sb_edac.c 68.0 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: 1.1.1 "
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

/* Devices 12 Function 6, Offsets 0x80 to 0xcc */
56
static const u32 sbridge_dram_rule[] = {
57 58 59 60
	0x80, 0x88, 0x90, 0x98, 0xa0,
	0xa8, 0xb0, 0xb8, 0xc0, 0xc8,
};

61 62 63 64 65 66
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,
};
67 68 69 70 71

#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)
72
#define A7MODE(reg)		GET_BITFIELD(reg, 26, 26)
73 74 75 76 77 78 79 80 81 82 83 84 85 86 87

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";
	}
}

88
static const u32 sbridge_interleave_list[] = {
89 90 91 92
	0x84, 0x8c, 0x94, 0x9c, 0xa4,
	0xac, 0xb4, 0xbc, 0xc4, 0xcc,
};

93 94 95 96 97 98 99
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
100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115
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 },
};

116 117 118 119 120 121 122 123 124 125 126
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
127 128
static inline int sad_pkg(const struct interleave_pkg *table, u32 reg,
			  int interleave)
129
{
A
Aristeu Rozanski 已提交
130 131
	return GET_BITFIELD(reg, table[interleave].start,
			    table[interleave].end);
132 133 134 135 136 137
}

/* Devices 12 Function 7 */

#define TOLM		0x80
#define	TOHM		0x84
138
#define HASWELL_TOLM	0xd0
139 140
#define HASWELL_TOHM_0	0xd4
#define HASWELL_TOHM_1	0xd8
141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248

#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 */

249
#define SB_RANK_CFG_A		0x0328
250

251
#define IB_RANK_CFG_A		0x0320
252 253 254 255 256

/*
 * sbridge structs
 */

257
#define NUM_CHANNELS		8	/* 2MC per socket, four chan per MC */
258 259
#define MAX_DIMMS		3	/* Max DIMMS per channel */
#define CHANNEL_UNSPECIFIED	0xf	/* Intel IA32 SDM 15-14 */
260

261 262 263
enum type {
	SANDY_BRIDGE,
	IVY_BRIDGE,
264
	HASWELL,
265
	BROADWELL,
266 267
};

A
Aristeu Rozanski 已提交
268
struct sbridge_pvt;
269
struct sbridge_info {
270
	enum type	type;
271 272 273 274
	u32		mcmtr;
	u32		rankcfgr;
	u64		(*get_tolm)(struct sbridge_pvt *pvt);
	u64		(*get_tohm)(struct sbridge_pvt *pvt);
275
	u64		(*rir_limit)(u32 reg);
276
	const u32	*dram_rule;
277
	const u32	*interleave_list;
A
Aristeu Rozanski 已提交
278
	const struct interleave_pkg *interleave_pkg;
279
	u8		max_sad;
280
	u8		max_interleave;
281
	u8		(*get_node_id)(struct sbridge_pvt *pvt);
282
	enum mem_type	(*get_memory_type)(struct sbridge_pvt *pvt);
283
	struct pci_dev	*pci_vtd;
284 285 286 287 288 289 290 291
};

struct sbridge_channel {
	u32		ranks;
	u32		dimms;
};

struct pci_id_descr {
292
	int			dev_id;
293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311
	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;
312 313 314
	struct pci_dev		*pci_sad0, *pci_sad1;
	struct pci_dev		*pci_ha0, *pci_ha1;
	struct pci_dev		*pci_br0, *pci_br1;
315
	struct pci_dev		*pci_ha1_ta;
316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339
	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;
};

340 341
#define PCI_DESCR(device_id, opt)	\
	.dev_id = (device_id),		\
342
	.optional = opt
343 344 345

static const struct pci_id_descr pci_dev_descr_sbridge[] = {
		/* Processor Home Agent */
346
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0, 0)	},
347 348

		/* Memory controller */
349 350 351 352 353 354 355
	{ 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)	},
356 357

		/* System Address Decoder */
358 359
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_SAD0, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_SAD1, 0)	},
360 361

		/* Broadcast Registers */
362
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_BR, 0)		},
363 364 365 366 367 368 369 370
};

#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. */
};

371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395
/* 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
396 397
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD2	0x0e6c
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD3	0x0e6d
398 399 400

static const struct pci_id_descr pci_dev_descr_ibridge[] = {
		/* Processor Home Agent */
401
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0, 0)		},
402 403

		/* Memory controller */
404 405 406 407 408 409
	{ 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)	},
410 411

		/* System Address Decoder */
412
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_SAD, 0)			},
413 414

		/* Broadcast Registers */
415 416
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_BR0, 1)			},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_BR1, 0)			},
417 418

		/* Optional, mode 2HA */
419
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1, 1)		},
420
#if 0
421 422
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TA, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_RAS, 1)	},
423
#endif
424 425
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD0, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD1, 1)	},
426 427
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD2, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD3, 1)	},
428

429 430
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_1HA_DDRIO0, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_2HA_DDRIO0, 1)	},
431 432 433 434 435 436 437
};

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

438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453
/* Haswell support */
/* EN processor:
 *	- 1 IMC
 *	- 3 DDR3 channels, 2 DPC per channel
 * EP processor:
 *	- 1 or 2 IMC
 *	- 4 DDR4 channels, 3 DPC per channel
 * EP 4S processor:
 *	- 2 IMC
 *	- 4 DDR4 channels, 3 DPC per channel
 * EX processor:
 *	- 2 IMC
 *	- each IMC interfaces with a SMI 2 channel
 *	- each SMI channel interfaces with a scalable memory buffer
 *	- each scalable memory buffer supports 4 DDR3/DDR4 channels, 3 DPC
 */
454
#define HASWELL_DDRCRCLKCONTROLS 0xa10 /* Ditto on Broadwell */
455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504
#define HASWELL_HASYSDEFEATURE2 0x84
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_VTD_MISC 0x2f28
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0	0x2fa0
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1	0x2f60
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TA	0x2fa8
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_THERMAL 0x2f71
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TA	0x2f68
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_THERMAL 0x2f79
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD0 0x2ffc
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD1 0x2ffd
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD0 0x2faa
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD1 0x2fab
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD2 0x2fac
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD3 0x2fad
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD0 0x2f6a
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD1 0x2f6b
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD2 0x2f6c
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD3 0x2f6d
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO0 0x2fbd
static const struct pci_id_descr pci_dev_descr_haswell[] = {
	/* first item must be the HA */
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0, 0)		},

	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD0, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD1, 0)	},

	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1, 1)		},

	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TA, 0)		},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_THERMAL, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD0, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD1, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD2, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD3, 1)	},

