hash_pgtable.c 13.5 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-or-later
2
3
4
5
6
7
8
/*
 * Copyright 2005, Paul Mackerras, IBM Corporation.
 * Copyright 2009, Benjamin Herrenschmidt, IBM Corporation.
 * Copyright 2015-2016, Aneesh Kumar K.V, IBM Corporation.
 */

#include <linux/sched.h>
9
#include <linux/mm_types.h>
10
#include <linux/mm.h>
11

12
#include <asm/pgalloc.h>
13
14
15
#include <asm/pgtable.h>
#include <asm/sections.h>
#include <asm/mmu.h>
16
17
#include <asm/tlb.h>

18
#include <mm/mmu_decl.h>
19

20
21
22
#define CREATE_TRACE_POINTS
#include <trace/events/thp.h>

23
24
25
26
#if H_PGTABLE_RANGE > (USER_VSID_RANGE * (TASK_SIZE_USER64 / TASK_CONTEXT_SIZE))
#warning Limited user VSID range means pagetable space is wasted
#endif

27
#ifdef CONFIG_SPARSEMEM_VMEMMAP
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
/*
 * vmemmap is the starting address of the virtual address space where
 * struct pages are allocated for all possible PFNs present on the system
 * including holes and bad memory (hence sparse). These virtual struct
 * pages are stored in sequence in this virtual address space irrespective
 * of the fact whether the corresponding PFN is valid or not. This achieves
 * constant relationship between address of struct page and its PFN.
 *
 * During boot or memory hotplug operation when a new memory section is
 * added, physical memory allocation (including hash table bolting) will
 * be performed for the set of struct pages which are part of the memory
 * section. This saves memory by not allocating struct pages for PFNs
 * which are not valid.
 *
 *		----------------------------------------------
 *		| PHYSICAL ALLOCATION OF VIRTUAL STRUCT PAGES|
 *		----------------------------------------------
 *
 *	   f000000000000000                  c000000000000000
 * vmemmap +--------------+                  +--------------+
 *  +      |  page struct | +--------------> |  page struct |
 *  |      +--------------+                  +--------------+
 *  |      |  page struct | +--------------> |  page struct |
 *  |      +--------------+ |                +--------------+
 *  |      |  page struct | +       +------> |  page struct |
 *  |      +--------------+         |        +--------------+
 *  |      |  page struct |         |   +--> |  page struct |
 *  |      +--------------+         |   |    +--------------+
 *  |      |  page struct |         |   |
 *  |      +--------------+         |   |
 *  |      |  page struct |         |   |
 *  |      +--------------+         |   |
 *  |      |  page struct |         |   |
 *  |      +--------------+         |   |
 *  |      |  page struct |         |   |
 *  |      +--------------+         |   |
 *  |      |  page struct | +-------+   |
 *  |      +--------------+             |
 *  |      |  page struct | +-----------+
 *  |      +--------------+
 *  |      |  page struct | No mapping
 *  |      +--------------+
 *  |      |  page struct | No mapping
 *  v      +--------------+
 *
 *		-----------------------------------------
 *		| RELATION BETWEEN STRUCT PAGES AND PFNS|
 *		-----------------------------------------
 *
 * vmemmap +--------------+                 +---------------+
 *  +      |  page struct | +-------------> |      PFN      |
 *  |      +--------------+                 +---------------+
 *  |      |  page struct | +-------------> |      PFN      |
 *  |      +--------------+                 +---------------+
 *  |      |  page struct | +-------------> |      PFN      |
 *  |      +--------------+                 +---------------+
 *  |      |  page struct | +-------------> |      PFN      |
 *  |      +--------------+                 +---------------+
 *  |      |              |
 *  |      +--------------+
 *  |      |              |
 *  |      +--------------+
 *  |      |              |
 *  |      +--------------+                 +---------------+
 *  |      |  page struct | +-------------> |      PFN      |
 *  |      +--------------+                 +---------------+
 *  |      |              |
 *  |      +--------------+
 *  |      |              |
 *  |      +--------------+                 +---------------+
 *  |      |  page struct | +-------------> |      PFN      |
 *  |      +--------------+                 +---------------+
 *  |      |  page struct | +-------------> |      PFN      |
 *  v      +--------------+                 +---------------+
 */
103
104
105
106
/*
 * On hash-based CPUs, the vmemmap is bolted in the hash table.
 *
 */
107
108
109
int __meminit hash__vmemmap_create_mapping(unsigned long start,
				       unsigned long page_size,
				       unsigned long phys)
110
{
111
112
113
	int rc;

