sb_edac.c 53.1 KB
Newer Older
1 2 3 4 5 6 7 8 9
/* Intel Sandy Bridge -EN/-EP/-EX Memory Controller kernel module
 *
 * This driver supports the memory controllers found on the Intel
 * processor family Sandy Bridge.
 *
 * This file may be distributed under the terms of the
 * GNU General Public License version 2 only.
 *
 * Copyright (c) 2011 by:
10
 *	 Mauro Carvalho Chehab
11 12 13 14 15 16 17 18 19 20 21 22
 */

#include <linux/module.h>
#include <linux/init.h>
#include <linux/pci.h>
#include <linux/pci_ids.h>
#include <linux/slab.h>
#include <linux/delay.h>
#include <linux/edac.h>
#include <linux/mmzone.h>
#include <linux/smp.h>
#include <linux/bitmap.h>
23
#include <linux/math64.h>
24
#include <asm/processor.h>
25
#include <asm/mce.h>
26 27 28 29 30 31 32 33 34 35 36

#include "edac_core.h"

/* Static vars */
static LIST_HEAD(sbridge_edac_list);
static DEFINE_MUTEX(sbridge_edac_lock);
static int probed;

/*
 * Alter this version for the module when modifications are made
 */
37
#define SBRIDGE_REVISION    " Ver: 1.1.0 "
38 39 40 41 42 43 44 45 46 47 48 49 50 51 52
#define EDAC_MOD_STR      "sbridge_edac"

/*
 * Debug macros
 */
#define sbridge_printk(level, fmt, arg...)			\
	edac_printk(level, "sbridge", fmt, ##arg)

#define sbridge_mc_printk(mci, level, fmt, arg...)		\
	edac_mc_chipset_printk(mci, level, "sbridge", fmt, ##arg)

/*
 * Get a bit field at register value <v>, from bit <lo> to bit <hi>
 */
#define GET_BITFIELD(v, lo, hi)	\
53
	(((v) & GENMASK_ULL(hi, lo)) >> (lo))
54 55 56 57 58 59 60

/*
 * sbridge Memory Controller Registers
 */

/*
 * FIXME: For now, let's order by device function, as it makes
D
David Mackey 已提交
61
 * easier for driver's development process. This table should be
62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
 * moved to pci_id.h when submitted upstream
 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_SAD0	0x3cf4	/* 12.6 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_SAD1	0x3cf6	/* 12.7 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_BR		0x3cf5	/* 13.6 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0	0x3ca0	/* 14.0 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TA	0x3ca8	/* 15.0 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_RAS	0x3c71	/* 15.1 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD0	0x3caa	/* 15.2 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD1	0x3cab	/* 15.3 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD2	0x3cac	/* 15.4 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD3	0x3cad	/* 15.5 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_DDRIO	0x3cb8	/* 17.0 */

	/*
	 * Currently, unused, but will be needed in the future
	 * implementations, as they hold the error counters
	 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_ERR0	0x3c72	/* 16.2 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_ERR1	0x3c73	/* 16.3 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_ERR2	0x3c76	/* 16.6 */
#define PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_ERR3	0x3c77	/* 16.7 */

/* Devices 12 Function 6, Offsets 0x80 to 0xcc */
86
static const u32 sbridge_dram_rule[] = {
87 88 89 90
	0x80, 0x88, 0x90, 0x98, 0xa0,
	0xa8, 0xb0, 0xb8, 0xc0, 0xc8,
};

91 92 93 94 95 96
static const u32 ibridge_dram_rule[] = {
	0x60, 0x68, 0x70, 0x78, 0x80,
	0x88, 0x90, 0x98, 0xa0,	0xa8,
	0xb0, 0xb8, 0xc0, 0xc8, 0xd0,
	0xd8, 0xe0, 0xe8, 0xf0, 0xf8,
};
97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116

#define SAD_LIMIT(reg)		((GET_BITFIELD(reg, 6, 25) << 26) | 0x3ffffff)
#define DRAM_ATTR(reg)		GET_BITFIELD(reg, 2,  3)
#define INTERLEAVE_MODE(reg)	GET_BITFIELD(reg, 1,  1)
#define DRAM_RULE_ENABLE(reg)	GET_BITFIELD(reg, 0,  0)

static char *get_dram_attr(u32 reg)
{
	switch(DRAM_ATTR(reg)) {
		case 0:
			return "DRAM";
		case 1:
			return "MMCFG";
		case 2:
			return "NXM";
		default:
			return "unknown";
	}
}

117
static const u32 sbridge_interleave_list[] = {
118 119 120 121
	0x84, 0x8c, 0x94, 0x9c, 0xa4,
	0xac, 0xb4, 0xbc, 0xc4, 0xcc,
};

122 123 124 125 126 127 128
static const u32 ibridge_interleave_list[] = {
	0x64, 0x6c, 0x74, 0x7c, 0x84,
	0x8c, 0x94, 0x9c, 0xa4, 0xac,
	0xb4, 0xbc, 0xc4, 0xcc, 0xd4,
	0xdc, 0xe4, 0xec, 0xf4, 0xfc,
};

A
Aristeu Rozanski 已提交
129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144
struct interleave_pkg {
	unsigned char start;
	unsigned char end;
};

static const struct interleave_pkg sbridge_interleave_pkg[] = {
	{ 0, 2 },
	{ 3, 5 },
	{ 8, 10 },
	{ 11, 13 },
	{ 16, 18 },
	{ 19, 21 },
	{ 24, 26 },
	{ 27, 29 },
};

145 146 147 148 149 150 151 152 153 154 155
static const struct interleave_pkg ibridge_interleave_pkg[] = {
	{ 0, 3 },
	{ 4, 7 },
	{ 8, 11 },
	{ 12, 15 },
	{ 16, 19 },
	{ 20, 23 },
	{ 24, 27 },
	{ 28, 31 },
};

A
Aristeu Rozanski 已提交
156 157
static inline int sad_pkg(const struct interleave_pkg *table, u32 reg,
			  int interleave)
158
{
A
Aristeu Rozanski 已提交
159 160
	return GET_BITFIELD(reg, table[interleave].start,
			    table[interleave].end);
161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274
}

/* Devices 12 Function 7 */

#define TOLM		0x80
#define	TOHM		0x84

#define GET_TOLM(reg)		((GET_BITFIELD(reg, 0,  3) << 28) | 0x3ffffff)
#define GET_TOHM(reg)		((GET_BITFIELD(reg, 0, 20) << 25) | 0x3ffffff)

/* Device 13 Function 6 */

#define SAD_TARGET	0xf0

#define SOURCE_ID(reg)		GET_BITFIELD(reg, 9, 11)

#define SAD_CONTROL	0xf4

/* Device 14 function 0 */

static const u32 tad_dram_rule[] = {
	0x40, 0x44, 0x48, 0x4c,
	0x50, 0x54, 0x58, 0x5c,
	0x60, 0x64, 0x68, 0x6c,
};
#define MAX_TAD	ARRAY_SIZE(tad_dram_rule)

#define TAD_LIMIT(reg)		((GET_BITFIELD(reg, 12, 31) << 26) | 0x3ffffff)
#define TAD_SOCK(reg)		GET_BITFIELD(reg, 10, 11)
#define TAD_CH(reg)		GET_BITFIELD(reg,  8,  9)
#define TAD_TGT3(reg)		GET_BITFIELD(reg,  6,  7)
#define TAD_TGT2(reg)		GET_BITFIELD(reg,  4,  5)
#define TAD_TGT1(reg)		GET_BITFIELD(reg,  2,  3)
#define TAD_TGT0(reg)		GET_BITFIELD(reg,  0,  1)

/* Device 15, function 0 */

#define MCMTR			0x7c

#define IS_ECC_ENABLED(mcmtr)		GET_BITFIELD(mcmtr, 2, 2)
#define IS_LOCKSTEP_ENABLED(mcmtr)	GET_BITFIELD(mcmtr, 1, 1)
#define IS_CLOSE_PG(mcmtr)		GET_BITFIELD(mcmtr, 0, 0)

/* Device 15, function 1 */

#define RASENABLES		0xac
#define IS_MIRROR_ENABLED(reg)		GET_BITFIELD(reg, 0, 0)

/* Device 15, functions 2-5 */

static const int mtr_regs[] = {
	0x80, 0x84, 0x88,
};

#define RANK_DISABLE(mtr)		GET_BITFIELD(mtr, 16, 19)
#define IS_DIMM_PRESENT(mtr)		GET_BITFIELD(mtr, 14, 14)
#define RANK_CNT_BITS(mtr)		GET_BITFIELD(mtr, 12, 13)
#define RANK_WIDTH_BITS(mtr)		GET_BITFIELD(mtr, 2, 4)
#define COL_WIDTH_BITS(mtr)		GET_BITFIELD(mtr, 0, 1)

static const u32 tad_ch_nilv_offset[] = {
	0x90, 0x94, 0x98, 0x9c,
	0xa0, 0xa4, 0xa8, 0xac,
	0xb0, 0xb4, 0xb8, 0xbc,
};
#define CHN_IDX_OFFSET(reg)		GET_BITFIELD(reg, 28, 29)
#define TAD_OFFSET(reg)			(GET_BITFIELD(reg,  6, 25) << 26)

static const u32 rir_way_limit[] = {
	0x108, 0x10c, 0x110, 0x114, 0x118,
};
#define MAX_RIR_RANGES ARRAY_SIZE(rir_way_limit)

#define IS_RIR_VALID(reg)	GET_BITFIELD(reg, 31, 31)
#define RIR_WAY(reg)		GET_BITFIELD(reg, 28, 29)