	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO0, 1)		},

	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TA, 1)		},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_THERMAL, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD0, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD1, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD2, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD3, 1)	},
};

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

505 506 507 508 509 510
/*
 * Broadwell support
 *
 * DE processor:
 *	- 1 IMC
 *	- 2 DDR3 channels, 2 DPC per channel
511 512 513 514 515 516 517 518 519 520 521
 * EP processor:
 *	- 1 or 2 IMC
 *	- 4 DDR4 channels, 3 DPC per channel
 * EP 4S processor:
 *	- 2 IMC
 *	- 4 DDR4 channels, 3 DPC per channel
 * EX processor:
 *	- 2 IMC
 *	- each IMC interfaces with a SMI 2 channel
 *	- each SMI channel interfaces with a scalable memory buffer
 *	- each scalable memory buffer supports 4 DDR3/DDR4 channels, 3 DPC
522 523 524
 */
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_VTD_MISC 0x6f28
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0	0x6fa0
525
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1	0x6f60
526 527
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TA	0x6fa8
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_THERMAL 0x6f71
528 529
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TA	0x6f68
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_THERMAL 0x6f79
530 531 532 533 534 535
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD0 0x6ffc
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD1 0x6ffd
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD0 0x6faa
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD1 0x6fab
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD2 0x6fac
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD3 0x6fad
536 537 538 539
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD0 0x6f6a
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD1 0x6f6b
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD2 0x6f6c
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD3 0x6f6d
540 541 542 543 544 545 546 547 548
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_DDRIO0 0x6faf

static const struct pci_id_descr pci_dev_descr_broadwell[] = {
	/* first item must be the HA */
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0, 0)		},

	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD0, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD1, 0)	},

549 550
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1, 1)		},

551 552 553 554
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TA, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_THERMAL, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD0, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD1, 0)	},
555 556 557
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD2, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD3, 1)	},

558
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_DDRIO0, 1)	},
559 560 561 562 563 564 565

	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TA, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_THERMAL, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD0, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD1, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD2, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD3, 1)	},
566 567 568 569 570 571 572
};

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

573 574 575
/*
 *	pci_device_id	table for which devices we are looking for
 */
576
static const struct pci_device_id sbridge_pci_tbl[] = {
577
	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0)},
578
	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA)},
579
	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0)},
580
	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0)},
581 582 583 584 585
	{0,}			/* 0 terminated list. */
};


/****************************************************************************
D
David Mackey 已提交
586
			Ancillary status routines
587 588
 ****************************************************************************/

589
static inline int numrank(enum type type, u32 mtr)
590 591
{
	int ranks = (1 << RANK_CNT_BITS(mtr));
592 593
	int max = 4;

594
	if (type == HASWELL || type == BROADWELL)
595
		max = 8;
596

597 598 599
	if (ranks > max) {
		edac_dbg(0, "Invalid number of ranks: %d (max = %i) raw value = %x (%04x)\n",
			 ranks, max, (unsigned int)RANK_CNT_BITS(mtr), mtr);
600 601 602 603 604 605 606 607 608 609 610
		return -EINVAL;
	}

	return ranks;
}

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

	if (rows < 13 || rows > 18) {
611 612
		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);
613 614 615 616 617 618 619 620 621 622 623
		return -EINVAL;
	}

	return 1 << rows;
}

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

	if (cols > 12) {
624 625
		edac_dbg(0, "Invalid number of cols: %d (max = 4) raw value = %x (%04x)\n",
			 cols, (unsigned int)COL_WIDTH_BITS(mtr), mtr);
626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673
		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
674 675 676 677 678 679 680 681 682
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
683 684 685 686 687 688 689 690
static u64 sbridge_get_tohm(struct sbridge_pvt *pvt)
{
	u32 reg;

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

691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708
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);
}

709 710 711 712 713
static u64 rir_limit(u32 reg)
{
	return ((u64)GET_BITFIELD(reg,  1, 10) << 29) | 0x1fffffff;
}

714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732
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;
}

733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764
static enum mem_type haswell_get_memory_type(struct sbridge_pvt *pvt)
{
	u32 reg;
	bool registered = false;
	enum mem_type mtype = MEM_UNKNOWN;

	if (!pvt->pci_ddrio)
		goto out;

	pci_read_config_dword(pvt->pci_ddrio,
			      HASWELL_DDRCRCLKCONTROLS, &reg);
	/* Is_Rdimm */
	if (GET_BITFIELD(reg, 16, 16))
		registered = true;

	pci_read_config_dword(pvt->pci_ta, MCMTR, &reg);
	if (GET_BITFIELD(reg, 14, 14)) {
		if (registered)
			mtype = MEM_RDDR4;
		else
			mtype = MEM_DDR4;
	} else {
		if (registered)
			mtype = MEM_RDDR3;
		else
			mtype = MEM_DDR3;
	}

out:
	return mtype;
}

765 766 767 768 769 770 771
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);
}

772 773 774 775 776 777 778 779 780 781 782 783
static u8 haswell_get_node_id(struct sbridge_pvt *pvt)
{
	u32 reg;

	pci_read_config_dword(pvt->pci_sad1, SAD_CONTROL, &reg);
	return GET_BITFIELD(reg, 0, 3);
}

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

784 785
	pci_read_config_dword(pvt->info.pci_vtd, HASWELL_TOLM, &reg);
	return (GET_BITFIELD(reg, 26, 31) << 26) | 0x3ffffff;
786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805
}

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

	pci_read_config_dword(pvt->info.pci_vtd, HASWELL_TOHM_0, &reg);
	rc = GET_BITFIELD(reg, 26, 31);
	pci_read_config_dword(pvt->info.pci_vtd, HASWELL_TOHM_1, &reg);
	rc = ((reg << 6) | rc) << 26;

	return rc | 0x1ffffff;
}

static u64 haswell_rir_limit(u32 reg)
{
	return (((u64)GET_BITFIELD(reg,  1, 11) + 1) << 29) - 1;
}

806 807 808
static inline u8 sad_pkg_socket(u8 pkg)
{
	/* on Ivy Bridge, nodeID is SASS, where A is HA and S is node id */
809
	return ((pkg >> 3) << 2) | (pkg & 0x3);
810 811 812 813 814 815 816
}

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

817 818 819
/****************************************************************************
			Memory check routines
 ****************************************************************************/
820
static struct pci_dev *get_pdev_same_bus(u8 bus, u32 id)
821
{
822
	struct pci_dev *pdev = NULL;
823

