This patch calls wcn36xx_smd_keep_alive_req() on the STA patch immediately
after associating with an AP.

This will cause the firmware to send a NULL packet out to the AP every 30
seconds, thus offloading keep-alive processing from the SoC to the
firmware.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

This patch switches on CONNECTION_MONITOR. Once done it is up to the
firmware to send keep alive and to monitor the link state.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

The firmware parameter LINK_FAIL_TX_CNT maps to the prima configuration
file parameter gLinkFailTxCnt and is described as:

quote: " If within gLinkFailTimeout period(values is mentioned in msec) if
         FW doesn't receive acks for gLinkFailTxCnt number of packets, then
         link will be disconnected."

The downstream description sets a minimum value of 1000 a maximum value of
60000 and a default value of 6000, however it appears that unless we
actually set this value deliberately firmware defaults it to 0.

Setting this value to non-zero results in the firmware doing link
monitoring. The working example from downstream paradoxically sets the
value to 200, here we opt to set the value to the minimum stipulated in the
configuration file 1000.

In conjunction with ieee80211_hw_set(wcn->hw, CONNECTION_MONITOR); this
change effects offload of link monitoring to the firmware.

Tested with:
'CNSS-PR-2-0-1-2-c1-74-130449-3' wcn3620
'CNSS-PR-2-0-1-2-c1-00083'       wcn3680

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Adds "qcom,pronto-v2-pd-pil" a binding that is similar to
"qcom,pronto-v2-pil" except vddmx and vddcx are implemented as power-domain
corners not via PMIC.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

This patch specifies the compatible string for the wcnss module as
"qcom,pronto-v2-pd-pil" which indicates to the PIL driver that it should
use power-domain logic when talking to vddmx and vddcx on the
MSM8939/APQ8039.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

This patch adds a pm domain specific regulator descriptor.
When using regulator corners on pronto-v2 hardware we need to map

- vddmx
- vddcx

as RPMD corner regulators. In contrast

- vddpx

is controlled via a PMIC as is the case on MSM8916 and MSM8997.

This patch adds a binding called "qcom,pronto-v2-pd-pil" and a probe time
table to go with that binding, describing pronto-v2 hardware with a
different regulator API.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Processors such as the MSM8936, APQ8036, MSM8939 and APQ8039 have voltage
rails that are based on fixed performance levels as opposed to specific
voltage or current load settings.

These corner regulators are presented via the RPMD interface.

In the case of WCNSS chips that would be paired with the above SoCs vddmx
and vddcx are attached to these corner regulators.

This patch provides a binding to represent the corner regulators in a
similar way to how the PMIC vregs are represented.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Loic Poulain authored 4 years ago and Bryan O'Donoghue committed 4 years ago

Instead of using the firmware generated sequence number, use the one
already allocated by the mac80211 layer. This allows better control
of the sequence number and avoid to rely on same sequence for Data,
QOS Data and QOS Null Data packets.

Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Loic Poulain authored 4 years ago and Bryan O'Donoghue committed 4 years ago

Some APs seems to have issues with too close wake/power-save
transition. Change the value to a less 'aggressive' one.

Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

qrtr_smd is required to communicate with the wcn36xx. Enable
CONFIG_QRTR_SMD=m now.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

To switch on the WiFi driver properly we need qrtr_smd.ko
qrtr_smd.ko depends on qrtr.ko - so switch on CONFIG_QRTR to get the base
qrtr.ko as a module now.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Ensure CONFIG_WCN36XX is switched on.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Benjamin Li authored 4 years ago and Bryan O'Donoghue committed 4 years ago

Signed-off-by: Benjamin Li <benl@squareup.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>

Shawn Guo authored 4 years ago and Bryan O'Donoghue committed 4 years ago

The square_defconfig was initially built from upstream defconfig where
quite many driver options are not needed for Square device.  Let's trim
it down for faster kernel build and smaller image size.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>

