sb_edac.c 94.2 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 72 73 74 75
static const u32 knl_dram_rule[] = {
	0x60, 0x68, 0x70, 0x78, 0x80, /* 0-4 */
	0x88, 0x90, 0x98, 0xa0, 0xa8, /* 5-9 */
	0xb0, 0xb8, 0xc0, 0xc8, 0xd0, /* 10-14 */
	0xd8, 0xe0, 0xe8, 0xf0, 0xf8, /* 15-19 */
	0x100, 0x108, 0x110, 0x118,   /* 20-23 */
};

76
#define DRAM_RULE_ENABLE(reg)	GET_BITFIELD(reg, 0,  0)
77
#define A7MODE(reg)		GET_BITFIELD(reg, 26, 26)
78

79
static char *show_dram_attr(u32 attr)
80
{
81
	switch (attr) {
82 83 84 85 86 87 88 89 90 91 92
		case 0:
			return "DRAM";
		case 1:
			return "MMCFG";
		case 2:
			return "NXM";
		default:
			return "unknown";
	}
}

93
static const u32 sbridge_interleave_list[] = {
94 95 96 97
	0x84, 0x8c, 0x94, 0x9c, 0xa4,
	0xac, 0xb4, 0xbc, 0xc4, 0xcc,
};

98 99 100 101 102 103 104
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,
};

105 106 107 108 109 110 111 112
static const u32 knl_interleave_list[] = {
	0x64, 0x6c, 0x74, 0x7c, 0x84, /* 0-4 */
	0x8c, 0x94, 0x9c, 0xa4, 0xac, /* 5-9 */
	0xb4, 0xbc, 0xc4, 0xcc, 0xd4, /* 10-14 */
	0xdc, 0xe4, 0xec, 0xf4, 0xfc, /* 15-19 */
	0x104, 0x10c, 0x114, 0x11c,   /* 20-23 */
};

A
Aristeu Rozanski 已提交
113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128
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 },
};

129 130 131 132 133 134 135 136 137 138 139
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
140 141
static inline int sad_pkg(const struct interleave_pkg *table, u32 reg,
			  int interleave)
142
{
A
Aristeu Rozanski 已提交
143 144
	return GET_BITFIELD(reg, table[interleave].start,
			    table[interleave].end);
145 146 147 148 149
}

/* Devices 12 Function 7 */

#define TOLM		0x80
150
#define TOHM		0x84
151
#define HASWELL_TOLM	0xd0
152 153
#define HASWELL_TOHM_0	0xd4
#define HASWELL_TOHM_1	0xd8
154 155 156
#define KNL_TOLM	0xd0
#define KNL_TOHM_0	0xd4
#define KNL_TOHM_1	0xd8
157 158 159 160 161 162 163 164 165 166

#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)

167 168
#define SOURCE_ID_KNL(reg)	GET_BITFIELD(reg, 12, 14)

169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190
#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
191
#define KNL_MCMTR		0x624
192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207

#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,
};

208 209
static const int knl_mtr_reg = 0xb60;

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
#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 */

270
#define SB_RANK_CFG_A		0x0328
271

272
#define IB_RANK_CFG_A		0x0320
273 274 275 276 277

/*
 * sbridge structs
 */

278
#define NUM_CHANNELS		8	/* 2MC per socket, four chan per MC */
279
#define MAX_DIMMS		3	/* Max DIMMS per channel */
280 281 282
#define KNL_MAX_CHAS		38	/* KNL max num. of Cache Home Agents */
#define KNL_MAX_CHANNELS	6	/* KNL max num. of PCI channels */
#define KNL_MAX_EDCS		8	/* Embedded DRAM controllers */
283
#define CHANNEL_UNSPECIFIED	0xf	/* Intel IA32 SDM 15-14 */
284

285 286 287
enum type {
	SANDY_BRIDGE,
	IVY_BRIDGE,
288
	HASWELL,
289
	BROADWELL,
290
	KNIGHTS_LANDING,
291 292
};

A
Aristeu Rozanski 已提交
293
struct sbridge_pvt;
294
struct sbridge_info {
295
	enum type	type;
296 297 298 299
	u32		mcmtr;
	u32		rankcfgr;
	u64		(*get_tolm)(struct sbridge_pvt *pvt);
	u64		(*get_tohm)(struct sbridge_pvt *pvt);
300
	u64		(*rir_limit)(u32 reg);
301 302 303 304
	u64		(*sad_limit)(u32 reg);
	u32		(*interleave_mode)(u32 reg);
	char*		(*show_interleave_mode)(u32 reg);
	u32		(*dram_attr)(u32 reg);
305
	const u32	*dram_rule;
306
	const u32	*interleave_list;
A
Aristeu Rozanski 已提交
307
	const struct interleave_pkg *interleave_pkg;
308
	u8		max_sad;
309
	u8		max_interleave;
310
	u8		(*get_node_id)(struct sbridge_pvt *pvt);
311
	enum mem_type	(*get_memory_type)(struct sbridge_pvt *pvt);
312
	enum dev_type	(*get_width)(struct sbridge_pvt *pvt, u32 mtr);
313
	struct pci_dev	*pci_vtd;
314 315 316 317 318 319 320 321
};

struct sbridge_channel {
	u32		ranks;
	u32		dimms;
};

struct pci_id_descr {
322
	int			dev_id;
323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339
	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;
};

340 341 342 343 344 345 346 347 348 349
struct knl_pvt {
	struct pci_dev          *pci_cha[KNL_MAX_CHAS];
	struct pci_dev          *pci_channel[KNL_MAX_CHANNELS];
	struct pci_dev          *pci_mc0;
	struct pci_dev          *pci_mc1;
	struct pci_dev          *pci_mc0_misc;
	struct pci_dev          *pci_mc1_misc;
	struct pci_dev          *pci_mc_info; /* tolm, tohm */
};

350 351
struct sbridge_pvt {
	struct pci_dev		*pci_ta, *pci_ddrio, *pci_ras;
352 353 354
	struct pci_dev		*pci_sad0, *pci_sad1;
	struct pci_dev		*pci_ha0, *pci_ha1;
	struct pci_dev		*pci_br0, *pci_br1;
355
	struct pci_dev		*pci_ha1_ta;
356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
	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;
378
	struct knl_pvt knl;
379 380
};

381 382
#define PCI_DESCR(device_id, opt)	\
	.dev_id = (device_id),		\
383
	.optional = opt
384 385 386

static const struct pci_id_descr pci_dev_descr_sbridge[] = {
		/* Processor Home Agent */
387
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0, 0)	},
388 389

		/* Memory controller */
390 391 392 393 394 395 396
	{ 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)	},
397 398

		/* System Address Decoder */
399 400
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_SAD0, 0)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_SAD1, 0)	},
401 402

		/* Broadcast Registers */
403
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_SBRIDGE_BR, 0)		},
404 405 406 407 408 409 410 411
};

#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. */
};

412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436
/* 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
437 438
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD2	0x0e6c
#define PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD3	0x0e6d
439 440 441

static const struct pci_id_descr pci_dev_descr_ibridge[] = {
		/* Processor Home Agent */
442
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0, 0)		},
443 444

		/* Memory controller */
445 446 447 448 449 450
	{ 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)	},
451 452

		/* System Address Decoder */
453
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_SAD, 0)			},
454 455

		/* Broadcast Registers */
456 457
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_BR0, 1)			},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_BR1, 0)			},
458 459

		/* Optional, mode 2HA */
460
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1, 1)		},
461
#if 0
462 463
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TA, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_RAS, 1)	},
464
#endif
465 466
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD0, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD1, 1)	},
467 468
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD2, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD3, 1)	},
469

470 471
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_1HA_DDRIO0, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_2HA_DDRIO0, 1)	},
472 473 474 475 476 477 478
};

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

479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494
/* 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
 */
495
#define HASWELL_DDRCRCLKCONTROLS 0xa10 /* Ditto on Broadwell */
496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514
#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
515 516 517
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO1 0x2fbf
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO2 0x2fb9
#define PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO3 0x2fbb
518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534
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)		},
535 536 537
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO1, 1)		},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO2, 1)		},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO3, 1)		},
538 539 540 541 542 543 544 545 546 547 548 549 550 551

	{ 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. */
};

552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595
/* Knight's Landing Support */
/*
 * KNL's memory channels are swizzled between memory controllers.
 * MC0 is mapped to CH3,5,6 and MC1 is mapped to CH0,1,2
 */
#define knl_channel_remap(channel) ((channel + 3) % 6)

/* Memory controller, TAD tables, error injection - 2-8-0, 2-9-0 (2 of these) */
#define PCI_DEVICE_ID_INTEL_KNL_IMC_MC       0x7840
/* DRAM channel stuff; bank addrs, dimmmtr, etc.. 2-8-2 - 2-9-4 (6 of these) */
#define PCI_DEVICE_ID_INTEL_KNL_IMC_CHANNEL  0x7843
/* kdrwdbu TAD limits/offsets, MCMTR - 2-10-1, 2-11-1 (2 of these) */
#define PCI_DEVICE_ID_INTEL_KNL_IMC_TA       0x7844
/* CHA broadcast registers, dram rules - 1-29-0 (1 of these) */
#define PCI_DEVICE_ID_INTEL_KNL_IMC_SAD0     0x782a
/* SAD target - 1-29-1 (1 of these) */
#define PCI_DEVICE_ID_INTEL_KNL_IMC_SAD1     0x782b
/* Caching / Home Agent */
#define PCI_DEVICE_ID_INTEL_KNL_IMC_CHA      0x782c
/* Device with TOLM and TOHM, 0-5-0 (1 of these) */
#define PCI_DEVICE_ID_INTEL_KNL_IMC_TOLHM    0x7810

/*
 * KNL differs from SB, IB, and Haswell in that it has multiple
 * instances of the same device with the same device ID, so we handle that
 * by creating as many copies in the table as we expect to find.
 * (Like device ID must be grouped together.)
 */

static const struct pci_id_descr pci_dev_descr_knl[] = {
	[0]         = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_SAD0, 0) },
	[1]         = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_SAD1, 0) },
	[2 ... 3]   = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_MC, 0)},
	[4 ... 41]  = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_CHA, 0) },
	[42 ... 47] = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_CHANNEL, 0) },
	[48]        = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_TA, 0) },
	[49]        = { PCI_DESCR(PCI_DEVICE_ID_INTEL_KNL_IMC_TOLHM, 0) },
};

static const struct pci_id_table pci_dev_descr_knl_table[] = {
	PCI_ID_TABLE_ENTRY(pci_dev_descr_knl),
	{0,}
};

