powerpc/mm: On PPC32, display 32 bits addresses in page table dump
[linux-2.6-block.git] / arch / powerpc / mm / dump_linuxpagetables.c
CommitLineData
8eb07b18
RG
1/*
2 * Copyright 2016, Rashmica Gupta, IBM Corp.
3 *
4 * This traverses the kernel pagetables and dumps the
5 * information about the used sections of memory to
6 * /sys/kernel/debug/kernel_pagetables.
7 *
8 * Derived from the arm64 implementation:
9 * Copyright (c) 2014, The Linux Foundation, Laura Abbott.
10 * (C) Copyright 2008 Intel Corporation, Arjan van de Ven.
11 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; version 2
15 * of the License.
16 */
17#include <linux/debugfs.h>
18#include <linux/fs.h>
19#include <linux/io.h>
20#include <linux/mm.h>
21#include <linux/sched.h>
22#include <linux/seq_file.h>
23#include <asm/fixmap.h>
24#include <asm/pgtable.h>
25#include <linux/const.h>
26#include <asm/page.h>
27#include <asm/pgalloc.h>
28
6c01bbd2
CL
29#ifdef CONFIG_PPC32
30#define KERN_VIRT_START 0
31#endif
32
8eb07b18
RG
33/*
34 * To visualise what is happening,
35 *
36 * - PTRS_PER_P** = how many entries there are in the corresponding P**
37 * - P**_SHIFT = how many bits of the address we use to index into the
38 * corresponding P**
39 * - P**_SIZE is how much memory we can access through the table - not the
40 * size of the table itself.
41 * P**={PGD, PUD, PMD, PTE}
42 *
43 *
44 * Each entry of the PGD points to a PUD. Each entry of a PUD points to a
45 * PMD. Each entry of a PMD points to a PTE. And every PTE entry points to
46 * a page.
47 *
48 * In the case where there are only 3 levels, the PUD is folded into the
49 * PGD: every PUD has only one entry which points to the PMD.
50 *
51 * The page dumper groups page table entries of the same type into a single
52 * description. It uses pg_state to track the range information while
53 * iterating over the PTE entries. When the continuity is broken it then
54 * dumps out a description of the range - ie PTEs that are virtually contiguous
55 * with the same PTE flags are chunked together. This is to make it clear how
56 * different areas of the kernel virtual memory are used.
57 *
58 */
59struct pg_state {
60 struct seq_file *seq;
61 const struct addr_marker *marker;
62 unsigned long start_address;
aaa22952
OH
63 unsigned long start_pa;
64 unsigned long last_pa;
8eb07b18
RG
65 unsigned int level;
66 u64 current_flags;
67};
68
69struct addr_marker {
70 unsigned long start_address;
71 const char *name;
72};
73
74static struct addr_marker address_markers[] = {
75 { 0, "Start of kernel VM" },
76 { 0, "vmalloc() Area" },
77 { 0, "vmalloc() End" },
6c01bbd2 78#ifdef CONFIG_PPC64
8eb07b18
RG
79 { 0, "isa I/O start" },
80 { 0, "isa I/O end" },
81 { 0, "phb I/O start" },
82 { 0, "phb I/O end" },
83 { 0, "I/O remap start" },
84 { 0, "I/O remap end" },
85 { 0, "vmemmap start" },
6c01bbd2
CL
86#else
87 { 0, "Early I/O remap start" },
88 { 0, "Early I/O remap end" },
89#ifdef CONFIG_NOT_COHERENT_CACHE
90 { 0, "Consistent mem start" },
91 { 0, "Consistent mem end" },
92#endif
93#ifdef CONFIG_HIGHMEM
94 { 0, "Highmem PTEs start" },
95 { 0, "Highmem PTEs end" },
96#endif
97 { 0, "Fixmap start" },
98 { 0, "Fixmap end" },
99#endif
8eb07b18
RG
100 { -1, NULL },
101};
102
103struct flag_info {
104 u64 mask;
105 u64 val;
106 const char *set;
107 const char *clear;
108 bool is_val;
109 int shift;
110};
111
112static const struct flag_info flag_array[] = {
113 {
114#ifdef CONFIG_PPC_STD_MMU_64
115 .mask = _PAGE_PRIVILEGED,
116 .val = 0,
117#else
118 .mask = _PAGE_USER,
119 .val = _PAGE_USER,
120#endif
121 .set = "user",
122 .clear = " ",
123 }, {
fd893fe5 124#if _PAGE_RO == 0
8eb07b18
RG
125 .mask = _PAGE_RW,
126 .val = _PAGE_RW,
fd893fe5
CL
127#else
