pat.c 14.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13
/*
 * Handle caching attributes in page tables (PAT)
 *
 * Authors: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
 *          Suresh B Siddha <suresh.b.siddha@intel.com>
 *
 * Loosely based on earlier PAT patchset from Eric Biederman and Andi Kleen.
 */

#include <linux/mm.h>
#include <linux/kernel.h>
#include <linux/gfp.h>
#include <linux/fs.h>
14
#include <linux/bootmem.h>
15 16
#include <linux/debugfs.h>
#include <linux/seq_file.h>
17 18 19 20

#include <asm/msr.h>
#include <asm/tlbflush.h>
#include <asm/processor.h>
21
#include <asm/page.h>
22 23 24 25 26 27
#include <asm/pgtable.h>
#include <asm/pat.h>
#include <asm/e820.h>
#include <asm/cacheflush.h>
#include <asm/fcntl.h>
#include <asm/mtrr.h>
28
#include <asm/io.h>
29

30
#ifdef CONFIG_X86_PAT
31
int __read_mostly pat_enabled = 1;
32

33
void __cpuinit pat_disable(char *reason)
34
{
35
	pat_enabled = 0;
36
	printk(KERN_INFO "%s\n", reason);
37 38
}

A
Andrew Morton 已提交
39
static int __init nopat(char *str)
40
{
41
	pat_disable("PAT support disabled.");
42 43
	return 0;
}
44 45 46
early_param("nopat", nopat);
#endif

47 48 49 50 51 52 53 54 55 56 57 58 59

static int debug_enable;
static int __init pat_debug_setup(char *str)
{
	debug_enable = 1;
	return 0;
}
__setup("debugpat", pat_debug_setup);

#define dprintk(fmt, arg...) \
	do { if (debug_enable) printk(KERN_INFO fmt, ##arg); } while (0)


60
static u64 __read_mostly boot_pat_state;
61 62 63 64 65 66 67 68 69 70

enum {
	PAT_UC = 0,		/* uncached */
	PAT_WC = 1,		/* Write combining */
	PAT_WT = 4,		/* Write Through */
	PAT_WP = 5,		/* Write Protected */
	PAT_WB = 6,		/* Write Back (default) */
	PAT_UC_MINUS = 7,	/* UC, but can be overriden by MTRR */
};

71
#define PAT(x, y)	((u64)PAT_ ## y << ((x)*8))
72 73 74 75 76

void pat_init(void)
{
	u64 pat;

77
	if (!pat_enabled)
78 79
		return;

80
	/* Paranoia check. */
81
	if (!cpu_has_pat && boot_pat_state) {
82
		/*
83
		 * If this happens we are on a secondary CPU, but
84 85
		 * switched to PAT on the boot CPU. We have no way to
		 * undo PAT.
86 87 88 89
		 */
		printk(KERN_ERR "PAT enabled, "
		       "but not supported by secondary CPU\n");
		BUG();
90
	}
91 92 93 94 95 96 97 98 99 100 101 102 103 104

	/* Set PWT to Write-Combining. All other bits stay the same */
	/*
	 * PTE encoding used in Linux:
	 *      PAT
	 *      |PCD
	 *      ||PWT
	 *      |||
	 *      000 WB		_PAGE_CACHE_WB
	 *      001 WC		_PAGE_CACHE_WC
	 *      010 UC-		_PAGE_CACHE_UC_MINUS
	 *      011 UC		_PAGE_CACHE_UC
	 * PAT bit unused
	 */
105 106
	pat = PAT(0, WB) | PAT(1, WC) | PAT(2, UC_MINUS) | PAT(3, UC) |
	      PAT(4, WB) | PAT(5, WC) | PAT(6, UC_MINUS) | PAT(7, UC);
107 108

	/* Boot CPU check */
109
	if (!boot_pat_state)
110 111 112 113 114 115 116 117 118 119 120 121
		rdmsrl(MSR_IA32_CR_PAT, boot_pat_state);

	wrmsrl(MSR_IA32_CR_PAT, pat);
	printk(KERN_INFO "x86 PAT enabled: cpu %d, old 0x%Lx, new 0x%Lx\n",
	       smp_processor_id(), boot_pat_state, pat);
}