596 597 598 599 600 601
/*
 * Broadwell support
 *
 * DE processor:
 *	- 1 IMC
 *	- 2 DDR3 channels, 2 DPC per channel
602 603 604 605 606 607 608 609 610 611 612
 * 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
613 614 615
 */
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_VTD_MISC 0x6f28
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0	0x6fa0
616
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1	0x6f60
617 618
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TA	0x6fa8
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_THERMAL 0x6f71
619 620
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_TA	0x6f68
#define PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1_THERMAL 0x6f79
621 622 623 624 625 626
#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
627 628 629 630
#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
631 632 633 634 635 636 637 638 639
#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)	},

640 641
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA1, 1)		},

642 643 644 645
	{ 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)	},
646 647 648
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD2, 1)	},
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TAD3, 1)	},

649
	{ PCI_DESCR(PCI_DEVICE_ID_INTEL_BROADWELL_IMC_DDRIO0, 1)	},
650 651 652 653 654 655 656

	{ 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)	},
657 658 659 660 661 662 663
};

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

664 665 666
/*
 *	pci_device_id	table for which devices we are looking for
 */
667
static const struct pci_device_id sbridge_pci_tbl[] = {
668
	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0)},
669
	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA)},
670
	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0)},
671
	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0)},
672
	{PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_KNL_IMC_SAD0)},
673 674 675 676 677
	{0,}			/* 0 terminated list. */
};


/****************************************************************************
D
David Mackey 已提交
678
			Ancillary status routines
679 680
 ****************************************************************************/

681
static inline int numrank(enum type type, u32 mtr)
682 683
{
	int ranks = (1 << RANK_CNT_BITS(mtr));
684 685
	int max = 4;

686
	if (type == HASWELL || type == BROADWELL || type == KNIGHTS_LANDING)
687
		max = 8;
688

689 690 691
	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);
692 693 694 695 696 697 698 699 700 701 702
		return -EINVAL;
	}

	return ranks;
}

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

	if (rows < 13 || rows > 18) {
703 704
		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);
705 706 707 708 709 710 711 712 713 714 715
		return -EINVAL;
	}

	return 1 << rows;
}

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

	if (cols > 12) {
716 717
		edac_dbg(0, "Invalid number of cols: %d (max = 4) raw value = %x (%04x)\n",
			 cols, (unsigned int)COL_WIDTH_BITS(mtr), mtr);
718 719 720 721 722 723
		return -EINVAL;
	}

	return 1 << cols;
}

724
static struct sbridge_dev *get_sbridge_dev(u8 bus, int multi_bus)
725 726 727
{
	struct sbridge_dev *sbridge_dev;

728 729 730 731 732 733 734 735 736
	/*
	 * If we have devices scattered across several busses that pertain
	 * to the same memory controller, we'll lump them all together.
	 */
	if (multi_bus) {
		return list_first_entry_or_null(&sbridge_edac_list,
				struct sbridge_dev, list);
	}

737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774
	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
775 776 777 778 779 780 781 782 783
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
784 785 786 787 788 789 790 791
static u64 sbridge_get_tohm(struct sbridge_pvt *pvt)
{
	u32 reg;

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

792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809
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);
}

810 811 812 813 814
static u64 rir_limit(u32 reg)
{
	return ((u64)GET_BITFIELD(reg,  1, 10) << 29) | 0x1fffffff;
}

815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834
static u64 sad_limit(u32 reg)
{
	return (GET_BITFIELD(reg, 6, 25) << 26) | 0x3ffffff;
}

static u32 interleave_mode(u32 reg)
{
	return GET_BITFIELD(reg, 1, 1);
}

char *show_interleave_mode(u32 reg)
{
	return interleave_mode(reg) ? "8:6" : "[8:6]XOR[18:16]";
}

static u32 dram_attr(u32 reg)
{
	return GET_BITFIELD(reg, 2, 3);
}

835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875
static u64 knl_sad_limit(u32 reg)
{
	return (GET_BITFIELD(reg, 7, 26) << 26) | 0x3ffffff;
}

static u32 knl_interleave_mode(u32 reg)
{
	return GET_BITFIELD(reg, 1, 2);
}

static char *knl_show_interleave_mode(u32 reg)
{
	char *s;

	switch (knl_interleave_mode(reg)) {
	case 0:
		s = "use address bits [8:6]";
		break;
	case 1:
		s = "use address bits [10:8]";
		break;
	case 2:
		s = "use address bits [14:12]";
		break;
	case 3:
		s = "use address bits [32:30]";
		break;
	default:
		WARN_ON(1);
		break;
	}

	return s;
}

static u32 dram_attr_knl(u32 reg)
{
	return GET_BITFIELD(reg, 3, 4);
}


876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894
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;
}

895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926
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;
}

927 928 929 930 931 932
static enum dev_type knl_get_width(struct sbridge_pvt *pvt, u32 mtr)
{
	/* for KNL value is fixed */
	return DEV_X16;
}

933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975
static enum dev_type sbridge_get_width(struct sbridge_pvt *pvt, u32 mtr)
{
	/* there's no way to figure out */
	return DEV_UNKNOWN;
}

static enum dev_type __ibridge_get_width(u32 mtr)
{
	enum dev_type type;

	switch (mtr) {
	case 3:
		type = DEV_UNKNOWN;
		break;
	case 2:
		type = DEV_X16;
		break;
	case 1:
		type = DEV_X8;
		break;
	case 0:
		type = DEV_X4;
		break;
	}

	return type;
}

static enum dev_type ibridge_get_width(struct sbridge_pvt *pvt, u32 mtr)
{
	/*
	 * ddr3_width on the documentation but also valid for DDR4 on
	 * Haswell
	 */
	return __ibridge_get_width(GET_BITFIELD(mtr, 7, 8));
}

static enum dev_type broadwell_get_width(struct sbridge_pvt *pvt, u32 mtr)
{
	/* ddr3_width on the documentation but also valid for DDR4 */
	return __ibridge_get_width(GET_BITFIELD(mtr, 8, 9));
}

976 977 978 979 980 981
static enum mem_type knl_get_memory_type(struct sbridge_pvt *pvt)
{
	/* DDR4 RDIMMS and LRDIMMS are supported */
	return MEM_RDDR4;
}

982 983 984 985 986 987 988
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);
}

989 990 991 992 993 994 995 996
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);
}

997 998 999 1000 1001 1002 1003 1004 1005
static u8 knl_get_node_id(struct sbridge_pvt *pvt)
{
	u32 reg;

	pci_read_config_dword(pvt->pci_sad1, SAD_CONTROL, &reg);
	return GET_BITFIELD(reg, 0, 2);
}


1006 1007 1008 1009
static u64 haswell_get_tolm(struct sbridge_pvt *pvt)
{
	u32 reg;

1010 1011
	pci_read_config_dword(pvt->info.pci_vtd, HASWELL_TOLM, &reg);
	return (GET_BITFIELD(reg, 26, 31) << 26) | 0x3ffffff;
1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026
}

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;
}

1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
static u64 knl_get_tolm(struct sbridge_pvt *pvt)
{
	u32 reg;

	pci_read_config_dword(pvt->knl.pci_mc_info, KNL_TOLM, &reg);
	return (GET_BITFIELD(reg, 26, 31) << 26) | 0x3ffffff;
}

static u64 knl_get_tohm(struct sbridge_pvt *pvt)
{
	u64 rc;
	u32 reg_lo, reg_hi;

	pci_read_config_dword(pvt->knl.pci_mc_info, KNL_TOHM_0, &reg_lo);
	pci_read_config_dword(pvt->knl.pci_mc_info, KNL_TOHM_1, &reg_hi);
	rc = ((u64)reg_hi << 32) | reg_lo;
	return rc | 0x3ffffff;
}


1047 1048 1049 1050 1051
static u64 haswell_rir_limit(u32 reg)
{
	return (((u64)GET_BITFIELD(reg,  1, 11) + 1) << 29) - 1;
}

1052 1053 1054
static inline u8 sad_pkg_socket(u8 pkg)
{
	/* on Ivy Bridge, nodeID is SASS, where A is HA and S is node id */
1055
	return ((pkg >> 3) << 2) | (pkg & 0x3);
1056 1057 1058 1059 1060 1061 1062
}

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

1063 1064 1065
/****************************************************************************
			Memory check routines
 ****************************************************************************/
1066
static struct pci_dev *get_pdev_same_bus(u8 bus, u32 id)
1067
{
1068
	struct pci_dev *pdev = NULL;
1069

1070 1071 1072 1073 1074
	do {
		pdev = pci_get_device(PCI_VENDOR_ID_INTEL, id, pdev);
		if (pdev && pdev->bus->number == bus)
			break;
	} while (pdev);
1075

1076
	return pdev;
1077 1078 1079
}

/**
1080
 * check_if_ecc_is_active() - Checks if ECC is active
1081 1082 1083 1084
 * @bus:	Device bus
 * @type:	Memory controller type
 * returns: 0 in case ECC is active, -ENODEV if it can't be determined or
 *	    disabled
1085
 */
1086
static int check_if_ecc_is_active(const u8 bus, enum type type)
1087 1088
{
	struct pci_dev *pdev = NULL;
1089
	u32 mcmtr, id;
1090

1091 1092
	switch (type) {
	case IVY_BRIDGE:
1093
		id = PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA;
1094 1095
		break;
	case HASWELL:
1096
		id = PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0_TA;
1097 1098
		break;
	case SANDY_BRIDGE:
1099
		id = PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TA;
1100 1101 1102 1103
		break;
	case BROADWELL:
		id = PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0_TA;
		break;
1104 1105 1106 1107 1108 1109 1110
	case KNIGHTS_LANDING:
		/*
		 * KNL doesn't group things by bus the same way
		 * SB/IB/Haswell does.
		 */
		id = PCI_DEVICE_ID_INTEL_KNL_IMC_TA;
		break;
1111 1112 1113
	default:
		return -ENODEV;
	}
1114

1115 1116 1117 1118 1119
	if (type != KNIGHTS_LANDING)
		pdev = get_pdev_same_bus(bus, id);
	else
		pdev = pci_get_device(PCI_VENDOR_ID_INTEL, id, 0);

1120 1121
	if (!pdev) {
		sbridge_printk(KERN_ERR, "Couldn't find PCI device "
1122 1123
					"%04x:%04x! on bus %02d\n",
					PCI_VENDOR_ID_INTEL, id, bus);
1124 1125 1126
		return -ENODEV;
	}

1127 1128
	pci_read_config_dword(pdev,
			type == KNIGHTS_LANDING ? KNL_MCMTR : MCMTR, &mcmtr);
1129 1130 1131 1132 1133 1134 1135
	if (!IS_ECC_ENABLED(mcmtr)) {
		sbridge_printk(KERN_ERR, "ECC is disabled. Aborting\n");
		return -ENODEV;
	}
	return 0;
}