128 .mask = _PAGE_RO,
129 .val = 0,
130#endif
8eb07b18
RG
131 .set = "rw",
132 .clear = "ro",
133 }, {
134 .mask = _PAGE_EXEC,
135 .val = _PAGE_EXEC,
136 .set = " X ",
137 .clear = " ",
138 }, {
139 .mask = _PAGE_PTE,
140 .val = _PAGE_PTE,
141 .set = "pte",
142 .clear = " ",
143 }, {
144 .mask = _PAGE_PRESENT,
145 .val = _PAGE_PRESENT,
146 .set = "present",
147 .clear = " ",
148 }, {
149#ifdef CONFIG_PPC_STD_MMU_64
150 .mask = H_PAGE_HASHPTE,
151 .val = H_PAGE_HASHPTE,
152#else
153 .mask = _PAGE_HASHPTE,
154 .val = _PAGE_HASHPTE,
155#endif
156 .set = "hpte",
157 .clear = " ",
158 }, {
159#ifndef CONFIG_PPC_STD_MMU_64
160 .mask = _PAGE_GUARDED,
161 .val = _PAGE_GUARDED,
162 .set = "guarded",
163 .clear = " ",
164 }, {
165#endif
166 .mask = _PAGE_DIRTY,
167 .val = _PAGE_DIRTY,
168 .set = "dirty",
169 .clear = " ",
170 }, {
171 .mask = _PAGE_ACCESSED,
172 .val = _PAGE_ACCESSED,
173 .set = "accessed",
174 .clear = " ",
175 }, {
176#ifndef CONFIG_PPC_STD_MMU_64
177 .mask = _PAGE_WRITETHRU,
178 .val = _PAGE_WRITETHRU,
179 .set = "write through",
180 .clear = " ",
181 }, {
182#endif
70538eaa 183#ifndef CONFIG_PPC_BOOK3S_64
8eb07b18
RG
184 .mask = _PAGE_NO_CACHE,
185 .val = _PAGE_NO_CACHE,
186 .set = "no cache",
187 .clear = " ",
188 }, {
70538eaa
OH
189#else
190 .mask = _PAGE_NON_IDEMPOTENT,
191 .val = _PAGE_NON_IDEMPOTENT,
192 .set = "non-idempotent",
193 .clear = " ",
194 }, {
195 .mask = _PAGE_TOLERANT,
196 .val = _PAGE_TOLERANT,
197 .set = "tolerant",
198 .clear = " ",
199 }, {
200#endif
dd5ac03e 201#ifdef CONFIG_PPC_BOOK3S_64
8eb07b18
RG
202 .mask = H_PAGE_BUSY,
203 .val = H_PAGE_BUSY,
204 .set = "busy",
205 }, {
206#ifdef CONFIG_PPC_64K_PAGES
207 .mask = H_PAGE_COMBO,
208 .val = H_PAGE_COMBO,
209 .set = "combo",
210 }, {
211 .mask = H_PAGE_4K_PFN,
212 .val = H_PAGE_4K_PFN,
213 .set = "4K_pfn",
214 }, {
215#endif
216 .mask = H_PAGE_F_GIX,
217 .val = H_PAGE_F_GIX,
218 .set = "f_gix",
219 .is_val = true,
220 .shift = H_PAGE_F_GIX_SHIFT,
221 }, {
222 .mask = H_PAGE_F_SECOND,
223 .val = H_PAGE_F_SECOND,
224 .set = "f_second",
225 }, {
dd5ac03e 226#endif
8eb07b18
RG
227 .mask = _PAGE_SPECIAL,
228 .val = _PAGE_SPECIAL,
229 .set = "special",
fd893fe5
CL
230 }, {
231 .mask = _PAGE_SHARED,
232 .val = _PAGE_SHARED,
233 .set = "shared",
8eb07b18
RG
234 }
235};
236
237struct pgtable_level {
238 const struct flag_info *flag;
239 size_t num;
240 u64 mask;
241};
242
243static struct pgtable_level pg_level[] = {
244 {
245 }, { /* pgd */
246 .flag = flag_array,
247 .num = ARRAY_SIZE(flag_array),
248 }, { /* pud */
249 .flag = flag_array,
250 .num = ARRAY_SIZE(flag_array),
251 }, { /* pmd */
252 .flag = flag_array,
253 .num = ARRAY_SIZE(flag_array),
254 }, { /* pte */
255 .flag = flag_array,
256 .num = ARRAY_SIZE(flag_array),
257 },
258};
259
260static void dump_flag_info(struct pg_state *st, const struct flag_info
261 *flag, u64 pte, int num)
262{
263 unsigned int i;
264
265 for (i = 0; i < num; i++, flag++) {
266 const char *s = NULL;
267 u64 val;
268
269 /* flag not defined so don't check it */
270 if (flag->mask == 0)
271 continue;
272 /* Some 'flags' are actually values */
273 if (flag->is_val) {
274 val = pte & flag->val;
275 if (flag->shift)
276 val = val >> flag->shift;
277 seq_printf(st->seq, " %s:%llx", flag->set, val);
278 } else {
279 if ((pte & flag->mask) == flag->val)
280 s = flag->set;
281 else
282 s = flag->clear;
283 if (s)
284 seq_printf(st->seq, " %s", s);
285 }
286 st->current_flags &= ~flag->mask;
287 }
288 if (st->current_flags != 0)
289 seq_printf(st->seq, " unknown flags:%llx", st->current_flags);
290}
291
292static void dump_addr(struct pg_state *st, unsigned long addr)
293{
294 static const char units[] = "KMGTPE";
295 const char *unit = units;
