Loading...
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * Debug helper to dump the current kernel pagetables of the system
4 * so that we can see what the various memory ranges are set to.
5 *
6 * Derived from x86 implementation:
7 * (C) Copyright 2008 Intel Corporation
8 *
9 * Author: Arjan van de Ven <arjan@linux.intel.com>
10 */
11#include <linux/debugfs.h>
12#include <linux/fs.h>
13#include <linux/mm.h>
14#include <linux/seq_file.h>
15
16#include <asm/domain.h>
17#include <asm/fixmap.h>
18#include <asm/memory.h>
19#include <asm/pgtable.h>
20#include <asm/ptdump.h>
21
22static struct addr_marker address_markers[] = {
23 { MODULES_VADDR, "Modules" },
24 { PAGE_OFFSET, "Kernel Mapping" },
25 { 0, "vmalloc() Area" },
26 { VMALLOC_END, "vmalloc() End" },
27 { FIXADDR_START, "Fixmap Area" },
28 { VECTORS_BASE, "Vectors" },
29 { VECTORS_BASE + PAGE_SIZE * 2, "Vectors End" },
30 { -1, NULL },
31};
32
33#define pt_dump_seq_printf(m, fmt, args...) \
34({ \
35 if (m) \
36 seq_printf(m, fmt, ##args); \
37})
38
39#define pt_dump_seq_puts(m, fmt) \
40({ \
41 if (m) \
42 seq_printf(m, fmt); \
43})
44
45struct pg_state {
46 struct seq_file *seq;
47 const struct addr_marker *marker;
48 unsigned long start_address;
49 unsigned level;
50 u64 current_prot;
51 bool check_wx;
52 unsigned long wx_pages;
53 const char *current_domain;
54};
55
56struct prot_bits {
57 u64 mask;
58 u64 val;
59 const char *set;
60 const char *clear;
61 bool ro_bit;
62 bool nx_bit;
63};
64
65static const struct prot_bits pte_bits[] = {
66 {
67 .mask = L_PTE_USER,
68 .val = L_PTE_USER,
69 .set = "USR",
70 .clear = " ",
71 }, {
72 .mask = L_PTE_RDONLY,
73 .val = L_PTE_RDONLY,
74 .set = "ro",
75 .clear = "RW",
76 .ro_bit = true,
77 }, {
78 .mask = L_PTE_XN,
79 .val = L_PTE_XN,
80 .set = "NX",
81 .clear = "x ",
82 .nx_bit = true,
83 }, {
84 .mask = L_PTE_SHARED,
85 .val = L_PTE_SHARED,
86 .set = "SHD",
87 .clear = " ",
88 }, {
89 .mask = L_PTE_MT_MASK,
90 .val = L_PTE_MT_UNCACHED,
91 .set = "SO/UNCACHED",
92 }, {
93 .mask = L_PTE_MT_MASK,
94 .val = L_PTE_MT_BUFFERABLE,
95 .set = "MEM/BUFFERABLE/WC",
96 }, {
97 .mask = L_PTE_MT_MASK,
98 .val = L_PTE_MT_WRITETHROUGH,
99 .set = "MEM/CACHED/WT",
100 }, {
101 .mask = L_PTE_MT_MASK,
102 .val = L_PTE_MT_WRITEBACK,
103 .set = "MEM/CACHED/WBRA",
104#ifndef CONFIG_ARM_LPAE
105 }, {
106 .mask = L_PTE_MT_MASK,
107 .val = L_PTE_MT_MINICACHE,
108 .set = "MEM/MINICACHE",
109#endif
110 }, {
111 .mask = L_PTE_MT_MASK,
112 .val = L_PTE_MT_WRITEALLOC,
113 .set = "MEM/CACHED/WBWA",
114 }, {
115 .mask = L_PTE_MT_MASK,
116 .val = L_PTE_MT_DEV_SHARED,
117 .set = "DEV/SHARED",
118#ifndef CONFIG_ARM_LPAE
119 }, {
120 .mask = L_PTE_MT_MASK,
121 .val = L_PTE_MT_DEV_NONSHARED,
122 .set = "DEV/NONSHARED",
123#endif
124 }, {
125 .mask = L_PTE_MT_MASK,
126 .val = L_PTE_MT_DEV_WC,
127 .set = "DEV/WC",
128 }, {
129 .mask = L_PTE_MT_MASK,
130 .val = L_PTE_MT_DEV_CACHED,
131 .set = "DEV/CACHED",
132 },
133};
134
135static const struct prot_bits section_bits[] = {
136#ifdef CONFIG_ARM_LPAE
137 {
138 .mask = PMD_SECT_USER,
139 .val = PMD_SECT_USER,
140 .set = "USR",
141 }, {