	if ((start + page_size) >= H_VMEMMAP_END) {
114
		pr_warn("Outside the supported range\n");
115
116
117
118
119
120
		return -1;
	}

	rc = htab_bolt_mapping(start, start + page_size, phys,
			       pgprot_val(PAGE_KERNEL),
			       mmu_vmemmap_psize, mmu_kernel_ssize);
121
122
123
124
125
126
127
128
129
130
	if (rc < 0) {
		int rc2 = htab_remove_mapping(start, start + page_size,
					      mmu_vmemmap_psize,
					      mmu_kernel_ssize);
		BUG_ON(rc2 && (rc2 != -ENOENT));
	}
	return rc;
}

#ifdef CONFIG_MEMORY_HOTPLUG
131
132
void hash__vmemmap_remove_mapping(unsigned long start,
			      unsigned long page_size)
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
{
	int rc = htab_remove_mapping(start, start + page_size,
				     mmu_vmemmap_psize,
				     mmu_kernel_ssize);
	BUG_ON((rc < 0) && (rc != -ENOENT));
	WARN_ON(rc == -ENOENT);
}
#endif
#endif /* CONFIG_SPARSEMEM_VMEMMAP */

/*
 * map_kernel_page currently only called by __ioremap
 * map_kernel_page adds an entry to the ioremap page table
 * and adds an entry to the HPT, possibly bolting it
 */
148
int hash__map_kernel_page(unsigned long ea, unsigned long pa, pgprot_t prot)
149
150
151
152
153
154
{
	pgd_t *pgdp;
	pud_t *pudp;
	pmd_t *pmdp;
	pte_t *ptep;

155
	BUILD_BUG_ON(TASK_SIZE_USER64 > H_PGTABLE_RANGE);
156
157
158
159
160
161
162
163
164
165
166
	if (slab_is_available()) {
		pgdp = pgd_offset_k(ea);
		pudp = pud_alloc(&init_mm, pgdp, ea);
		if (!pudp)
			return -ENOMEM;
		pmdp = pmd_alloc(&init_mm, pudp, ea);
		if (!pmdp)
			return -ENOMEM;
		ptep = pte_alloc_kernel(pmdp, ea);
		if (!ptep)
			return -ENOMEM;
167
		set_pte_at(&init_mm, ea, ptep, pfn_pte(pa >> PAGE_SHIFT, prot));
168
169
170
171
172
173
174
	} else {
		/*
		 * If the mm subsystem is not fully up, we cannot create a
		 * linux page table entry for this mapping.  Simply bolt an
		 * entry in the hardware page table.
		 *
		 */
175
		if (htab_bolt_mapping(ea, ea + PAGE_SIZE, pa, pgprot_val(prot),
176
177
178
179
180
181
182
183
184
185
				      mmu_io_psize, mmu_kernel_ssize)) {
			printk(KERN_ERR "Failed to do bolted mapping IO "
			       "memory at %016lx !\n", pa);
			return -ENOMEM;
		}
	}

	smp_wmb();
	return 0;
}
186
187
188

#ifdef CONFIG_TRANSPARENT_HUGEPAGE

189
190
191
unsigned long hash__pmd_hugepage_update(struct mm_struct *mm, unsigned long addr,
				    pmd_t *pmdp, unsigned long clr,
				    unsigned long set)
192
193
194
195
196
{
	__be64 old_be, tmp;
	unsigned long old;

#ifdef CONFIG_DEBUG_VM
197
	WARN_ON(!hash__pmd_trans_huge(*pmdp) && !pmd_devmap(*pmdp));
198
	assert_spin_locked(pmd_lockptr(mm, pmdp));
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
#endif

	__asm__ __volatile__(
	"1:	ldarx	%0,0,%3\n\
		and.	%1,%0,%6\n\
		bne-	1b \n\
		andc	%1,%0,%4 \n\
		or	%1,%1,%7\n\
		stdcx.	%1,0,%3 \n\
		bne-	1b"
	: "=&r" (old_be), "=&r" (tmp), "=m" (*pmdp)
	: "r" (pmdp), "r" (cpu_to_be64(clr)), "m" (*pmdp),
	  "r" (cpu_to_be64(H_PAGE_BUSY)), "r" (cpu_to_be64(set))
	: "cc" );

	old = be64_to_cpu(old_be);

	trace_hugepage_update(addr, old, clr, set);
	if (old & H_PAGE_HASHPTE)
		hpte_do_hugepage_flush(mm, addr, pmdp, old);
	return old;
}

