A few functions in HIF component returns QDF status value with return
type as non QDF STATUS. For such functions, update the return type as
QDF_STATUS.
Change-Id: I69644a2206266ffe2c2d105056f9fec452f5d972
CRs-Fixed: 2734818
If disable pktlog for hasting/genoa, we need to
mark the ce5/ce7 not used separately to save
some memory. But currently, if set target ce
entries to 0, it has wlan fw crash issue,
so reduce it from 32 to 2 first. Once wlan
fw fix the crash issue, and then change to 0
Change-Id: I7ab11f202fe353737a4104e71fc79ee749b80fc8
hif_ahb_enable_bus returns -EIO, PTR_ERR and QDF_STATUS. PTR_ERR can
return zero for NULL. This will get interpreted as success resulting in
invalid access later. Fix return status to QDF_STATUS so that caller can
cosistently use error codes for error handling.
Change-Id: Ibd4d6e52c930272febf9a8b76bcfd2df4a4a4705
CRs-Fixed: 2726180
Use sync wake when trying to awake device for register
accessing.
This can avoid device re-enter low power mode after
device_wake asserted.
Change-Id: Iba62f5753ccb7e11d9746a445db4cce48e5e3a14
CRs-Fixed: 2703609
Fix WOW wake up interrupt from WLAN FW, in system
suspend/resume for Moselle platform.
Change-Id: I5f71c65e0aa1eeef8225f3fcfae5410233df8b78
CRs-Fixed: 2724142
Move the srng history to hif context.
Process the monitor ring only for the lmac
corresponding to the configured monitor channel.
Add the timer and srng history for monitor mode.
Change-Id: I4e5e49ad5e657b55bfafbb40ef5f59496277cb40
CRs-Fixed: 2690530
From kernel version 4.18 onward, the API to fetch pci_id
is deprecated.Assign "0" to pci_id for such case.
Change-Id: Ifb370999402ddc0889d982e251a7f8dea877b17d
The API of_get_pci_domain_nr() is deprecated from
kernel version 4.8 onwards.Hence, alternate API should
be used.
Change-Id: I95decbd2a855ec16f013168c8678816c8a017c81
Move all the Runtime PM related HIF code changes to the
common section, from if_pci.
So that, could be leveraged for other HIF interfaces
like if_ipci to use.
Change-Id: Id8303acd519ab07f307311971fb55c45476f30bf
CRs-Fixed: 2699419
Currently hif_info is used for printing irq details
in hif_pci_configure_grp_irq, this may cause excessive
logging so use hif_debug log type instead.
CRs-Fixed: 2696837
Change-Id: I75d79f6040ecb1fe30b2192ce3a34685af5370ab
Since cds hif context was already NULL at the time of
hif_pm_runtime_close and qdf_runtime_lock_deinit tries to get
cds hif context, replacing it with hif_runtime_lock_deinit
and corresponding hif_runtime_lock_init function.
Change-Id: I6226d15d2930978fdb178624cc0ffd807e9b256c
CRs-Fixed: 2714590
Since DP interrupts are not disabled at source, and enabled and
disabled only at GIC level, set IRQ_DISABLE_UNLAZY flag to avoid
spurious interrupt detection and avoid interrupt affinity failures.
Change-Id: Ie98d8cba6f93bccd4d508a28bebc216d2b74fe1f
On nbuf allocation failure in hif_post_recv_buffers_for_pipe
for the last buffer in the CE destination ring, oom_allocation
work is not getting scheduled since there is a mismatch
between the recv_buffers_needed and the num_entries - 1 of the
destination ring.
Fix is to schedule the oom_allocation work for srng
based CE with appropriate check of recv_buffers_needed
and destination ring num_entries.
Change-Id: Ie20d5b21358a401e60776db8a3a8a8644279f1fe
CRs-Fixed: 2705475
ic_irqnum saves the irq number given by platform,
existing max irq which can be saved per ce is 255,
if platform return irq greater than 255, value is
misinterpreted with that saved in a byte, this will
eventually have problem when host driver suspends/stop,
during which freeing irq will fail.
But in scenarios such as IPI mechanisms such as whunt,
since free irq fails, it would still assume irq is enabled
and call registered handlers, which will lead to
null pointer access.
Inorder to fix this changed data type to uint16_t.
Change-Id: I0d92f965462a0a6416ee3dce880c4218e5ab09e8
CRs-Fixed: 2700577
One parameter of proc_create_data is changed from struct
file_operations to struct proc_ops.
Add support for it under the kernel version compilation flag.
Change-Id: Ic87222f9c267debef56dcc66fc629854058ffabd
CRs-Fixed: 2666492
Add NULL pointer check in function hif_send_single() and
htc_dump_counter_info() for KW issues.
Change-Id: I4b020c566b14b28a825c13752dccf67b5f8bb837
CRs-Fixed: 2675205
Firmware dump is necessary to debug host panic issues because
host and firmware are out of sync.