824 825 826 827 828
	do {
		pdev = pci_get_device(PCI_VENDOR_ID_INTEL, id, pdev);
		if (pdev && pdev->bus->number == bus)
			break;
	} while (pdev);
829

830
	return pdev;
831 832 833
}

/**
834
 * check_if_ecc_is_active() - Checks if ECC is active
835 836 837 838
 * @bus:	Device bus
 * @type:	Memory controller type
 * returns: 0 in case ECC is active, -ENODEV if it can't be determined or
 *	    disabled
839
 */
840
static int check_if_ecc_is_active(const u8 bus, enum type type)
841 842
{
	struct pci_dev *pdev = NULL;
843
	u32 mcmtr, id;
844

845 846
	switch (type) {
	case IVY_BRIDGE:
847
		id = PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA;
848 849
		break;
	case HASWELL:
850
		id = PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TA;
851 852
		break;
	case SANDY_BRIDGE:
853
		id = PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TA;
854 855 856 857 858 859 860
		break;
	case BROADWELL:
		id = PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TA;
		break;
	default:
		return -ENODEV;
	}
861 862

	pdev = get_pdev_same_bus(bus, id);
863 864
	if (!pdev) {
		sbridge_printk(KERN_ERR, "Couldn't find PCI device "
865 866
					"%04x:%04x! on bus %02d\n",
					PCI_VENDOR_ID_INTEL, id, bus);
867 868 869 870 871 872 873 874 875 876 877
		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;
}

878
static int get_dimm_config(struct mem_ctl_info *mci)
879 880
{
	struct sbridge_pvt *pvt = mci->pvt_info;
881
	struct dimm_info *dimm;
882 883
	unsigned i, j, banks, ranks, rows, cols, npages;
	u64 size;
884 885
	u32 reg;
	enum edac_type mode;
886
	enum mem_type mtype;
887

888
	if (pvt->info.type == HASWELL || pvt->info.type == BROADWELL)
889 890 891 892
		pci_read_config_dword(pvt->pci_sad1, SAD_TARGET, &reg);
	else
		pci_read_config_dword(pvt->pci_br0, SAD_TARGET, &reg);

893 894
	pvt->sbridge_dev->source_id = SOURCE_ID(reg);

895
	pvt->sbridge_dev->node_id = pvt->info.get_node_id(pvt);
896 897 898 899
	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);
900 901 902

	pci_read_config_dword(pvt->pci_ras, RASENABLES, &reg);
	if (IS_MIRROR_ENABLED(reg)) {
903
		edac_dbg(0, "Memory mirror is enabled\n");
904 905
		pvt->is_mirrored = true;
	} else {
906
		edac_dbg(0, "Memory mirror is disabled\n");
907 908 909 910 911
		pvt->is_mirrored = false;
	}

	pci_read_config_dword(pvt->pci_ta, MCMTR, &pvt->info.mcmtr);
	if (IS_LOCKSTEP_ENABLED(pvt->info.mcmtr)) {
912
		edac_dbg(0, "Lockstep is enabled\n");
913 914 915
		mode = EDAC_S8ECD8ED;
		pvt->is_lockstep = true;
	} else {
916
		edac_dbg(0, "Lockstep is disabled\n");
917 918 919 920
		mode = EDAC_S4ECD4ED;
		pvt->is_lockstep = false;
	}
	if (IS_CLOSE_PG(pvt->info.mcmtr)) {
921
		edac_dbg(0, "address map is on closed page mode\n");
922 923
		pvt->is_close_pg = true;
	} else {
924
		edac_dbg(0, "address map is on open page mode\n");
925 926 927
		pvt->is_close_pg = false;
	}

928
	mtype = pvt->info.get_memory_type(pvt);
929
	if (mtype == MEM_RDDR3 || mtype == MEM_RDDR4)
930 931
		edac_dbg(0, "Memory is registered\n");
	else if (mtype == MEM_UNKNOWN)
932
		edac_dbg(0, "Cannot determine memory type\n");
933 934
	else
		edac_dbg(0, "Memory is unregistered\n");
935

936
	if (mtype == MEM_DDR4 || mtype == MEM_RDDR4)
937 938 939
		banks = 16;
	else
		banks = 8;
940 941 942 943

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

944 945
		if (!pvt->pci_tad[i])
			continue;
946
		for (j = 0; j < ARRAY_SIZE(mtr_regs); j++) {
947 948
			dimm = EDAC_DIMM_PTR(mci->layers, mci->dimms, mci->n_layers,
				       i, j, 0);
949 950
			pci_read_config_dword(pvt->pci_tad[i],
					      mtr_regs[j], &mtr);
951
			edac_dbg(4, "Channel #%d  MTR%d = %x\n", i, j, mtr);
952 953 954
			if (IS_DIMM_PRESENT(mtr)) {
				pvt->channel[i].dimms++;

955
				ranks = numrank(pvt->info.type, mtr);
956 957 958
				rows = numrow(mtr);
				cols = numcol(mtr);

959
				size = ((u64)rows * cols * banks * ranks) >> (20 - 3);
960 961
				npages = MiB_TO_PAGES(size);

962 963
				edac_dbg(0, "mc#%d: ha %d channel %d, dimm %d, %lld Mb (%d pages) bank: %d, rank: %d, row: %#x, col: %#x\n",
					 pvt->sbridge_dev->mc, i/4, i%4, j,
964 965
					 size, npages,
					 banks, ranks, rows, cols);
966

967
				dimm->nr_pages = npages;
968
				dimm->grain = 32;
969 970 971 972 973 974 975 976 977 978 979
				switch (banks) {
				case 16:
					dimm->dtype = DEV_X16;
					break;
				case 8:
					dimm->dtype = DEV_X8;
					break;
				case 4:
					dimm->dtype = DEV_X4;
					break;
				}
980 981 982
				dimm->mtype = mtype;
				dimm->edac_mode = mode;
				snprintf(dimm->label, sizeof(dimm->label),
983 984
					 "CPU_SrcID#%u_Ha#%u_Chan#%u_DIMM#%u",
					 pvt->sbridge_dev->source_id, i/4, i%4, j);
985 986 987 988 989 990 991 992 993 994 995 996 997 998
			}
		}
	}

	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;
999
	u32 gb, mb;
1000 1001 1002 1003 1004 1005
	u32 rir_way;

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

A
Aristeu Rozanski 已提交
1006
	pvt->tolm = pvt->info.get_tolm(pvt);
1007 1008
	tmp_mb = (1 + pvt->tolm) >> 20;

1009 1010 1011
	gb = div_u64_rem(tmp_mb, 1024, &mb);
	edac_dbg(0, "TOLM: %u.%03u GB (0x%016Lx)\n",
		gb, (mb*1000)/1024, (u64)pvt->tolm);
1012 1013