#define MAX_RIR_WAY	8

static const u32 rir_offset[MAX_RIR_RANGES][MAX_RIR_WAY] = {
	{ 0x120, 0x124, 0x128, 0x12c, 0x130, 0x134, 0x138, 0x13c },
	{ 0x140, 0x144, 0x148, 0x14c, 0x150, 0x154, 0x158, 0x15c },
	{ 0x160, 0x164, 0x168, 0x16c, 0x170, 0x174, 0x178, 0x17c },
	{ 0x180, 0x184, 0x188, 0x18c, 0x190, 0x194, 0x198, 0x19c },
	{ 0x1a0, 0x1a4, 0x1a8, 0x1ac, 0x1b0, 0x1b4, 0x1b8, 0x1bc },
};

#define RIR_RNK_TGT(reg)		GET_BITFIELD(reg, 16, 19)
#define RIR_OFFSET(reg)		GET_BITFIELD(reg,  2, 14)

/* Device 16, functions 2-7 */

/*
 * FIXME: Implement the error count reads directly
 */

static const u32 correrrcnt[] = {
	0x104, 0x108, 0x10c, 0x110,
};

#define RANK_ODD_OV(reg)		GET_BITFIELD(reg, 31, 31)
#define RANK_ODD_ERR_CNT(reg)		GET_BITFIELD(reg, 16, 30)
#define RANK_EVEN_OV(reg)		GET_BITFIELD(reg, 15, 15)
#define RANK_EVEN_ERR_CNT(reg)		GET_BITFIELD(reg,  0, 14)

static const u32 correrrthrsld[] = {
	0x11c, 0x120, 0x124, 0x128,
};

#define RANK_ODD_ERR_THRSLD(reg)	GET_BITFIELD(reg, 16, 30)
#define RANK_EVEN_ERR_THRSLD(reg)	GET_BITFIELD(reg,  0, 14)


/* Device 17, function 0 */

275
#define SB_RANK_CFG_A		0x0328
276

277
#define IB_RANK_CFG_A		0x0320
278 279 280 281 282 283 284 285

/*
 * sbridge structs
 */

#define NUM_CHANNELS	4
#define MAX_DIMMS	3		/* Max DIMMS per channel */

286 287 288 289 290
enum type {
	SANDY_BRIDGE,
	IVY_BRIDGE,
};

A
Aristeu Rozanski 已提交
291
struct sbridge_pvt;
292
struct sbridge_info {
293
	enum type	type;
294 295 296 297
	u32		mcmtr;
	u32		rankcfgr;
	u64		(*get_tolm)(struct sbridge_pvt *pvt);
	u64		(*get_tohm)(struct sbridge_pvt *pvt);
298
	u64		(*rir_limit)(u32 reg);
299
	const u32	*dram_rule;
300
	const u32	*interleave_list;
A
Aristeu Rozanski 已提交
301
	const struct interleave_pkg *interleave_pkg;
302
	u8		max_sad;
303
	u8		max_interleave;
304
	u8		(*get_node_id)(struct sbridge_pvt *pvt);
305
	enum mem_type	(*get_memory_type)(struct sbridge_pvt *pvt);
306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333
};

struct sbridge_channel {
	u32		ranks;
	u32		dimms;
};

struct pci_id_descr {
	int 			dev_id;
	int			optional;
};

struct pci_id_table {
	const struct pci_id_descr	*descr;
	int				n_devs;
};

struct sbridge_dev {
	struct list_head	list;
	u8			bus, mc;
	u8			node_id, source_id;
	struct pci_dev		**pdev;
	int			n_devs;
	struct mem_ctl_info	*mci;
};

struct sbridge_pvt {
	struct pci_dev		*pci_ta, *pci_ddrio, *pci_ras;
334 335 336
	struct pci_dev		*pci_sad0, *pci_sad1;
	struct pci_dev		*pci_ha0, *pci_ha1;
	struct pci_dev		*pci_br0, *pci_br1;
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360
	struct pci_dev		*pci_tad[NUM_CHANNELS];

	struct sbridge_dev	*sbridge_dev;

	struct sbridge_info	info;
	struct sbridge_channel	channel[NUM_CHANNELS];

	/* Memory type detection */
	bool			is_mirrored, is_lockstep, is_close_pg;

	/* Fifo double buffers */
	struct mce		mce_entry[MCE_LOG_LEN];
	struct mce		mce_outentry[MCE_LOG_LEN];

	/* Fifo in/out counters */
	unsigned		mce_in, mce_out;

	/* Count indicator to show errors not got */
	unsigned		mce_overrun;

	/* Memory description */
	u64			tolm, tohm;
};

361 362
#define PCI_DESCR(device_id, opt)	\
	.dev_id = (device_id),		\
363
	.optional = opt
364 365 366

static const struct pci_id_descr pci_dev_descr_sbridge[] = {
		/* Processor Home Agent */
367
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0, 0)	},
368 369

		/* Memory controller */
370 371 372 373 374 375 376
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TA, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_RAS, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD0, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD1, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD2, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD3, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_DDRIO, 1)	},
377 378

		/* System Address Decoder */
379 380
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_SAD0, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_SAD1, 0)	},
381 382

		/* Broadcast Registers */
383
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_BR, 0)		},
384 385 386 387 388 389 390 391
};

#define PCI_ID_TABLE_ENTRY(A) { .descr=A, .n_devs = ARRAY_SIZE(A) }
static const struct pci_id_table pci_dev_descr_sbridge_table[] = {
	PCI_ID_TABLE_ENTRY(pci_dev_descr_sbridge),
	{0,}			/* 0 terminated list. */
};

392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419
/* This changes depending if 1HA or 2HA:
 * 1HA:
 *	0x0eb8 (17.0) is DDRIO0
 * 2HA:
 *	0x0ebc (17.4) is DDRIO0
 */
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_1HA_DDRIO0	0x0eb8
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_2HA_DDRIO0	0x0ebc

/* pci ids */
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0		0x0ea0
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA		0x0ea8
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_RAS		0x0e71
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD0	0x0eaa
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD1	0x0eab
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD2	0x0eac
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD3	0x0ead
#define PCI_DEVICE_ID_INTEL_IBRIDGE_SAD			0x0ec8
#define PCI_DEVICE_ID_INTEL_IBRIDGE_BR0			0x0ec9
#define PCI_DEVICE_ID_INTEL_IBRIDGE_BR1			0x0eca
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1		0x0e60
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TA		0x0e68
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_RAS		0x0e79
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD0	0x0e6a
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD1	0x0e6b

static const struct pci_id_descr pci_dev_descr_ibridge[] = {
		/* Processor Home Agent */
420
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0, 0)		},
421 422

		/* Memory controller */
423 424 425 426 427 428
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA, 0)		},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_RAS, 0)		},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD0, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD1, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD2, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD3, 0)	},
429 430

		/* System Address Decoder */
431
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_SAD, 0)			},
432 433

		/* Broadcast Registers */
434 435
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_BR0, 1)			},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_BR1, 0)			},
436 437

		/* Optional, mode 2HA */
438
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1, 1)		},
439
#if 0
440 441
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TA, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_RAS, 1)	},
442
#endif
443 444
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD0, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD1, 1)	},
445

446 447
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_1HA_DDRIO0, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_2HA_DDRIO0, 1)	},
448 449 450 451 452 453 454
};

static const struct pci_id_table pci_dev_descr_ibridge_table[] = {
	PCI_ID_TABLE_ENTRY(pci_dev_descr_ibridge),
	{0,}			/* 0 terminated list. */
};

455 456 457
/*
 *	pci_device_id	table for which devices we are looking for
 */
458
static const struct pci_device_id sbridge_pci_tbl[] = {
459
	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TA)},
460
	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA)},
461 462 463 464 465
	{0,}			/* 0 terminated list. */
};


/****************************************************************************
D
David Mackey 已提交
466
			Ancillary status routines
467 468 469 470 471 472 473
 ****************************************************************************/

static inline int numrank(u32 mtr)
{
	int ranks = (1 << RANK_CNT_BITS(mtr));

	if (ranks > 4) {
474 475
		edac_dbg(0, "Invalid number of ranks: %d (max = 4) raw value = %x (%04x)\n",
			 ranks, (unsigned int)RANK_CNT_BITS(mtr), mtr);
476 477 478 479 480 481 482 483 484 485 486
		return -EINVAL;
	}

	return ranks;
}

static inline int numrow(u32 mtr)
{
	int rows = (RANK_WIDTH_BITS(mtr) + 12);

	if (rows < 13 || rows > 18) {
487 488
		edac_dbg(0, "Invalid number of rows: %d (should be between 14 and 17) raw value = %x (%04x)\n",
			 rows, (unsigned int)RANK_WIDTH_BITS(mtr), mtr);
489 490 491 492 493 494 495 496 497 498 499
		return -EINVAL;
	}

	return 1 << rows;
}

static inline int numcol(u32 mtr)
{
	int cols = (COL_WIDTH_BITS(mtr) + 10);

	if (cols > 12) {
500 501
		edac_dbg(0, "Invalid number of cols: %d (max = 4) raw value = %x (%04x)\n",
			 cols, (unsigned int)COL_WIDTH_BITS(mtr), mtr);
502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
		return -EINVAL;
	}

	return 1 << cols;
}

static struct sbridge_dev *get_sbridge_dev(u8 bus)
{
	struct sbridge_dev *sbridge_dev;

	list_for_each_entry(sbridge_dev, &sbridge_edac_list, list) {
		if (sbridge_dev->bus == bus)
			return sbridge_dev;
	}

	return NULL;
}

static struct sbridge_dev *alloc_sbridge_dev(u8 bus,
					   const struct pci_id_table *table)
{
	struct sbridge_dev *sbridge_dev;

	sbridge_dev = kzalloc(sizeof(*sbridge_dev), GFP_KERNEL);
	if (!sbridge_dev)
		return NULL;

	sbridge_dev->pdev = kzalloc(sizeof(*sbridge_dev->pdev) * table->n_devs,
				   GFP_KERNEL);
	if (!sbridge_dev->pdev) {
		kfree(sbridge_dev);
		return NULL;
	}

	sbridge_dev->bus = bus;
	sbridge_dev->n_devs = table->n_devs;
	list_add_tail(&sbridge_dev->list, &sbridge_edac_list);

	return sbridge_dev;
}

static void free_sbridge_dev(struct sbridge_dev *sbridge_dev)
{
	list_del(&sbridge_dev->list);
	kfree(sbridge_dev->pdev);
	kfree(sbridge_dev);
}