1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 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 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 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 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 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 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576
/* Low bits of TAD limit, and some metadata. */
static const u32 knl_tad_dram_limit_lo[] = {
	0x400, 0x500, 0x600, 0x700,
	0x800, 0x900, 0xa00, 0xb00,
};

/* Low bits of TAD offset. */
static const u32 knl_tad_dram_offset_lo[] = {
	0x404, 0x504, 0x604, 0x704,
	0x804, 0x904, 0xa04, 0xb04,
};

/* High 16 bits of TAD limit and offset. */
static const u32 knl_tad_dram_hi[] = {
	0x408, 0x508, 0x608, 0x708,
	0x808, 0x908, 0xa08, 0xb08,
};

/* Number of ways a tad entry is interleaved. */
static const u32 knl_tad_ways[] = {
	8, 6, 4, 3, 2, 1,
};

/*
 * Retrieve the n'th Target Address Decode table entry
 * from the memory controller's TAD table.
 *
 * @pvt:	driver private data
 * @entry:	which entry you want to retrieve
 * @mc:		which memory controller (0 or 1)
 * @offset:	output tad range offset
 * @limit:	output address of first byte above tad range
 * @ways:	output number of interleave ways
 *
 * The offset value has curious semantics.  It's a sort of running total
 * of the sizes of all the memory regions that aren't mapped in this
 * tad table.
 */
static int knl_get_tad(const struct sbridge_pvt *pvt,
		const int entry,
		const int mc,
		u64 *offset,
		u64 *limit,
		int *ways)
{
	u32 reg_limit_lo, reg_offset_lo, reg_hi;
	struct pci_dev *pci_mc;
	int way_id;

	switch (mc) {
	case 0:
		pci_mc = pvt->knl.pci_mc0;
		break;
	case 1:
		pci_mc = pvt->knl.pci_mc1;
		break;
	default:
		WARN_ON(1);
		return -EINVAL;
	}

	pci_read_config_dword(pci_mc,
			knl_tad_dram_limit_lo[entry], &reg_limit_lo);
	pci_read_config_dword(pci_mc,
			knl_tad_dram_offset_lo[entry], &reg_offset_lo);
	pci_read_config_dword(pci_mc,
			knl_tad_dram_hi[entry], &reg_hi);

	/* Is this TAD entry enabled? */
	if (!GET_BITFIELD(reg_limit_lo, 0, 0))
		return -ENODEV;

	way_id = GET_BITFIELD(reg_limit_lo, 3, 5);

	if (way_id < ARRAY_SIZE(knl_tad_ways)) {
		*ways = knl_tad_ways[way_id];
	} else {
		*ways = 0;
		sbridge_printk(KERN_ERR,
				"Unexpected value %d in mc_tad_limit_lo wayness field\n",
				way_id);
		return -ENODEV;
	}

	/*
	 * The least significant 6 bits of base and limit are truncated.
	 * For limit, we fill the missing bits with 1s.
	 */
	*offset = ((u64) GET_BITFIELD(reg_offset_lo, 6, 31) << 6) |
				((u64) GET_BITFIELD(reg_hi, 0,  15) << 32);
	*limit = ((u64) GET_BITFIELD(reg_limit_lo,  6, 31) << 6) | 63 |
				((u64) GET_BITFIELD(reg_hi, 16, 31) << 32);

	return 0;
}

/* Determine which memory controller is responsible for a given channel. */
static int knl_channel_mc(int channel)
{
	WARN_ON(channel < 0 || channel >= 6);

	return channel < 3 ? 1 : 0;
}

/*
 * Get the Nth entry from EDC_ROUTE_TABLE register.
 * (This is the per-tile mapping of logical interleave targets to
 *  physical EDC modules.)
 *
 * entry 0: 0:2
 *       1: 3:5
 *       2: 6:8
 *       3: 9:11
 *       4: 12:14
 *       5: 15:17
 *       6: 18:20
 *       7: 21:23
 * reserved: 24:31
 */
static u32 knl_get_edc_route(int entry, u32 reg)
{
	WARN_ON(entry >= KNL_MAX_EDCS);
	return GET_BITFIELD(reg, entry*3, (entry*3)+2);
}

/*
 * Get the Nth entry from MC_ROUTE_TABLE register.
 * (This is the per-tile mapping of logical interleave targets to
 *  physical DRAM channels modules.)
 *
 * entry 0: mc 0:2   channel 18:19
 *       1: mc 3:5   channel 20:21
 *       2: mc 6:8   channel 22:23
 *       3: mc 9:11  channel 24:25
 *       4: mc 12:14 channel 26:27
 *       5: mc 15:17 channel 28:29
 * reserved: 30:31
 *
 * Though we have 3 bits to identify the MC, we should only see
 * the values 0 or 1.
 */

static u32 knl_get_mc_route(int entry, u32 reg)
{
	int mc, chan;

	WARN_ON(entry >= KNL_MAX_CHANNELS);

	mc = GET_BITFIELD(reg, entry*3, (entry*3)+2);
	chan = GET_BITFIELD(reg, (entry*2) + 18, (entry*2) + 18 + 1);

	return knl_channel_remap(mc*3 + chan);
}

/*
 * Render the EDC_ROUTE register in human-readable form.
 * Output string s should be at least KNL_MAX_EDCS*2 bytes.
 */
static void knl_show_edc_route(u32 reg, char *s)
{
	int i;

	for (i = 0; i < KNL_MAX_EDCS; i++) {
		s[i*2] = knl_get_edc_route(i, reg) + '0';
		s[i*2+1] = '-';
	}

	s[KNL_MAX_EDCS*2 - 1] = '\0';
}

/*
 * Render the MC_ROUTE register in human-readable form.
 * Output string s should be at least KNL_MAX_CHANNELS*2 bytes.
 */
static void knl_show_mc_route(u32 reg, char *s)
{
	int i;

	for (i = 0; i < KNL_MAX_CHANNELS; i++) {
		s[i*2] = knl_get_mc_route(i, reg) + '0';
		s[i*2+1] = '-';
	}

	s[KNL_MAX_CHANNELS*2 - 1] = '\0';
}

#define KNL_EDC_ROUTE 0xb8
#define KNL_MC_ROUTE 0xb4

/* Is this dram rule backed by regular DRAM in flat mode? */
#define KNL_EDRAM(reg) GET_BITFIELD(reg, 29, 29)

/* Is this dram rule cached? */
#define KNL_CACHEABLE(reg) GET_BITFIELD(reg, 28, 28)

/* Is this rule backed by edc ? */
#define KNL_EDRAM_ONLY(reg) GET_BITFIELD(reg, 29, 29)

/* Is this rule backed by DRAM, cacheable in EDRAM? */
#define KNL_CACHEABLE(reg) GET_BITFIELD(reg, 28, 28)

/* Is this rule mod3? */
#define KNL_MOD3(reg) GET_BITFIELD(reg, 27, 27)

/*
 * Figure out how big our RAM modules are.
 *
 * The DIMMMTR register in KNL doesn't tell us the size of the DIMMs, so we
 * have to figure this out from the SAD rules, interleave lists, route tables,
 * and TAD rules.
 *
 * SAD rules can have holes in them (e.g. the 3G-4G hole), so we have to
 * inspect the TAD rules to figure out how large the SAD regions really are.
 *
 * When we know the real size of a SAD region and how many ways it's
 * interleaved, we know the individual contribution of each channel to
 * TAD is size/ways.
 *
 * Finally, we have to check whether each channel participates in each SAD
 * region.
 *
 * Fortunately, KNL only supports one DIMM per channel, so once we know how
 * much memory the channel uses, we know the DIMM is at least that large.
 * (The BIOS might possibly choose not to map all available memory, in which
 * case we will underreport the size of the DIMM.)
 *
 * In theory, we could try to determine the EDC sizes as well, but that would
 * only work in flat mode, not in cache mode.
 *
 * @mc_sizes: Output sizes of channels (must have space for KNL_MAX_CHANNELS
 *            elements)
 */
static int knl_get_dimm_capacity(struct sbridge_pvt *pvt, u64 *mc_sizes)
{
	u64 sad_base, sad_size, sad_limit = 0;
	u64 tad_base, tad_size, tad_limit, tad_deadspace, tad_livespace;
	int sad_rule = 0;
	int tad_rule = 0;
	int intrlv_ways, tad_ways;
	u32 first_pkg, pkg;
	int i;
	u64 sad_actual_size[2]; /* sad size accounting for holes, per mc */
	u32 dram_rule, interleave_reg;
	u32 mc_route_reg[KNL_MAX_CHAS];
	u32 edc_route_reg[KNL_MAX_CHAS];
	int edram_only;
	char edc_route_string[KNL_MAX_EDCS*2];
	char mc_route_string[KNL_MAX_CHANNELS*2];
	int cur_reg_start;
	int mc;
	int channel;
	int way;
	int participants[KNL_MAX_CHANNELS];
	int participant_count = 0;

	for (i = 0; i < KNL_MAX_CHANNELS; i++)
		mc_sizes[i] = 0;

	/* Read the EDC route table in each CHA. */
	cur_reg_start = 0;
	for (i = 0; i < KNL_MAX_CHAS; i++) {
		pci_read_config_dword(pvt->knl.pci_cha[i],
				KNL_EDC_ROUTE, &edc_route_reg[i]);

		if (i > 0 && edc_route_reg[i] != edc_route_reg[i-1]) {
			knl_show_edc_route(edc_route_reg[i-1],
					edc_route_string);
			if (cur_reg_start == i-1)
				edac_dbg(0, "edc route table for CHA %d: %s\n",
					cur_reg_start, edc_route_string);
			else
				edac_dbg(0, "edc route table for CHA %d-%d: %s\n",
					cur_reg_start, i-1, edc_route_string);
			cur_reg_start = i;
		}
	}
	knl_show_edc_route(edc_route_reg[i-1], edc_route_string);
	if (cur_reg_start == i-1)
		edac_dbg(0, "edc route table for CHA %d: %s\n",
			cur_reg_start, edc_route_string);
	else
		edac_dbg(0, "edc route table for CHA %d-%d: %s\n",
			cur_reg_start, i-1, edc_route_string);

	/* Read the MC route table in each CHA. */
	cur_reg_start = 0;
	for (i = 0; i < KNL_MAX_CHAS; i++) {
		pci_read_config_dword(pvt->knl.pci_cha[i],
			KNL_MC_ROUTE, &mc_route_reg[i]);

		if (i > 0 && mc_route_reg[i] != mc_route_reg[i-1]) {
			knl_show_mc_route(mc_route_reg[i-1], mc_route_string);
			if (cur_reg_start == i-1)
				edac_dbg(0, "mc route table for CHA %d: %s\n",
					cur_reg_start, mc_route_string);
			else
				edac_dbg(0, "mc route table for CHA %d-%d: %s\n",
					cur_reg_start, i-1, mc_route_string);
			cur_reg_start = i;
		}
	}
	knl_show_mc_route(mc_route_reg[i-1], mc_route_string);
	if (cur_reg_start == i-1)
		edac_dbg(0, "mc route table for CHA %d: %s\n",
			cur_reg_start, mc_route_string);
	else
		edac_dbg(0, "mc route table for CHA %d-%d: %s\n",
			cur_reg_start, i-1, mc_route_string);