	/* Address range is already 45:25 */
A
Aristeu Rozanski 已提交
1014
	pvt->tohm = pvt->info.get_tohm(pvt);
1015 1016
	tmp_mb = (1 + pvt->tohm) >> 20;

1017 1018 1019
	gb = div_u64_rem(tmp_mb, 1024, &mb);
	edac_dbg(0, "TOHM: %u.%03u GB (0x%016Lx)\n",
		gb, (mb*1000)/1024, (u64)pvt->tohm);
1020 1021 1022 1023 1024 1025 1026 1027

	/*
	 * 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;
1028
	for (n_sads = 0; n_sads < pvt->info.max_sad; n_sads++) {
1029
		/* SAD_LIMIT Address range is 45:26 */
1030
		pci_read_config_dword(pvt->pci_sad0, pvt->info.dram_rule[n_sads],
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040
				      &reg);
		limit = SAD_LIMIT(reg);

		if (!DRAM_RULE_ENABLE(reg))
			continue;

		if (limit <= prv)
			break;

		tmp_mb = (limit + 1) >> 20;
1041
		gb = div_u64_rem(tmp_mb, 1024, &mb);
1042 1043 1044
		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),
1045
			 gb, (mb*1000)/1024,
1046 1047 1048
			 ((u64)tmp_mb) << 20L,
			 INTERLEAVE_MODE(reg) ? "8:6" : "[8:6]XOR[18:16]",
			 reg);
1049 1050
		prv = limit;

1051
		pci_read_config_dword(pvt->pci_sad0, pvt->info.interleave_list[n_sads],
1052
				      &reg);
A
Aristeu Rozanski 已提交
1053
		sad_interl = sad_pkg(pvt->info.interleave_pkg, reg, 0);
1054
		for (j = 0; j < 8; j++) {
A
Aristeu Rozanski 已提交
1055 1056
			u32 pkg = sad_pkg(pvt->info.interleave_pkg, reg, j);
			if (j > 0 && sad_interl == pkg)
1057 1058
				break;

1059
			edac_dbg(0, "SAD#%d, interleave #%d: %d\n",
A
Aristeu Rozanski 已提交
1060
				 n_sads, j, pkg);
1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075
		}
	}

	/*
	 * 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;

1076
		gb = div_u64_rem(tmp_mb, 1024, &mb);
1077
		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",
1078
			 n_tads, gb, (mb*1000)/1024,
1079 1080 1081 1082 1083 1084 1085 1086
			 ((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);
1087
		prv = limit;
1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100
	}

	/*
	 * 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;
1101
			gb = div_u64_rem(tmp_mb, 1024, &mb);
1102 1103
			edac_dbg(0, "TAD CH#%d, offset #%d: %u.%03u GB (0x%016Lx), reg=0x%08x\n",
				 i, j,
1104
				 gb, (mb*1000)/1024,
1105 1106
				 ((u64)tmp_mb) << 20L,
				 reg);
1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123
		}
	}

	/*
	 * 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;

1124
			tmp_mb = pvt->info.rir_limit(reg) >> 20;
1125
			rir_way = 1 << RIR_WAY(reg);
1126
			gb = div_u64_rem(tmp_mb, 1024, &mb);
1127 1128
			edac_dbg(0, "CH#%d RIR#%d, limit: %u.%03u GB (0x%016Lx), way: %d, reg=0x%08x\n",
				 i, j,
1129
				 gb, (mb*1000)/1024,
1130 1131 1132
				 ((u64)tmp_mb) << 20L,
				 rir_way,
				 reg);
1133 1134 1135 1136 1137 1138 1139

			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;

1140
				gb = div_u64_rem(tmp_mb, 1024, &mb);
1141 1142
				edac_dbg(0, "CH#%d RIR#%d INTL#%d, offset %u.%03u GB (0x%016Lx), tgt: %d, reg=0x%08x\n",
					 i, j, k,
1143
					 gb, (mb*1000)/1024,
1144 1145 1146
					 ((u64)tmp_mb) << 20L,
					 (u32)RIR_RNK_TGT(reg),
					 reg);
1147 1148 1149 1150 1151
			}
		}
	}
}

1152
static struct mem_ctl_info *get_mci_for_node_id(u8 node_id)
1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164
{
	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,
1165
				 u8 *socket, u8 *ha,
1166 1167
				 long *channel_mask,
				 u8 *rank,
1168
				 char **area_type, char *msg)
1169 1170 1171
{
	struct mem_ctl_info	*new_mci;
	struct sbridge_pvt *pvt = mci->pvt_info;
1172
	struct pci_dev		*pci_ha;
1173
	int			n_rir, n_sads, n_tads, sad_way, sck_xch;
1174
	int			sad_interl, idx, base_ch;
1175
	int			interleave_mode, shiftup = 0;
1176
	unsigned		sad_interleave[pvt->info.max_interleave];
1177
	u32			reg, dram_rule;
1178
	u8			ch_way, sck_way, pkg, sad_ha = 0, ch_add = 0;
1179 1180
	u32			tad_offset;
	u32			rir_way;
1181
	u32			mb, gb;
1182
	u64			ch_addr, offset, limit = 0, prv = 0;
1183 1184 1185 1186 1187 1188 1189 1190 1191


	/*
	 * 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.
	 */
1192
	if ((addr > (u64) pvt->tolm) && (addr < (1LL << 32))) {
1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203
		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
	 */
1204 1205
	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],
1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219
				      &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;
	}
1220
	if (n_sads == pvt->info.max_sad) {
1221 1222 1223
		sprintf(msg, "Can't discover the memory socket");
		return -EINVAL;
	}
1224 1225 1226
	dram_rule = reg;
	*area_type = get_dram_attr(dram_rule);
	interleave_mode = INTERLEAVE_MODE(dram_rule);
1227

1228
	pci_read_config_dword(pvt->pci_sad0, pvt->info.interleave_list[n_sads],
1229
			      &reg);
1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254

	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;
1255
			break;
1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
		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);
1271
	} else if (pvt->info.type == HASWELL || pvt->info.type == BROADWELL) {
1272 1273 1274 1275 1276 1277 1278
		int bits, a7mode = A7MODE(dram_rule);

		if (a7mode) {
			/* A7 mode swaps P9 with P6 */
			bits = GET_BITFIELD(addr, 7, 8) << 1;
			bits |= GET_BITFIELD(addr, 9, 9);
		} else
1279
			bits = GET_BITFIELD(addr, 6, 8);
1280