A
Aristeu Rozanski 已提交
550 551 552 553 554 555 556 557 558
static u64 sbridge_get_tolm(struct sbridge_pvt *pvt)
{
	u32 reg;

	/* Address range is 32:28 */
	pci_read_config_dword(pvt->pci_sad1, TOLM, &reg);
	return GET_TOLM(reg);
}

A
Aristeu Rozanski 已提交
559 560 561 562 563 564 565 566
static u64 sbridge_get_tohm(struct sbridge_pvt *pvt)
{
	u32 reg;

	pci_read_config_dword(pvt->pci_sad1, TOHM, &reg);
	return GET_TOHM(reg);
}

567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584
static u64 ibridge_get_tolm(struct sbridge_pvt *pvt)
{
	u32 reg;

	pci_read_config_dword(pvt->pci_br1, TOLM, &reg);

	return GET_TOLM(reg);
}

static u64 ibridge_get_tohm(struct sbridge_pvt *pvt)
{
	u32 reg;

	pci_read_config_dword(pvt->pci_br1, TOHM, &reg);

	return GET_TOHM(reg);
}

585 586 587 588 589
static u64 rir_limit(u32 reg)
{
	return ((u64)GET_BITFIELD(reg,  1, 10) << 29) | 0x1fffffff;
}

590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608
static enum mem_type get_memory_type(struct sbridge_pvt *pvt)
{
	u32 reg;
	enum mem_type mtype;

	if (pvt->pci_ddrio) {
		pci_read_config_dword(pvt->pci_ddrio, pvt->info.rankcfgr,
				      &reg);
		if (GET_BITFIELD(reg, 11, 11))
			/* FIXME: Can also be LRDIMM */
			mtype = MEM_RDDR3;
		else
			mtype = MEM_DDR3;
	} else
		mtype = MEM_UNKNOWN;

	return mtype;
}

609 610 611 612 613 614 615
static u8 get_node_id(struct sbridge_pvt *pvt)
{
	u32 reg;
	pci_read_config_dword(pvt->pci_br0, SAD_CONTROL, &reg);
	return GET_BITFIELD(reg, 0, 2);
}

616 617 618
static inline u8 sad_pkg_socket(u8 pkg)
{
	/* on Ivy Bridge, nodeID is SASS, where A is HA and S is node id */
619
	return ((pkg >> 3) << 2) | (pkg & 0x3);
620 621 622 623 624 625 626
}

static inline u8 sad_pkg_ha(u8 pkg)
{
	return (pkg >> 2) & 0x1;
}

627 628 629
/****************************************************************************
			Memory check routines
 ****************************************************************************/
630
static struct pci_dev *get_pdev_same_bus(u8 bus, u32 id)
631
{
632
	struct pci_dev *pdev = NULL;
633

634 635 636 637 638
	do {
		pdev = pci_get_device(PCI_VENDOR_ID_INTEL, id, pdev);
		if (pdev && pdev->bus->number == bus)
			break;
	} while (pdev);
639

640
	return pdev;
641 642 643
}

/**
644
 * check_if_ecc_is_active() - Checks if ECC is active
645 646
 * bus:		Device bus
 */
647
static int check_if_ecc_is_active(const u8 bus, enum type type)
648 649
{
	struct pci_dev *pdev = NULL;
650
	u32 mcmtr, id;
651

652 653 654 655 656 657
	if (type == IVY_BRIDGE)
		id = PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA;
	else
		id = PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TA;

	pdev = get_pdev_same_bus(bus, id);
658 659
	if (!pdev) {
		sbridge_printk(KERN_ERR, "Couldn't find PCI device "
660 661
					"%04x:%04x! on bus %02d\n",
					PCI_VENDOR_ID_INTEL, id, bus);
662 663 664 665 666 667 668 669 670 671 672
		return -ENODEV;
	}

	pci_read_config_dword(pdev, MCMTR, &mcmtr);
	if (!IS_ECC_ENABLED(mcmtr)) {
		sbridge_printk(KERN_ERR, "ECC is disabled. Aborting\n");
		return -ENODEV;
	}
	return 0;
}

673
static int get_dimm_config(struct mem_ctl_info *mci)
674 675
{
	struct sbridge_pvt *pvt = mci->pvt_info;
676
	struct dimm_info *dimm;
677 678
	unsigned i, j, banks, ranks, rows, cols, npages;
	u64 size;
679 680
	u32 reg;
	enum edac_type mode;
681
	enum mem_type mtype;
682

A
Aristeu Rozanski 已提交
683
	pci_read_config_dword(pvt->pci_br0, SAD_TARGET, &reg);
684 685
	pvt->sbridge_dev->source_id = SOURCE_ID(reg);

686
	pvt->sbridge_dev->node_id = pvt->info.get_node_id(pvt);
687 688 689 690
	edac_dbg(0, "mc#%d: Node ID: %d, source ID: %d\n",
		 pvt->sbridge_dev->mc,
		 pvt->sbridge_dev->node_id,
		 pvt->sbridge_dev->source_id);
691 692 693

	pci_read_config_dword(pvt->pci_ras, RASENABLES, &reg);
	if (IS_MIRROR_ENABLED(reg)) {
694
		edac_dbg(0, "Memory mirror is enabled\n");
695 696
		pvt->is_mirrored = true;
	} else {
697
		edac_dbg(0, "Memory mirror is disabled\n");
698 699 700 701 702
		pvt->is_mirrored = false;
	}

	pci_read_config_dword(pvt->pci_ta, MCMTR, &pvt->info.mcmtr);
	if (IS_LOCKSTEP_ENABLED(pvt->info.mcmtr)) {
703
		edac_dbg(0, "Lockstep is enabled\n");
704 705 706
		mode = EDAC_S8ECD8ED;
		pvt->is_lockstep = true;
	} else {
707
		edac_dbg(0, "Lockstep is disabled\n");
708 709 710 711
		mode = EDAC_S4ECD4ED;
		pvt->is_lockstep = false;
	}
	if (IS_CLOSE_PG(pvt->info.mcmtr)) {
712
		edac_dbg(0, "address map is on closed page mode\n");
713 714
		pvt->is_close_pg = true;
	} else {
715
		edac_dbg(0, "address map is on open page mode\n");
716 717 718
		pvt->is_close_pg = false;
	}

719 720 721 722
	mtype = pvt->info.get_memory_type(pvt);
	if (mtype == MEM_RDDR3)
		edac_dbg(0, "Memory is registered\n");
	else if (mtype == MEM_UNKNOWN)
723
		edac_dbg(0, "Cannot determine memory type\n");
724 725
	else
		edac_dbg(0, "Memory is unregistered\n");
726 727 728 729 730 731 732 733

	/* On all supported DDR3 DIMM types, there are 8 banks available */
	banks = 8;

	for (i = 0; i < NUM_CHANNELS; i++) {
		u32 mtr;

		for (j = 0; j < ARRAY_SIZE(mtr_regs); j++) {
734 735
			dimm = EDAC_DIMM_PTR(mci->layers, mci->dimms, mci->n_layers,
				       i, j, 0);
736 737
			pci_read_config_dword(pvt->pci_tad[i],
					      mtr_regs[j], &mtr);
738
			edac_dbg(4, "Channel #%d  MTR%d = %x\n", i, j, mtr);
739 740 741 742 743 744 745 746
			if (IS_DIMM_PRESENT(mtr)) {
				pvt->channel[i].dimms++;

				ranks = numrank(mtr);
				rows = numrow(mtr);
				cols = numcol(mtr);

				/* DDR3 has 8 I/O banks */
747
				size = ((u64)rows * cols * banks * ranks) >> (20 - 3);
748 749
				npages = MiB_TO_PAGES(size);

750
				edac_dbg(0, "mc#%d: channel %d, dimm %d, %Ld Mb (%d pages) bank: %d, rank: %d, row: %#x, col: %#x\n",
751 752 753
					 pvt->sbridge_dev->mc, i, j,
					 size, npages,
					 banks, ranks, rows, cols);
754

755
				dimm->nr_pages = npages;
756 757 758 759 760
				dimm->grain = 32;
				dimm->dtype = (banks == 8) ? DEV_X8 : DEV_X4;
				dimm->mtype = mtype;
				dimm->edac_mode = mode;
				snprintf(dimm->label, sizeof(dimm->label),
761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776
					 "CPU_SrcID#%u_Channel#%u_DIMM#%u",
					 pvt->sbridge_dev->source_id, i, j);
			}
		}
	}

	return 0;
}

static void get_memory_layout(const struct mem_ctl_info *mci)
{
	struct sbridge_pvt *pvt = mci->pvt_info;
	int i, j, k, n_sads, n_tads, sad_interl;
	u32 reg;
	u64 limit, prv = 0;
	u64 tmp_mb;
777
	u32 mb, kb;
778 779 780 781 782 783
	u32 rir_way;

	/*
	 * Step 1) Get TOLM/TOHM ranges
	 */

A
Aristeu Rozanski 已提交
784
	pvt->tolm = pvt->info.get_tolm(pvt);
785 786
	tmp_mb = (1 + pvt->tolm) >> 20;

787
	mb = div_u64_rem(tmp_mb, 1000, &kb);
788
	edac_dbg(0, "TOLM: %u.%03u GB (0x%016Lx)\n", mb, kb, (u64)pvt->tolm);
789 790

