d0aa
CRs-Fixed: 2965601
Four new CEs are introduced in QCN9224. Update MSI mapping
for 16 CEs in QCN9224.
Change-Id: I781e2b42453a4ec4f31f4076be1f5813d3637eff
CRs-Fixed: 2969994