1281
		if (interleave_mode == 0) {
1282 1283 1284 1285 1286 1287 1288 1289 1290
			/* interleave mode will XOR {8,7,6} with {18,17,16} */
			idx = GET_BITFIELD(addr, 16, 18);
			idx ^= bits;
		} else
			idx = bits;

		pkg = sad_pkg(pvt->info.interleave_pkg, reg, idx);
		*socket = sad_pkg_socket(pkg);
		sad_ha = sad_pkg_ha(pkg);
1291 1292
		if (sad_ha)
			ch_add = 4;
1293 1294 1295 1296 1297 1298 1299 1300 1301 1302

		if (a7mode) {
			/* MCChanShiftUpEnable */
			pci_read_config_dword(pvt->pci_ha0,
					      HASWELL_HASYSDEFEATURE2, &reg);
			shiftup = GET_BITFIELD(reg, 22, 22);
		}

		edac_dbg(0, "SAD interleave package: %d = CPU socket %d, HA %i, shiftup: %i\n",
			 idx, *socket, sad_ha, shiftup);
1303 1304
	} else {
		/* Ivy Bridge's SAD mode doesn't support XOR interleave mode */
1305
		idx = (addr >> 6) & 7;
1306 1307 1308
		pkg = sad_pkg(pvt->info.interleave_pkg, reg, idx);
		*socket = sad_pkg_socket(pkg);
		sad_ha = sad_pkg_ha(pkg);
1309 1310
		if (sad_ha)
			ch_add = 4;
1311 1312
		edac_dbg(0, "SAD interleave package: %d = CPU socket %d, HA %d\n",
			 idx, *socket, sad_ha);
1313 1314
	}

1315 1316
	*ha = sad_ha;

1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333
	/*
	 * 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;
1334 1335 1336 1337 1338 1339 1340 1341
	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;
	}
1342
	for (n_tads = 0; n_tads < MAX_TAD; n_tads++) {
1343
		pci_read_config_dword(pci_ha, tad_dram_rule[n_tads], &reg);
1344 1345 1346 1347 1348 1349 1350 1351 1352
		limit = TAD_LIMIT(reg);
		if (limit <= prv) {
			sprintf(msg, "Can't discover the memory channel");
			return -EINVAL;
		}
		if  (addr <= limit)
			break;
		prv = limit;
	}
1353 1354 1355 1356 1357
	if (n_tads == MAX_TAD) {
		sprintf(msg, "Can't discover the memory channel");
		return -EINVAL;
	}

1358 1359 1360 1361 1362 1363
	ch_way = TAD_CH(reg) + 1;
	sck_way = TAD_SOCK(reg) + 1;

	if (ch_way == 3)
		idx = addr >> 6;
	else
1364
		idx = (addr >> (6 + sck_way + shiftup)) & 0x3;
1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388
	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;

1389
	pci_read_config_dword(pvt->pci_tad[ch_add + base_ch],
1390 1391 1392
				tad_ch_nilv_offset[n_tads],
				&tad_offset);

1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411
	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);

1412 1413 1414 1415 1416 1417 1418 1419 1420 1421
	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);
1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435

	/* 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;
1436
	addr = div_u64(addr, sck_xch);
1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447
#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++) {
1448
		pci_read_config_dword(pvt->pci_tad[ch_add + base_ch],
1449 1450 1451 1452 1453 1454
				      rir_way_limit[n_rir],
				      &reg);

		if (!IS_RIR_VALID(reg))
			continue;

1455
		limit = pvt->info.rir_limit(reg);
1456
		gb = div_u64_rem(limit >> 20, 1024, &mb);
1457 1458
		edac_dbg(0, "RIR#%d, limit: %u.%03u GB (0x%016Lx), way: %d\n",
			 n_rir,
1459
			 gb, (mb*1000)/1024,
1460 1461
			 limit,
			 1 << RIR_WAY(reg));
1462 1463 1464 1465 1466 1467 1468 1469 1470
		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);
1471

1472 1473 1474 1475 1476 1477
	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;

1478
	pci_read_config_dword(pvt->pci_tad[ch_add + base_ch],
1479 1480 1481 1482
			      rir_offset[n_rir][idx],
			      &reg);
	*rank = RIR_RNK_TGT(reg);

1483 1484 1485 1486 1487 1488
	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);
1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504

	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;

1505
	edac_dbg(0, "\n");
1506 1507 1508 1509
	for (i = 0; i < sbridge_dev->n_devs; i++) {
		struct pci_dev *pdev = sbridge_dev->pdev[i];
		if (!pdev)
			continue;
1510 1511 1512
		edac_dbg(0, "Removing dev %02x:%02x.%d\n",
			 pdev->bus->number,
			 PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536
		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;

1537
	sbridge_printk(KERN_DEBUG,
1538
		"Seeking for: PCI ID %04x:%04x\n",
1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552
		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;

1553
		/* if the HA wasn't found */
1554 1555 1556 1557
		if (devno == 0)
			return -ENODEV;

		sbridge_printk(KERN_INFO,
1558
			"Device not found: %04x:%04x\n",
1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577
			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,
1578
			"Duplicated device for %04x:%04x\n",
1579 1580 1581 1582 1583 1584 1585 1586 1587 1588
			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,
1589
			"Couldn't enable %04x:%04x\n",
1590 1591 1592 1593
			PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
		return -ENODEV;
	}

1594
	edac_dbg(0, "Detected %04x:%04x\n",
1595
		 PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608

	/*
	 * 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;
}

1609 1610
/*
 * sbridge_get_all_devices - Find and perform 'get' operation on the MCH's
1611
 *			     devices we want to reference for this driver.
1612
 * @num_mc: pointer to the memory controllers count, to be incremented in case
1613
 *	    of success.
1614 1615 1616 1617 1618 1619
 * @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)
1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645
{
	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
1646 1647
static int sbridge_mci_bind_devs(struct mem_ctl_info *mci,
				 struct sbridge_dev *sbridge_dev)
1648 1649 1650
{
	struct sbridge_pvt *pvt = mci->pvt_info;
	struct pci_dev *pdev;
1651
	int i;
1652 1653 1654 1655 1656

	for (i = 0; i < sbridge_dev->n_devs; i++) {
		pdev = sbridge_dev->pdev[i];
		if (!pdev)
			continue;
1657 1658 1659 1660

		switch (pdev->device) {
		case PCI_DEVICE_ID_INTEL_SBRIDGE_SAD0:
			pvt->pci_sad0 = pdev;
1661
			break;
1662 1663
		case PCI_DEVICE_ID_INTEL_SBRIDGE_SAD1:
			pvt->pci_sad1 = pdev;
1664
			break;
1665 1666
		case PCI_DEVICE_ID_INTEL_SBRIDGE_BR:
			pvt->pci_br0 = pdev;
1667
			break;
1668 1669
		case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0:
			pvt->pci_ha0 = pdev;
1670
			break;
1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687
		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;
1688 1689 1690 1691 1692
			break;
		default:
			goto error;
		}