	/* Address range is already 45:25 */
A
Aristeu Rozanski 已提交
791
	pvt->tohm = pvt->info.get_tohm(pvt);
792 793
	tmp_mb = (1 + pvt->tohm) >> 20;

794
	mb = div_u64_rem(tmp_mb, 1000, &kb);
795
	edac_dbg(0, "TOHM: %u.%03u GB (0x%016Lx)\n", mb, kb, (u64)pvt->tohm);
796 797 798 799 800 801 802 803

	/*
	 * Step 2) Get SAD range and SAD Interleave list
	 * TAD registers contain the interleave wayness. However, it
	 * seems simpler to just discover it indirectly, with the
	 * algorithm bellow.
	 */
	prv = 0;
804
	for (n_sads = 0; n_sads < pvt->info.max_sad; n_sads++) {
805
		/* SAD_LIMIT Address range is 45:26 */
806
		pci_read_config_dword(pvt->pci_sad0, pvt->info.dram_rule[n_sads],
807 808 809 810 811 812 813 814 815 816
				      &reg);
		limit = SAD_LIMIT(reg);

		if (!DRAM_RULE_ENABLE(reg))
			continue;

		if (limit <= prv)
			break;

		tmp_mb = (limit + 1) >> 20;
817
		mb = div_u64_rem(tmp_mb, 1000, &kb);
818 819 820 821 822 823 824
		edac_dbg(0, "SAD#%d %s up to %u.%03u GB (0x%016Lx) Interleave: %s reg=0x%08x\n",
			 n_sads,
			 get_dram_attr(reg),
			 mb, kb,
			 ((u64)tmp_mb) << 20L,
			 INTERLEAVE_MODE(reg) ? "8:6" : "[8:6]XOR[18:16]",
			 reg);
825 826
		prv = limit;

827
		pci_read_config_dword(pvt->pci_sad0, pvt->info.interleave_list[n_sads],
828
				      &reg);
A
Aristeu Rozanski 已提交
829
		sad_interl = sad_pkg(pvt->info.interleave_pkg, reg, 0);
830
		for (j = 0; j < 8; j++) {
A
Aristeu Rozanski 已提交
831 832
			u32 pkg = sad_pkg(pvt->info.interleave_pkg, reg, j);
			if (j > 0 && sad_interl == pkg)
833 834
				break;

835
			edac_dbg(0, "SAD#%d, interleave #%d: %d\n",
A
Aristeu Rozanski 已提交
836
				 n_sads, j, pkg);
837 838 839 840 841 842 843 844 845 846 847 848 849 850 851
		}
	}

	/*
	 * Step 3) Get TAD range
	 */
	prv = 0;
	for (n_tads = 0; n_tads < MAX_TAD; n_tads++) {
		pci_read_config_dword(pvt->pci_ha0, tad_dram_rule[n_tads],
				      &reg);
		limit = TAD_LIMIT(reg);
		if (limit <= prv)
			break;
		tmp_mb = (limit + 1) >> 20;

852
		mb = div_u64_rem(tmp_mb, 1000, &kb);
853 854 855 856 857 858 859 860 861 862
		edac_dbg(0, "TAD#%d: up to %u.%03u GB (0x%016Lx), socket interleave %d, memory interleave %d, TGT: %d, %d, %d, %d, reg=0x%08x\n",
			 n_tads, mb, kb,
			 ((u64)tmp_mb) << 20L,
			 (u32)TAD_SOCK(reg),
			 (u32)TAD_CH(reg),
			 (u32)TAD_TGT0(reg),
			 (u32)TAD_TGT1(reg),
			 (u32)TAD_TGT2(reg),
			 (u32)TAD_TGT3(reg),
			 reg);
863
		prv = limit;
864 865 866 867 868 869 870 871 872 873 874 875 876
	}

	/*
	 * Step 4) Get TAD offsets, per each channel
	 */
	for (i = 0; i < NUM_CHANNELS; i++) {
		if (!pvt->channel[i].dimms)
			continue;
		for (j = 0; j < n_tads; j++) {
			pci_read_config_dword(pvt->pci_tad[i],
					      tad_ch_nilv_offset[j],
					      &reg);
			tmp_mb = TAD_OFFSET(reg) >> 20;
877
			mb = div_u64_rem(tmp_mb, 1000, &kb);
878 879 880 881 882
			edac_dbg(0, "TAD CH#%d, offset #%d: %u.%03u GB (0x%016Lx), reg=0x%08x\n",
				 i, j,
				 mb, kb,
				 ((u64)tmp_mb) << 20L,
				 reg);
883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899
		}
	}

	/*
	 * Step 6) Get RIR Wayness/Limit, per each channel
	 */
	for (i = 0; i < NUM_CHANNELS; i++) {
		if (!pvt->channel[i].dimms)
			continue;
		for (j = 0; j < MAX_RIR_RANGES; j++) {
			pci_read_config_dword(pvt->pci_tad[i],
					      rir_way_limit[j],
					      &reg);

			if (!IS_RIR_VALID(reg))
				continue;

900
			tmp_mb = pvt->info.rir_limit(reg) >> 20;
901
			rir_way = 1 << RIR_WAY(reg);
902
			mb = div_u64_rem(tmp_mb, 1000, &kb);
903 904 905 906 907 908
			edac_dbg(0, "CH#%d RIR#%d, limit: %u.%03u GB (0x%016Lx), way: %d, reg=0x%08x\n",
				 i, j,
				 mb, kb,
				 ((u64)tmp_mb) << 20L,
				 rir_way,
				 reg);
909 910 911 912 913 914 915

			for (k = 0; k < rir_way; k++) {
				pci_read_config_dword(pvt->pci_tad[i],
						      rir_offset[j][k],
						      &reg);
				tmp_mb = RIR_OFFSET(reg) << 6;

916
				mb = div_u64_rem(tmp_mb, 1000, &kb);
917 918 919 920 921 922
				edac_dbg(0, "CH#%d RIR#%d INTL#%d, offset %u.%03u GB (0x%016Lx), tgt: %d, reg=0x%08x\n",
					 i, j, k,
					 mb, kb,
					 ((u64)tmp_mb) << 20L,
					 (u32)RIR_RNK_TGT(reg),
					 reg);
923 924 925 926 927
			}
		}
	}
}

928
static struct mem_ctl_info *get_mci_for_node_id(u8 node_id)
929 930 931 932 933 934 935 936 937 938 939 940 941 942 943
{
	struct sbridge_dev *sbridge_dev;

	list_for_each_entry(sbridge_dev, &sbridge_edac_list, list) {
		if (sbridge_dev->node_id == node_id)
			return sbridge_dev->mci;
	}
	return NULL;
}

static int get_memory_error_data(struct mem_ctl_info *mci,
				 u64 addr,
				 u8 *socket,
				 long *channel_mask,
				 u8 *rank,
944
				 char **area_type, char *msg)
945 946 947
{
	struct mem_ctl_info	*new_mci;
	struct sbridge_pvt *pvt = mci->pvt_info;
948
	struct pci_dev		*pci_ha;
949 950 951
	int 			n_rir, n_sads, n_tads, sad_way, sck_xch;
	int			sad_interl, idx, base_ch;
	int			interleave_mode;
952
	unsigned		sad_interleave[pvt->info.max_interleave];
953
	u32			reg;
954
	u8			ch_way, sck_way, pkg, sad_ha = 0;
955 956
	u32			tad_offset;
	u32			rir_way;
957
	u32			mb, kb;
958
	u64			ch_addr, offset, limit = 0, prv = 0;
959 960 961 962 963 964 965 966 967


	/*
	 * Step 0) Check if the address is at special memory ranges
	 * The check bellow is probably enough to fill all cases where
	 * the error is not inside a memory, except for the legacy
	 * range (e. g. VGA addresses). It is unlikely, however, that the
	 * memory controller would generate an error on that range.
	 */
968
	if ((addr > (u64) pvt->tolm) && (addr < (1LL << 32))) {
969 970 971 972 973 974 975 976 977 978 979
		sprintf(msg, "Error at TOLM area, on addr 0x%08Lx", addr);
		return -EINVAL;
	}
	if (addr >= (u64)pvt->tohm) {
		sprintf(msg, "Error at MMIOH area, on addr 0x%016Lx", addr);
		return -EINVAL;
	}

	/*
	 * Step 1) Get socket
	 */
980 981
	for (n_sads = 0; n_sads < pvt->info.max_sad; n_sads++) {
		pci_read_config_dword(pvt->pci_sad0, pvt->info.dram_rule[n_sads],
982 983 984 985 986 987 988 989 990 991 992 993 994 995
				      &reg);

		if (!DRAM_RULE_ENABLE(reg))
			continue;

		limit = SAD_LIMIT(reg);
		if (limit <= prv) {
			sprintf(msg, "Can't discover the memory socket");
			return -EINVAL;
		}
		if  (addr <= limit)
			break;
		prv = limit;
	}
996
	if (n_sads == pvt->info.max_sad) {
997 998 999
		sprintf(msg, "Can't discover the memory socket");
		return -EINVAL;
	}
1000
	*area_type = get_dram_attr(reg);
1001 1002
	interleave_mode = INTERLEAVE_MODE(reg);