MACRO DEBUG is undefined by default, after remove it,
hif_crash_shutdown->ol_copy_ramdump is called to collect firmware
dump.
Since dump depend on functions in bmi, add MACRO
WLAN_FEATURE_BMI check.
TargetFailure event wont't be received after inject assert due to
irqs disabled, so can't set target RESET state which is checked
when dump IRAM
Change-Id: Ia75f5032a2817e35e4c7e777857b137ba3998ebd
CRs-Fixed: 2682677
Sometimes, BMI download FW fail, cause destroy this uncreated spinlock.
Because fw download fail, hif_start() not called, so
&pipe_info->completion_freeq_lock not be created.But hif_ce_stop() call
qdf_spinlock_destroy(&pipe_info->completion_freeq_lock);Cause WLAN Panic
@ qdf_lock_stats_cookie_destroy:824: Lock destroyed twice or never created
hif_state->started be set to true in hif_start( )This fixes use
hif_state->started to judge hif_start() be called or not.If false,
hif_start() not be called, so needn’t destroy uncreated
completion_freeq_lock
Change-Id: Id6aae8024fe9df2ac6cf7976047ebd7a850438e4
Use %zu instead of %lu for type size_t to fix the compiling error
on 32bit system when enable the HIF_CE_DEBUG_DATA_BUF.
Change-Id: Ib11cf7a798514bf654141269b2cafc6fc198a48c
CRs-Fixed: 2688568
In ipq5018 CE registers(0x08400000) kept outside WCSS(0x0C000000) block.
As both regions are more than 60MB apart, not feasible to allocate
single resource which include both.
So, using a separate I/O region to access CE registers.
Change-Id: I67bb6d5ac82a1c0ed1d3e13f7776f9d69ee19956
There may be a race between the usage count check and pending MHI
activity which may result false panic. Remove this check from host
driver and add it back to platform driver after powering off MHI.
Change-Id: I9a87c596dc1eea5d2242099b870361a62494abcc
CRs-fixed: 2687834
Added unique irq name per pci slot so that affinity settings
can be altered as needed. Pci id will be read from dts, and
when there is no entry in dts, default value 0 will be used.
Also for all other target types except qcn9000, default value
0 will be used.
If WLAN_MAX_PDEVS is defined to 1, then always return pci id as
0 since there is only one pdev present
Change-Id: I8159ec8d8daecae687fcc286d620bfe9861dff98
Add support for HW version soc id and HW name string
for Hastings and HastingsPrime platform.
Change-Id: I1ba57339b2f9afd425e1ecc9f312ad484340d226
CRs-Fixed: 2681134
Currently as part of runtime PM, only the active
tasklets are being drained. For chips eg. QCA6390,
QCA6490 etc, there are grp_tasklets and delayed reg
write work which has to be drained before entering
runtime PM.
Add the logic to drain all the possible tasks
before entering runtime PM.
Change-Id: Ieb486f00fffd7346dcdc1faea6fed5850ef6daf7
CRs-Fixed: 2676000
Reduce the number of CEs used in IPQ5018 to 6 since it has only
one mac and requires only 6 CEs. CE7 and CE8 are reserved for FW
and not needed to be configured from host.
Change-Id: Ibe262c32d8b3234873391559d939bc198c658c60
Commit c81a71e488 ("qcacmn: Cleanup and refine runtime PM open/close
logic") addes the change to align default usage count for both runtime
PM enabled and disabled cases. So active usage count for runtime PM
enable case is 0 and for disabled case is 2. Refine the usage count
checking logic based on that change.
Change-Id: I4e78089760a28d10e6b37bb1a7b2d5242de6c111
CRs-fixed: 2676761
Do not vote for PCIe link for QCN7605 to allow PCIe
bus to go to low power mode.
Change-Id: I14e1a5d142497e3cbe8ed12874a428ab7d90afed
CRs-Fixed: 2649897
Change the error level logging which are called by holding
the spin_lock to debug level to avoid printing to console.
Change-Id: If057087e808ed49000eba8b9795397a0cf35a357
CRs-Fixed: 2676235
As part of HIF runtime PM close add assert to make sure no runtime
PM user has missed any put which may lead to runtime PM getting
disabled at runtime. This assert ensures all runtime PM users
have get and put in sync.
Change-Id: I6663be013e85b8a262ccc0f60c8bd9c59bac25ef
CRs-Fixed: 2665231
Runtime PM code has been refactoring multiple times so some open/close
logic may be duplicated and messed up. Cleanup and refine the logic.
In other to align both runtime PM enable and disable cases, make sure
default usage count is always 2 and sanitize the same when runtime PM
exit. hif_pm_runtime_stop() should always happen from HDD disable power
management APIs so clean it up from hif_pm_runtime_close().
Change-Id: I3b64369608c2d9aa4b3353503c12e0d05a57b592
CRs-fixed: 2666196
Add prefetch_timer configuration for CE rings.