	/* Process DRAM rules */
	for (sad_rule = 0; sad_rule < pvt->info.max_sad; sad_rule++) {
		/* previous limit becomes the new base */
		sad_base = sad_limit;

		pci_read_config_dword(pvt->pci_sad0,
			pvt->info.dram_rule[sad_rule], &dram_rule);

		if (!DRAM_RULE_ENABLE(dram_rule))
			break;

		edram_only = KNL_EDRAM_ONLY(dram_rule);

		sad_limit = pvt->info.sad_limit(dram_rule)+1;
		sad_size = sad_limit - sad_base;

		pci_read_config_dword(pvt->pci_sad0,
			pvt->info.interleave_list[sad_rule], &interleave_reg);

		/*
		 * Find out how many ways this dram rule is interleaved.
		 * We stop when we see the first channel again.
		 */
		first_pkg = sad_pkg(pvt->info.interleave_pkg,
						interleave_reg, 0);
		for (intrlv_ways = 1; intrlv_ways < 8; intrlv_ways++) {
			pkg = sad_pkg(pvt->info.interleave_pkg,
						interleave_reg, intrlv_ways);

			if ((pkg & 0x8) == 0) {
				/*
				 * 0 bit means memory is non-local,
				 * which KNL doesn't support
				 */
				edac_dbg(0, "Unexpected interleave target %d\n",
					pkg);
				return -1;
			}

			if (pkg == first_pkg)
				break;
		}
		if (KNL_MOD3(dram_rule))
			intrlv_ways *= 3;

		edac_dbg(3, "dram rule %d (base 0x%llx, limit 0x%llx), %d way interleave%s\n",
			sad_rule,
			sad_base,
			sad_limit,
			intrlv_ways,
			edram_only ? ", EDRAM" : "");

		/*
		 * Find out how big the SAD region really is by iterating
		 * over TAD tables (SAD regions may contain holes).
		 * Each memory controller might have a different TAD table, so
		 * we have to look at both.
		 *
		 * Livespace is the memory that's mapped in this TAD table,
		 * deadspace is the holes (this could be the MMIO hole, or it
		 * could be memory that's mapped by the other TAD table but
		 * not this one).
		 */
		for (mc = 0; mc < 2; mc++) {
			sad_actual_size[mc] = 0;
			tad_livespace = 0;
			for (tad_rule = 0;
					tad_rule < ARRAY_SIZE(
						knl_tad_dram_limit_lo);
					tad_rule++) {
				if (knl_get_tad(pvt,
						tad_rule,
						mc,
						&tad_deadspace,
						&tad_limit,
						&tad_ways))
					break;

				tad_size = (tad_limit+1) -
					(tad_livespace + tad_deadspace);
				tad_livespace += tad_size;
				tad_base = (tad_limit+1) - tad_size;

				if (tad_base < sad_base) {
					if (tad_limit > sad_base)
						edac_dbg(0, "TAD region overlaps lower SAD boundary -- TAD tables may be configured incorrectly.\n");
				} else if (tad_base < sad_limit) {
					if (tad_limit+1 > sad_limit) {
						edac_dbg(0, "TAD region overlaps upper SAD boundary -- TAD tables may be configured incorrectly.\n");
					} else {
						/* TAD region is completely inside SAD region */
						edac_dbg(3, "TAD region %d 0x%llx - 0x%llx (%lld bytes) table%d\n",
							tad_rule, tad_base,
							tad_limit, tad_size,
							mc);
						sad_actual_size[mc] += tad_size;
					}
				}
				tad_base = tad_limit+1;
			}
		}

		for (mc = 0; mc < 2; mc++) {
			edac_dbg(3, " total TAD DRAM footprint in table%d : 0x%llx (%lld bytes)\n",
				mc, sad_actual_size[mc], sad_actual_size[mc]);
		}

		/* Ignore EDRAM rule */
		if (edram_only)
			continue;

		/* Figure out which channels participate in interleave. */
		for (channel = 0; channel < KNL_MAX_CHANNELS; channel++)
			participants[channel] = 0;

		/* For each channel, does at least one CHA have
		 * this channel mapped to the given target?
		 */
		for (channel = 0; channel < KNL_MAX_CHANNELS; channel++) {
			for (way = 0; way < intrlv_ways; way++) {
				int target;
				int cha;

				if (KNL_MOD3(dram_rule))
					target = way;
				else
					target = 0x7 & sad_pkg(
				pvt->info.interleave_pkg, interleave_reg, way);

				for (cha = 0; cha < KNL_MAX_CHAS; cha++) {
					if (knl_get_mc_route(target,
						mc_route_reg[cha]) == channel
1577
						&& !participants[channel]) {
1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605
						participant_count++;
						participants[channel] = 1;
						break;
					}
				}
			}
		}

		if (participant_count != intrlv_ways)
			edac_dbg(0, "participant_count (%d) != interleave_ways (%d): DIMM size may be incorrect\n",
				participant_count, intrlv_ways);

		for (channel = 0; channel < KNL_MAX_CHANNELS; channel++) {
			mc = knl_channel_mc(channel);
			if (participants[channel]) {
				edac_dbg(4, "mc channel %d contributes %lld bytes via sad entry %d\n",
					channel,
					sad_actual_size[mc]/intrlv_ways,
					sad_rule);
				mc_sizes[channel] +=
					sad_actual_size[mc]/intrlv_ways;
			}
		}
	}

	return 0;
}

1606
static int get_dimm_config(struct mem_ctl_info *mci)
1607 1608
{
	struct sbridge_pvt *pvt = mci->pvt_info;
1609
	struct dimm_info *dimm;
1610 1611
	unsigned i, j, banks, ranks, rows, cols, npages;
	u64 size;
1612 1613
	u32 reg;
	enum edac_type mode;
1614
	enum mem_type mtype;
1615 1616 1617
	int channels = pvt->info.type == KNIGHTS_LANDING ?
		KNL_MAX_CHANNELS : NUM_CHANNELS;
	u64 knl_mc_sizes[KNL_MAX_CHANNELS];
1618

1619 1620
	if (pvt->info.type == HASWELL || pvt->info.type == BROADWELL ||
			pvt->info.type == KNIGHTS_LANDING)
1621 1622 1623 1624
		pci_read_config_dword(pvt->pci_sad1, SAD_TARGET, &reg);
	else
		pci_read_config_dword(pvt->pci_br0, SAD_TARGET, &reg);

1625 1626 1627 1628
	if (pvt->info.type == KNIGHTS_LANDING)
		pvt->sbridge_dev->source_id = SOURCE_ID_KNL(reg);
	else
		pvt->sbridge_dev->source_id = SOURCE_ID(reg);
1629

1630
	pvt->sbridge_dev->node_id = pvt->info.get_node_id(pvt);
1631 1632 1633 1634
	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);
1635

1636 1637 1638 1639 1640
	/* KNL doesn't support mirroring or lockstep,
	 * and is always closed page
	 */
	if (pvt->info.type == KNIGHTS_LANDING) {
		mode = EDAC_S4ECD4ED;
1641 1642
		pvt->is_mirrored = false;

1643 1644
		if (knl_get_dimm_capacity(pvt, knl_mc_sizes) != 0)
			return -1;
1645
	} else {
1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671
		pci_read_config_dword(pvt->pci_ras, RASENABLES, &reg);
		if (IS_MIRROR_ENABLED(reg)) {
			edac_dbg(0, "Memory mirror is enabled\n");
			pvt->is_mirrored = true;
		} else {
			edac_dbg(0, "Memory mirror is disabled\n");
			pvt->is_mirrored = false;
		}

		pci_read_config_dword(pvt->pci_ta, MCMTR, &pvt->info.mcmtr);
		if (IS_LOCKSTEP_ENABLED(pvt->info.mcmtr)) {
			edac_dbg(0, "Lockstep is enabled\n");
			mode = EDAC_S8ECD8ED;
			pvt->is_lockstep = true;
		} else {
			edac_dbg(0, "Lockstep is disabled\n");
			mode = EDAC_S4ECD4ED;
			pvt->is_lockstep = false;
		}
		if (IS_CLOSE_PG(pvt->info.mcmtr)) {
			edac_dbg(0, "address map is on closed page mode\n");
			pvt->is_close_pg = true;
		} else {
			edac_dbg(0, "address map is on open page mode\n");
			pvt->is_close_pg = false;
		}
1672 1673
	}

1674
	mtype = pvt->info.get_memory_type(pvt);
1675
	if (mtype == MEM_RDDR3 || mtype == MEM_RDDR4)
1676 1677
		edac_dbg(0, "Memory is registered\n");
	else if (mtype == MEM_UNKNOWN)
1678
		edac_dbg(0, "Cannot determine memory type\n");
1679 1680
	else
		edac_dbg(0, "Memory is unregistered\n");
1681

1682
	if (mtype == MEM_DDR4 || mtype == MEM_RDDR4)
1683 1684 1685
		banks = 16;
	else
		banks = 8;
1686

1687
	for (i = 0; i < channels; i++) {
1688 1689
		u32 mtr;

1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702
		int max_dimms_per_channel;

		if (pvt->info.type == KNIGHTS_LANDING) {
			max_dimms_per_channel = 1;
			if (!pvt->knl.pci_channel[i])
				continue;
		} else {
			max_dimms_per_channel = ARRAY_SIZE(mtr_regs);
			if (!pvt->pci_tad[i])
				continue;
		}

		for (j = 0; j < max_dimms_per_channel; j++) {
1703 1704
			dimm = EDAC_DIMM_PTR(mci->layers, mci->dimms, mci->n_layers,
				       i, j, 0);
1705 1706 1707 1708 1709 1710 1711
			if (pvt->info.type == KNIGHTS_LANDING) {
				pci_read_config_dword(pvt->knl.pci_channel[i],
					knl_mtr_reg, &mtr);
			} else {
				pci_read_config_dword(pvt->pci_tad[i],
					mtr_regs[j], &mtr);
			}
1712
			edac_dbg(4, "Channel #%d  MTR%d = %x\n", i, j, mtr);
1713 1714 1715
			if (IS_DIMM_PRESENT(mtr)) {
				pvt->channel[i].dimms++;