296 unsigned long delta;
297
78a18dbf 298#ifdef CONFIG_PPC64
aaa22952
OH
299 seq_printf(st->seq, "0x%016lx-0x%016lx ", st->start_address, addr-1);
300 seq_printf(st->seq, "0x%016lx ", st->start_pa);
78a18dbf
CL
301#else
302 seq_printf(st->seq, "0x%08lx-0x%08lx ", st->start_address, addr - 1);
303 seq_printf(st->seq, "0x%08lx ", st->start_pa);
304#endif
aaa22952 305
8eb07b18
RG
306 delta = (addr - st->start_address) >> 10;
307 /* Work out what appropriate unit to use */
308 while (!(delta & 1023) && unit[1]) {
309 delta >>= 10;
310 unit++;
311 }
312 seq_printf(st->seq, "%9lu%c", delta, *unit);
313
314}
315
316static void note_page(struct pg_state *st, unsigned long addr,
317 unsigned int level, u64 val)
318{
319 u64 flag = val & pg_level[level].mask;
aaa22952
OH
320 u64 pa = val & PTE_RPN_MASK;
321
8eb07b18
RG
322 /* At first no level is set */
323 if (!st->level) {
324 st->level = level;
325 st->current_flags = flag;
326 st->start_address = addr;
aaa22952
OH
327 st->start_pa = pa;
328 st->last_pa = pa;
8eb07b18
RG
329 seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
330 /*
331 * Dump the section of virtual memory when:
332 * - the PTE flags from one entry to the next differs.
333 * - we change levels in the tree.
334 * - the address is in a different section of memory and is thus
335 * used for a different purpose, regardless of the flags.
aaa22952 336 * - the pa of this page is not adjacent to the last inspected page
8eb07b18
RG
337 */
338 } else if (flag != st->current_flags || level != st->level ||
aaa22952
OH
339 addr >= st->marker[1].start_address ||
340 pa != st->last_pa + PAGE_SIZE) {
8eb07b18
RG
341
342 /* Check the PTE flags */
343 if (st->current_flags) {
344 dump_addr(st, addr);
345
346 /* Dump all the flags */
347 if (pg_level[st->level].flag)
348 dump_flag_info(st, pg_level[st->level].flag,
349 st->current_flags,
350 pg_level[st->level].num);
351
352 seq_puts(st->seq, "\n");
353 }
354
355 /*
356 * Address indicates we have passed the end of the
357 * current section of virtual memory
358 */
359 while (addr >= st->marker[1].start_address) {
360 st->marker++;
361 seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
362 }
363 st->start_address = addr;
aaa22952
OH
364 st->start_pa = pa;
365 st->last_pa = pa;
8eb07b18
RG
366 st->current_flags = flag;
367 st->level = level;
aaa22952
OH
368 } else {
369 st->last_pa = pa;
8eb07b18
RG
370 }
371}
372
373static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start)
374{
375 pte_t *pte = pte_offset_kernel(pmd, 0);
376 unsigned long addr;
377 unsigned int i;
378
379 for (i = 0; i < PTRS_PER_PTE; i++, pte++) {
380 addr = start + i * PAGE_SIZE;
381 note_page(st, addr, 4, pte_val(*pte));
382
383 }
384}
385
386static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
387{
388 pmd_t *pmd = pmd_offset(pud, 0);
389 unsigned long addr;
390 unsigned int i;
391
392 for (i = 0; i < PTRS_PER_PMD; i++, pmd++) {
393 addr = start + i * PMD_SIZE;
394 if (!pmd_none(*pmd))
395 /* pmd exists */
396 walk_pte(st, pmd, addr);
397 else
398 note_page(st, addr, 3, pmd_val(*pmd));
399 }
400}
401
402static void walk_pud(struct pg_state *st, pgd_t *pgd, unsigned long start)
403{
404 pud_t *pud = pud_offset(pgd, 0);
405 unsigned long addr;
406 unsigned int i;
407
408 for (i = 0; i < PTRS_PER_PUD; i++, pud++) {
409 addr = start + i * PUD_SIZE;
410 if (!pud_none(*pud))
411 /* pud exists */
412 walk_pmd(st, pud, addr);
413 else
414 note_page(st, addr, 2, pud_val(*pud));
415 }
416}
417
418static void walk_pagetables(struct pg_state *st)
419{
420 pgd_t *pgd = pgd_offset_k(0UL);
421 unsigned int i;
422 unsigned long addr;
423
424 /*
425 * Traverse the linux pagetable structure and dump pages that are in
426 * the hash pagetable.