#undef PAT

static char *cattr_name(unsigned long flags)
{
	switch (flags & _PAGE_CACHE_MASK) {
122 123 124 125 126
	case _PAGE_CACHE_UC:		return "uncached";
	case _PAGE_CACHE_UC_MINUS:	return "uncached-minus";
	case _PAGE_CACHE_WB:		return "write-back";
	case _PAGE_CACHE_WC:		return "write-combining";
	default:			return "broken";
127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163
	}
}

/*
 * The global memtype list keeps track of memory type for specific
 * physical memory areas. Conflicting memory types in different
 * mappings can cause CPU cache corruption. To avoid this we keep track.
 *
 * The list is sorted based on starting address and can contain multiple
 * entries for each address (this allows reference counting for overlapping
 * areas). All the aliases have the same cache attributes of course.
 * Zero attributes are represented as holes.
 *
 * Currently the data structure is a list because the number of mappings
 * are expected to be relatively small. If this should be a problem
 * it could be changed to a rbtree or similar.
 *
 * memtype_lock protects the whole list.
 */

struct memtype {
	u64 start;
	u64 end;
	unsigned long type;
	struct list_head nd;
};

static LIST_HEAD(memtype_list);
static DEFINE_SPINLOCK(memtype_lock); 	/* protects memtype list */

/*
 * Does intersection of PAT memory type and MTRR memory type and returns
 * the resulting memory type as PAT understands it.
 * (Type in pat and mtrr will not have same value)
 * The intersection is based on "Effective Memory Type" tables in IA-32
 * SDM vol 3a
 */
164
static unsigned long pat_x_mtrr_type(u64 start, u64 end, unsigned long req_type)
165
{
166 167 168 169
	/*
	 * Look for MTRR hint to get the effective type in case where PAT
	 * request is for WB.
	 */
170 171 172 173 174 175 176 177 178 179 180
	if (req_type == _PAGE_CACHE_WB) {
		u8 mtrr_type;

		mtrr_type = mtrr_type_lookup(start, end);
		if (mtrr_type == MTRR_TYPE_UNCACHABLE)
			return _PAGE_CACHE_UC;
		if (mtrr_type == MTRR_TYPE_WRCOMB)
			return _PAGE_CACHE_WC;
	}

	return req_type;
181 182
}

183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
static int chk_conflict(struct memtype *new, struct memtype *entry,
			unsigned long *type)
{
	if (new->type != entry->type) {
		if (type) {
			new->type = entry->type;
			*type = entry->type;
		} else
			goto conflict;
	}

	 /* check overlaps with more than one entry in the list */
	list_for_each_entry_continue(entry, &memtype_list, nd) {
		if (new->end <= entry->start)
			break;
		else if (new->type != entry->type)
			goto conflict;
	}
	return 0;

 conflict:
	printk(KERN_INFO "%s:%d conflicting memory types "
	       "%Lx-%Lx %s<->%s\n", current->comm, current->pid, new->start,
	       new->end, cattr_name(new->type), cattr_name(entry->type));
	return -EBUSY;
}

210 211 212
static struct memtype *cached_entry;
static u64 cached_start;

213 214 215 216 217 218 219 220 221 222
/*
 * req_type typically has one of the:
 * - _PAGE_CACHE_WB
 * - _PAGE_CACHE_WC
 * - _PAGE_CACHE_UC_MINUS
 * - _PAGE_CACHE_UC
 *
 * req_type will have a special case value '-1', when requester want to inherit
 * the memory type from mtrr (if WB), existing PAT, defaulting to UC_MINUS.
 *
223 224 225
 * If new_type is NULL, function will return an error if it cannot reserve the
 * region with req_type. If new_type is non-NULL, function will return
 * available type in new_type in case of no error. In case of any error
226 227
 * it will return a negative return value.
 */
228
int reserve_memtype(u64 start, u64 end, unsigned long req_type,
229
			unsigned long *new_type)
230
{
231
	struct memtype *new, *entry;
232
	unsigned long actual_type;
233
	struct list_head *where;
234 235
	int err = 0;

236 237
 	BUG_ON(start >= end); /* end is exclusive */

238
	if (!pat_enabled) {
239
		/* This is identical to page table setting without PAT */
240 241 242 243 244
		if (new_type) {
			if (req_type == -1)
				*new_type = _PAGE_CACHE_WB;
			else
				*new_type = req_type & _PAGE_CACHE_MASK;
245
		}
246 247 248 249
		return 0;
	}