1716
				ranks = numrank(pvt->info.type, mtr);
1717 1718 1719 1720 1721 1722 1723 1724 1725 1726

				if (pvt->info.type == KNIGHTS_LANDING) {
					/* For DDR4, this is fixed. */
					cols = 1 << 10;
					rows = knl_mc_sizes[i] /
						((u64) cols * ranks * banks * 8);
				} else {
					rows = numrow(mtr);
					cols = numcol(mtr);
				}
1727

1728
				size = ((u64)rows * cols * banks * ranks) >> (20 - 3);
1729 1730
				npages = MiB_TO_PAGES(size);

1731 1732
				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,
1733 1734
					 size, npages,
					 banks, ranks, rows, cols);
1735

1736
				dimm->nr_pages = npages;
1737
				dimm->grain = 32;
1738
				dimm->dtype = pvt->info.get_width(pvt, mtr);
1739 1740 1741
				dimm->mtype = mtype;
				dimm->edac_mode = mode;
				snprintf(dimm->label, sizeof(dimm->label),
1742 1743
					 "CPU_SrcID#%u_Ha#%u_Chan#%u_DIMM#%u",
					 pvt->sbridge_dev->source_id, i/4, i%4, j);
1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757
			}
		}
	}

	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;
1758
	u32 gb, mb;
1759 1760 1761 1762 1763 1764
	u32 rir_way;

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

A
Aristeu Rozanski 已提交
1765
	pvt->tolm = pvt->info.get_tolm(pvt);
1766 1767
	tmp_mb = (1 + pvt->tolm) >> 20;

1768 1769 1770
	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);
1771 1772

	/* Address range is already 45:25 */
A
Aristeu Rozanski 已提交
1773
	pvt->tohm = pvt->info.get_tohm(pvt);
1774 1775
	tmp_mb = (1 + pvt->tohm) >> 20;

1776 1777 1778
	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);
1779 1780 1781 1782 1783 1784 1785 1786

	/*
	 * 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;
1787
	for (n_sads = 0; n_sads < pvt->info.max_sad; n_sads++) {
1788
		/* SAD_LIMIT Address range is 45:26 */
1789
		pci_read_config_dword(pvt->pci_sad0, pvt->info.dram_rule[n_sads],
1790
				      &reg);
1791
		limit = pvt->info.sad_limit(reg);
1792 1793 1794 1795 1796 1797 1798 1799

		if (!DRAM_RULE_ENABLE(reg))
			continue;

		if (limit <= prv)
			break;

		tmp_mb = (limit + 1) >> 20;
1800
		gb = div_u64_rem(tmp_mb, 1024, &mb);
1801 1802
		edac_dbg(0, "SAD#%d %s up to %u.%03u GB (0x%016Lx) Interleave: %s reg=0x%08x\n",
			 n_sads,
1803
			 show_dram_attr(pvt->info.dram_attr(reg)),
1804
			 gb, (mb*1000)/1024,
1805
			 ((u64)tmp_mb) << 20L,
1806
			 pvt->info.show_interleave_mode(reg),
1807
			 reg);
1808 1809
		prv = limit;

1810
		pci_read_config_dword(pvt->pci_sad0, pvt->info.interleave_list[n_sads],
1811
				      &reg);
A
Aristeu Rozanski 已提交
1812
		sad_interl = sad_pkg(pvt->info.interleave_pkg, reg, 0);
1813
		for (j = 0; j < 8; j++) {
A
Aristeu Rozanski 已提交
1814 1815
			u32 pkg = sad_pkg(pvt->info.interleave_pkg, reg, j);
			if (j > 0 && sad_interl == pkg)
1816 1817
				break;

1818
			edac_dbg(0, "SAD#%d, interleave #%d: %d\n",
A
Aristeu Rozanski 已提交
1819
				 n_sads, j, pkg);
1820 1821 1822
		}
	}

1823 1824 1825
	if (pvt->info.type == KNIGHTS_LANDING)
		return;

1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837
	/*
	 * 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;

1838
		gb = div_u64_rem(tmp_mb, 1024, &mb);
1839
		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",
1840
			 n_tads, gb, (mb*1000)/1024,
1841
			 ((u64)tmp_mb) << 20L,
1842 1843
			 (u32)(1 << TAD_SOCK(reg)),
			 (u32)TAD_CH(reg) + 1,
1844 1845 1846 1847 1848
			 (u32)TAD_TGT0(reg),
			 (u32)TAD_TGT1(reg),
			 (u32)TAD_TGT2(reg),
			 (u32)TAD_TGT3(reg),
			 reg);
1849
		prv = limit;
1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862
	}

	/*
	 * 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;
1863
			gb = div_u64_rem(tmp_mb, 1024, &mb);
1864 1865
			edac_dbg(0, "TAD CH#%d, offset #%d: %u.%03u GB (0x%016Lx), reg=0x%08x\n",
				 i, j,
1866
				 gb, (mb*1000)/1024,
1867 1868
				 ((u64)tmp_mb) << 20L,
				 reg);
1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885
		}
	}

	/*
	 * 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;

1886
			tmp_mb = pvt->info.rir_limit(reg) >> 20;
1887
			rir_way = 1 << RIR_WAY(reg);
1888
			gb = div_u64_rem(tmp_mb, 1024, &mb);
1889 1890
			edac_dbg(0, "CH#%d RIR#%d, limit: %u.%03u GB (0x%016Lx), way: %d, reg=0x%08x\n",
				 i, j,
1891
				 gb, (mb*1000)/1024,
1892 1893 1894
				 ((u64)tmp_mb) << 20L,
				 rir_way,
				 reg);
1895 1896 1897 1898 1899 1900 1901

			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;

1902
				gb = div_u64_rem(tmp_mb, 1024, &mb);
1903 1904
				edac_dbg(0, "CH#%d RIR#%d INTL#%d, offset %u.%03u GB (0x%016Lx), tgt: %d, reg=0x%08x\n",
					 i, j, k,
1905
					 gb, (mb*1000)/1024,
1906 1907 1908
					 ((u64)tmp_mb) << 20L,
					 (u32)RIR_RNK_TGT(reg),
					 reg);
1909 1910 1911 1912 1913
			}
		}
	}
}

1914
static struct mem_ctl_info *get_mci_for_node_id(u8 node_id)
1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926
{
	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,
1927
				 u8 *socket, u8 *ha,
1928 1929
				 long *channel_mask,
				 u8 *rank,
1930
				 char **area_type, char *msg)
1931 1932 1933
{
	struct mem_ctl_info	*new_mci;
	struct sbridge_pvt *pvt = mci->pvt_info;
1934
	struct pci_dev		*pci_ha;
1935
	int			n_rir, n_sads, n_tads, sad_way, sck_xch;
1936
	int			sad_interl, idx, base_ch;
1937
	int			interleave_mode, shiftup = 0;
1938
	unsigned		sad_interleave[pvt->info.max_interleave];
1939
	u32			reg, dram_rule;
1940
	u8			ch_way, sck_way, pkg, sad_ha = 0, ch_add = 0;
1941 1942
	u32			tad_offset;
	u32			rir_way;
1943
	u32			mb, gb;
1944
	u64			ch_addr, offset, limit = 0, prv = 0;
1945 1946 1947 1948 1949 1950 1951 1952 1953


	/*
	 * 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.
	 */
1954
	if ((addr > (u64) pvt->tolm) && (addr < (1LL << 32))) {
1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965
		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
	 */
1966 1967
	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],
1968 1969 1970 1971 1972
				      &reg);

		if (!DRAM_RULE_ENABLE(reg))
			continue;

1973
		limit = pvt->info.sad_limit(reg);
1974 1975 1976 1977 1978 1979 1980 1981
		if (limit <= prv) {
			sprintf(msg, "Can't discover the memory socket");
			return -EINVAL;
		}
		if  (addr <= limit)
			break;
		prv = limit;
	}
1982
	if (n_sads == pvt->info.max_sad) {
1983 1984 1985
		sprintf(msg, "Can't discover the memory socket");
		return -EINVAL;
	}
1986
	dram_rule = reg;
1987 1988
	*area_type = show_dram_attr(pvt->info.dram_attr(dram_rule));
	interleave_mode = pvt->info.interleave_mode(dram_rule);
1989

1990
	pci_read_config_dword(pvt->pci_sad0, pvt->info.interleave_list[n_sads],
1991
			      &reg);
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

	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;
2017
			break;
2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032
		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);
2033
	} else if (pvt->info.type == HASWELL || pvt->info.type == BROADWELL) {
2034 2035 2036 2037 2038 2039 2040
		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
2041
			bits = GET_BITFIELD(addr, 6, 8);
2042

2043
		if (interleave_mode == 0) {
2044 2045 2046 2047 2048 2049 2050 2051 2052
			/* 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);
2053 2054
		if (sad_ha)
			ch_add = 4;
2055 2056 2057 2058 2059 2060 2061 2062 2063 2064

		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);
2065 2066
	} else {
		/* Ivy Bridge's SAD mode doesn't support XOR interleave mode */
2067
		idx = (addr >> 6) & 7;
2068 2069 2070
		pkg = sad_pkg(pvt->info.interleave_pkg, reg, idx);
		*socket = sad_pkg_socket(pkg);
		sad_ha = sad_pkg_ha(pkg);
2071 2072
		if (sad_ha)
			ch_add = 4;
2073 2074
		edac_dbg(0, "SAD interleave package: %d = CPU socket %d, HA %d\n",
			 idx, *socket, sad_ha);
2075 2076
	}

2077 2078
	*ha = sad_ha;

2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095
	/*
	 * 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;
2096 2097 2098 2099 2100 2101 2102 2103
	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;
	}
2104
	for (n_tads = 0; n_tads < MAX_TAD; n_tads++) {
2105
		pci_read_config_dword(pci_ha, tad_dram_rule[n_tads], &reg);
2106 2107 2108 2109 2110 2111 2112 2113 2114
		limit = TAD_LIMIT(reg);
		if (limit <= prv) {
			sprintf(msg, "Can't discover the memory channel");
			return -EINVAL;
		}
		if  (addr <= limit)
			break;
		prv = limit;
	}
2115 2116 2117 2118 2119
	if (n_tads == MAX_TAD) {
		sprintf(msg, "Can't discover the memory channel");
		return -EINVAL;
	}

2120
	ch_way = TAD_CH(reg) + 1;
2121
	sck_way = TAD_SOCK(reg);
2122 2123 2124 2125

	if (ch_way == 3)
		idx = addr >> 6;
	else
2126
		idx = (addr >> (6 + sck_way + shiftup)) & 0x3;
2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150
	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;

2151
	pci_read_config_dword(pvt->pci_tad[ch_add + base_ch],
2152 2153 2154
				tad_ch_nilv_offset[n_tads],
				&tad_offset);