427 */
428 for (i = 0; i < PTRS_PER_PGD; i++, pgd++) {
429 addr = KERN_VIRT_START + i * PGDIR_SIZE;
430 if (!pgd_none(*pgd))
431 /* pgd exists */
432 walk_pud(st, pgd, addr);
433 else
434 note_page(st, addr, 1, pgd_val(*pgd));
435 }
436}
437
438static void populate_markers(void)
439{
6c01bbd2
CL
440 int i = 0;
441
442 address_markers[i++].start_address = PAGE_OFFSET;
443 address_markers[i++].start_address = VMALLOC_START;
444 address_markers[i++].start_address = VMALLOC_END;
445#ifdef CONFIG_PPC64
446 address_markers[i++].start_address = ISA_IO_BASE;
447 address_markers[i++].start_address = ISA_IO_END;
448 address_markers[i++].start_address = PHB_IO_BASE;
449 address_markers[i++].start_address = PHB_IO_END;
450 address_markers[i++].start_address = IOREMAP_BASE;
451 address_markers[i++].start_address = IOREMAP_END;
8eb07b18 452#ifdef CONFIG_PPC_STD_MMU_64
6c01bbd2 453 address_markers[i++].start_address = H_VMEMMAP_BASE;
8eb07b18 454#else
6c01bbd2
CL
455 address_markers[i++].start_address = VMEMMAP_BASE;
456#endif
457#else /* !CONFIG_PPC64 */
458 address_markers[i++].start_address = ioremap_bot;
459 address_markers[i++].start_address = IOREMAP_TOP;
460#ifdef CONFIG_NOT_COHERENT_CACHE
461 address_markers[i++].start_address = IOREMAP_TOP;
462 address_markers[i++].start_address = IOREMAP_TOP +
463 CONFIG_CONSISTENT_SIZE;
464#endif
465#ifdef CONFIG_HIGHMEM
466 address_markers[i++].start_address = PKMAP_BASE;
467 address_markers[i++].start_address = PKMAP_ADDR(LAST_PKMAP);
8eb07b18 468#endif
6c01bbd2
CL
469 address_markers[i++].start_address = FIXADDR_START;
470 address_markers[i++].start_address = FIXADDR_TOP;
471#endif /* CONFIG_PPC64 */
8eb07b18
RG
472}
473
474static int ptdump_show(struct seq_file *m, void *v)
475{
476 struct pg_state st = {
477 .seq = m,
478 .start_address = KERN_VIRT_START,
479 .marker = address_markers,
480 };
481 /* Traverse kernel page tables */
482 walk_pagetables(&st);
483 note_page(&st, 0, 0, 0);
484 return 0;
485}
486
487
488static int ptdump_open(struct inode *inode, struct file *file)
489{
490 return single_open(file, ptdump_show, NULL);
491}
492
493static const struct file_operations ptdump_fops = {
494 .open = ptdump_open,
495 .read = seq_read,
496 .llseek = seq_lseek,
497 .release = single_release,
498};
499
500static void build_pgtable_complete_mask(void)
501{
502 unsigned int i, j;
503
504 for (i = 0; i < ARRAY_SIZE(pg_level); i++)
505 if (pg_level[i].flag)
506 for (j = 0; j < pg_level[i].num; j++)
507 pg_level[i].mask |= pg_level[i].flag[j].mask;
508}
509
510static int ptdump_init(void)
511{
512 struct dentry *debugfs_file;
513
514 populate_markers();
515 build_pgtable_complete_mask();
516 debugfs_file = debugfs_create_file("kernel_pagetables", 0400, NULL,
517 NULL, &ptdump_fops);
518 return debugfs_file ? 0 : -ENOMEM;
519}
520device_initcall(ptdump_init);