src/intel/xeon_sp: add hardware error support (HEST)

This patch adds the ACPI hardware error source table (HEST) support.
This involves a few different parts: (1) The ACPI HEST table which is filled
with the appropriate fields (2) Reserved memory which is used by runtime
SW to provide error information. OS will not accept a HEST table with
this memory set to 0.

The ASL code to enable APEI bit will be submitted in a separate patch.

Tested on DeltaLake mainboard with following options enabled
SOC_INTEL_XEON_RAS

After boot to Linux, the following will show in dmesg:
HEST: Table parsing has been initialized

Change-Id: If76b2af153616182cc053ca878f30fe056e9c8bd
Signed-off-by: Rocky Phagura <rphagura@fb.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/52090
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>

3 files changed, 123 insertions, 0 deletions

diff --git a/src/soc/intel/xeon_sp/ras/Kconfig b/src/soc/intel/xeon_sp/ras/Kconfig
new file mode 100644
index 0000000000..79c599c2fb
--- /dev/null
+++ b/src/soc/intel/xeon_sp/ras/Kconfig
@@ -0,0 +1,20 @@
+# SPDX-License-Identifier: GPL-2.0-or-later
+
+config SOC_ACPI_HEST
+	def_bool n
+	depends on HAVE_ACPI_TABLES
+	help
+	  This variable provides control for ACPI hardware error source table (HEST)
+
+config SOC_RAS_ELOG
+	def_bool n
+	depends on SOC_ACPI_HEST
+	help
+	  This variable provides enhanced error logging support used with HEST
+
+config ERROR_LOG_BUFFER_SIZE
+	hex
+	default 0x4000
+	depends on SOC_RAS_ELOG
+	help
+	  This variable allows a configurable error log based on system requirements
diff --git a/src/soc/intel/xeon_sp/ras/Makefile.inc b/src/soc/intel/xeon_sp/ras/Makefile.inc
new file mode 100644
index 0000000000..93c8705f94
--- /dev/null
+++ b/src/soc/intel/xeon_sp/ras/Makefile.inc
@@ -0,0 +1,3 @@
+## SPDX-License-Identifier: GPL-2.0-or-later
+
+ramstage-$(CONFIG_SOC_ACPI_HEST) += hest.c
diff --git a/src/soc/intel/xeon_sp/ras/hest.c b/src/soc/intel/xeon_sp/ras/hest.c
new file mode 100644
index 0000000000..6164edfe19
--- /dev/null
+++ b/src/soc/intel/xeon_sp/ras/hest.c
@@ -0,0 +1,100 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+
+#include <soc/acpi.h>
+#include <acpi/acpi_gnvs.h>
+#include <cbmem.h>
+#include <console/console.h>
+#include <soc/hest.h>
+#include <intelblocks/nvs.h>
+
+static u64 hest_get_elog_addr(void)
+{
+	/* The elog address comes from reserved memory */
+	struct global_nvs *gnvs;
+	gnvs = acpi_get_gnvs();
+	if (!gnvs) {
+		printk(BIOS_ERR, "Unable to get gnvs\n");
+		return 0;
+	}
+
+	/* Runtime logging address */
+	printk(BIOS_DEBUG, "\t status blk start addr = %llx\n", gnvs->hest_log_addr);
+	printk(BIOS_DEBUG, "\t size = %x\n", CONFIG_ERROR_LOG_BUFFER_SIZE);
+	return gnvs->hest_log_addr;
+}
+
+static u32 acpi_hest_add_ghes(void *current)
+{
+	ghes_record_t *rec = (ghes_record_t *)current;
+	u32 size = sizeof(ghes_record_t);
+
+	/* Fill GHES error source descriptor  */
+	memset(rec, 0, size);
+	rec->esd.type = HEST_GHES_DESC_TYPE;
+	rec->esd.source_id = 0; /* 0 for MCE check exception source */
+	rec->esd.enabled = 1;
+	rec->esd.related_src_id = 0xffff;
+	rec->esd.prealloc_erecords = 1;
+	rec->esd.max_section_per_record = 0xf;
+	rec->max_raw_data_length = GHES_MAX_RAW_DATA_LENGTH;
+
+	/* Add error_status_address */
+	rec->sts_addr.space_id = 0;
+	rec->sts_addr.bit_width = 0x40;
+	rec->sts_addr.bit_offset = 0;
+	rec->sts_addr.access_size = QWORD_ACCESS;
+
+	/* Add notification structure */
+	rec->notify.type = NOTIFY_TYPE_SCI;
+	rec->notify.length = sizeof(acpi_hest_hen_t);
+	rec->err_sts_blk_len = GHEST_ERROR_STATUS_BLOCK_LENGTH;
+
+	/* error status block entries start address */
+	if (CONFIG(SOC_ACPI_HEST))
+		rec->sts_addr.addr = hest_get_elog_addr();
+
+	return size;
+}
+
+static unsigned long acpi_fill_hest(acpi_hest_t *hest)
+{
+	acpi_header_t *header = &(hest->header);
+	void *current;
+	current = (void *)(hest);
+	void *next = current;
+	next = hest + 1;
+	next += acpi_hest_add_ghes(next);
+	hest->error_source_count += 1;
+	header->length += next - current;
+	return header->length;
+}
+
+unsigned long hest_create(unsigned long current, struct acpi_rsdp *rsdp)
+{
+	struct global_nvs *gnvs;
+	acpi_hest_t *hest;
+
+	/* Reserve memory for Enhanced error logging */
+	void *mem = cbmem_add(CMBMEM_ID_ACPI_HEST, CONFIG_ERROR_LOG_BUFFER_SIZE);
+	if (!mem) {
+		printk(BIOS_ERR, "Unable to allocate HEST memory\n");
+		return current;
+	}
+
+	printk(BIOS_DEBUG, "HEST memory created: %p\n", mem);
+	gnvs = acpi_get_gnvs();
+	if (!gnvs) {
+		printk(BIOS_ERR, "Unable to get gnvs\n");
+		return current;
+	}
+	gnvs->hest_log_addr = (uintptr_t)mem;
+	printk(BIOS_DEBUG, "elog_addr: %llx, size:%x\n", gnvs->hest_log_addr,
+		CONFIG_ERROR_LOG_BUFFER_SIZE);
+
+	current = ALIGN(current, 8);
+	hest = (acpi_hest_t *)current;
+	acpi_write_hest(hest, acpi_fill_hest);
+	acpi_add_table(rsdp, (void *)current);
+	current += hest->header.length;
+	return current;
+}