	/* Low ISA region is always mapped WB in page table. No need to track */
250
	if (is_ISA_range(start, end - 1)) {
251 252
		if (new_type)
			*new_type = _PAGE_CACHE_WB;
253 254 255
		return 0;
	}

256 257
	if (req_type == -1) {
		/*
258 259 260 261
		 * Call mtrr_lookup to get the type hint. This is an
		 * optimization for /dev/mem mmap'ers into WB memory (BIOS
		 * tools and ACPI tools). Use WB request for WB memory and use
		 * UC_MINUS otherwise.
262 263 264
		 */
		u8 mtrr_type = mtrr_type_lookup(start, end);

265
		if (mtrr_type == MTRR_TYPE_WRBACK)
266
			actual_type = _PAGE_CACHE_WB;
267
		else
268
			actual_type = _PAGE_CACHE_UC_MINUS;
269 270 271
	} else
		actual_type = pat_x_mtrr_type(start, end,
					      req_type & _PAGE_CACHE_MASK);
272

273 274
	new  = kmalloc(sizeof(struct memtype), GFP_KERNEL);
	if (!new)
275 276
		return -ENOMEM;

277 278 279
	new->start = start;
	new->end = end;
	new->type = actual_type;
280

281 282
	if (new_type)
		*new_type = actual_type;
283 284 285

	spin_lock(&memtype_lock);

286 287 288 289 290
	if (cached_entry && start >= cached_start)
		entry = cached_entry;
	else
		entry = list_entry(&memtype_list, struct memtype, nd);

291
	/* Search for existing mapping that overlaps the current range */
292
	where = NULL;
293
	list_for_each_entry_continue(entry, &memtype_list, nd) {
294
		if (end <= entry->start) {
295
			where = entry->nd.prev;
296
			cached_entry = list_entry(where, struct memtype, nd);
297
			break;
298
		} else if (start <= entry->start) { /* end > entry->start */
299
			err = chk_conflict(new, entry, new_type);
300 301 302 303
			if (!err) {
				dprintk("Overlap at 0x%Lx-0x%Lx\n",
					entry->start, entry->end);
				where = entry->nd.prev;
304 305
				cached_entry = list_entry(where,
							struct memtype, nd);
306 307
			}
			break;
308
		} else if (start < entry->end) { /* start > entry->start */
309
			err = chk_conflict(new, entry, new_type);
310 311 312
			if (!err) {
				dprintk("Overlap at 0x%Lx-0x%Lx\n",
					entry->start, entry->end);
313 314 315 316 317 318 319 320 321 322 323 324 325 326
				cached_entry = list_entry(entry->nd.prev,
							struct memtype, nd);

				/*
				 * Move to right position in the linked
				 * list to add this new entry
				 */
				list_for_each_entry_continue(entry,
							&memtype_list, nd) {
					if (start <= entry->start) {
						where = entry->nd.prev;
						break;
					}
				}
327 328 329 330 331 332
			}
			break;
		}
	}

	if (err) {
333 334 335
		printk(KERN_INFO "reserve_memtype failed 0x%Lx-0x%Lx, "
		       "track %s, req %s\n",
		       start, end, cattr_name(new->type), cattr_name(req_type));
336
		kfree(new);
337 338 339 340
		spin_unlock(&memtype_lock);
		return err;
	}

341 342
	cached_start = start;

343 344 345
	if (where)
		list_add(&new->nd, where);
	else
346
		list_add_tail(&new->nd, &memtype_list);
347

348
	spin_unlock(&memtype_lock);
349 350 351 352 353

	dprintk("reserve_memtype added 0x%Lx-0x%Lx, track %s, req %s, ret %s\n",
		start, end, cattr_name(new->type), cattr_name(req_type),
		new_type ? cattr_name(*new_type) : "-");

354 355 356 357 358
	return err;
}

int free_memtype(u64 start, u64 end)
{
359
	struct memtype *entry;
360 361
	int err = -EINVAL;

362
	if (!pat_enabled)
363 364 365
		return 0;