2155 2156 2157 2158 2159
	if (pvt->is_mirrored) {
		*channel_mask |= 1 << ((base_ch + 2) % 4);
		switch(ch_way) {
		case 2:
		case 4:
2160
			sck_xch = (1 << sck_way) * (ch_way >> 1);
2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173
			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);

2174 2175 2176 2177
	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,
2178
		 sck_way,
2179 2180 2181 2182 2183
		 ch_way,
		 offset,
		 idx,
		 base_ch,
		 *channel_mask);
2184 2185 2186 2187 2188 2189 2190 2191 2192

	/* 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;
	}
2193 2194 2195

	ch_addr = addr - offset;
	ch_addr >>= (6 + shiftup);
2196
	ch_addr /= sck_xch;
2197 2198
	ch_addr <<= (6 + shiftup);
	ch_addr |= addr & ((1 << (6 + shiftup)) - 1);
2199 2200 2201 2202 2203

	/*
	 * Step 3) Decode rank
	 */
	for (n_rir = 0; n_rir < MAX_RIR_RANGES; n_rir++) {
2204
		pci_read_config_dword(pvt->pci_tad[ch_add + base_ch],
2205 2206 2207 2208 2209 2210
				      rir_way_limit[n_rir],
				      &reg);

		if (!IS_RIR_VALID(reg))
			continue;

2211
		limit = pvt->info.rir_limit(reg);
2212
		gb = div_u64_rem(limit >> 20, 1024, &mb);
2213 2214
		edac_dbg(0, "RIR#%d, limit: %u.%03u GB (0x%016Lx), way: %d\n",
			 n_rir,
2215
			 gb, (mb*1000)/1024,
2216 2217
			 limit,
			 1 << RIR_WAY(reg));
2218 2219 2220 2221 2222 2223 2224 2225 2226
		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);
2227

2228 2229 2230 2231 2232 2233
	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;

2234
	pci_read_config_dword(pvt->pci_tad[ch_add + base_ch],
2235 2236 2237 2238
			      rir_offset[n_rir][idx],
			      &reg);
	*rank = RIR_RNK_TGT(reg);

2239 2240 2241 2242 2243 2244
	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);
2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260

	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;

2261
	edac_dbg(0, "\n");
2262 2263 2264 2265
	for (i = 0; i < sbridge_dev->n_devs; i++) {
		struct pci_dev *pdev = sbridge_dev->pdev[i];
		if (!pdev)
			continue;
2266 2267 2268
		edac_dbg(0, "Removing dev %02x:%02x.%d\n",
			 pdev->bus->number,
			 PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285
		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,
2286 2287
				 const unsigned devno,
				 const int multi_bus)
2288 2289 2290 2291 2292 2293
{
	struct sbridge_dev *sbridge_dev;
	const struct pci_id_descr *dev_descr = &table->descr[devno];
	struct pci_dev *pdev = NULL;
	u8 bus = 0;

2294
	sbridge_printk(KERN_DEBUG,
2295
		"Seeking for: PCI ID %04x:%04x\n",
2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309
		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;

2310
		/* if the HA wasn't found */
2311 2312 2313 2314
		if (devno == 0)
			return -ENODEV;

		sbridge_printk(KERN_INFO,
2315
			"Device not found: %04x:%04x\n",
2316 2317 2318 2319 2320 2321 2322
			PCI_VENDOR_ID_INTEL, dev_descr->dev_id);

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

2323
	sbridge_dev = get_sbridge_dev(bus, multi_bus);
2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334
	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,
2335
			"Duplicated device for %04x:%04x\n",
2336 2337 2338 2339 2340 2341 2342 2343 2344 2345
			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,
2346
			"Couldn't enable %04x:%04x\n",
2347 2348 2349 2350
			PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
		return -ENODEV;
	}

2351
	edac_dbg(0, "Detected %04x:%04x\n",
2352
		 PCI_VENDOR_ID_INTEL, dev_descr->dev_id);
2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365

	/*
	 * 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;
}

2366 2367
/*
 * sbridge_get_all_devices - Find and perform 'get' operation on the MCH's
2368
 *			     devices we want to reference for this driver.
2369
 * @num_mc: pointer to the memory controllers count, to be incremented in case
2370
 *	    of success.
2371
 * @table: model specific table
2372 2373 2374 2375 2376
 * @allow_dups: allow for multiple devices to exist with the same device id
 *              (as implemented, this isn't expected to work correctly in the
 *              multi-socket case).
 * @multi_bus: don't assume devices on different buses belong to different
 *             memory controllers.
2377 2378 2379
 *
 * returns 0 in case of success or error code
 */
2380 2381 2382 2383
static int sbridge_get_all_devices_full(u8 *num_mc,
					const struct pci_id_table *table,
					int allow_dups,
					int multi_bus)
2384 2385 2386 2387 2388 2389
{
	int i, rc;
	struct pci_dev *pdev = NULL;

	while (table && table->descr) {
		for (i = 0; i < table->n_devs; i++) {
2390 2391 2392 2393 2394
			if (!allow_dups || i == 0 ||
					table->descr[i].dev_id !=
						table->descr[i-1].dev_id) {
				pdev = NULL;
			}
2395 2396
			do {
				rc = sbridge_get_onedevice(&pdev, num_mc,
2397
							   table, i, multi_bus);
2398 2399 2400 2401 2402 2403 2404 2405
				if (rc < 0) {
					if (i == 0) {
						i = table->n_devs;
						break;
					}
					sbridge_put_all_devices();
					return -ENODEV;
				}
2406
			} while (pdev && !allow_dups);
2407 2408 2409 2410 2411 2412 2413
		}
		table++;
	}

	return 0;
}

2414 2415
#define sbridge_get_all_devices(num_mc, table) \
		sbridge_get_all_devices_full(num_mc, table, 0, 0)
2416 2417
#define sbridge_get_all_devices_knl(num_mc, table) \
		sbridge_get_all_devices_full(num_mc, table, 1, 1)
2418

A
Aristeu Rozanski 已提交
2419 2420
static int sbridge_mci_bind_devs(struct mem_ctl_info *mci,
				 struct sbridge_dev *sbridge_dev)
2421 2422 2423
{
	struct sbridge_pvt *pvt = mci->pvt_info;
	struct pci_dev *pdev;
2424
	u8 saw_chan_mask = 0;
2425
	int i;
2426 2427 2428 2429 2430

	for (i = 0; i < sbridge_dev->n_devs; i++) {
		pdev = sbridge_dev->pdev[i];
		if (!pdev)
			continue;
2431 2432 2433 2434

		switch (pdev->device) {
		case PCI_DEVICE_ID_INTEL_SBRIDGE_SAD0:
			pvt->pci_sad0 = pdev;
2435
			break;
2436 2437
		case PCI_DEVICE_ID_INTEL_SBRIDGE_SAD1:
			pvt->pci_sad1 = pdev;
2438
			break;
2439 2440
		case PCI_DEVICE_ID_INTEL_SBRIDGE_BR:
			pvt->pci_br0 = pdev;
2441
			break;
2442 2443
		case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0:
			pvt->pci_ha0 = pdev;
2444
			break;
2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457
		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;
2458
			saw_chan_mask |= 1 << id;
2459 2460 2461 2462
		}
			break;
		case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_DDRIO:
			pvt->pci_ddrio = pdev;
2463 2464 2465 2466 2467
			break;
		default:
			goto error;
		}

2468 2469
		edac_dbg(0, "Associated PCI %02x:%02x, bus %d with dev = %p\n",
			 pdev->vendor, pdev->device,
2470 2471
			 sbridge_dev->bus,
			 pdev);
2472 2473 2474 2475
	}

	/* Check if everything were registered */
	if (!pvt->pci_sad0 || !pvt->pci_sad1 || !pvt->pci_ha0 ||
2476
	    !pvt-> pci_tad || !pvt->pci_ras  || !pvt->pci_ta)
2477 2478
		goto enodev;

2479 2480
	if (saw_chan_mask != 0x0f)
		goto enodev;
2481 2482 2483 2484 2485 2486 2487
	return 0;

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

error:
2488 2489
	sbridge_printk(KERN_ERR, "Unexpected device %02x:%02x\n",
		       PCI_VENDOR_ID_INTEL, pdev->device);
2490 2491 2492
	return -EINVAL;
}

2493 2494 2495 2496
static int ibridge_mci_bind_devs(struct mem_ctl_info *mci,
				 struct sbridge_dev *sbridge_dev)
{
	struct sbridge_pvt *pvt = mci->pvt_info;
2497 2498
	struct pci_dev *pdev;
	u8 saw_chan_mask = 0;
2499
	int i;
2500 2501 2502 2503 2504 2505

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

2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516
		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:
2517 2518
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD2:
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD3:
2519 2520 2521
		{
			int id = pdev->device - PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TAD0;
			pvt->pci_tad[id] = pdev;
2522
			saw_chan_mask |= 1 << id;
2523
		}
2524
			break;
2525 2526 2527 2528
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_2HA_DDRIO0:
			pvt->pci_ddrio = pdev;
			break;
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_1HA_DDRIO0:
2529
			pvt->pci_ddrio = pdev;
2530
			break;
2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544
		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:
2545 2546
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD2:
		case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD3:
2547
		{
2548
			int id = pdev->device - PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA1_TAD0 + 4;
2549
			pvt->pci_tad[id] = pdev;
2550
			saw_chan_mask |= 1 << id;
2551 2552
		}
			break;
2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568
		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;

2569 2570 2571 2572
	if (saw_chan_mask != 0x0f && /* -EN */
	    saw_chan_mask != 0x33 && /* -EP */
	    saw_chan_mask != 0xff)   /* -EX */
		goto enodev;
2573 2574 2575 2576 2577 2578 2579 2580
	return 0;

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

error:
	sbridge_printk(KERN_ERR,
2581 2582
		       "Unexpected device %02x:%02x\n", PCI_VENDOR_ID_INTEL,
			pdev->device);
2583 2584 2585
	return -EINVAL;
}

2586 2587 2588 2589
static int haswell_mci_bind_devs(struct mem_ctl_info *mci,
				 struct sbridge_dev *sbridge_dev)
{
	struct sbridge_pvt *pvt = mci->pvt_info;
2590 2591
	struct pci_dev *pdev;
	u8 saw_chan_mask = 0;
2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625
	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:
2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642
		{
			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;
		}
2643 2644
			break;
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO0:
2645 2646 2647 2648 2649
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO1:
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO2:
		case PCI_DEVICE_ID_INTEL_HASWELL_IMC_DDRIO3:
			if (!pvt->pci_ddrio)
				pvt->pci_ddrio = pdev;
2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671
			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;