1003
	pci_read_config_dword(pvt->pci_sad0, pvt->info.interleave_list[n_sads],
1004
			      &reg);
1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029

	if (pvt->info.type == SANDY_BRIDGE) {
		sad_interl = sad_pkg(pvt->info.interleave_pkg, reg, 0);
		for (sad_way = 0; sad_way < 8; sad_way++) {
			u32 pkg = sad_pkg(pvt->info.interleave_pkg, reg, sad_way);
			if (sad_way > 0 && sad_interl == pkg)
				break;
			sad_interleave[sad_way] = pkg;
			edac_dbg(0, "SAD interleave #%d: %d\n",
				 sad_way, sad_interleave[sad_way]);
		}
		edac_dbg(0, "mc#%d: Error detected on SAD#%d: address 0x%016Lx < 0x%016Lx, Interleave [%d:6]%s\n",
			 pvt->sbridge_dev->mc,
			 n_sads,
			 addr,
			 limit,
			 sad_way + 7,
			 !interleave_mode ? "" : "XOR[18:16]");
		if (interleave_mode)
			idx = ((addr >> 6) ^ (addr >> 16)) & 7;
		else
			idx = (addr >> 6) & 7;
		switch (sad_way) {
		case 1:
			idx = 0;
1030
			break;
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047
		case 2:
			idx = idx & 1;
			break;
		case 4:
			idx = idx & 3;
			break;
		case 8:
			break;
		default:
			sprintf(msg, "Can't discover socket interleave");
			return -EINVAL;
		}
		*socket = sad_interleave[idx];
		edac_dbg(0, "SAD interleave index: %d (wayness %d) = CPU socket %d\n",
			 idx, sad_way, *socket);
	} else {
		/* Ivy Bridge's SAD mode doesn't support XOR interleave mode */
1048
		idx = (addr >> 6) & 7;
1049 1050 1051 1052 1053
		pkg = sad_pkg(pvt->info.interleave_pkg, reg, idx);
		*socket = sad_pkg_socket(pkg);
		sad_ha = sad_pkg_ha(pkg);
		edac_dbg(0, "SAD interleave package: %d = CPU socket %d, HA %d\n",
			 idx, *socket, sad_ha);
1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072
	}

	/*
	 * Move to the proper node structure, in order to access the
	 * right PCI registers
	 */
	new_mci = get_mci_for_node_id(*socket);
	if (!new_mci) {
		sprintf(msg, "Struct for socket #%u wasn't initialized",
			*socket);
		return -EINVAL;
	}
	mci = new_mci;
	pvt = mci->pvt_info;

	/*
	 * Step 2) Get memory channel
	 */
	prv = 0;
1073 1074 1075 1076 1077 1078 1079 1080
	if (pvt->info.type == SANDY_BRIDGE)
		pci_ha = pvt->pci_ha0;
	else {
		if (sad_ha)
			pci_ha = pvt->pci_ha1;
		else
			pci_ha = pvt->pci_ha0;
	}
1081
	for (n_tads = 0; n_tads < MAX_TAD; n_tads++) {
1082
		pci_read_config_dword(pci_ha, tad_dram_rule[n_tads], &reg);
1083 1084 1085 1086 1087 1088 1089 1090 1091
		limit = TAD_LIMIT(reg);
		if (limit <= prv) {
			sprintf(msg, "Can't discover the memory channel");
			return -EINVAL;
		}
		if  (addr <= limit)
			break;
		prv = limit;
	}
1092 1093 1094 1095 1096
	if (n_tads == MAX_TAD) {
		sprintf(msg, "Can't discover the memory channel");
		return -EINVAL;
	}

1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127
	ch_way = TAD_CH(reg) + 1;
	sck_way = TAD_SOCK(reg) + 1;

	if (ch_way == 3)
		idx = addr >> 6;
	else
		idx = addr >> (6 + sck_way);
	idx = idx % ch_way;

	/*
	 * FIXME: Shouldn't we use CHN_IDX_OFFSET() here, when ch_way == 3 ???
	 */
	switch (idx) {
	case 0:
		base_ch = TAD_TGT0(reg);
		break;
	case 1:
		base_ch = TAD_TGT1(reg);
		break;
	case 2:
		base_ch = TAD_TGT2(reg);
		break;
	case 3:
		base_ch = TAD_TGT3(reg);
		break;
	default:
		sprintf(msg, "Can't discover the TAD target");
		return -EINVAL;
	}
	*channel_mask = 1 << base_ch;

1128 1129 1130 1131
	pci_read_config_dword(pvt->pci_tad[base_ch],
				tad_ch_nilv_offset[n_tads],
				&tad_offset);

1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150
	if (pvt->is_mirrored) {
		*channel_mask |= 1 << ((base_ch + 2) % 4);
		switch(ch_way) {
		case 2:
		case 4:
			sck_xch = 1 << sck_way * (ch_way >> 1);
			break;
		default:
			sprintf(msg, "Invalid mirror set. Can't decode addr");
			return -EINVAL;
		}
	} else
		sck_xch = (1 << sck_way) * ch_way;

	if (pvt->is_lockstep)
		*channel_mask |= 1 << ((base_ch + 1) % 4);

	offset = TAD_OFFSET(tad_offset);

1151 1152 1153 1154 1155 1156 1157 1158 1159 1160
	edac_dbg(0, "TAD#%d: address 0x%016Lx < 0x%016Lx, socket interleave %d, channel interleave %d (offset 0x%08Lx), index %d, base ch: %d, ch mask: 0x%02lx\n",
		 n_tads,
		 addr,
		 limit,
		 (u32)TAD_SOCK(reg),
		 ch_way,
		 offset,
		 idx,
		 base_ch,
		 *channel_mask);
1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174

	/* Calculate channel address */
	/* Remove the TAD offset */

	if (offset > addr) {
		sprintf(msg, "Can't calculate ch addr: TAD offset 0x%08Lx is too high for addr 0x%08Lx!",
			offset, addr);
		return -EINVAL;
	}
	addr -= offset;
	/* Store the low bits [0:6] of the addr */
	ch_addr = addr & 0x7f;
	/* Remove socket wayness and remove 6 bits */
	addr >>= 6;
1175
	addr = div_u64(addr, sck_xch);
1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193
#if 0
	/* Divide by channel way */
	addr = addr / ch_way;
#endif
	/* Recover the last 6 bits */
	ch_addr |= addr << 6;

	/*
	 * Step 3) Decode rank
	 */
	for (n_rir = 0; n_rir < MAX_RIR_RANGES; n_rir++) {
		pci_read_config_dword(pvt->pci_tad[base_ch],
				      rir_way_limit[n_rir],
				      &reg);

		if (!IS_RIR_VALID(reg))
			continue;

1194
		limit = pvt->info.rir_limit(reg);
1195
		mb = div_u64_rem(limit >> 20, 1000, &kb);
1196 1197 1198 1199 1200
		edac_dbg(0, "RIR#%d, limit: %u.%03u GB (0x%016Lx), way: %d\n",
			 n_rir,
			 mb, kb,
			 limit,
			 1 << RIR_WAY(reg));
1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
		if  (ch_addr <= limit)
			break;
	}
	if (n_rir == MAX_RIR_RANGES) {
		sprintf(msg, "Can't discover the memory rank for ch addr 0x%08Lx",
			ch_addr);
		return -EINVAL;
	}
	rir_way = RIR_WAY(reg);
	if (pvt->is_close_pg)
		idx = (ch_addr >> 6);
	else
		idx = (ch_addr >> 13);	/* FIXME: Datasheet says to shift by 15 */
	idx %= 1 << rir_way;

	pci_read_config_dword(pvt->pci_tad[base_ch],
			      rir_offset[n_rir][idx],
			      &reg);
	*rank = RIR_RNK_TGT(reg);

1221 1222 1223 1224 1225 1226
	edac_dbg(0, "RIR#%d: channel address 0x%08Lx < 0x%08Lx, RIR interleave %d, index %d\n",
		 n_rir,
		 ch_addr,
		 limit,
		 rir_way,
		 idx);
1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242

	return 0;
}

/****************************************************************************
	Device initialization routines: put/get, init/exit
 ****************************************************************************/

/*
 *	sbridge_put_all_devices	'put' all the devices that we have
 *				reserved via 'get'
 */
static void sbridge_put_devices(struct sbridge_dev *sbridge_dev)
{
	int i;

1243
	edac_dbg(0, "\n");
1244 1245 1246 1247
	for (i = 0; i < sbridge_dev->n_devs; i++) {
		struct pci_dev *pdev = sbridge_dev->pdev[i];
		if (!pdev)
			continue;
1248 1249 1250
		edac_dbg(0, "Removing dev %02x:%02x.%d\n",
			 pdev->bus->number,
			 PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274
		pci_dev_put(pdev);
	}
}

static void sbridge_put_all_devices(void)
{
	struct sbridge_dev *sbridge_dev, *tmp;

	list_for_each_entry_safe(sbridge_dev, tmp, &sbridge_edac_list, list) {
		sbridge_put_devices(sbridge_dev);
		free_sbridge_dev(sbridge_dev);
	}
}

static int sbridge_get_onedevice(struct pci_dev **prev,
				 u8 *num_mc,
				 const struct pci_id_table *table,
				 const unsigned devno)
{
	struct sbridge_dev *sbridge_dev;
	const struct pci_id_descr *dev_descr = &table->descr[devno];
	struct pci_dev *pdev = NULL;
	u8 bus = 0;

1275
	sbridge_printk(KERN_DEBUG,
1276
		"Seeking for: PCI ID %04x:%04x\n",
1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290
		PCI_VENDOR_ID_INTEL, dev_descr->dev_id);

	pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
			      dev_descr->dev_id, *prev);

	if (!pdev) {
		if (*prev) {
			*prev = pdev;
			return 0;
		}

		if (dev_descr->optional)
			return 0;

1291
		/* if the HA wasn't found */
1292 1293 1294 1295
		if (devno == 0)
			return -ENODEV;

		sbridge_printk(KERN_INFO,
1296
			"Device not found: %04x:%04x\n",
1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315
			PCI_VENDOR_ID_INTEL, dev_descr->dev_id);