	/* Low ISA region is always mapped WB. No need to track */
366
	if (is_ISA_range(start, end - 1))
367 368 369
		return 0;

	spin_lock(&memtype_lock);
370 371
	list_for_each_entry(entry, &memtype_list, nd) {
		if (entry->start == start && entry->end == end) {
372 373 374
			if (cached_entry == entry || cached_start == start)
				cached_entry = NULL;

375 376
			list_del(&entry->nd);
			kfree(entry);
377 378 379 380 381 382 383
			err = 0;
			break;
		}
	}
	spin_unlock(&memtype_lock);

	if (err) {
I
Ingo Molnar 已提交
384
		printk(KERN_INFO "%s:%d freeing invalid memtype %Lx-%Lx\n",
385 386
			current->comm, current->pid, start, end);
	}
387

388
	dprintk("free_memtype request 0x%Lx-0x%Lx\n", start, end);
389 390 391
	return err;
}

392

393 394 395 396 397 398 399 400
/*
 * /dev/mem mmap interface. The memtype used for mapping varies:
 * - Use UC for mappings with O_SYNC flag
 * - Without O_SYNC flag, if there is any conflict in reserve_memtype,
 *   inherit the memtype from existing mapping.
 * - Else use UC_MINUS memtype (for backward compatibility with existing
 *   X drivers.
 */
401 402 403 404 405 406
pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
				unsigned long size, pgprot_t vma_prot)
{
	return vma_prot;
}

407 408
#ifdef CONFIG_STRICT_DEVMEM
/* This check is done in drivers/char/mem.c in case of STRICT_DEVMEM*/
409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431
static inline int range_is_allowed(unsigned long pfn, unsigned long size)
{
	return 1;
}
#else
static inline int range_is_allowed(unsigned long pfn, unsigned long size)
{
	u64 from = ((u64)pfn) << PAGE_SHIFT;
	u64 to = from + size;
	u64 cursor = from;

	while (cursor < to) {
		if (!devmem_is_allowed(pfn)) {
			printk(KERN_INFO
		"Program %s tried to access /dev/mem between %Lx->%Lx.\n",
				current->comm, from, to);
			return 0;
		}
		cursor += PAGE_SIZE;
		pfn++;
	}
	return 1;
}
432
#endif /* CONFIG_STRICT_DEVMEM */
433

434 435 436
int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
				unsigned long size, pgprot_t *vma_prot)
{
437 438 439
	u64 offset = ((u64) pfn) << PAGE_SHIFT;
	unsigned long flags = _PAGE_CACHE_UC_MINUS;
	int retval;
440

441 442 443
	if (!range_is_allowed(pfn, size))
		return 0;

444
	if (file->f_flags & O_SYNC) {
445
		flags = _PAGE_CACHE_UC;
446 447 448 449 450 451 452 453 454 455 456
	}

#ifdef CONFIG_X86_32
	/*
	 * On the PPro and successors, the MTRRs are used to set
	 * memory types for physical addresses outside main memory,
	 * so blindly setting UC or PWT on those pages is wrong.
	 * For Pentiums and earlier, the surround logic should disable
	 * caching for the high addresses through the KEN pin, but
	 * we maintain the tradition of paranoia in this code.
	 */
457
	if (!pat_enabled &&
458 459 460 461 462
	    !(boot_cpu_has(X86_FEATURE_MTRR) ||
	      boot_cpu_has(X86_FEATURE_K6_MTRR) ||
	      boot_cpu_has(X86_FEATURE_CYRIX_ARR) ||
	      boot_cpu_has(X86_FEATURE_CENTAUR_MCR)) &&
	    (pfn << PAGE_SHIFT) >= __pa(high_memory)) {
463
		flags = _PAGE_CACHE_UC;
464 465 466
	}
#endif

467 468 469 470 471 472 473 474 475 476
	/*
	 * With O_SYNC, we can only take UC mapping. Fail if we cannot.
	 * Without O_SYNC, we want to get
	 * - WB for WB-able memory and no other conflicting mappings
	 * - UC_MINUS for non-WB-able memory with no other conflicting mappings
	 * - Inherit from confliting mappings otherwise
	 */
	if (flags != _PAGE_CACHE_UC_MINUS) {
		retval = reserve_memtype(offset, offset + size, flags, NULL);
	} else {
I
Ingo Molnar 已提交
477
		retval = reserve_memtype(offset, offset + size, -1, &flags);
478 479 480 481 482
	}

	if (retval < 0)
		return 0;