1693 1694
		edac_dbg(0, "Associated PCI %02x:%02x, bus %d with dev = %p\n",
			 pdev->vendor, pdev->device,
1695 1696
			 sbridge_dev->bus,
			 pdev);
1697 1698 1699 1700
	}

	/* Check if everything were registered */
	if (!pvt->pci_sad0 || !pvt->pci_sad1 || !pvt->pci_ha0 ||
1701
	    !pvt-> pci_tad || !pvt->pci_ras  || !pvt->pci_ta)
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714
		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:
1715 1716
	sbridge_printk(KERN_ERR, "Unexpected device %02x:%02x\n",
		       PCI_VENDOR_ID_INTEL, pdev->device);
1717 1718 1719
	return -EINVAL;
}

1720 1721 1722 1723
static int ibridge_mci_bind_devs(struct mem_ctl_info *mci,
				 struct sbridge_dev *sbridge_dev)
{
	struct sbridge_pvt *pvt = mci->pvt_info;
1724 1725
	struct pci_dev *pdev;
	u8 saw_chan_mask = 0;
1726
	int i;
1727 1728 1729 1730 1731 1732

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

1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743
		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_TAD0:
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD1:
1744 1745
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD2:
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD3:
1746 1747 1748
		{
			int id = pdev->device - PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD0;
			pvt->pci_tad[id] = pdev;
1749
			saw_chan_mask |= 1 << id;
1750
		}
1751
			break;
1752 1753 1754 1755
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_2HA_DDRIO0:
			pvt->pci_ddrio = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_1HA_DDRIO0:
1756
			pvt->pci_ddrio = pdev;
1757
			break;
1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771
		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:
1772 1773
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD2:
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD3:
1774
		{
1775
			int id = pdev->device - PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD0 + 4;
1776
			pvt->pci_tad[id] = pdev;
1777
			saw_chan_mask |= 1 << id;
1778 1779
		}
			break;
1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795
		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;

1796 1797 1798 1799
	if (saw_chan_mask != 0x0f && /* -EN */
	    saw_chan_mask != 0x33 && /* -EP */
	    saw_chan_mask != 0xff)   /* -EX */
		goto enodev;
1800 1801 1802 1803 1804 1805 1806 1807
	return 0;

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

error:
	sbridge_printk(KERN_ERR,
1808 1809
		       "Unexpected device %02x:%02x\n", PCI_VENDOR_ID_INTEL,
			pdev->device);
1810 1811 1812
	return -EINVAL;
}

1813 1814 1815 1816
static int haswell_mci_bind_devs(struct mem_ctl_info *mci,
				 struct sbridge_dev *sbridge_dev)
{
	struct sbridge_pvt *pvt = mci->pvt_info;
1817 1818
	struct pci_dev *pdev;
	u8 saw_chan_mask = 0;
1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
	int i;

	/* there's only one device per system; not tied to any bus */
	if (pvt->info.pci_vtd == NULL)
		/* result will be checked later */
		pvt->info.pci_vtd = pci_get_device(PCI_VENDOR_ID_INTEL,
						   PCI_DEVICE_ID_INTEL_HASWELL_IMC_VTD_MISC,
						   NULL);

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

		switch (pdev->device) {
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD0:
			pvt->pci_sad0 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_CBO_SAD1:
			pvt->pci_sad1 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0:
			pvt->pci_ha0 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TA:
			pvt->pci_ta = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_THERMAL:
			pvt->pci_ras = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD0:
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD1:
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD2:
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD3:
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869
		{
			int id = pdev->device - PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TAD0;

			pvt->pci_tad[id] = pdev;
			saw_chan_mask |= 1 << id;
		}
			break;
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD0:
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD1:
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD2:
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD3:
		{
			int id = pdev->device - PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TAD0 + 4;

			pvt->pci_tad[id] = pdev;
			saw_chan_mask |= 1 << id;
		}
1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894
			break;
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO0:
			pvt->pci_ddrio = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1:
			pvt->pci_ha1 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA1_TA:
			pvt->pci_ha1_ta = pdev;
			break;
		default:
			break;
		}

		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_sad1 ||
	    !pvt->pci_ras  || !pvt->pci_ta || !pvt->info.pci_vtd)
		goto enodev;

1895 1896 1897 1898
	if (saw_chan_mask != 0x0f && /* -EN */
	    saw_chan_mask != 0x33 && /* -EP */
	    saw_chan_mask != 0xff)   /* -EX */
		goto enodev;
1899 1900 1901 1902 1903 1904 1905
	return 0;

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

1906 1907 1908 1909 1910
static int broadwell_mci_bind_devs(struct mem_ctl_info *mci,
				 struct sbridge_dev *sbridge_dev)
{
	struct sbridge_pvt *pvt = mci->pvt_info;
	struct pci_dev *pdev;
1911
	u8 saw_chan_mask = 0;
1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945
	int i;

	/* there's only one device per system; not tied to any bus */
	if (pvt->info.pci_vtd == NULL)
		/* result will be checked later */
		pvt->info.pci_vtd = pci_get_device(PCI_VENDOR_ID_INTEL,
						   PCI_DEVICE_ID_INTEL_BROADWELL_IMC_VTD_MISC,
						   NULL);

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

		switch (pdev->device) {
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD0:
			pvt->pci_sad0 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_CBO_SAD1:
			pvt->pci_sad1 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0:
			pvt->pci_ha0 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TA:
			pvt->pci_ta = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_THERMAL:
			pvt->pci_ras = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD0:
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD1:
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD2:
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD3:
1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960
		{
			int id = pdev->device - PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD0;
			pvt->pci_tad[id] = pdev;
			saw_chan_mask |= 1 << id;
		}
			break;
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD0:
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD1:
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD2:
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD3:
		{
			int id = pdev->device - PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TAD0 + 4;
			pvt->pci_tad[id] = pdev;
			saw_chan_mask |= 1 << id;
		}
1961 1962 1963 1964
			break;
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_DDRIO0:
			pvt->pci_ddrio = pdev;
			break;
1965 1966 1967 1968 1969 1970
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1:
			pvt->pci_ha1 = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TA:
			pvt->pci_ha1_ta = pdev;
			break;
1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985
		default:
			break;
		}