2672 2673 2674 2675
	if (saw_chan_mask != 0x0f && /* -EN */
	    saw_chan_mask != 0x33 && /* -EP */
	    saw_chan_mask != 0xff)   /* -EX */
		goto enodev;
2676 2677 2678 2679 2680 2681 2682
	return 0;

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

2683 2684 2685 2686 2687
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;
2688
	u8 saw_chan_mask = 0;
2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722
	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:
2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737
		{
			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;
		}
2738 2739 2740 2741
			break;
		case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_DDRIO0:
			pvt->pci_ddrio = pdev;
			break;
2742 2743 2744 2745 2746 2747
		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;
2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762
		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;

2763 2764 2765 2766
	if (saw_chan_mask != 0x0f && /* -EN */
	    saw_chan_mask != 0x33 && /* -EP */
	    saw_chan_mask != 0xff)   /* -EX */
		goto enodev;
2767 2768 2769 2770 2771 2772 2773
	return 0;

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

2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898
static int knl_mci_bind_devs(struct mem_ctl_info *mci,
			struct sbridge_dev *sbridge_dev)
{
	struct sbridge_pvt *pvt = mci->pvt_info;
	struct pci_dev *pdev;
	int dev, func;

	int i;
	int devidx;

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

		/* Extract PCI device and function. */
		dev = (pdev->devfn >> 3) & 0x1f;
		func = pdev->devfn & 0x7;

		switch (pdev->device) {
		case PCI_DEVICE_ID_INTEL_KNL_IMC_MC:
			if (dev == 8)
				pvt->knl.pci_mc0 = pdev;
			else if (dev == 9)
				pvt->knl.pci_mc1 = pdev;
			else {
				sbridge_printk(KERN_ERR,
					"Memory controller in unexpected place! (dev %d, fn %d)\n",
					dev, func);
				continue;
			}
			break;

		case PCI_DEVICE_ID_INTEL_KNL_IMC_SAD0:
			pvt->pci_sad0 = pdev;
			break;

		case PCI_DEVICE_ID_INTEL_KNL_IMC_SAD1:
			pvt->pci_sad1 = pdev;
			break;

		case PCI_DEVICE_ID_INTEL_KNL_IMC_CHA:
			/* There are one of these per tile, and range from
			 * 1.14.0 to 1.18.5.
			 */
			devidx = ((dev-14)*8)+func;

			if (devidx < 0 || devidx >= KNL_MAX_CHAS) {
				sbridge_printk(KERN_ERR,
					"Caching and Home Agent in unexpected place! (dev %d, fn %d)\n",
					dev, func);
				continue;
			}

			WARN_ON(pvt->knl.pci_cha[devidx] != NULL);

			pvt->knl.pci_cha[devidx] = pdev;
			break;

		case PCI_DEVICE_ID_INTEL_KNL_IMC_CHANNEL:
			devidx = -1;

			/*
			 *  MC0 channels 0-2 are device 9 function 2-4,
			 *  MC1 channels 3-5 are device 8 function 2-4.
			 */

			if (dev == 9)
				devidx = func-2;
			else if (dev == 8)
				devidx = 3 + (func-2);

			if (devidx < 0 || devidx >= KNL_MAX_CHANNELS) {
				sbridge_printk(KERN_ERR,
					"DRAM Channel Registers in unexpected place! (dev %d, fn %d)\n",
					dev, func);
				continue;
			}

			WARN_ON(pvt->knl.pci_channel[devidx] != NULL);
			pvt->knl.pci_channel[devidx] = pdev;
			break;

		case PCI_DEVICE_ID_INTEL_KNL_IMC_TOLHM:
			pvt->knl.pci_mc_info = pdev;
			break;

		case PCI_DEVICE_ID_INTEL_KNL_IMC_TA:
			pvt->pci_ta = pdev;
			break;

		default:
			sbridge_printk(KERN_ERR, "Unexpected device %d\n",
				pdev->device);
			break;
		}
	}

	if (!pvt->knl.pci_mc0  || !pvt->knl.pci_mc1 ||
	    !pvt->pci_sad0     || !pvt->pci_sad1    ||
	    !pvt->pci_ta) {
		goto enodev;
	}

	for (i = 0; i < KNL_MAX_CHANNELS; i++) {
		if (!pvt->knl.pci_channel[i]) {
			sbridge_printk(KERN_ERR, "Missing channel %d\n", i);
			goto enodev;
		}
	}

	for (i = 0; i < KNL_MAX_CHAS; i++) {
		if (!pvt->knl.pci_cha[i]) {
			sbridge_printk(KERN_ERR, "Missing CHA %d\n", i);
			goto enodev;
		}
	}

	return 0;

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

2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913
/****************************************************************************
			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;
2914
	enum hw_event_mc_err_type tp_event;
2915
	char *type, *optype, msg[256];
2916 2917 2918
	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);
2919
	bool recoverable;
2920 2921 2922 2923 2924 2925
	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;
2926
	u8  rank, socket, ha;
2927
	int rc, dimm;
2928
	char *area_type = NULL;
2929

2930
	if (pvt->info.type != SANDY_BRIDGE)
2931 2932 2933 2934
		recoverable = true;
	else
		recoverable = GET_BITFIELD(m->status, 56, 56);

2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946
	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;
	}
2947 2948

	/*
D
David Mackey 已提交
2949
	 * According with Table 15-9 of the Intel Architecture spec vol 3A,
2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963
	 * 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:
2964
			optype = "generic undef request error";
2965 2966
			break;
		case 1:
2967
			optype = "memory read error";
2968 2969
			break;
		case 2:
2970
			optype = "memory write error";
2971 2972
			break;
		case 3:
2973
			optype = "addr/cmd error";
2974 2975
			break;
		case 4:
2976
			optype = "memory scrubbing error";
2977 2978 2979 2980 2981 2982 2983
			break;
		default:
			optype = "reserved";
			break;
		}
	}

2984 2985 2986 2987
	/* Only decode errors with an valid address (ADDRV) */
	if (!GET_BITFIELD(m->status, 58, 58))
		return;

2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017
	if (pvt->info.type == KNIGHTS_LANDING) {
		if (channel == 14) {
			edac_dbg(0, "%s%s err_code:%04x:%04x EDRAM bank %d\n",
				overflow ? " OVERFLOW" : "",
				(uncorrected_error && recoverable)
				? " recoverable" : "",
				mscod, errcode,
				m->bank);
		} else {
			char A = *("A");

			channel = knl_channel_remap(channel);
			channel_mask = 1 << channel;
			snprintf(msg, sizeof(msg),
				"%s%s err_code:%04x:%04x channel:%d (DIMM_%c)",
				overflow ? " OVERFLOW" : "",
				(uncorrected_error && recoverable)
				? " recoverable" : " ",
				mscod, errcode, channel, A + channel);
			edac_mc_handle_error(tp_event, mci, core_err_cnt,
				m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0,
				channel, 0, -1,
				optype, msg);
		}
		return;
	} else {
		rc = get_memory_error_data(mci, m->addr, &socket, &ha,
				&channel_mask, &rank, &area_type, msg);
	}

3018
	if (rc < 0)
3019
		goto err_parsing;
3020 3021
	new_mci = get_mci_for_node_id(socket);
	if (!new_mci) {
3022 3023
		strcpy(msg, "Error: socket got corrupted!");
		goto err_parsing;
3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038
	}
	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;


	/*
3039 3040 3041 3042
	 * 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.
3043
	 */
3044 3045 3046
	if (!pvt->is_lockstep && !pvt->is_mirrored && !pvt->is_close_pg)
		channel = first_channel;

3047
	snprintf(msg, sizeof(msg),
3048
		 "%s%s area:%s err_code:%04x:%04x socket:%d ha:%d channel_mask:%ld rank:%d",
3049 3050 3051 3052
		 overflow ? " OVERFLOW" : "",
		 (uncorrected_error && recoverable) ? " recoverable" : "",
		 area_type,
		 mscod, errcode,
3053
		 socket, ha,
3054 3055
		 channel_mask,
		 rank);
3056

3057
	edac_dbg(0, "%s\n", msg);
3058

3059 3060
	/* FIXME: need support for channel mask */

3061 3062 3063
	if (channel == CHANNEL_UNSPECIFIED)
		channel = -1;

3064
	/* Call the helper to output message */
3065
	edac_mc_handle_error(tp_event, mci, core_err_cnt,
3066
			     m->addr >> PAGE_SHIFT, m->addr & ~PAGE_MASK, 0,
3067
			     4*ha+channel, dimm, -1,
3068
			     optype, msg);
3069 3070
	return;
err_parsing:
3071
	edac_mc_handle_error(tp_event, mci, core_err_cnt, 0, 0, 0,
3072
			     -1, -1, -1,
3073
			     msg, "");
3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135

}

/*
 *	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.
 */
3136 3137
static int sbridge_mce_check_error(struct notifier_block *nb, unsigned long val,
				   void *data)
3138
{
3139 3140 3141
	struct mce *mce = (struct mce *)data;
	struct mem_ctl_info *mci;
	struct sbridge_pvt *pvt;
3142
	char *type;
3143

3144 3145 3146
	if (get_edac_report_status() == EDAC_REPORTING_DISABLED)
		return NOTIFY_DONE;

3147 3148 3149 3150
	mci = get_mci_for_node_id(mce->socketid);
	if (!mci)
		return NOTIFY_BAD;
	pvt = mci->pvt_info;
3151 3152 3153 3154 3155 3156 3157 3158

	/*
	 * 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)
3159
		return NOTIFY_DONE;
3160

3161 3162 3163 3164 3165
	if (mce->mcgstatus & MCG_STATUS_MCIP)
		type = "Exception";
	else
		type = "Event";

3166
	sbridge_mc_printk(mci, KERN_DEBUG, "HANDLING MCE MEMORY ERROR\n");
3167

3168 3169 3170 3171 3172 3173
	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);
3174

3175 3176 3177
	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);
3178 3179 3180 3181 3182

	smp_rmb();
	if ((pvt->mce_out + 1) % MCE_LOG_LEN == pvt->mce_in) {
		smp_wmb();
		pvt->mce_overrun++;
3183
		return NOTIFY_DONE;
3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195
	}

	/* 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 */
3196
	return NOTIFY_STOP;
3197 3198
}

3199 3200 3201 3202
static struct notifier_block sbridge_mce_dec = {
	.notifier_call      = sbridge_mce_check_error,
};

3203 3204 3205 3206 3207 3208 3209 3210 3211 3212
/****************************************************************************
			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)) {
3213
		edac_dbg(0, "MC: dev = %p\n", &sbridge_dev->pdev[0]->dev);
3214 3215 3216 3217 3218 3219 3220

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

	pvt = mci->pvt_info;