Shawn Guo authored 4 years ago and Bryan O'Donoghue committed 4 years ago

Follow commit 70a39be6 ("arm64: dts: msm8916: Update coresight
bindings for hardware ports") to apply the same update for msm8939
coresight devices.  The bonus point is that the following DTC warnings
will be gone.

 arch/arm64/boot/dts/qcom/msm8939-coresight.dtsi:58.11-63.6: Warning (graph_port): /soc/funnel@821000/ports/port@8: graph node unit address error, expected "0"
 arch/arm64/boot/dts/qcom/msm8939-coresight.dtsi:91.11-97.6: Warning (graph_port): /soc/replicator@824000/ports/port@2: graph node unit address error, expected "0"
 arch/arm64/boot/dts/qcom/msm8939-coresight.dtsi:119.11-124.6: Warning (graph_port): /soc/etf@825000/ports/port@1: graph node unit address error, expected "0"
 arch/arm64/boot/dts/qcom/msm8939-coresight.dtsi:304.11-309.6: Warning (graph_port): /soc/funnel@8e1000/ports/port@8: graph node unit address error, expected "0"

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>

Shawn Guo authored 4 years ago and Bryan O'Donoghue committed 4 years ago

With the msm8939 clock landed on mainline, we need to have the input
clocks to GCC controller "xo" and "sleep_clk" parsed from DT.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>

Shawn Guo authored 4 years ago and Bryan O'Donoghue committed 4 years ago

Commit e6717dba ("arm64: dts: qcom: msm8916: avoid using _ in node
names") changes pinctrl device node name.  Let's update uart1 pinctrl
node name in apq8039-t2-pinctl.dtsi accordingly.  Otherwise, we get
'duplicate_label' error from DTC.

While at it, add blsp1_uart1_default label, so that we will be reminded
by DTC if the label points to different node, just like sleep node.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>

Benjamin Li authored 4 years ago and Bryan O'Donoghue committed 4 years ago

Use the default BLSP function of GPIOs 0-1 (as is inherited by default
from msm8916-pins.dtsi), not the GPIO function.

Signed-off-by: Benjamin Li <benl@squareup.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>

Jun Nie authored 4 years ago and Bryan O'Donoghue committed 4 years ago

Switch on interconnect to build bus interconnect driver to adjust
bus bandwidth.

Signed-off-by: Jun Nie <jun.nie@linaro.org>

Jun Nie authored 5 years ago and Bryan O'Donoghue committed 4 years ago

Add interconnect controller for bus bandwidth management.

Signed-off-by: Jun Nie <jun.nie@linaro.org>

Jun Nie authored 4 years ago and Bryan O'Donoghue committed 4 years ago

Add driver for the Qualcomm interconnect buses found in MSM8939 based
platforms. The topology consists of four NoCs that are controlled by
a remote processor that collects the aggregated bandwidth for each
master-slave pairs.

Signed-off-by: Jun Nie <jun.nie@linaro.org>

Jun Nie authored 4 years ago and Bryan O'Donoghue committed 4 years ago

The Qualcomm MSM8939 platform has several bus fabrics that could be
controlled and tuned dynamically according to the bandwidth demand.

Signed-off-by: Jun Nie <jun.nie@linaro.org>

For the bluetooth stack to properly initialize we need to provide a default
bluetooth MAC.

For T2 production we can accomplish this on a device-specific basis with a
DT overlay.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Unlike the msm8916 the msm8939 uses power-domain corners for vddmx and
vddcx respectively.

Specify both of those in terms of the SMD power-domain driver now.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Enables the WiFi wcn36xx pronto driver on T2.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

This patch defaults the WiFi chipset for the MSM8939 to the WCN3680.
It is possible that WCN3660 or WCN3620 was paired with a shipping MSM8939
however that's unlikely to be the case and if it is the case, is the
outlier.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

When dumping the a DTB for the 8939 I saw the following error.