		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_sad1 ||
	    !pvt->pci_ras  || !pvt->pci_ta || !pvt->info.pci_vtd)
		goto enodev;

1986 1987 1988 1989
	if (saw_chan_mask != 0x0f && /* -EN */
	    saw_chan_mask != 0x33 && /* -EP */
	    saw_chan_mask != 0xff)   /* -EX */
		goto enodev;
1990 1991 1992 1993 1994 1995 1996
	return 0;

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

1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
/****************************************************************************
			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;
2012
	enum hw_event_mc_err_type tp_event;
2013
	char *type, *optype, msg[256];
2014 2015 2016
	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);
2017
	bool recoverable;
2018 2019 2020 2021 2022 2023
	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;
2024
	u8  rank, socket, ha;
2025
	int rc, dimm;
2026
	char *area_type = NULL;
2027

2028
	if (pvt->info.type != SANDY_BRIDGE)
2029 2030 2031 2032
		recoverable = true;
	else
		recoverable = GET_BITFIELD(m->status, 56, 56);

2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044
	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;
	}
2045 2046

	/*
D
David Mackey 已提交
2047
	 * According with Table 15-9 of the Intel Architecture spec vol 3A,
2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061
	 * 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:
2062
			optype = "generic undef request error";
2063 2064
			break;
		case 1:
2065
			optype = "memory read error";
2066 2067
			break;
		case 2:
2068
			optype = "memory write error";
2069 2070
			break;
		case 3:
2071
			optype = "addr/cmd error";
2072 2073
			break;
		case 4:
2074
			optype = "memory scrubbing error";
2075 2076 2077 2078 2079 2080 2081
			break;
		default:
			optype = "reserved";
			break;
		}
	}

2082 2083 2084 2085
	/* Only decode errors with an valid address (ADDRV) */
	if (!GET_BITFIELD(m->status, 58, 58))
		return;

2086
	rc = get_memory_error_data(mci, m->addr, &socket, &ha,
2087
				   &channel_mask, &rank, &area_type, msg);
2088
	if (rc < 0)
2089
		goto err_parsing;
2090 2091
	new_mci = get_mci_for_node_id(socket);
	if (!new_mci) {
2092 2093
		strcpy(msg, "Error: socket got corrupted!");
		goto err_parsing;
2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108
	}
	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;


	/*
2109 2110 2111 2112
	 * 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.
2113
	 */
2114 2115 2116
	if (!pvt->is_lockstep && !pvt->is_mirrored && !pvt->is_close_pg)
		channel = first_channel;

2117
	snprintf(msg, sizeof(msg),
2118
		 "%s%s area:%s err_code:%04x:%04x socket:%d ha:%d channel_mask:%ld rank:%d",
2119 2120 2121 2122
		 overflow ? " OVERFLOW" : "",
		 (uncorrected_error && recoverable) ? " recoverable" : "",
		 area_type,
		 mscod, errcode,
2123
		 socket, ha,
2124 2125
		 channel_mask,
		 rank);
2126

2127
	edac_dbg(0, "%s\n", msg);
2128

2129 2130
	/* FIXME: need support for channel mask */

2131 2132 2133
	if (channel == CHANNEL_UNSPECIFIED)
		channel = -1;

2134
	/* Call the helper to output message */
2135
	edac_mc_handle_error(tp_event, mci, core_err_cnt,
2136
			     m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0,
2137
			     4*ha+channel, dimm, -1,
2138
			     optype, msg);
2139 2140
	return;
err_parsing:
2141
	edac_mc_handle_error(tp_event, mci, core_err_cnt, 0, 0, 0,
2142
			     -1, -1, -1,
2143
			     msg, "");
2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205

}

/*
 *	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.
 */
2206 2207
static int sbridge_mce_check_error(struct notifier_block *nb, unsigned long val,
				   void *data)
2208
{
2209 2210 2211
	struct mce *mce = (struct mce *)data;
	struct mem_ctl_info *mci;
	struct sbridge_pvt *pvt;
2212
	char *type;
2213

2214 2215 2216
	if (get_edac_report_status() == EDAC_REPORTING_DISABLED)
		return NOTIFY_DONE;

2217 2218 2219 2220
	mci = get_mci_for_node_id(mce->socketid);
	if (!mci)
		return NOTIFY_BAD;
	pvt = mci->pvt_info;
2221 2222 2223 2224 2225 2226 2227 2228

	/*
	 * 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)
2229
		return NOTIFY_DONE;
2230

2231 2232 2233 2234 2235
	if (mce->mcgstatus & MCG_STATUS_MCIP)
		type = "Exception";
	else
		type = "Event";

2236
	sbridge_mc_printk(mci, KERN_DEBUG, "HANDLING MCE MEMORY ERROR\n");
2237

2238 2239 2240 2241 2242 2243
	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);
2244

2245 2246 2247
	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);
2248 2249 2250 2251 2252

	smp_rmb();
	if ((pvt->mce_out + 1) % MCE_LOG_LEN == pvt->mce_in) {
		smp_wmb();
		pvt->mce_overrun++;
2253
		return NOTIFY_DONE;
2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265
	}

	/* 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 */
2266
	return NOTIFY_STOP;
2267 2268
}

2269 2270 2271 2272
static struct notifier_block sbridge_mce_dec = {
	.notifier_call      = sbridge_mce_check_error,
};

2273 2274 2275 2276 2277 2278 2279 2280 2281 2282
/****************************************************************************
			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)) {
2283
		edac_dbg(0, "MC: dev = %p\n", &sbridge_dev->pdev[0]->dev);
2284 2285 2286 2287 2288 2289 2290

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

	pvt = mci->pvt_info;

2291 2292
	edac_dbg(0, "MC: mci = %p, dev = %p\n",
		 mci, &sbridge_dev->pdev[0]->dev);
2293 2294

	/* Remove MC sysfs nodes */
2295
	edac_mc_del_mc(mci->pdev);
2296

2297
	edac_dbg(1, "%s: free mci struct\n", mci->ctl_name);
2298 2299 2300 2301 2302
	kfree(mci->ctl_name);
	edac_mc_free(mci);
	sbridge_dev->mci = NULL;
}

2303
static int sbridge_register_mci(struct sbridge_dev *sbridge_dev, enum type type)
2304 2305
{
	struct mem_ctl_info *mci;
2306
	struct edac_mc_layer layers[2];
2307
	struct sbridge_pvt *pvt;
2308
	struct pci_dev *pdev = sbridge_dev->pdev[0];
2309
	int rc;
2310 2311