222
223
pmd_t hash__pmdp_collapse_flush(struct vm_area_struct *vma, unsigned long address,
			    pmd_t *pmdp)
224
225
226
227
228
{
	pmd_t pmd;

	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
	VM_BUG_ON(pmd_trans_huge(*pmdp));
229
	VM_BUG_ON(pmd_devmap(*pmdp));
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247

	pmd = *pmdp;
	pmd_clear(pmdp);
	/*
	 * Wait for all pending hash_page to finish. This is needed
	 * in case of subpage collapse. When we collapse normal pages
	 * to hugepage, we first clear the pmd, then invalidate all
	 * the PTE entries. The assumption here is that any low level
	 * page fault will see a none pmd and take the slow path that
	 * will wait on mmap_sem. But we could very well be in a
	 * hash_page with local ptep pointer value. Such a hash page
	 * can result in adding new HPTE entries for normal subpages.
	 * That means we could be modifying the page content as we
	 * copy them to a huge page. So wait for parallel hash_page
	 * to finish before invalidating HPTE entries. We can do this
	 * by sending an IPI to all the cpus and executing a dummy
	 * function there.
	 */
248
	serialize_against_pte_lookup(vma->vm_mm);
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
	/*
	 * Now invalidate the hpte entries in the range
	 * covered by pmd. This make sure we take a
	 * fault and will find the pmd as none, which will
	 * result in a major fault which takes mmap_sem and
	 * hence wait for collapse to complete. Without this
	 * the __collapse_huge_page_copy can result in copying
	 * the old content.
	 */
	flush_tlb_pmd_range(vma->vm_mm, &pmd, address);
	return pmd;
}

/*
 * We want to put the pgtable in pmd and use pgtable for tracking
 * the base page size hptes
 */
266
267
void hash__pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
				  pgtable_t pgtable)
268
269
{
	pgtable_t *pgtable_slot;
270
271

	assert_spin_locked(pmd_lockptr(mm, pmdp));
272
273
274
275
276
277
278
279
280
281
282
283
284
285
	/*
	 * we store the pgtable in the second half of PMD
	 */
	pgtable_slot = (pgtable_t *)pmdp + PTRS_PER_PMD;
	*pgtable_slot = pgtable;
	/*
	 * expose the deposited pgtable to other cpus.
	 * before we set the hugepage PTE at pmd level
	 * hash fault code looks at the deposted pgtable
	 * to store hash index values.
	 */
	smp_wmb();
}

286
pgtable_t hash__pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp)
287
288
289
290
{
	pgtable_t pgtable;
	pgtable_t *pgtable_slot;

291
292
	assert_spin_locked(pmd_lockptr(mm, pmdp));

293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
	pgtable_slot = (pgtable_t *)pmdp + PTRS_PER_PMD;
	pgtable = *pgtable_slot;
	/*
	 * Once we withdraw, mark the entry NULL.
	 */
	*pgtable_slot = NULL;
	/*
	 * We store HPTE information in the deposited PTE fragment.
	 * zero out the content on withdraw.
	 */
	memset(pgtable, 0, PTE_FRAG_SIZE);
	return pgtable;
}

/*
 * A linux hugepage PMD was changed and the corresponding hash table entries
 * neesd to be flushed.
 */
void hpte_do_hugepage_flush(struct mm_struct *mm, unsigned long addr,
			    pmd_t *pmdp, unsigned long old_pmd)
{
	int ssize;
	unsigned int psize;
	unsigned long vsid;
	unsigned long flags = 0;

	/* get the base page size,vsid and segment size */
#ifdef CONFIG_DEBUG_VM
	psize = get_slice_psize(mm, addr);
	BUG_ON(psize == MMU_PAGE_16M);
#endif
	if (old_pmd & H_PAGE_COMBO)
		psize = MMU_PAGE_4K;
	else
		psize = MMU_PAGE_64K;

	if (!is_kernel_addr(addr)) {
		ssize = user_segment_size(addr);
331
		vsid = get_user_vsid(&mm->context, addr, ssize);
332
333
334
335
336
337
		WARN_ON(vsid == 0);
	} else {
		vsid = get_kernel_vsid(addr, mmu_kernel_ssize);
		ssize = mmu_kernel_ssize;
	}