		/* End of list, leave */
		return -ENODEV;
	}
	bus = pdev->bus->number;

	sbridge_dev = get_sbridge_dev(bus);
	if (!sbridge_dev) {
		sbridge_dev = alloc_sbridge_dev(bus, table);
		if (!sbridge_dev) {
			pci_dev_put(pdev);
			return -ENOMEM;
		}
		(*num_mc)++;
	}

	if (sbridge_dev->pdev[devno]) {
		sbridge_printk(KERN_ERR,
1316
			"Duplicated device for %04x:%04x\n",
1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
			PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
		pci_dev_put(pdev);
		return -ENODEV;
	}

	sbridge_dev->pdev[devno] = pdev;

	/* Be sure that the device is enabled */
	if (unlikely(pci_enable_device(pdev) < 0)) {
		sbridge_printk(KERN_ERR,
1327
			"Couldn't enable %04x:%04x\n",
1328 1329 1330 1331
			PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
		return -ENODEV;
	}

1332
	edac_dbg(0, "Detected %04x:%04x\n",
1333
		 PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346

	/*
	 * As stated on drivers/pci/search.c, the reference count for
	 * @from is always decremented if it is not %NULL. So, as we need
	 * to get all devices up to null, we need to do a get for the device
	 */
	pci_dev_get(pdev);

	*prev = pdev;

	return 0;
}

1347 1348
/*
 * sbridge_get_all_devices - Find and perform 'get' operation on the MCH's
1349
 *			     devices we want to reference for this driver.
1350 1351 1352 1353 1354 1355 1356 1357
 * @num_mc: pointer to the memory controllers count, to be incremented in case
 * 	    of success.
 * @table: model specific table
 *
 * returns 0 in case of success or error code
 */
static int sbridge_get_all_devices(u8 *num_mc,
				   const struct pci_id_table *table)
1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383
{
	int i, rc;
	struct pci_dev *pdev = NULL;

	while (table && table->descr) {
		for (i = 0; i < table->n_devs; i++) {
			pdev = NULL;
			do {
				rc = sbridge_get_onedevice(&pdev, num_mc,
							   table, i);
				if (rc < 0) {
					if (i == 0) {
						i = table->n_devs;
						break;
					}
					sbridge_put_all_devices();
					return -ENODEV;
				}
			} while (pdev);
		}
		table++;
	}

	return 0;
}

A
Aristeu Rozanski 已提交
1384 1385
static int sbridge_mci_bind_devs(struct mem_ctl_info *mci,
				 struct sbridge_dev *sbridge_dev)
1386 1387 1388
{
	struct sbridge_pvt *pvt = mci->pvt_info;
	struct pci_dev *pdev;
1389
	int i;
1390 1391 1392 1393 1394

	for (i = 0; i < sbridge_dev->n_devs; i++) {
		pdev = sbridge_dev->pdev[i];
		if (!pdev)
			continue;
1395 1396 1397 1398

		switch (pdev->device) {
		case PCI_DEVICE_ID_INTEL_SBRIDGE_SAD0:
			pvt->pci_sad0 = pdev;
1399
			break;
1400 1401
		case PCI_DEVICE_ID_INTEL_SBRIDGE_SAD1:
			pvt->pci_sad1 = pdev;
1402
			break;
1403 1404
		case PCI_DEVICE_ID_INTEL_SBRIDGE_BR:
			pvt->pci_br0 = pdev;
1405
			break;
1406 1407
		case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0:
			pvt->pci_ha0 = pdev;
1408
			break;
1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425
		case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TA:
			pvt->pci_ta = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_RAS:
			pvt->pci_ras = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD0:
		case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD1:
		case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD2:
		case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD3:
		{
			int id = pdev->device - PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TAD0;
			pvt->pci_tad[id] = pdev;
		}
			break;
		case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_DDRIO:
			pvt->pci_ddrio = pdev;
1426 1427 1428 1429 1430
			break;
		default:
			goto error;
		}

1431 1432
		edac_dbg(0, "Associated PCI %02x:%02x, bus %d with dev = %p\n",
			 pdev->vendor, pdev->device,
1433 1434
			 sbridge_dev->bus,
			 pdev);
1435 1436 1437 1438
	}

	/* Check if everything were registered */
	if (!pvt->pci_sad0 || !pvt->pci_sad1 || !pvt->pci_ha0 ||
1439
	    !pvt-> pci_tad || !pvt->pci_ras  || !pvt->pci_ta)
1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452
		goto enodev;

	for (i = 0; i < NUM_CHANNELS; i++) {
		if (!pvt->pci_tad[i])
			goto enodev;
	}
	return 0;

enodev:
	sbridge_printk(KERN_ERR, "Some needed devices are missing\n");
	return -ENODEV;

error:
1453 1454
	sbridge_printk(KERN_ERR, "Unexpected device %02x:%02x\n",
		       PCI_VENDOR_ID_INTEL, pdev->device);
1455 1456 1457
	return -EINVAL;
}

1458 1459 1460 1461 1462
static int ibridge_mci_bind_devs(struct mem_ctl_info *mci,
				 struct sbridge_dev *sbridge_dev)
{
	struct sbridge_pvt *pvt = mci->pvt_info;
	struct pci_dev *pdev, *tmp;
1463
	int i;
1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477
	bool mode_2ha = false;

	tmp = pci_get_device(PCI_VENDOR_ID_INTEL,
			     PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1, NULL);
	if (tmp) {
		mode_2ha = true;
		pci_dev_put(tmp);
	}

	for (i = 0; i < sbridge_dev->n_devs; i++) {
		pdev = sbridge_dev->pdev[i];
		if (!pdev)
			continue;

1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491
		switch (pdev->device) {
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0:
			pvt->pci_ha0 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA:
			pvt->pci_ta = pdev;
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_RAS:
			pvt->pci_ras = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD2:
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD3:
			/* if we have 2 HAs active, channels 2 and 3
			 * are in other device */
			if (mode_2ha)
1492
				break;
1493 1494 1495 1496 1497 1498 1499
			/* fall through */
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD0:
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD1:
		{
			int id = pdev->device - PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD0;
			pvt->pci_tad[id] = pdev;
		}
1500
			break;
1501 1502 1503 1504 1505
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_2HA_DDRIO0:
			pvt->pci_ddrio = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_1HA_DDRIO0:
			if (!mode_2ha)
1506 1507
				pvt->pci_ddrio = pdev;
			break;
1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524
		case PCI_DEVICE_ID_INTEL_IBRIDGE_SAD:
			pvt->pci_sad0 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_IBRIDGE_BR0:
			pvt->pci_br0 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_IBRIDGE_BR1:
			pvt->pci_br1 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1:
			pvt->pci_ha1 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD0:
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD1:
		{
			int id = pdev->device - PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD0 + 2;

1525 1526 1527
			/* we shouldn't have this device if we have just one
			 * HA present */
			WARN_ON(!mode_2ha);
1528 1529 1530
			pvt->pci_tad[id] = pdev;
		}
			break;
1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558
		default:
			goto error;
		}

		edac_dbg(0, "Associated PCI %02x.%02d.%d with dev = %p\n",
			 sbridge_dev->bus,
			 PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn),
			 pdev);
	}

	/* Check if everything were registered */
	if (!pvt->pci_sad0 || !pvt->pci_ha0 || !pvt->pci_br0 ||
	    !pvt->pci_br1 || !pvt->pci_tad || !pvt->pci_ras  ||
	    !pvt->pci_ta)
		goto enodev;

	for (i = 0; i < NUM_CHANNELS; i++) {
		if (!pvt->pci_tad[i])
			goto enodev;
	}
	return 0;

enodev:
	sbridge_printk(KERN_ERR, "Some needed devices are missing\n");
	return -ENODEV;

error:
	sbridge_printk(KERN_ERR,
1559 1560
		       "Unexpected device %02x:%02x\n", PCI_VENDOR_ID_INTEL,
			pdev->device);
1561 1562 1563
	return -EINVAL;
}

1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578
/****************************************************************************
			Error check routines
 ****************************************************************************/

/*
 * While Sandy Bridge has error count registers, SMI BIOS read values from
 * and resets the counters. So, they are not reliable for the OS to read
 * from them. So, we have no option but to just trust on whatever MCE is
 * telling us about the errors.
 */
static void sbridge_mce_output_error(struct mem_ctl_info *mci,
				    const struct mce *m)
{
	struct mem_ctl_info *new_mci;
	struct sbridge_pvt *pvt = mci->pvt_info;
1579
	enum hw_event_mc_err_type tp_event;
1580
	char *type, *optype, msg[256];
1581 1582 1583
	bool ripv = GET_BITFIELD(m->mcgstatus, 0, 0);
	bool overflow = GET_BITFIELD(m->status, 62, 62);
	bool uncorrected_error = GET_BITFIELD(m->status, 61, 61);
1584
	bool recoverable;
1585 1586 1587 1588 1589 1590 1591
	u32 core_err_cnt = GET_BITFIELD(m->status, 38, 52);
	u32 mscod = GET_BITFIELD(m->status, 16, 31);
	u32 errcode = GET_BITFIELD(m->status, 0, 15);
	u32 channel = GET_BITFIELD(m->status, 0, 3);
	u32 optypenum = GET_BITFIELD(m->status, 4, 6);
	long channel_mask, first_channel;
	u8  rank, socket;
1592
	int rc, dimm;
1593
	char *area_type = NULL;
1594

1595 1596 1597 1598 1599
	if (pvt->info.type == IVY_BRIDGE)
		recoverable = true;
	else
		recoverable = GET_BITFIELD(m->status, 56, 56);

1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611
	if (uncorrected_error) {
		if (ripv) {
			type = "FATAL";
			tp_event = HW_EVENT_ERR_FATAL;
		} else {
			type = "NON_FATAL";
			tp_event = HW_EVENT_ERR_UNCORRECTED;
		}
	} else {
		type = "CORRECTED";
		tp_event = HW_EVENT_ERR_CORRECTED;
	}
1612 1613

