Patch set updated for coreboot: e72c3ee hp/pavilion_m6_1035dx: Add SMI handler and handle EC requests
Alexandru Gagniuc (mr.nuke.me@gmail.com) just uploaded a new patch set to gerrit, which you can find at http://review.coreboot.org/5512
-gerrit
commit e72c3eec14b5772936d2608ede4e1dbed8198949 Author: Alexandru Gagniuc mr.nuke.me@gmail.com Date: Wed Apr 9 12:24:39 2014 -0500
hp/pavilion_m6_1035dx: Add SMI handler and handle EC requests
The EC may disable some functionality, such as Caps Lock LED and battery charging if it never receives a command to go in APM mode. If we start it in APM mode, then immediately switch to ACPI mode, it will not get its SCIs serviced until an ACPI OS boots. If its SCIs are not serviced, it may assume the OS has hung.
The way we solve this is to initalize the EC in APM mode, and only switch it to ACPI when an ACPI-capable OS issues the ACPI_ENABLE command. The switch has to be handled in SMM.
Although we aren't yet processing SMIs from the EC, we are reading the status in order to satisfy the EC that the event is handled.
Change-Id: Iffaeb9a6f57841f456c4bce8337dc09b287f8758 Signed-off-by: Alexandru Gagniuc mr.nuke.me@gmail.com --- src/mainboard/hp/pavilion_m6_1035dx/Kconfig | 1 + src/mainboard/hp/pavilion_m6_1035dx/Makefile.inc | 2 + src/mainboard/hp/pavilion_m6_1035dx/ec.c | 18 +++-- src/mainboard/hp/pavilion_m6_1035dx/ec.h | 4 ++ src/mainboard/hp/pavilion_m6_1035dx/mainboard.c | 5 ++ .../hp/pavilion_m6_1035dx/mainboard_smi.c | 77 ++++++++++++++++++++++ 6 files changed, 102 insertions(+), 5 deletions(-)
diff --git a/src/mainboard/hp/pavilion_m6_1035dx/Kconfig b/src/mainboard/hp/pavilion_m6_1035dx/Kconfig index 87e8212..d6cdb10 100644 --- a/src/mainboard/hp/pavilion_m6_1035dx/Kconfig +++ b/src/mainboard/hp/pavilion_m6_1035dx/Kconfig @@ -30,6 +30,7 @@ config BOARD_SPECIFIC_OPTIONS # dummy select HAVE_PIRQ_TABLE select HAVE_MP_TABLE select HAVE_ACPI_RESUME + select HAVE_SMI_HANDLER select SERIAL_CPU_INIT select HAVE_ACPI_TABLES select BOARD_ROMSIZE_KB_4096 diff --git a/src/mainboard/hp/pavilion_m6_1035dx/Makefile.inc b/src/mainboard/hp/pavilion_m6_1035dx/Makefile.inc index 07ea765..7039c72 100644 --- a/src/mainboard/hp/pavilion_m6_1035dx/Makefile.inc +++ b/src/mainboard/hp/pavilion_m6_1035dx/Makefile.inc @@ -27,3 +27,5 @@ ramstage-y += agesawrapper.c ramstage-y += BiosCallOuts.c ramstage-y += PlatformGnbPcie.c ramstage-y += ec.c + +smm-y += mainboard_smi.c diff --git a/src/mainboard/hp/pavilion_m6_1035dx/ec.c b/src/mainboard/hp/pavilion_m6_1035dx/ec.c index d61a2e5..2dd0009 100644 --- a/src/mainboard/hp/pavilion_m6_1035dx/ec.c +++ b/src/mainboard/hp/pavilion_m6_1035dx/ec.c @@ -13,16 +13,24 @@ static void set_keyboard_matrix_us(void) ec_kbc_write_ib(0xE5); }
-/* Tell EC to operate in ACPI mode, thus generating SCIs on events, not SMIs */ -static void enter_acpi_mode(void) +/* Tell EC to operate in APM mode. Events generate SMIs instead of SCIs */ +static void enter_apm_mode(void) { ec_kbc_write_cmd(0x59); - ec_kbc_write_ib(0xE8); + ec_kbc_write_ib(0xE9); }
