diff options
author | Raul E Rangel <rrangel@chromium.org> | 2020-07-14 14:32:02 -0600 |
---|---|---|
committer | Aaron Durbin <adurbin@chromium.org> | 2020-07-23 21:26:14 +0000 |
commit | 728c0787f2ad742da287791bf8f606b7d5921b0b (patch) | |
tree | 1be27a7c186915ee1043fb3d51c222b435093bcc /src/mainboard/google/zork/variants/dalboz | |
parent | be2d6541b98639de8e8e5492af526c022366df59 (diff) |
mb/google/zork: Don't expose reset GPIO for touchscreen to OS
The Raydium ACPI entry currently provides a reset GPIO and an _ON/_OFF
method to the kernel. These are contradictory. The ownership of the GPIO
should be mutually exclusive between either the OS or the FW. Since we
have two methods exposed this causes the OS to reset the TS twice. Once
using the _ON method, and once using the GPIO. Additionally the _ON
method is waiting for 20ms after reset while the OS driver uses a 50ms
delay. The Raydium TS datasheet specifies 20ms for FW ready time, so the
OS driver is adding additional padding.
The reference design has a 32ms rise time on the reset line. So without
this patch, the OS tries to reset the TS using the _ON method and it
waits for 20ms. This is not enough time for the reset line to reach
high, let alone account for the FW ready time. The OS driver then tries
to reset the device by toggling the GPIO. It waits 50ms which is still
2ms less than required.
This CL removes the GPIO from being exported in the _CRS so the OS
driver won't try and reset the device. It also increases the reset delay
by 32ms to account for the rise time.
This isn't a complete fix. I think that the slow rise time is causing
some kind of metastability in the TS reset hardware. Using a script to
bind and unbind the TS driver, the TS device becomes unresponsive after
~200 iterations. The only way to reset the device is to power cycle.
The TS power is also not currently controlled by the power resource.
This means that we have no guarantee over when the reset line is
toggled. This will lead to issues while spending and resuming.
BUG=b:160854397
TEST=Boot trembyle and make sure TS works. Suspend/Resume trembyle 300+
times.
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Change-Id: I23131be5d7109eed660a8bd6e2c156c015aa3c4e
Reviewed-on: https://review.coreboot.org/c/coreboot/+/43467
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-by: Furquan Shaikh <furquan@google.com>
Diffstat (limited to 'src/mainboard/google/zork/variants/dalboz')
-rw-r--r-- | src/mainboard/google/zork/variants/dalboz/overridetree.cb | 5 |
1 files changed, 4 insertions, 1 deletions
diff --git a/src/mainboard/google/zork/variants/dalboz/overridetree.cb b/src/mainboard/google/zork/variants/dalboz/overridetree.cb index bcfff1f4e0..9fe15c5d34 100644 --- a/src/mainboard/google/zork/variants/dalboz/overridetree.cb +++ b/src/mainboard/google/zork/variants/dalboz/overridetree.cb @@ -49,8 +49,11 @@ chip soc/amd/picasso register "probed" = "1" register "irq_gpio" = "ACPI_GPIO_IRQ_EDGE_LOW(GPIO_12)" register "reset_gpio" = "ACPI_GPIO_OUTPUT_ACTIVE_HIGH(GPIO_140)" - register "reset_delay_ms" = "20" + # 32ms: Rise time of the reset line + # 20ms: Firmware ready time + register "reset_delay_ms" = "32 + 20" register "has_power_resource" = "1" + register "disable_gpio_export_in_crs" = "1" device i2c 39 on end end chip drivers/i2c/generic |