Revert "bloonchipper: Reintroduce sleep lines to RO"

This reverts commit 5e2a0808b7dfc736676daf8ba4d6cf683076d961.

Reason for revert: It may broke CQ
BUG=b:195718112

Original change's description:
> bloonchipper: Reintroduce sleep lines to RO
>
> Change https://crrev.com/c/2673909 removed the sleep lines from RO
> in an effort to minimize the RO complexity.
> Most notably, this isolated the deep-sleep/low-power-idle
> logic to RW only.
>
> Unfortunately, the sleep lines also control whether the SPI host interface
> is listening, which allows it to ignore spurious communication.
>
> It seems safer to reinstate the the sleep line with low power idle
> active and directly disable CONFIG_LOW_POWER_IDLE in subsequent CL.
>
> We reinstate the sleep line gpio logic from the following:
> https://crrev.com/c5545464431669029f42829d542fa491d767ee5f/board/hatch_fp/board.c
> This is the parent commit to the CL that refactors the sleep lines.
>
> BRANCH=none
> BUG=b:178746753
> TEST=# Connect servo_micro to a dragonclaw board.
>      make proj-bloonchipper -j
>      sudo servod --board=dragonclaw
>      ./util/flash_ec --board=bloonchipper
>
>      dut-control fpmcu_slp:off fpmcu_slp_alt:off
>      dut-control pp3300_dx_mcu_mw # Should be around 20mw
>      dut-control fpmcu_slp:on fpmcu_slp_alt:off
>      dut-control pp3300_dx_mcu_mw # Should be less the 5mw
>      dut-control fpmcu_slp:off fpmcu_slp_alt:on
>      dut-control pp3300_dx_mcu_mw # Should be less the 5mw
>      dut-control fpmcu_slp:on fpmcu_slp_alt:on
>      dut-control pp3300_dx_mcu_mw # Should be less the 5mw
>      dut-control fpmcu_slp:off fpmcu_slp_alt:off
>
>      minicom -D$(dut-control -o raw_fpmcu_console_uart_pty)
>      > reboot ro
>      # Ctrl-A Q
>
>      dut-control fpmcu_slp:off fpmcu_slp_alt:off
>      dut-control pp3300_dx_mcu_mw # Should be around 20mw
>      dut-control fpmcu_slp:on fpmcu_slp_alt:off
>      dut-control pp3300_dx_mcu_mw # Should be less the 5mw
>      dut-control fpmcu_slp:off fpmcu_slp_alt:on
>      dut-control pp3300_dx_mcu_mw # Should be less the 5mw
>      dut-control fpmcu_slp:on fpmcu_slp_alt:on
>      dut-control pp3300_dx_mcu_mw # Should be less the 5mw
>      dut-control fpmcu_slp:off fpmcu_slp_alt:off
>
>      minicom -D$(dut-control -o raw_fpmcu_console_uart_pty)
>      > reboot
>      > fpenroll
>      > fpmatch
>      > reboot
>      # Ctrl-A Q
>
> Signed-off-by: Craig Hesling <hesling@chromium.org>
> Change-Id: I503cb3b62740300b265a4ddb165e29d9e36727fd
> Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3012241
> Commit-Queue: Josie Nordrum <josienordrum@google.com>
> Reviewed-by: Josie Nordrum <josienordrum@google.com>

Bug: b:178746753
Change-Id: If610fdffaef778f26c712bbad360e84651df61bf
Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/3077589
Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Owners-Override: Hsu Wei-Cheng <mojahsu@chromium.org>
Auto-Submit: Hsu Wei-Cheng <mojahsu@chromium.org>