Set prefetch_timer=1 configuration for qca6490 destination CEs,
prefetch_timer=0 configuration for other targets CEs.
Basically setting to 1us asking CE hw to update ring tail pointer to
update within 1us. FW side CE SW sets all rings to 1us already.
Idea behind this change is, we have seen pre-silicon issue where SRC
ring TP read by SW was not seen updated value when prefetch was set
to 8us. Changing prefetch timer value to 1us helps to resolve
pre-silicon issue.
So host side rings need to update the prefetch timer to 1us.
Change-Id: I0830c73517c29cf39e6b2974bf3faa44e5673741
CRs-Fixed: 2669762
During initialization of driver, the runtime pm usage
count is incremented to 1 when the pm state is NONE.
Runtime get is done as part of htc send packet for
a wmi command in scheduler thread context. In kworker,
runtime_start is processed as part of driver load and
runtime put is done prior to updating the pm state to
ON. Runtime put triggered as part of htc send packet
causes a panic since the state is NONE and usage
count is 1.
Fix is to set runtime pm state to ON prior to doing
a runtime put as part of initialization.
Change-Id: I52cca5240f2f0872c681aab3a58a382f3fa1df0e
CRs-Fixed: 2669029
In order to avoid panic in hif_pci_runtime_pm_warn(), call
qdf_is_fw_down() to check if fw is down. If fw is down, return without
panicking.
Change-Id: I3fd3e0334220c1cf3ae6b3ec3fd5b1ede1705013
CRs-Fixed: 2664133
Change DP interrupt names to unique values so that
interrupt affinity can be adjusted based on this. This is
similar to what is being done for Copy Engine interrupts.
Change-Id: I7f1789c8c0103e7c01b0a2956a7a37149c7462d1
Currently, athdiag can't work with QCN7605.
Thus made below change to make it work:
1. Skip pci address boundary check for QCN7605 as QCA6390.
2. Make QCN7605 access target memory via QMI.
Change-Id: Ibfcccb57486c99d78c02e322c4e2ae1a7c8dab81
CRs-Fixed: 2653814
Currently there are cases where the CE status ring
head pointer is incremented, but the corresponding
buffer in the CE DST ring has nbytes as zero. This
can be a possible case where the head pointer was
updated before the DMA copy was completed and before
the nbytes could be updated with the number of bytes
copied to the CE DST ring buffer.In such cases we
haven't processed this entry, and hence should not
increment the tail pointer.
Hence we do not increment the tail pointer if
there was no data in the CE DST ring. Also
enable the logging of CE STATUS RING REAP
even if the CE DST reap was not successful.
This will help identify the cases where the reap
was not successful because of nbytes being zero.
In Success case, the CE desc history will show
STATUS_RING_REAP and DST_RING_REAP events.
In nbytes=0 failure case, the CE desc history will
show STATUS_RING_REAP event with HP moving ahead,
but the TP will not have moved.
Change-Id: Ibc3145142b6c88f6da3e12748d0ac8090486e8e3
CRs-Fixed: 2657285
Update QCA_NAPI_DEF_SCALE_BIN_SHIFT for defconfig builds to
the same value as the perf build.
Move the WLAN ext irqs to gold cores for defconfig builds.
Disable cpu isolation before moving the IRQs to gold cores.
Add the ability to move IRQs to gold cores when the cpus
hotplug in.
Change-Id: I4cfecd02a1a2200dc99adee9a324b90c877c13fd
CRs-Fixed: 2638820
This is to enhance statics for runtime put/get, which is
to detect if there is mismatch for usage_count.
Change-Id: I24cddb9d10e4cb675c8375cbd0f589c7718bd680
CRs-Fixed: 2647972
Macro DMA_COHERENT_MASK_IPA_VER_3_AND_ABOVE is used in non
IPA offload case also so bring this macro definition out of
IPA offload feature and change it to DMA_COHERENT_MASK_DEFAULT.
Change-Id: Ib03aa8ae2938e1de1a957660923a711d452a9337
CRs-Fixed: 2649909
Add HIF_INFO to log the linkstate_vote value in the hif_vote_link_up()
and hif_vote_link_down() functions in order to log the occurences of
the rare event when we vote for link up/down.
Change-Id: I7549b3400977583bd2cd4ec4fcb3f16dd0865ce7
CRs-Fixed: 2648860
Currently HIF runtime PM APIs will check runtime PM enablement using
pm_runtime_enabled() by checking for disable_depth which can be changed
at any time. This may cause usage count imbalance since the HIF API may
skip calling the corresponding runtime PM framework API to increase or
decrease usage count. Enhance runtime PM enablement check so that it
will always honor WLAN ini config first.
Change-Id: I4b2863ef7fa0dcd6fd61b776e6cb0ab109ffb3ec
CRs-fixed: 2647986