142 .mask = L_PMD_SECT_RDONLY | PMD_SECT_AP2,
143 .val = L_PMD_SECT_RDONLY | PMD_SECT_AP2,
144 .set = "ro",
145 .clear = "RW",
146 .ro_bit = true,
147#elif __LINUX_ARM_ARCH__ >= 6
148 {
149 .mask = PMD_SECT_APX | PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
150 .val = PMD_SECT_APX | PMD_SECT_AP_WRITE,
151 .set = " ro",
152 .ro_bit = true,
153 }, {
154 .mask = PMD_SECT_APX | PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
155 .val = PMD_SECT_AP_WRITE,
156 .set = " RW",
157 }, {
158 .mask = PMD_SECT_APX | PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
159 .val = PMD_SECT_AP_READ,
160 .set = "USR ro",
161 }, {
162 .mask = PMD_SECT_APX | PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
163 .val = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
164 .set = "USR RW",
165#else /* ARMv4/ARMv5 */
166 /* These are approximate */
167 {
168 .mask = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
169 .val = 0,
170 .set = " ro",
171 .ro_bit = true,
172 }, {
173 .mask = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
174 .val = PMD_SECT_AP_WRITE,
175 .set = " RW",
176 }, {
177 .mask = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
178 .val = PMD_SECT_AP_READ,
179 .set = "USR ro",
180 }, {
181 .mask = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
182 .val = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
183 .set = "USR RW",
184#endif
185 }, {
186 .mask = PMD_SECT_XN,
187 .val = PMD_SECT_XN,
188 .set = "NX",
189 .clear = "x ",
190 .nx_bit = true,
191 }, {
192 .mask = PMD_SECT_S,
193 .val = PMD_SECT_S,
194 .set = "SHD",
195 .clear = " ",
196 },
197};
198
199struct pg_level {
200 const struct prot_bits *bits;
201 size_t num;
202 u64 mask;
203 const struct prot_bits *ro_bit;
204 const struct prot_bits *nx_bit;
205};
206
207static struct pg_level pg_level[] = {
208 {
209 }, { /* pgd */
210 }, { /* pud */
211 }, { /* pmd */
212 .bits = section_bits,
213 .num = ARRAY_SIZE(section_bits),
214 }, { /* pte */
215 .bits = pte_bits,
216 .num = ARRAY_SIZE(pte_bits),
217 },
218};
219
220static void dump_prot(struct pg_state *st, const struct prot_bits *bits, size_t num)
221{
222 unsigned i;
223
224 for (i = 0; i < num; i++, bits++) {
225 const char *s;
226
227 if ((st->current_prot & bits->mask) == bits->val)
228 s = bits->set;
229 else
230 s = bits->clear;
231
232 if (s)
233 pt_dump_seq_printf(st->seq, " %s", s);
234 }
235}
236
237static void note_prot_wx(struct pg_state *st, unsigned long addr)
238{
239 if (!st->check_wx)
240 return;
241 if ((st->current_prot & pg_level[st->level].ro_bit->mask) ==
242 pg_level[st->level].ro_bit->val)
243 return;
244 if ((st->current_prot & pg_level[st->level].nx_bit->mask) ==
245 pg_level[st->level].nx_bit->val)
246 return;
247
248 WARN_ONCE(1, "arm/mm: Found insecure W+X mapping at address %pS\n",
249 (void *)st->start_address);
250
251 st->wx_pages += (addr - st->start_address) / PAGE_SIZE;
252}
253
254static void note_page(struct pg_state *st, unsigned long addr,
255 unsigned int level, u64 val, const char *domain)
256{
257 static const char units[] = "KMGTPE";