	/*
D
David Mackey 已提交
1614
	 * According with Table 15-9 of the Intel Architecture spec vol 3A,
1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628
	 * memory errors should fit in this mask:
	 *	000f 0000 1mmm cccc (binary)
	 * where:
	 *	f = Correction Report Filtering Bit. If 1, subsequent errors
	 *	    won't be shown
	 *	mmm = error type
	 *	cccc = channel
	 * If the mask doesn't match, report an error to the parsing logic
	 */
	if (! ((errcode & 0xef80) == 0x80)) {
		optype = "Can't parse: it is not a mem";
	} else {
		switch (optypenum) {
		case 0:
1629
			optype = "generic undef request error";
1630 1631
			break;
		case 1:
1632
			optype = "memory read error";
1633 1634
			break;
		case 2:
1635
			optype = "memory write error";
1636 1637
			break;
		case 3:
1638
			optype = "addr/cmd error";
1639 1640
			break;
		case 4:
1641
			optype = "memory scrubbing error";
1642 1643 1644 1645 1646 1647 1648
			break;
		default:
			optype = "reserved";
			break;
		}
	}

1649 1650 1651 1652
	/* Only decode errors with an valid address (ADDRV) */
	if (!GET_BITFIELD(m->status, 58, 58))
		return;

1653
	rc = get_memory_error_data(mci, m->addr, &socket,
1654
				   &channel_mask, &rank, &area_type, msg);
1655
	if (rc < 0)
1656
		goto err_parsing;
1657 1658
	new_mci = get_mci_for_node_id(socket);
	if (!new_mci) {
1659 1660
		strcpy(msg, "Error: socket got corrupted!");
		goto err_parsing;
1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675
	}
	mci = new_mci;
	pvt = mci->pvt_info;

	first_channel = find_first_bit(&channel_mask, NUM_CHANNELS);

	if (rank < 4)
		dimm = 0;
	else if (rank < 8)
		dimm = 1;
	else
		dimm = 2;


	/*
1676 1677 1678 1679
	 * FIXME: On some memory configurations (mirror, lockstep), the
	 * Memory Controller can't point the error to a single DIMM. The
	 * EDAC core should be handling the channel mask, in order to point
	 * to the group of dimm's where the error may be happening.
1680
	 */
1681 1682 1683
	if (!pvt->is_lockstep && !pvt->is_mirrored && !pvt->is_close_pg)
		channel = first_channel;

1684
	snprintf(msg, sizeof(msg),
1685
		 "%s%s area:%s err_code:%04x:%04x socket:%d channel_mask:%ld rank:%d",
1686 1687 1688 1689 1690 1691 1692
		 overflow ? " OVERFLOW" : "",
		 (uncorrected_error && recoverable) ? " recoverable" : "",
		 area_type,
		 mscod, errcode,
		 socket,
		 channel_mask,
		 rank);
1693

1694
	edac_dbg(0, "%s\n", msg);
1695

1696 1697
	/* FIXME: need support for channel mask */

1698
	/* Call the helper to output message */
1699
	edac_mc_handle_error(tp_event, mci, core_err_cnt,
1700 1701
			     m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0,
			     channel, dimm, -1,
1702
			     optype, msg);
1703 1704
	return;
err_parsing:
1705
	edac_mc_handle_error(tp_event, mci, core_err_cnt, 0, 0, 0,
1706
			     -1, -1, -1,
1707
			     msg, "");
1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769

}

/*
 *	sbridge_check_error	Retrieve and process errors reported by the
 *				hardware. Called by the Core module.
 */
static void sbridge_check_error(struct mem_ctl_info *mci)
{
	struct sbridge_pvt *pvt = mci->pvt_info;
	int i;
	unsigned count = 0;
	struct mce *m;

	/*
	 * MCE first step: Copy all mce errors into a temporary buffer
	 * We use a double buffering here, to reduce the risk of
	 * loosing an error.
	 */
	smp_rmb();
	count = (pvt->mce_out + MCE_LOG_LEN - pvt->mce_in)
		% MCE_LOG_LEN;
	if (!count)
		return;

	m = pvt->mce_outentry;
	if (pvt->mce_in + count > MCE_LOG_LEN) {
		unsigned l = MCE_LOG_LEN - pvt->mce_in;

		memcpy(m, &pvt->mce_entry[pvt->mce_in], sizeof(*m) * l);
		smp_wmb();
		pvt->mce_in = 0;
		count -= l;
		m += l;
	}
	memcpy(m, &pvt->mce_entry[pvt->mce_in], sizeof(*m) * count);
	smp_wmb();
	pvt->mce_in += count;

	smp_rmb();
	if (pvt->mce_overrun) {
		sbridge_printk(KERN_ERR, "Lost %d memory errors\n",
			      pvt->mce_overrun);
		smp_wmb();
		pvt->mce_overrun = 0;
	}

	/*
	 * MCE second step: parse errors and display
	 */
	for (i = 0; i < count; i++)
		sbridge_mce_output_error(mci, &pvt->mce_outentry[i]);
}

/*
 * sbridge_mce_check_error	Replicates mcelog routine to get errors
 *				This routine simply queues mcelog errors, and
 *				return. The error itself should be handled later
 *				by sbridge_check_error.
 * WARNING: As this routine should be called at NMI time, extra care should
 * be taken to avoid deadlocks, and to be as fast as possible.
 */
1770 1771
static int sbridge_mce_check_error(struct notifier_block *nb, unsigned long val,
				   void *data)
1772
{
1773 1774 1775
	struct mce *mce = (struct mce *)data;
	struct mem_ctl_info *mci;
	struct sbridge_pvt *pvt;
1776
	char *type;
1777

1778 1779 1780
	if (get_edac_report_status() == EDAC_REPORTING_DISABLED)
		return NOTIFY_DONE;

1781 1782 1783 1784
	mci = get_mci_for_node_id(mce->socketid);
	if (!mci)
		return NOTIFY_BAD;
	pvt = mci->pvt_info;
1785 1786 1787 1788 1789 1790 1791 1792

	/*
	 * Just let mcelog handle it if the error is
	 * outside the memory controller. A memory error
	 * is indicated by bit 7 = 1 and bits = 8-11,13-15 = 0.
	 * bit 12 has an special meaning.
	 */
	if ((mce->status & 0xefff) >> 7 != 1)
1793
		return NOTIFY_DONE;
1794

1795 1796 1797 1798 1799
	if (mce->mcgstatus & MCG_STATUS_MCIP)
		type = "Exception";
	else
		type = "Event";

1800
	sbridge_mc_printk(mci, KERN_DEBUG, "HANDLING MCE MEMORY ERROR\n");
1801

1802 1803 1804 1805 1806 1807
	sbridge_mc_printk(mci, KERN_DEBUG, "CPU %d: Machine Check %s: %Lx "
			  "Bank %d: %016Lx\n", mce->extcpu, type,
			  mce->mcgstatus, mce->bank, mce->status);
	sbridge_mc_printk(mci, KERN_DEBUG, "TSC %llx ", mce->tsc);
	sbridge_mc_printk(mci, KERN_DEBUG, "ADDR %llx ", mce->addr);
	sbridge_mc_printk(mci, KERN_DEBUG, "MISC %llx ", mce->misc);
1808

1809 1810 1811
	sbridge_mc_printk(mci, KERN_DEBUG, "PROCESSOR %u:%x TIME %llu SOCKET "
			  "%u APIC %x\n", mce->cpuvendor, mce->cpuid,
			  mce->time, mce->socketid, mce->apicid);
1812 1813 1814 1815 1816

	smp_rmb();
	if ((pvt->mce_out + 1) % MCE_LOG_LEN == pvt->mce_in) {
		smp_wmb();
		pvt->mce_overrun++;
1817
		return NOTIFY_DONE;
1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829
	}

	/* Copy memory error at the ringbuffer */
	memcpy(&pvt->mce_entry[pvt->mce_out], mce, sizeof(*mce));
	smp_wmb();
	pvt->mce_out = (pvt->mce_out + 1) % MCE_LOG_LEN;

	/* Handle fatal errors immediately */
	if (mce->mcgstatus & 1)
		sbridge_check_error(mci);

	/* Advice mcelog that the error were handled */
1830
	return NOTIFY_STOP;
1831 1832
}

1833 1834 1835 1836
static struct notifier_block sbridge_mce_dec = {
	.notifier_call      = sbridge_mce_check_error,
};

1837 1838 1839 1840 1841 1842 1843 1844 1845 1846
/****************************************************************************
			EDAC register/unregister logic
 ****************************************************************************/

static void sbridge_unregister_mci(struct sbridge_dev *sbridge_dev)
{
	struct mem_ctl_info *mci = sbridge_dev->mci;
	struct sbridge_pvt *pvt;

	if (unlikely(!mci || !mci->pvt_info)) {
1847
		edac_dbg(0, "MC: dev = %p\n", &sbridge_dev->pdev[0]->dev);
1848 1849 1850 1851 1852 1853 1854

		sbridge_printk(KERN_ERR, "Couldn't find mci handler\n");
		return;
	}

	pvt = mci->pvt_info;

1855 1856
	edac_dbg(0, "MC: mci = %p, dev = %p\n",
		 mci, &sbridge_dev->pdev[0]->dev);
1857 1858

	/* Remove MC sysfs nodes */
1859
	edac_mc_del_mc(mci->pdev);
1860

1861
	edac_dbg(1, "%s: free mci struct\n", mci->ctl_name);
1862 1863 1864 1865 1866
	kfree(mci->ctl_name);
	edac_mc_free(mci);
	sbridge_dev->mci = NULL;
}

1867
static int sbridge_register_mci(struct sbridge_dev *sbridge_dev, enum type type)
1868 1869
{
	struct mem_ctl_info *mci;
1870
	struct edac_mc_layer layers[2];
1871
	struct sbridge_pvt *pvt;
1872
	struct pci_dev *pdev = sbridge_dev->pdev[0];
1873
	int rc;
1874 1875