void pavilion_m6_1035dx_ec_init(void) { set_keyboard_matrix_us(); - /* This could also be done in an SMI, should we decide to use SMM */ - enter_acpi_mode(); + + /* + * The EC has a special "blinking Caps Lock LED" mode which it normally + * enters when it believes the OS is not responding. It occasionally + * disables battery charging when in this mode, although other + * functionality is unaffected. Although the EC starts in APM mode by + * default, it only leaves the "blinking Caps Lock LED" mode after + * receiving the following command. + */ + enter_apm_mode(); } diff --git a/src/mainboard/hp/pavilion_m6_1035dx/ec.h b/src/mainboard/hp/pavilion_m6_1035dx/ec.h index 0c00d7a..579bf7e 100644 --- a/src/mainboard/hp/pavilion_m6_1035dx/ec.h +++ b/src/mainboard/hp/pavilion_m6_1035dx/ec.h @@ -6,6 +6,10 @@ #ifndef _MAINBOARD_HP_PAVILION_M6_1035DX_EC_H #define _MAINBOARD_HP_PAVILION_M6_1035DX_EC_H
+#define EC_SMI_GEVENT 23 + +#ifndef __SMM__ void pavilion_m6_1035dx_ec_init(void); +#endif
#endif /* _MAINBOARD_HP_PAVILION_M6_1035DX_EC_H */ diff --git a/src/mainboard/hp/pavilion_m6_1035dx/mainboard.c b/src/mainboard/hp/pavilion_m6_1035dx/mainboard.c index 2e2b72b..756b443 100644 --- a/src/mainboard/hp/pavilion_m6_1035dx/mainboard.c +++ b/src/mainboard/hp/pavilion_m6_1035dx/mainboard.c @@ -30,6 +30,8 @@ #include <device/pci.h> #include <device/pci_def.h>
+#include <southbridge/amd/agesa/hudson/smi.h> + /************************************************* * enable the dedicated function in parmer board. *************************************************/ @@ -39,6 +41,9 @@ static void mainboard_enable(device_t dev)
pavilion_m6_1035dx_ec_init();
+ hudson_enable_gevent_smi(EC_SMI_GEVENT); + hudson_enable_smi_generation(); + /* * The mainboard is the first place that we get control in ramstage. Check * for S3 resume and call the approriate AGESA/CIMx resume functions. diff --git a/src/mainboard/hp/pavilion_m6_1035dx/mainboard_smi.c b/src/mainboard/hp/pavilion_m6_1035dx/mainboard_smi.c new file mode 100644 index 0000000..bb9cc2e --- /dev/null +++ b/src/mainboard/hp/pavilion_m6_1035dx/mainboard_smi.c @@ -0,0 +1,77 @@ +/* + * SMI handler -- mostly takes care of SMIs from the EC + * + * Copyright (C) 2014 Alexandru Gagniuc mr.nuke.me@gmail.com + * Subject to the GNU GPL v2, or (at your option) any later version. + */ +#include "ec.h" + +#include <console/console.h> +#include <cpu/x86/smm.h> +#include <delay.h> +#include <ec/compal/ene932/ec.h> +#include <southbridge/amd/agesa/hudson/hudson.h> + +enum ec_smi_event { + EC_SMI_EVENT_IDLE = 0x80, +}; + +/* Tell EC to operate in APM mode. Events generate SMIs instead of SCIs */ +static void ec_enter_apm_mode(void) +{ + ec_kbc_write_cmd(0x59); + ec_kbc_write_ib(0xE9); +} +/* Tell EC to operate in ACPI mode, thus generating SCIs on events, not SMIs */ +static void ec_enter_acpi_mode(void) +{ + ec_kbc_write_cmd(0x59); + ec_kbc_write_ib(0xE8); +} + +static uint8_t ec_get_smi_event(void) +{ + ec_kbc_write_cmd(0x56); + return ec_kbc_read_ob(); +} + +static void ec_process_smi(uint8_t src) +{ + /* + * Stub: We aren't processing any events yet, but reading the SMI source + * satisfies the EC in terms of responding to the event. + */ + + printk(BIOS_DEBUG, "EC_SMI event 0x%x\n", src); +} + +static void handle_ec_smi(void) +{ + uint8_t src; + + while ((src = ec_get_smi_event()) != EC_SMI_EVENT_IDLE) + ec_process_smi(src); +} + +int mainboard_smi_apmc(uint8_t data) +{ + switch (data) { + case ACPI_SMI_CMD_ENABLE: + printk(BIOS_DEBUG, "Enable ACPI mode\n"); + ec_enter_acpi_mode(); + break; + case ACPI_SMI_CMD_DISABLE: + printk(BIOS_DEBUG, "Disable ACPI mode\n"); + ec_enter_apm_mode(); + break; + default: + printk(BIOS_DEBUG, "Unhandled ACPI command: 0x%x\n", data); + } + return 0; +} + +void mainboard_smi_gpi(uint32_t gpi_sts) +{ + if (gpi_sts & (1 << EC_SMI_GEVENT)) + handle_ec_smi(); +}
-
Alexandru Gagniuc