1 files changed, 74 insertions, 0 deletions

diff --git a/board/hatch_fp/board_rw.c b/board/hatch_fp/board_rw.c
index 35709860e3..09c78c5e5c 100644
--- a/board/hatch_fp/board_rw.c
+++ b/board/hatch_fp/board_rw.c
@@ -7,8 +7,10 @@
 #include "console.h"
 #include "fpsensor_detect.h"
 #include "gpio.h"
+#include "hooks.h"
 #include "registers.h"
 #include "spi.h"
+#include "system.h"
 #include "task.h"
 #include "usart_host_command.h"
 #include "util.h"
@@ -17,6 +19,45 @@
 #error "This file should only be built for RW."
 #endif
 
+/*
+ * Some platforms have a broken SLP_S0_L signal (stuck to 0 in S0)
+ * if set, ignore it and only uses SLP_S3_L for the AP state.
+ */
+static bool broken_slp;
+
+static void ap_deferred(void)
+{
+	/*
+	 * Behavior:
+	 * AP Active  (ex. Intel S0):   SLP_L is 1
+	 * AP Suspend (ex. Intel S0ix): SLP_L is 0
+	 * The alternative SLP_ALT_L should be pulled high at all the times.
+	 *
+	 * Legacy Intel behavior:
+	 * in S3:   SLP_ALT_L is 0 and SLP_L is X.
+	 * in S0ix: SLP_ALT_L is 1 and SLP_L is 0.
+	 * in S0:   SLP_ALT_L is 1 and SLP_L is 1.
+	 * in S5/G3, the FP MCU should not be running.
+	 */
+	int running = gpio_get_level(GPIO_SLP_ALT_L) &&
+		      (gpio_get_level(GPIO_SLP_L) || broken_slp);
+
+	if (running) { /* S0 */
+		disable_sleep(SLEEP_MASK_AP_RUN);
+		hook_notify(HOOK_CHIPSET_RESUME);
+	} else { /* S0ix/S3 */
+		hook_notify(HOOK_CHIPSET_SUSPEND);
+		enable_sleep(SLEEP_MASK_AP_RUN);
+	}
+}
+DECLARE_DEFERRED(ap_deferred);
+
+/* PCH power state changes */
+void slp_event(enum gpio_signal signal)
+{
+	hook_call_deferred(&ap_deferred_data, 0);
+}
+
 /* SPI devices */
 struct spi_device_t spi_devices[] = {
 	/* Fingerprint sensor (SCLK at 4Mhz) */
@@ -46,6 +87,39 @@ static void configure_fp_sensor_spi(void)
 
 void board_init_rw(void)
 {
+	enum fp_transport_type ret_transport = get_fp_transport_type();
+
+	/*
+	 * FP_RST_ODL pin is defined in gpio_rw.inc (with GPIO_OUT_HIGH
+	 * flag) but not in gpio.inc, so RO leaves this pin set to 0 (reset
+	 * default), but RW doesn't initialize this pin to 1 because sysjump
+	 * to RW is a warm reset (see gpio_pre_init() in chip/stm32/gpio.c).
+	 * Explicitly reset FP_RST_ODL pin to default value.
+	 */
+	gpio_reset(GPIO_FP_RST_ODL);
+
+	if (ret_transport == FP_TRANSPORT_TYPE_UART) {
+		/*
+		 * The Zork variants currently have a broken SLP_S0_L signal
+		 * (stuck to 0 in S0). For now, unconditionally ignore it here
+		 * as they are the only UART users and the AP has no S0ix state.
+		 * TODO(b/174695987) once the RW AP firmware has been updated
+		 * on all those machines, remove this workaround.
+		 */
+		broken_slp = true;
+	}
+
 	/* Configure and enable SPI as master for FP sensor */
 	configure_fp_sensor_spi();
+
+	/* Enable interrupt on PCH power signals */
+	gpio_enable_interrupt(GPIO_SLP_ALT_L);
+	gpio_enable_interrupt(GPIO_SLP_L);
+
+	/*
+	 * Enable the SPI slave interface if the PCH is up.
+	 * Do not use hook_call_deferred(), because ap_deferred() will be
+	 * called after tasks with priority higher than HOOK task (very late).
+	 */
+	ap_deferred();
 }