	/* Check the number of active and not disabled channels */
1876
	rc = check_if_ecc_is_active(sbridge_dev->bus, type);
1877 1878 1879 1880
	if (unlikely(rc < 0))
		return rc;

	/* allocate a new MC control structure */
1881 1882 1883 1884 1885 1886
	layers[0].type = EDAC_MC_LAYER_CHANNEL;
	layers[0].size = NUM_CHANNELS;
	layers[0].is_virt_csrow = false;
	layers[1].type = EDAC_MC_LAYER_SLOT;
	layers[1].size = MAX_DIMMS;
	layers[1].is_virt_csrow = true;
1887
	mci = edac_mc_alloc(sbridge_dev->mc, ARRAY_SIZE(layers), layers,
1888 1889
			    sizeof(*pvt));

1890 1891 1892
	if (unlikely(!mci))
		return -ENOMEM;

1893
	edac_dbg(0, "MC: mci = %p, dev = %p\n",
1894
		 mci, &pdev->dev);
1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907

	pvt = mci->pvt_info;
	memset(pvt, 0, sizeof(*pvt));

	/* Associate sbridge_dev and mci for future usage */
	pvt->sbridge_dev = sbridge_dev;
	sbridge_dev->mci = mci;

	mci->mtype_cap = MEM_FLAG_DDR3;
	mci->edac_ctl_cap = EDAC_FLAG_NONE;
	mci->edac_cap = EDAC_FLAG_NONE;
	mci->mod_name = "sbridge_edac.c";
	mci->mod_ver = SBRIDGE_REVISION;
1908
	mci->dev_name = pci_name(pdev);
1909 1910 1911 1912 1913
	mci->ctl_page_to_phys = NULL;

	/* Set the function pointer to an actual operation function */
	mci->edac_check = sbridge_check_error;

1914 1915 1916 1917 1918 1919
	pvt->info.type = type;
	if (type == IVY_BRIDGE) {
		pvt->info.rankcfgr = IB_RANK_CFG_A;
		pvt->info.get_tolm = ibridge_get_tolm;
		pvt->info.get_tohm = ibridge_get_tohm;
		pvt->info.dram_rule = ibridge_dram_rule;
1920
		pvt->info.get_memory_type = get_memory_type;
1921
		pvt->info.get_node_id = get_node_id;
1922
		pvt->info.rir_limit = rir_limit;
1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937
		pvt->info.max_sad = ARRAY_SIZE(ibridge_dram_rule);
		pvt->info.interleave_list = ibridge_interleave_list;
		pvt->info.max_interleave = ARRAY_SIZE(ibridge_interleave_list);
		pvt->info.interleave_pkg = ibridge_interleave_pkg;
		mci->ctl_name = kasprintf(GFP_KERNEL, "Ivy Bridge Socket#%d", mci->mc_idx);

		/* Store pci devices at mci for faster access */
		rc = ibridge_mci_bind_devs(mci, sbridge_dev);
		if (unlikely(rc < 0))
			goto fail0;
	} else {
		pvt->info.rankcfgr = SB_RANK_CFG_A;
		pvt->info.get_tolm = sbridge_get_tolm;
		pvt->info.get_tohm = sbridge_get_tohm;
		pvt->info.dram_rule = sbridge_dram_rule;
1938
		pvt->info.get_memory_type = get_memory_type;
1939
		pvt->info.get_node_id = get_node_id;
1940
		pvt->info.rir_limit = rir_limit;
1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952
		pvt->info.max_sad = ARRAY_SIZE(sbridge_dram_rule);
		pvt->info.interleave_list = sbridge_interleave_list;
		pvt->info.max_interleave = ARRAY_SIZE(sbridge_interleave_list);
		pvt->info.interleave_pkg = sbridge_interleave_pkg;
		mci->ctl_name = kasprintf(GFP_KERNEL, "Sandy Bridge Socket#%d", mci->mc_idx);

		/* Store pci devices at mci for faster access */
		rc = sbridge_mci_bind_devs(mci, sbridge_dev);
		if (unlikely(rc < 0))
			goto fail0;
	}

1953 1954 1955 1956 1957 1958

	/* Get dimm basic config and the memory layout */
	get_dimm_config(mci);
	get_memory_layout(mci);

	/* record ptr to the generic device */
1959
	mci->pdev = &pdev->dev;
1960 1961 1962

	/* add this new MC control structure to EDAC's list of MCs */
	if (unlikely(edac_mc_add_mc(mci))) {
1963
		edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
		rc = -EINVAL;
		goto fail0;
	}

	return 0;

fail0:
	kfree(mci->ctl_name);
	edac_mc_free(mci);
	sbridge_dev->mci = NULL;
	return rc;
}

/*
 *	sbridge_probe	Probe for ONE instance of device to see if it is
 *			present.
 *	return:
 *		0 for FOUND a device
 *		< 0 for error code
 */

1985
static int sbridge_probe(struct pci_dev *pdev, const struct pci_device_id *id)
1986 1987 1988 1989
{
	int rc;
	u8 mc, num_mc = 0;
	struct sbridge_dev *sbridge_dev;
1990
	enum type type;
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

	/* get the pci devices we want to reserve for our use */
	mutex_lock(&sbridge_edac_lock);

	/*
	 * All memory controllers are allocated at the first pass.
	 */
	if (unlikely(probed >= 1)) {
		mutex_unlock(&sbridge_edac_lock);
		return -ENODEV;
	}
	probed++;

2004 2005 2006 2007 2008 2009 2010
	if (pdev->device == PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA) {
		rc = sbridge_get_all_devices(&num_mc, pci_dev_descr_ibridge_table);
		type = IVY_BRIDGE;
	} else {
		rc = sbridge_get_all_devices(&num_mc, pci_dev_descr_sbridge_table);
		type = SANDY_BRIDGE;
	}
2011 2012 2013 2014 2015
	if (unlikely(rc < 0))
		goto fail0;
	mc = 0;

	list_for_each_entry(sbridge_dev, &sbridge_edac_list, list) {
2016 2017
		edac_dbg(0, "Registering MC#%d (%d of %d)\n",
			 mc, mc + 1, num_mc);
2018
		sbridge_dev->mc = mc++;
2019
		rc = sbridge_register_mci(sbridge_dev, type);
2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042
		if (unlikely(rc < 0))
			goto fail1;
	}

	sbridge_printk(KERN_INFO, "Driver loaded.\n");

	mutex_unlock(&sbridge_edac_lock);
	return 0;

fail1:
	list_for_each_entry(sbridge_dev, &sbridge_edac_list, list)
		sbridge_unregister_mci(sbridge_dev);

	sbridge_put_all_devices();
fail0:
	mutex_unlock(&sbridge_edac_lock);
	return rc;
}

/*
 *	sbridge_remove	destructor for one instance of device
 *
 */
2043
static void sbridge_remove(struct pci_dev *pdev)
2044 2045 2046
{
	struct sbridge_dev *sbridge_dev;

2047
	edac_dbg(0, "\n");
2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083

	/*
	 * we have a trouble here: pdev value for removal will be wrong, since
	 * it will point to the X58 register used to detect that the machine
	 * is a Nehalem or upper design. However, due to the way several PCI
	 * devices are grouped together to provide MC functionality, we need
	 * to use a different method for releasing the devices
	 */

	mutex_lock(&sbridge_edac_lock);

	if (unlikely(!probed)) {
		mutex_unlock(&sbridge_edac_lock);
		return;
	}

	list_for_each_entry(sbridge_dev, &sbridge_edac_list, list)
		sbridge_unregister_mci(sbridge_dev);

	/* Release PCI resources */
	sbridge_put_all_devices();

	probed--;

	mutex_unlock(&sbridge_edac_lock);
}

MODULE_DEVICE_TABLE(pci, sbridge_pci_tbl);

/*
 *	sbridge_driver	pci_driver structure for this module
 *
 */
static struct pci_driver sbridge_driver = {
	.name     = "sbridge_edac",
	.probe    = sbridge_probe,
2084
	.remove   = sbridge_remove,
2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095
	.id_table = sbridge_pci_tbl,
};

/*
 *	sbridge_init		Module entry function
 *			Try to initialize this module for its devices
 */
static int __init sbridge_init(void)
{
	int pci_rc;

2096
	edac_dbg(2, "\n");
2097 2098 2099 2100 2101

	/* Ensure that the OPSTATE is set correctly for POLL or NMI */
	opstate_init();

	pci_rc = pci_register_driver(&sbridge_driver);
2102 2103
	if (pci_rc >= 0) {
		mce_register_decode_chain(&sbridge_mce_dec);
2104 2105
		if (get_edac_report_status() == EDAC_REPORTING_DISABLED)
			sbridge_printk(KERN_WARNING, "Loading driver, error reporting disabled.\n");
2106
		return 0;
2107
	}
2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120

	sbridge_printk(KERN_ERR, "Failed to register device with error %d.\n",
		      pci_rc);

	return pci_rc;
}

/*
 *	sbridge_exit()	Module exit function
 *			Unregister the driver
 */
static void __exit sbridge_exit(void)
{
2121
	edac_dbg(2, "\n");
2122
	pci_unregister_driver(&sbridge_driver);
2123
	mce_unregister_decode_chain(&sbridge_mce_dec);
2124 2125 2126 2127 2128 2129 2130 2131 2132
}

module_init(sbridge_init);
module_exit(sbridge_exit);

module_param(edac_op_state, int, 0444);
MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");

MODULE_LICENSE("GPL");
2133
MODULE_AUTHOR("Mauro Carvalho Chehab");
2134
MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)");
2135
MODULE_DESCRIPTION("MC Driver for Intel Sandy Bridge and Ivy Bridge memory controllers - "
2136
		   SBRIDGE_REVISION);