	/* Check the number of active and not disabled channels */
2312
	rc = check_if_ecc_is_active(sbridge_dev->bus, type);
2313 2314 2315 2316
	if (unlikely(rc < 0))
		return rc;

	/* allocate a new MC control structure */
2317 2318 2319 2320 2321 2322
	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;
2323
	mci = edac_mc_alloc(sbridge_dev->mc, ARRAY_SIZE(layers), layers,
2324 2325
			    sizeof(*pvt));

2326 2327 2328
	if (unlikely(!mci))
		return -ENOMEM;

2329
	edac_dbg(0, "MC: mci = %p, dev = %p\n",
2330
		 mci, &pdev->dev);
2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343

	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;
2344
	mci->dev_name = pci_name(pdev);
2345 2346 2347 2348 2349
	mci->ctl_page_to_phys = NULL;

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

2350
	pvt->info.type = type;
2351 2352
	switch (type) {
	case IVY_BRIDGE:
2353 2354 2355 2356
		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;
2357
		pvt->info.get_memory_type = get_memory_type;
2358
		pvt->info.get_node_id = get_node_id;
2359
		pvt->info.rir_limit = rir_limit;
2360 2361 2362 2363 2364 2365 2366 2367 2368 2369
		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;
2370 2371
		break;
	case SANDY_BRIDGE:
2372 2373 2374 2375
		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;
2376
		pvt->info.get_memory_type = get_memory_type;
2377
		pvt->info.get_node_id = get_node_id;
2378
		pvt->info.rir_limit = rir_limit;
2379 2380 2381 2382 2383 2384 2385 2386 2387 2388
		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;
2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402
		break;
	case HASWELL:
		/* rankcfgr isn't used */
		pvt->info.get_tolm = haswell_get_tolm;
		pvt->info.get_tohm = haswell_get_tohm;
		pvt->info.dram_rule = ibridge_dram_rule;
		pvt->info.get_memory_type = haswell_get_memory_type;
		pvt->info.get_node_id = haswell_get_node_id;
		pvt->info.rir_limit = haswell_rir_limit;
		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, "Haswell Socket#%d", mci->mc_idx);
2403

2404 2405 2406 2407 2408
		/* Store pci devices at mci for faster access */
		rc = haswell_mci_bind_devs(mci, sbridge_dev);
		if (unlikely(rc < 0))
			goto fail0;
		break;
2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427
	case BROADWELL:
		/* rankcfgr isn't used */
		pvt->info.get_tolm = haswell_get_tolm;
		pvt->info.get_tohm = haswell_get_tohm;
		pvt->info.dram_rule = ibridge_dram_rule;
		pvt->info.get_memory_type = haswell_get_memory_type;
		pvt->info.get_node_id = haswell_get_node_id;
		pvt->info.rir_limit = haswell_rir_limit;
		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, "Broadwell Socket#%d", mci->mc_idx);

		/* Store pci devices at mci for faster access */
		rc = broadwell_mci_bind_devs(mci, sbridge_dev);
		if (unlikely(rc < 0))
			goto fail0;
		break;
2428
	}
2429 2430 2431 2432 2433 2434

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

	/* record ptr to the generic device */
2435
	mci->pdev = &pdev->dev;
2436 2437 2438

	/* add this new MC control structure to EDAC's list of MCs */
	if (unlikely(edac_mc_add_mc(mci))) {
2439
		edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460
		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
 */

2461
static int sbridge_probe(struct pci_dev *pdev, const struct pci_device_id *id)
2462
{
2463
	int rc = -ENODEV;
2464 2465
	u8 mc, num_mc = 0;
	struct sbridge_dev *sbridge_dev;
2466
	enum type type = SANDY_BRIDGE;
2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479

	/* 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++;

2480 2481
	switch (pdev->device) {
	case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA:
2482 2483
		rc = sbridge_get_all_devices(&num_mc, pci_dev_descr_ibridge_table);
		type = IVY_BRIDGE;
2484
		break;
2485
	case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0:
2486 2487
		rc = sbridge_get_all_devices(&num_mc, pci_dev_descr_sbridge_table);
		type = SANDY_BRIDGE;
2488 2489 2490 2491 2492
		break;
	case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0:
		rc = sbridge_get_all_devices(&num_mc, pci_dev_descr_haswell_table);
		type = HASWELL;
		break;
2493 2494 2495 2496
	case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0:
		rc = sbridge_get_all_devices(&num_mc, pci_dev_descr_broadwell_table);
		type = BROADWELL;
		break;
2497
	}
2498 2499
	if (unlikely(rc < 0)) {
		edac_dbg(0, "couldn't get all devices for 0x%x\n", pdev->device);
2500
		goto fail0;
2501 2502
	}

2503 2504 2505
	mc = 0;

	list_for_each_entry(sbridge_dev, &sbridge_edac_list, list) {
2506 2507
		edac_dbg(0, "Registering MC#%d (%d of %d)\n",
			 mc, mc + 1, num_mc);
2508

2509
		sbridge_dev->mc = mc++;
2510
		rc = sbridge_register_mci(sbridge_dev, type);
2511 2512 2513 2514
		if (unlikely(rc < 0))
			goto fail1;
	}

2515
	sbridge_printk(KERN_INFO, "%s\n", SBRIDGE_REVISION);
2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533

	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
 *
 */
2534
static void sbridge_remove(struct pci_dev *pdev)
2535 2536 2537
{
	struct sbridge_dev *sbridge_dev;

2538
	edac_dbg(0, "\n");
2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574

	/*
	 * 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,
2575
	.remove   = sbridge_remove,
2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586
	.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;

2587
	edac_dbg(2, "\n");
2588 2589 2590 2591 2592

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

	pci_rc = pci_register_driver(&sbridge_driver);
2593 2594
	if (pci_rc >= 0) {
		mce_register_decode_chain(&sbridge_mce_dec);
2595 2596
		if (get_edac_report_status() == EDAC_REPORTING_DISABLED)
			sbridge_printk(KERN_WARNING, "Loading driver, error reporting disabled.\n");
2597
		return 0;
2598
	}
2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611

	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)
{
2612
	edac_dbg(2, "\n");
2613
	pci_unregister_driver(&sbridge_driver);
2614
	mce_unregister_decode_chain(&sbridge_mce_dec);
2615 2616 2617 2618 2619 2620 2621 2622 2623
}

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");
2624
MODULE_AUTHOR("Mauro Carvalho Chehab");
2625
MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)");
2626
MODULE_DESCRIPTION("MC Driver for Intel Sandy Bridge and Ivy Bridge memory controllers - "
2627
		   SBRIDGE_REVISION);