258 u64 prot = val & pg_level[level].mask;
259
260 if (!st->level) {
261 st->level = level;
262 st->current_prot = prot;
263 st->current_domain = domain;
264 pt_dump_seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
265 } else if (prot != st->current_prot || level != st->level ||
266 domain != st->current_domain ||
267 addr >= st->marker[1].start_address) {
268 const char *unit = units;
269 unsigned long delta;
270
271 if (st->current_prot) {
272 note_prot_wx(st, addr);
273 pt_dump_seq_printf(st->seq, "0x%08lx-0x%08lx ",
274 st->start_address, addr);
275
276 delta = (addr - st->start_address) >> 10;
277 while (!(delta & 1023) && unit[1]) {
278 delta >>= 10;
279 unit++;
280 }
281 pt_dump_seq_printf(st->seq, "%9lu%c", delta, *unit);
282 if (st->current_domain)
283 pt_dump_seq_printf(st->seq, " %s",
284 st->current_domain);
285 if (pg_level[st->level].bits)
286 dump_prot(st, pg_level[st->level].bits, pg_level[st->level].num);
287 pt_dump_seq_printf(st->seq, "\n");
288 }
289
290 if (addr >= st->marker[1].start_address) {
291 st->marker++;
292 pt_dump_seq_printf(st->seq, "---[ %s ]---\n",
293 st->marker->name);
294 }
295 st->start_address = addr;
296 st->current_prot = prot;
297 st->current_domain = domain;
298 st->level = level;
299 }
300}
301
302static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start,
303 const char *domain)
304{
305 pte_t *pte = pte_offset_kernel(pmd, 0);
306 unsigned long addr;
307 unsigned i;
308
309 for (i = 0; i < PTRS_PER_PTE; i++, pte++) {
310 addr = start + i * PAGE_SIZE;
311 note_page(st, addr, 4, pte_val(*pte), domain);
312 }
313}
314
315static const char *get_domain_name(pmd_t *pmd)
316{
317#ifndef CONFIG_ARM_LPAE
318 switch (pmd_val(*pmd) & PMD_DOMAIN_MASK) {
319 case PMD_DOMAIN(DOMAIN_KERNEL):
320 return "KERNEL ";
321 case PMD_DOMAIN(DOMAIN_USER):
322 return "USER ";
323 case PMD_DOMAIN(DOMAIN_IO):
324 return "IO ";
325 case PMD_DOMAIN(DOMAIN_VECTORS):
326 return "VECTORS";
327 default:
328 return "unknown";
329 }
330#endif
331 return NULL;
332}
333
334static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
335{
336 pmd_t *pmd = pmd_offset(pud, 0);
337 unsigned long addr;
338 unsigned i;
339 const char *domain;
340
341 for (i = 0; i < PTRS_PER_PMD; i++, pmd++) {
342 addr = start + i * PMD_SIZE;
343 domain = get_domain_name(pmd);
344 if (pmd_none(*pmd) || pmd_large(*pmd) || !pmd_present(*pmd))
345 note_page(st, addr, 3, pmd_val(*pmd), domain);
346 else
347 walk_pte(st, pmd, addr, domain);
348
349 if (SECTION_SIZE < PMD_SIZE && pmd_large(pmd[1])) {
350 addr += SECTION_SIZE;
351 pmd++;
352 domain = get_domain_name(pmd);
353 note_page(st, addr, 3, pmd_val(*pmd), domain);
354 }
355 }
356}
357
358static void walk_pud(struct pg_state *st, pgd_t *pgd, unsigned long start)
359{
360 pud_t *pud = pud_offset(pgd, 0);
361 unsigned long addr;
362 unsigned i;
363
364 for (i = 0; i < PTRS_PER_PUD; i++, pud++) {
365 addr = start + i * PUD_SIZE;
366 if (!pud_none(*pud)) {
367 walk_pmd(st, pud, addr);
368 } else {
369 note_page(st, addr, 2, pud_val(*pud), NULL);
370 }
371 }
372}
373
374static void walk_pgd(struct pg_state *st, struct mm_struct *mm,
