mce-intel-sb.c

// SPDX-License-Identifier: GPL-2.0

/*
 * The code below came from Andi Kleen/Intel/SUSE mcelog code,
 * released under GNU Public General License, v.2
 */

#include <stdio.h>
#include <string.h>

#include "bitfield.h"
#include "ras-mce-handler.h"

/* See IA32 SDM Vol3B Table 16.4.1 */

static char *pcu_1[] = {
	[0] = "No error",
	[1] = "Non_IMem_Sel",
	[2] = "I_Parity_Error",
	[3] = "Bad_OpCode",
	[4] = "I_Stack_Underflow",
	[5] = "I_Stack_Overflow",
	[6] = "D_Stack_Underflow",
	[7] = "D_Stack_Overflow",
	[8] = "Non-DMem_Sel",
	[9] = "D_Parity_Error"
};

static char *pcu_2[] = {
	[0x00] = "No Error",
	[0x0D] = "MC_IMC_FORCE_SR_S3_TIMEOUT",
	[0x0E] = "MC_MC_CPD_UNCPD_ST_TIMEOUT",
	[0x0F] = "MC_PKGS_SAFE_WP_TIMEOUT",
	[0x43] = "MC_PECI_MAILBOX_QUIESCE_TIMEOUT",
	[0x5C] = "MC_MORE_THAN_ONE_LT_AGENT",
	[0x60] = "MC_INVALID_PKGS_REQ_PCH",
	[0x61] = "MC_INVALID_PKGS_REQ_QPI",
	[0x62] = "MC_INVALID_PKGS_RES_QPI",
	[0x63] = "MC_INVALID_PKGC_RES_PCH",
	[0x64] = "MC_INVALID_PKG_STATE_CONFIG",
	[0x70] = "MC_WATCHDG_TIMEOUT_PKGC_SLAVE",
	[0x71] = "MC_WATCHDG_TIMEOUT_PKGC_MASTER",
	[0x72] = "MC_WATCHDG_TIMEOUT_PKGS_MASTER",
	[0x7A] = "MC_HA_FAILSTS_CHANGE_DETECTED",
	[0x81] = "MC_RECOVERABLE_DIE_THERMAL_TOO_HOT",
};

static struct field pcu_mc4[] = {
	FIELD(16, pcu_1),
	FIELD(24, pcu_2),
	{}
};

static char *memctrl_1[] = {
	[0x001] = "Address parity error",
	[0x002] = "HA Wrt buffer Data parity error",
	[0x004] = "HA Wrt byte enable parity error",
	[0x008] = "Corrected patrol scrub error",
	[0x010] = "Uncorrected patrol scrub error",
	[0x020] = "Corrected spare error",
	[0x040] = "Uncorrected spare error",
};

static struct field memctrl_mc8[] = {
	FIELD(16, memctrl_1),
	{}
};

void snb_decode_model(struct ras_events *ras, struct mce_event *e)
{
	struct mce_priv *mce = ras->mce_priv;
	uint32_t mca = e->status & 0xffff;
	unsigned int rank0 = -1, rank1 = -1, chan;

	switch (e->bank) {
	case 4:
		decode_bitfield(e, e->status, pcu_mc4);
		break;
	case 6:
	case 7:
		/* MCACOD already decoded */
		if (mce->cputype == CPU_SANDY_BRIDGE_EP)
			mce_snprintf(e->bank_name, "QPI");
		break;
	case 8:
	case 9:
	case 10:
	case 11:
//		Wprintf("MemCtrl: ");
		decode_bitfield(e, e->status, memctrl_mc8);
		break;
	}

	/*
	 * Memory error specific code. Returns if the error is not a MC one
	 */

	/* Check if the error is at the memory controller */
	if ((mca >> 7) != 1)
		return;

	/* Ignore unless this is an corrected extended error from an iMC bank */
	if (e->bank < 8 || e->bank > 11 || (e->status & MCI_STATUS_UC) ||
	    !test_prefix(7, e->status & 0xefff))
		return;

	/*
	 * Parse the reported channel and ranks
	 */

	chan = EXTRACT(e->status, 0, 3);
	if (chan == 0xf)
		return;

	mce_snprintf(e->mc_location, "memory_channel=%d", chan);

	if (EXTRACT(e->misc, 62, 62))
		rank0 = EXTRACT(e->misc, 46, 50);

	if (EXTRACT(e->misc, 63, 63))
		rank1 = EXTRACT(e->misc, 51, 55);

	/*
	 * FIXME: The conversion from rank to dimm requires to parse the
	 * DMI tables and call failrank2dimm().
	 */
	if (rank0 >= 0 && rank1 >= 0)
		mce_snprintf(e->mc_location, "ranks=%d and %d",
			     rank0, rank1);
	else if (rank0 >= 0)
		mce_snprintf(e->mc_location, "rank=%d", rank0);
	else
		mce_snprintf(e->mc_location, "rank=%d", rank1);
}

#if 0
/*
 * Sandy Bridge EP and EP4S processors (family 6, model 45) support additional
 * logging for corrected errors in the integrated memory controller (IMC)
 * banks. The mode is off by default, but can be enabled by setting the
 * "MemError Log Enable" * bit in MSR_ERROR_CONTROL (MSR 0x17f).
 * The documentation in the August 2012 edition of Intel's Software developer
 * manual has some minor errors because the wrong version of table 16-16
 * "Intel IMC MC Error Codes for IA32_MCi_MISC (i= 8, 11)" was included.
 * Corrections are:
 *  Bit 62 is the "VALID" bit for the "first-device" bits in MISC and STATUS
 *  Bit 63 is the "VALID" bit for the "second-device" bits in MISC
 *  Bits 58:56 and 61:59 should be marked as "reserved".
 * There should also be a footnote explaining how the "failing rank" fields
 * can be converted to a DIMM number within a channel for systems with either
 * two or three DIMMs per channel.
 */
static int failrank2dimm(unsigned int failrank, int socket, int channel)
{
	switch (failrank) {
	case 0: case 1: case 2: case 3:
		return 0;
	case 4: case 5:
		return 1;
	case 6: case 7:
		if (get_memdimm(socket, channel, 2, 0))
			return 2;
		else
			return 1;
	}
	return -1;
}
#endif