338
	if (mm_is_thread_local(mm))
339
340
341
342
343
		flags |= HPTE_LOCAL_UPDATE;

	return flush_hash_hugepage(vsid, addr, pmdp, psize, ssize, flags);
}

344
345
pmd_t hash__pmdp_huge_get_and_clear(struct mm_struct *mm,
				unsigned long addr, pmd_t *pmdp)
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
{
	pmd_t old_pmd;
	pgtable_t pgtable;
	unsigned long old;
	pgtable_t *pgtable_slot;

	old = pmd_hugepage_update(mm, addr, pmdp, ~0UL, 0);
	old_pmd = __pmd(old);
	/*
	 * We have pmd == none and we are holding page_table_lock.
	 * So we can safely go and clear the pgtable hash
	 * index info.
	 */
	pgtable_slot = (pgtable_t *)pmdp + PTRS_PER_PMD;
	pgtable = *pgtable_slot;
	/*
	 * Let's zero out old valid and hash index details
	 * hash fault look at them.
	 */
	memset(pgtable, 0, PTE_FRAG_SIZE);
	/*
367
	 * Serialize against find_current_mm_pte variants which does lock-less
368
369
370
371
372
373
	 * lookup in page tables with local interrupts disabled. For huge pages
	 * it casts pmd_t to pte_t. Since format of pte_t is different from
	 * pmd_t we want to prevent transit from pmd pointing to page table
	 * to pmd pointing to huge page (and back) while interrupts are disabled.
	 * We clear pmd to possibly replace it with page table pointer in
	 * different code paths. So make sure we wait for the parallel
374
	 * find_curren_mm_pte to finish.
375
	 */
376
	serialize_against_pte_lookup(mm);
377
378
379
	return old_pmd;
}

380
int hash__has_transparent_hugepage(void)
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
{

	if (!mmu_has_feature(MMU_FTR_16M_PAGE))
		return 0;
	/*
	 * We support THP only if PMD_SIZE is 16MB.
	 */
	if (mmu_psize_defs[MMU_PAGE_16M].shift != PMD_SHIFT)
		return 0;
	/*
	 * We need to make sure that we support 16MB hugepage in a segement
	 * with base page size 64K or 4K. We only enable THP with a PAGE_SIZE
	 * of 64K.
	 */
	/*
	 * If we have 64K HPTE, we will be using that by default
	 */
	if (mmu_psize_defs[MMU_PAGE_64K].shift &&
	    (mmu_psize_defs[MMU_PAGE_64K].penc[MMU_PAGE_16M] == -1))
		return 0;
	/*
	 * Ok we only have 4K HPTE
	 */
	if (mmu_psize_defs[MMU_PAGE_4K].penc[MMU_PAGE_16M] == -1)
		return 0;

	return 1;
}
409
410
EXPORT_SYMBOL_GPL(hash__has_transparent_hugepage);

411
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
412
413

#ifdef CONFIG_STRICT_KERNEL_RWX
414
415
static bool hash__change_memory_range(unsigned long start, unsigned long end,
				      unsigned long newpp)
416
417
418
419
420
421
422
{
	unsigned long idx;
	unsigned int step, shift;

	shift = mmu_psize_defs[mmu_linear_psize].shift;
	step = 1 << shift;

423
424
	start = ALIGN_DOWN(start, step);
	end = ALIGN(end, step); // aligns up
425

426
427
	if (start >= end)
		return false;
428

429
430
	pr_debug("Changing page protection on range 0x%lx-0x%lx, to 0x%lx, step 0x%x\n",
		 start, end, newpp, step);
431
432
433
434
435
436

	for (idx = start; idx < end; idx += step)
		/* Not sure if we can do much with the return value */
		mmu_hash_ops.hpte_updateboltedpp(newpp, idx, mmu_linear_psize,
							mmu_kernel_ssize);

437
438
439
440
441
442
443
444
445
446
447
	return true;
}

void hash__mark_rodata_ro(void)
{
	unsigned long start, end;

	start = (unsigned long)_stext;
	end = (unsigned long)__init_begin;

	WARN_ON(!hash__change_memory_range(start, end, PP_RXXX));
448
}
449
450
451
452
453
454
455
456
457
458
459
460

void hash__mark_initmem_nx(void)
{
	unsigned long start, end, pp;

	start = (unsigned long)__init_begin;
	end = (unsigned long)__init_end;

	pp = htab_convert_pte_flags(pgprot_val(PAGE_KERNEL));

	WARN_ON(!hash__change_memory_range(start, end, pp));
}
461
#endif