375 unsigned long start)
376{
377 pgd_t *pgd = pgd_offset(mm, 0UL);
378 unsigned i;
379 unsigned long addr;
380
381 for (i = 0; i < PTRS_PER_PGD; i++, pgd++) {
382 addr = start + i * PGDIR_SIZE;
383 if (!pgd_none(*pgd)) {
384 walk_pud(st, pgd, addr);
385 } else {
386 note_page(st, addr, 1, pgd_val(*pgd), NULL);
387 }
388 }
389}
390
391void ptdump_walk_pgd(struct seq_file *m, struct ptdump_info *info)
392{
393 struct pg_state st = {
394 .seq = m,
395 .marker = info->markers,
396 .check_wx = false,
397 };
398
399 walk_pgd(&st, info->mm, info->base_addr);
400 note_page(&st, 0, 0, 0, NULL);
401}
402
403static void ptdump_initialize(void)
404{
405 unsigned i, j;
406
407 for (i = 0; i < ARRAY_SIZE(pg_level); i++)
408 if (pg_level[i].bits)
409 for (j = 0; j < pg_level[i].num; j++) {
410 pg_level[i].mask |= pg_level[i].bits[j].mask;
411 if (pg_level[i].bits[j].ro_bit)
412 pg_level[i].ro_bit = &pg_level[i].bits[j];
413 if (pg_level[i].bits[j].nx_bit)
414 pg_level[i].nx_bit = &pg_level[i].bits[j];
415 }
416
417 address_markers[2].start_address = VMALLOC_START;
418}
419
420static struct ptdump_info kernel_ptdump_info = {
421 .mm = &init_mm,
422 .markers = address_markers,
423 .base_addr = 0,
424};
425
426void ptdump_check_wx(void)
427{
428 struct pg_state st = {
429 .seq = NULL,
430 .marker = (struct addr_marker[]) {
431 { 0, NULL},
432 { -1, NULL},
433 },
434 .check_wx = true,
435 };
436
437 walk_pgd(&st, &init_mm, 0);
438 note_page(&st, 0, 0, 0, NULL);
439 if (st.wx_pages)
440 pr_warn("Checked W+X mappings: FAILED, %lu W+X pages found\n",
441 st.wx_pages);
442 else
443 pr_info("Checked W+X mappings: passed, no W+X pages found\n");
444}
445
446static int ptdump_init(void)
447{
448 ptdump_initialize();
449 ptdump_debugfs_register(&kernel_ptdump_info, "kernel_page_tables");
450 return 0;
451}
452__initcall(ptdump_init);
1/*
2 * Debug helper to dump the current kernel pagetables of the system
3 * so that we can see what the various memory ranges are set to.
4 *
5 * Derived from x86 implementation:
6 * (C) Copyright 2008 Intel Corporation
7 *
8 * Author: Arjan van de Ven <arjan@linux.intel.com>
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; version 2
13 * of the License.
14 */
15#include <linux/debugfs.h>
16#include <linux/fs.h>
17#include <linux/mm.h>
18#include <linux/seq_file.h>
19
20#include <asm/fixmap.h>
21#include <asm/pgtable.h>
22
23struct addr_marker {
24 unsigned long start_address;
25 const char *name;
26};
27
28static struct addr_marker address_markers[] = {
29 { MODULES_VADDR, "Modules" },
30 { PAGE_OFFSET, "Kernel Mapping" },
31 { 0, "vmalloc() Area" },
32 { VMALLOC_END, "vmalloc() End" },
33 { FIXADDR_START, "Fixmap Area" },
34 { CONFIG_VECTORS_BASE, "Vectors" },
35 { CONFIG_VECTORS_BASE + PAGE_SIZE * 2, "Vectors End" },
36 { -1, NULL },
37};
38
39struct pg_state {
40 struct seq_file *seq;
41 const struct addr_marker *marker;
42 unsigned long start_address;
43 unsigned level;
44 u64 current_prot;
45};
46
47struct prot_bits {
48 u64 mask;
49 u64 val;
50 const char *set;
51 const char *clear;
52};
53
54static const struct prot_bits pte_bits[] = {