Y
Yinghai Lu 已提交
483 484
	if (((pfn < max_low_pfn_mapped) ||
	     (pfn >= (1UL<<(32 - PAGE_SHIFT)) && pfn < max_pfn_mapped)) &&
485
	    ioremap_change_attr((unsigned long)__va(offset), size, flags) < 0) {
486
		free_memtype(offset, offset + size);
I
Ingo Molnar 已提交
487
		printk(KERN_INFO
488 489 490
		"%s:%d /dev/mem ioremap_change_attr failed %s for %Lx-%Lx\n",
			current->comm, current->pid,
			cattr_name(flags),
491
			offset, (unsigned long long)(offset + size));
492 493 494 495 496
		return 0;
	}

	*vma_prot = __pgprot((pgprot_val(*vma_prot) & ~_PAGE_CACHE_MASK) |
			     flags);
497 498
	return 1;
}
499 500 501 502 503 504 505 506 507

void map_devmem(unsigned long pfn, unsigned long size, pgprot_t vma_prot)
{
	u64 addr = (u64)pfn << PAGE_SHIFT;
	unsigned long flags;
	unsigned long want_flags = (pgprot_val(vma_prot) & _PAGE_CACHE_MASK);

	reserve_memtype(addr, addr + size, want_flags, &flags);
	if (flags != want_flags) {
I
Ingo Molnar 已提交
508
		printk(KERN_INFO
509 510 511
		"%s:%d /dev/mem expected mapping type %s for %Lx-%Lx, got %s\n",
			current->comm, current->pid,
			cattr_name(want_flags),
512
			addr, (unsigned long long)(addr + size),
513 514 515 516 517 518 519 520 521 522 523
			cattr_name(flags));
	}
}

void unmap_devmem(unsigned long pfn, unsigned long size, pgprot_t vma_prot)
{
	u64 addr = (u64)pfn << PAGE_SHIFT;

	free_memtype(addr, addr + size);
}

524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608
#if defined(CONFIG_DEBUG_FS)

/* get Nth element of the linked list */
static struct memtype *memtype_get_idx(loff_t pos)
{
	struct memtype *list_node, *print_entry;
	int i = 1;

	print_entry  = kmalloc(sizeof(struct memtype), GFP_KERNEL);
	if (!print_entry)
		return NULL;

	spin_lock(&memtype_lock);
	list_for_each_entry(list_node, &memtype_list, nd) {
		if (pos == i) {
			*print_entry = *list_node;
			spin_unlock(&memtype_lock);
			return print_entry;
		}
		++i;
	}
	spin_unlock(&memtype_lock);
	kfree(print_entry);
	return NULL;
}

static void *memtype_seq_start(struct seq_file *seq, loff_t *pos)
{
	if (*pos == 0) {
		++*pos;
		seq_printf(seq, "PAT memtype list:\n");
	}

	return memtype_get_idx(*pos);
}

static void *memtype_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
	++*pos;
	return memtype_get_idx(*pos);
}

static void memtype_seq_stop(struct seq_file *seq, void *v)
{
}

static int memtype_seq_show(struct seq_file *seq, void *v)
{
	struct memtype *print_entry = (struct memtype *)v;

	seq_printf(seq, "%s @ 0x%Lx-0x%Lx\n", cattr_name(print_entry->type),
			print_entry->start, print_entry->end);
	kfree(print_entry);
	return 0;
}

static struct seq_operations memtype_seq_ops = {
	.start = memtype_seq_start,
	.next  = memtype_seq_next,
	.stop  = memtype_seq_stop,
	.show  = memtype_seq_show,
};

static int memtype_seq_open(struct inode *inode, struct file *file)
{
	return seq_open(file, &memtype_seq_ops);
}

static const struct file_operations memtype_fops = {
	.open    = memtype_seq_open,
	.read    = seq_read,
	.llseek  = seq_lseek,
	.release = seq_release,
};

static int __init pat_memtype_list_init(void)
{
	debugfs_create_file("pat_memtype_list", S_IRUSR, arch_debugfs_dir,
				NULL, &memtype_fops);
	return 0;
}

late_initcall(pat_memtype_list_init);

#endif /* CONFIG_DEBUG_FS */