3221 3222
	edac_dbg(0, "MC: mci = %p, dev = %p\n",
		 mci, &sbridge_dev->pdev[0]->dev);
3223 3224

	/* Remove MC sysfs nodes */
3225
	edac_mc_del_mc(mci->pdev);
3226

3227
	edac_dbg(1, "%s: free mci struct\n", mci->ctl_name);
3228 3229 3230 3231 3232
	kfree(mci->ctl_name);
	edac_mc_free(mci);
	sbridge_dev->mci = NULL;
}

3233
static int sbridge_register_mci(struct sbridge_dev *sbridge_dev, enum type type)
3234 3235
{
	struct mem_ctl_info *mci;
3236
	struct edac_mc_layer layers[2];
3237
	struct sbridge_pvt *pvt;
3238
	struct pci_dev *pdev = sbridge_dev->pdev[0];
3239
	int rc;
3240 3241

	/* Check the number of active and not disabled channels */
3242
	rc = check_if_ecc_is_active(sbridge_dev->bus, type);
3243 3244 3245 3246
	if (unlikely(rc < 0))
		return rc;

	/* allocate a new MC control structure */
3247
	layers[0].type = EDAC_MC_LAYER_CHANNEL;
3248 3249
	layers[0].size = type == KNIGHTS_LANDING ?
		KNL_MAX_CHANNELS : NUM_CHANNELS;
3250 3251
	layers[0].is_virt_csrow = false;
	layers[1].type = EDAC_MC_LAYER_SLOT;
3252
	layers[1].size = type == KNIGHTS_LANDING ? 1 : MAX_DIMMS;
3253
	layers[1].is_virt_csrow = true;
3254
	mci = edac_mc_alloc(sbridge_dev->mc, ARRAY_SIZE(layers), layers,
3255 3256
			    sizeof(*pvt));

3257 3258 3259
	if (unlikely(!mci))
		return -ENOMEM;

3260
	edac_dbg(0, "MC: mci = %p, dev = %p\n",
3261
		 mci, &pdev->dev);
3262 3263 3264 3265 3266 3267 3268 3269

	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;

3270 3271
	mci->mtype_cap = type == KNIGHTS_LANDING ?
		MEM_FLAG_DDR4 : MEM_FLAG_DDR3;
3272 3273 3274 3275
	mci->edac_ctl_cap = EDAC_FLAG_NONE;
	mci->edac_cap = EDAC_FLAG_NONE;
	mci->mod_name = "sbridge_edac.c";
	mci->mod_ver = SBRIDGE_REVISION;
3276
	mci->dev_name = pci_name(pdev);
3277 3278 3279 3280 3281
	mci->ctl_page_to_phys = NULL;

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

3282
	pvt->info.type = type;
3283 3284
	switch (type) {
	case IVY_BRIDGE:
3285 3286 3287 3288
		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;
3289
		pvt->info.get_memory_type = get_memory_type;
3290
		pvt->info.get_node_id = get_node_id;
3291
		pvt->info.rir_limit = rir_limit;
3292 3293 3294 3295
		pvt->info.sad_limit = sad_limit;
		pvt->info.interleave_mode = interleave_mode;
		pvt->info.show_interleave_mode = show_interleave_mode;
		pvt->info.dram_attr = dram_attr;
3296 3297 3298 3299
		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;
3300
		pvt->info.get_width = ibridge_get_width;
3301 3302 3303 3304 3305 3306
		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;
3307 3308
		break;
	case SANDY_BRIDGE:
3309 3310 3311 3312
		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;
3313
		pvt->info.get_memory_type = get_memory_type;
3314
		pvt->info.get_node_id = get_node_id;
3315
		pvt->info.rir_limit = rir_limit;
3316 3317 3318 3319
		pvt->info.sad_limit = sad_limit;
		pvt->info.interleave_mode = interleave_mode;
		pvt->info.show_interleave_mode = show_interleave_mode;
		pvt->info.dram_attr = dram_attr;
3320 3321 3322 3323
		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;
3324
		pvt->info.get_width = sbridge_get_width;
3325 3326 3327 3328 3329 3330
		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;
3331 3332 3333 3334 3335 3336 3337 3338 3339
		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;
3340 3341 3342 3343
		pvt->info.sad_limit = sad_limit;
		pvt->info.interleave_mode = interleave_mode;
		pvt->info.show_interleave_mode = show_interleave_mode;
		pvt->info.dram_attr = dram_attr;
3344 3345 3346 3347
		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;
3348
		pvt->info.get_width = ibridge_get_width;
3349
		mci->ctl_name = kasprintf(GFP_KERNEL, "Haswell Socket#%d", mci->mc_idx);
3350

3351 3352 3353 3354 3355
		/* Store pci devices at mci for faster access */
		rc = haswell_mci_bind_devs(mci, sbridge_dev);
		if (unlikely(rc < 0))
			goto fail0;
		break;
3356 3357 3358 3359 3360 3361 3362 3363
	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;
3364 3365 3366 3367
		pvt->info.sad_limit = sad_limit;
		pvt->info.interleave_mode = interleave_mode;
		pvt->info.show_interleave_mode = show_interleave_mode;
		pvt->info.dram_attr = dram_attr;
3368 3369 3370 3371
		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;
3372
		pvt->info.get_width = broadwell_get_width;
3373 3374 3375 3376 3377 3378 3379
		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;
3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395
	case KNIGHTS_LANDING:
		/* pvt->info.rankcfgr == ??? */
		pvt->info.get_tolm = knl_get_tolm;
		pvt->info.get_tohm = knl_get_tohm;
		pvt->info.dram_rule = knl_dram_rule;
		pvt->info.get_memory_type = knl_get_memory_type;
		pvt->info.get_node_id = knl_get_node_id;
		pvt->info.rir_limit = NULL;
		pvt->info.sad_limit = knl_sad_limit;
		pvt->info.interleave_mode = knl_interleave_mode;
		pvt->info.show_interleave_mode = knl_show_interleave_mode;
		pvt->info.dram_attr = dram_attr_knl;
		pvt->info.max_sad = ARRAY_SIZE(knl_dram_rule);
		pvt->info.interleave_list = knl_interleave_list;
		pvt->info.max_interleave = ARRAY_SIZE(knl_interleave_list);
		pvt->info.interleave_pkg = ibridge_interleave_pkg;
3396
		pvt->info.get_width = knl_get_width;
3397 3398 3399 3400 3401 3402 3403
		mci->ctl_name = kasprintf(GFP_KERNEL,
			"Knights Landing Socket#%d", mci->mc_idx);

		rc = knl_mci_bind_devs(mci, sbridge_dev);
		if (unlikely(rc < 0))
			goto fail0;
		break;
3404
	}
3405 3406 3407 3408 3409 3410

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

	/* record ptr to the generic device */
3411
	mci->pdev = &pdev->dev;
3412 3413 3414

	/* add this new MC control structure to EDAC's list of MCs */
	if (unlikely(edac_mc_add_mc(mci))) {
3415
		edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436
		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
 */

3437
static int sbridge_probe(struct pci_dev *pdev, const struct pci_device_id *id)
3438
{
3439
	int rc = -ENODEV;
3440 3441
	u8 mc, num_mc = 0;
	struct sbridge_dev *sbridge_dev;
3442
	enum type type = SANDY_BRIDGE;
3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455

	/* 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++;

3456 3457
	switch (pdev->device) {
	case PCI_DEVICE_ID_INTEL_IBRIDGE_IMC_HA0_TA:
3458 3459
		rc = sbridge_get_all_devices(&num_mc,
					pci_dev_descr_ibridge_table);
3460
		type = IVY_BRIDGE;
3461
		break;
3462
	case PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0:
3463 3464
		rc = sbridge_get_all_devices(&num_mc,
					pci_dev_descr_sbridge_table);
3465
		type = SANDY_BRIDGE;
3466 3467
		break;
	case PCI_DEVICE_ID_INTEL_HASWELL_IMC_HA0:
3468 3469
		rc = sbridge_get_all_devices(&num_mc,
					pci_dev_descr_haswell_table);
3470 3471
		type = HASWELL;
		break;
3472
	case PCI_DEVICE_ID_INTEL_BROADWELL_IMC_HA0:
3473 3474
		rc = sbridge_get_all_devices(&num_mc,
					pci_dev_descr_broadwell_table);
3475
		type = BROADWELL;
3476 3477 3478 3479 3480
	    break;
	case PCI_DEVICE_ID_INTEL_KNL_IMC_SAD0:
		rc = sbridge_get_all_devices_knl(&num_mc,
					pci_dev_descr_knl_table);
		type = KNIGHTS_LANDING;
3481
		break;
3482
	}
3483 3484
	if (unlikely(rc < 0)) {
		edac_dbg(0, "couldn't get all devices for 0x%x\n", pdev->device);
3485
		goto fail0;
3486 3487
	}

3488 3489 3490
	mc = 0;

	list_for_each_entry(sbridge_dev, &sbridge_edac_list, list) {
3491 3492
		edac_dbg(0, "Registering MC#%d (%d of %d)\n",
			 mc, mc + 1, num_mc);
3493

3494
		sbridge_dev->mc = mc++;
3495
		rc = sbridge_register_mci(sbridge_dev, type);
3496 3497 3498 3499
		if (unlikely(rc < 0))
			goto fail1;
	}

3500
	sbridge_printk(KERN_INFO, "%s\n", SBRIDGE_REVISION);
3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518

	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
 *
 */
3519
static void sbridge_remove(struct pci_dev *pdev)
3520 3521 3522
{
	struct sbridge_dev *sbridge_dev;

3523
	edac_dbg(0, "\n");
3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559

	/*
	 * 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,
3560
	.remove   = sbridge_remove,
3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571
	.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;

3572
	edac_dbg(2, "\n");
3573 3574 3575 3576 3577

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

	pci_rc = pci_register_driver(&sbridge_driver);
3578 3579
	if (pci_rc >= 0) {
		mce_register_decode_chain(&sbridge_mce_dec);
3580 3581
		if (get_edac_report_status() == EDAC_REPORTING_DISABLED)
			sbridge_printk(KERN_WARNING, "Loading driver, error reporting disabled.\n");
3582
		return 0;
3583
	}
3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596

	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)
{
3597
	edac_dbg(2, "\n");
3598
	pci_unregister_driver(&sbridge_driver);
3599
	mce_unregister_decode_chain(&sbridge_mce_dec);
3600 3601 3602 3603 3604 3605 3606 3607 3608
}

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");
3609
MODULE_AUTHOR("Mauro Carvalho Chehab");
3610
MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)");
3611
MODULE_DESCRIPTION("MC Driver for Intel Sandy Bridge and Ivy Bridge memory controllers - "
3612
		   SBRIDGE_REVISION);