[U-Boot] [PATCH 1/2] x86: baytrail: Fix unstable ACPI S3 resume

It was observed that when booting a Ubuntu 16.04 kernel, doing ACPI S3 suspend/resume sometimes causes the Ubuntu kernel hang forever. The issue is however not reproduced with a kernel built from i386/ x86_64 defconfig configuration.
The unstability is actually caused by unexpected interrupts being generated during the S3 resume. For some unknown reason, FSP (gold4) for BayTrail configures the GPIO DFX5 PAD to enable level interrupt (bit 24 and 25). As this pin keeps generating interrupts during an S3 resume, and there is no IRQ requester in the kernel to handle it, the kernel seems to hang and does not continue resuming.
Clear the mysterious interrupt bits for this pin.
Reported-by: Stefan Roese sr@denx.de Signed-off-by: Bin Meng bmeng.cn@gmail.com ---
arch/x86/cpu/baytrail/valleyview.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)
diff --git a/arch/x86/cpu/baytrail/valleyview.c b/arch/x86/cpu/baytrail/valleyview.c index c58f6a8..9af1bda 100644 --- a/arch/x86/cpu/baytrail/valleyview.c +++ b/arch/x86/cpu/baytrail/valleyview.c @@ -10,6 +10,13 @@ #include <asm/irq.h> #include <asm/mrccache.h> #include <asm/post.h> +#include <asm/arch/iomap.h> + +/* GPIO SUS */ +#define GPIO_SUS_PAD_BASE (IO_BASE_ADDRESS + IO_BASE_OFFSET_GPSSUS) +#define GPIO_SUS_DFX5_CONF0 0x150 +#define BYT_TRIG_LVL BIT(24) +#define BYT_TRIG_POS BIT(25)
#ifndef CONFIG_EFI_APP int arch_cpu_init(void) @@ -33,6 +40,21 @@ int arch_misc_init(void) mrccache_save(); #endif
+ /* + * For some unknown reason, FSP (gold4) for BayTrail configures + * the GPIO DFX5 PAD to enable level interrupt (bit 24 and 25). + * This does not cause any issue when Linux kernel runs w/ or w/o + * the pinctrl driver for BayTrail. However this causes unstable + * S3 resume if the pinctrl driver is included in the kernel build. + * As this pin keeps generating interrupts during an S3 resume, + * and there is no IRQ requester in the kernel to handle it, the + * kernel seems to hang and does not continue resuming. + * + * Clear the mysterious interrupt bits for this pin. + */ + clrbits_le32(GPIO_SUS_PAD_BASE + GPIO_SUS_DFX5_CONF0, + BYT_TRIG_LVL | BYT_TRIG_POS); + return 0; }

This is only needed when graphics console is used. For kernel with native graphics driver, this can be turned off to speed up.
Change this option's default to n in the Kconfig.
Signed-off-by: Bin Meng bmeng.cn@gmail.com ---
arch/x86/Kconfig | 1 - 1 file changed, 1 deletion(-)
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index 38a6187..c869ae2 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -677,7 +677,6 @@ config HAVE_ACPI_RESUME config S3_VGA_ROM_RUN bool "Re-run VGA option ROMs on S3 resume" depends on HAVE_ACPI_RESUME - default y if HAVE_ACPI_RESUME help Execute VGA option ROMs in U-Boot when resuming from S3. Normally this is needed when graphics console is being used in the kernel.

On 12.10.2017 14:07, Bin Meng wrote:
This is only needed when graphics console is used. For kernel with native graphics driver, this can be turned off to speed up.
Change this option's default to n in the Kconfig.
Signed-off-by: Bin Meng bmeng.cn@gmail.com
arch/x86/Kconfig | 1 - 1 file changed, 1 deletion(-)
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index 38a6187..c869ae2 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -677,7 +677,6 @@ config HAVE_ACPI_RESUME config S3_VGA_ROM_RUN bool "Re-run VGA option ROMs on S3 resume" depends on HAVE_ACPI_RESUME
- default y if HAVE_ACPI_RESUME help Execute VGA option ROMs in U-Boot when resuming from S3. Normally this is needed when graphics console is being used in the kernel.
Reviewed-by: Stefan Roese sr@denx.de
Thanks, Stefan

On Fri, Oct 13, 2017 at 7:09 PM, Stefan Roese sr@denx.de wrote:
On 12.10.2017 14:07, Bin Meng wrote:
This is only needed when graphics console is used. For kernel with native graphics driver, this can be turned off to speed up.
Change this option's default to n in the Kconfig.
Signed-off-by: Bin Meng bmeng.cn@gmail.com
arch/x86/Kconfig | 1 - 1 file changed, 1 deletion(-)
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index 38a6187..c869ae2 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -677,7 +677,6 @@ config HAVE_ACPI_RESUME config S3_VGA_ROM_RUN bool "Re-run VGA option ROMs on S3 resume" depends on HAVE_ACPI_RESUME
default y if HAVE_ACPI_RESUME help Execute VGA option ROMs in U-Boot when resuming from S3.
Normally this is needed when graphics console is being used in the kernel.
Reviewed-by: Stefan Roese sr@denx.de
applied to u-boot-x86, thanks!

On 12.10.2017 14:07, Bin Meng wrote:
It was observed that when booting a Ubuntu 16.04 kernel, doing ACPI S3 suspend/resume sometimes causes the Ubuntu kernel hang forever. The issue is however not reproduced with a kernel built from i386/ x86_64 defconfig configuration.
The unstability is actually caused by unexpected interrupts being generated during the S3 resume. For some unknown reason, FSP (gold4) for BayTrail configures the GPIO DFX5 PAD to enable level interrupt (bit 24 and 25). As this pin keeps generating interrupts during an S3 resume, and there is no IRQ requester in the kernel to handle it, the kernel seems to hang and does not continue resuming.
Clear the mysterious interrupt bits for this pin.
Thanks for working on this. This must have been very tough to debug and fix.
Reported-by: Stefan Roese sr@denx.de Signed-off-by: Bin Meng bmeng.cn@gmail.com
Tested-by: Stefan Roese sr@denx.de Reviewed-by: Stefan Roese sr@denx.de
Thanks, Stefan

On Fri, Oct 13, 2017 at 7:09 PM, Stefan Roese sr@denx.de wrote:
On 12.10.2017 14:07, Bin Meng wrote:
It was observed that when booting a Ubuntu 16.04 kernel, doing ACPI S3 suspend/resume sometimes causes the Ubuntu kernel hang forever. The issue is however not reproduced with a kernel built from i386/ x86_64 defconfig configuration.
The unstability is actually caused by unexpected interrupts being generated during the S3 resume. For some unknown reason, FSP (gold4) for BayTrail configures the GPIO DFX5 PAD to enable level interrupt (bit 24 and 25). As this pin keeps generating interrupts during an S3 resume, and there is no IRQ requester in the kernel to handle it, the kernel seems to hang and does not continue resuming.
Clear the mysterious interrupt bits for this pin.
Thanks for working on this. This must have been very tough to debug and fix.
Reported-by: Stefan Roese sr@denx.de Signed-off-by: Bin Meng bmeng.cn@gmail.com
Tested-by: Stefan Roese sr@denx.de Reviewed-by: Stefan Roese sr@denx.de
applied to u-boot-x86, thanks!
participants (2)
-
Bin Meng
-
Stefan Roese