/soc/wcnss@a21b000: simple-bus unit address format error, expected "a204000"

This is because the first address given in the regs field is a204000 not
a21b000. Make dtc happy by specifying the base address correctly.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Enable TTY emulation over bluetooth RFCOMM.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Enables bluetooth specific networking.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Connects the bluetooth hardware to the WCNSS driver via SMD.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Switch on wireless extensions so we can have consistent behavior between T2
and db410c in terms of iwconfig output.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Switch on debugfs for this part.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Enables the CONFIG_WCN36XX option along with associated dependencies
automatically selected by it.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

In order to actually download the WCNSS firmware we need to enable an SMD
driver associated with the right channel/memory for that task.

CONFIG_QCOM_WCNSS_CTRL is the right driver for that, switch it on now.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Switch on CONFIG_QCOM_WCNSS_PIL to enable loading of Wireless Connectivity
Subsystem firmware.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Shawn Guo authored 4 years ago and Bryan O'Donoghue committed 4 years ago

We now have qfprom device for msm8939 added by commit 'arm64: dts:
msm8939: Add nvmem-cpufreq and cpr'.  Let's drop the one which was
initially copied from msm8916, to avoid the following error during boot.

[    3.744910] sysfs: cannot create duplicate filename '/bus/nvmem/devices/qfprom0'
[    3.745943] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.19.107-00369-gd85d7131c675-dirty #490
[    3.753403] Hardware name: Square, Inc. T2 Devkit (DT)
[    3.761821] Call trace:
[    3.766888]  dump_backtrace+0x0/0x158
[    3.769200]  show_stack+0x14/0x20
[    3.773029]  dump_stack+0x98/0xbc
[    3.776325]  sysfs_warn_dup+0x60/0x78
[    3.779612]  sysfs_do_create_link_sd.isra.0+0xdc/0xe8
[    3.783259]  sysfs_create_link+0x20/0x40
[    3.788302]  bus_add_device+0x68/0x130
[    3.792286]  device_add+0x3f8/0x628
[    3.795857]  nvmem_register.part.4+0x150/0x348
[    3.799233]  devm_nvmem_register+0x4c/0xa8
[    3.803745]  qfprom_probe+0x94/0xb8
[    3.807829]  platform_drv_probe+0x50/0xa0
[    3.811209]  really_probe+0x1b8/0x298
[    3.815377]  driver_probe_device+0x58/0x100
[    3.819022]  __driver_attach+0xe0/0xe8
[    3.823016]  bus_for_each_dev+0x74/0xc8
[    3.826835]  driver_attach+0x20/0x28
[    3.830567]  bus_add_driver+0x1ac/0x218
[    3.834386]  driver_register+0x60/0x110
[    3.837946]  __platform_driver_register+0x40/0x48
[    3.841777]  qfprom_driver_init+0x18/0x20
[    3.846629]  do_one_initcall+0x5c/0x178
[    3.850625]  kernel_init_freeable+0x198/0x244
[    3.854271]  kernel_init+0x10/0x108
[    3.858779]  ret_from_fork+0x10/0x18
[    3.862975] qcom,qfprom: probe of 5c000.qfprom failed with error -17

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>

Shawn Guo authored 4 years ago and Bryan O'Donoghue committed 4 years ago

Set ‘ret’ properly for error case to fix the build warning below.

drivers/clk/clk.c: In function ‘clk_register’:
drivers/clk/clk.c:3374:9: warning: ‘ret’ may be used uninitialized in this function [-Wmaybe-uninitialized]
  int i, ret;
         ^~~

Fixes: a494c5b9 ("clk: Aggregate power operation per clk")
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>

Jun Nie authored 4 years ago and Bryan O'Donoghue committed 4 years ago

Select CONFIG_REGULATOR_ONSEMI_NCP6335D as it is required by CPR
as backend power supply.

Signed-off-by: Jun Nie <jun.nie@linaro.org>