55 {
56 .mask = L_PTE_USER,
57 .val = L_PTE_USER,
58 .set = "USR",
59 .clear = " ",
60 }, {
61 .mask = L_PTE_RDONLY,
62 .val = L_PTE_RDONLY,
63 .set = "ro",
64 .clear = "RW",
65 }, {
66 .mask = L_PTE_XN,
67 .val = L_PTE_XN,
68 .set = "NX",
69 .clear = "x ",
70 }, {
71 .mask = L_PTE_SHARED,
72 .val = L_PTE_SHARED,
73 .set = "SHD",
74 .clear = " ",
75 }, {
76 .mask = L_PTE_MT_MASK,
77 .val = L_PTE_MT_UNCACHED,
78 .set = "SO/UNCACHED",
79 }, {
80 .mask = L_PTE_MT_MASK,
81 .val = L_PTE_MT_BUFFERABLE,
82 .set = "MEM/BUFFERABLE/WC",
83 }, {
84 .mask = L_PTE_MT_MASK,
85 .val = L_PTE_MT_WRITETHROUGH,
86 .set = "MEM/CACHED/WT",
87 }, {
88 .mask = L_PTE_MT_MASK,
89 .val = L_PTE_MT_WRITEBACK,
90 .set = "MEM/CACHED/WBRA",
91#ifndef CONFIG_ARM_LPAE
92 }, {
93 .mask = L_PTE_MT_MASK,
94 .val = L_PTE_MT_MINICACHE,
95 .set = "MEM/MINICACHE",
96#endif
97 }, {
98 .mask = L_PTE_MT_MASK,
99 .val = L_PTE_MT_WRITEALLOC,
100 .set = "MEM/CACHED/WBWA",
101 }, {
102 .mask = L_PTE_MT_MASK,
103 .val = L_PTE_MT_DEV_SHARED,
104 .set = "DEV/SHARED",
105#ifndef CONFIG_ARM_LPAE
106 }, {
107 .mask = L_PTE_MT_MASK,
108 .val = L_PTE_MT_DEV_NONSHARED,
109 .set = "DEV/NONSHARED",
110#endif
111 }, {
112 .mask = L_PTE_MT_MASK,
113 .val = L_PTE_MT_DEV_WC,
114 .set = "DEV/WC",
115 }, {
116 .mask = L_PTE_MT_MASK,
117 .val = L_PTE_MT_DEV_CACHED,
118 .set = "DEV/CACHED",
119 },
120};
121
122static const struct prot_bits section_bits[] = {
123#ifdef CONFIG_ARM_LPAE
124 {
125 .mask = PMD_SECT_USER,
126 .val = PMD_SECT_USER,
127 .set = "USR",
128 }, {
129 .mask = L_PMD_SECT_RDONLY,
130 .val = L_PMD_SECT_RDONLY,
131 .set = "ro",
132 .clear = "RW",
133#elif __LINUX_ARM_ARCH__ >= 6
134 {
135 .mask = PMD_SECT_APX | PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
136 .val = PMD_SECT_APX | PMD_SECT_AP_WRITE,
137 .set = " ro",
138 }, {
139 .mask = PMD_SECT_APX | PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
140 .val = PMD_SECT_AP_WRITE,
141 .set = " RW",
142 }, {
143 .mask = PMD_SECT_APX | PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
144 .val = PMD_SECT_AP_READ,
145 .set = "USR ro",
146 }, {
147 .mask = PMD_SECT_APX | PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
148 .val = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
149 .set = "USR RW",
150#else /* ARMv4/ARMv5 */
151 /* These are approximate */
152 {
153 .mask = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
154 .val = 0,
155 .set = " ro",
156 }, {
157 .mask = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
158 .val = PMD_SECT_AP_WRITE,
159 .set = " RW",
160 }, {
161 .mask = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
162 .val = PMD_SECT_AP_READ,
163 .set = "USR ro",
164 }, {
165 .mask = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
166 .val = PMD_SECT_AP_READ | PMD_SECT_AP_WRITE,
167 .set = "USR RW",
168#endif
169 }, {
170 .mask = PMD_SECT_XN,
171 .val = PMD_SECT_XN,
172 .set = "NX",
173 .clear = "x ",
174 }, {
175 .mask = PMD_SECT_S,
176 .val = PMD_SECT_S,
177 .set = "SHD",
178 .clear = " ",
179 },
180};
181
182struct pg_level {
183 const struct prot_bits *bits;
184 size_t num;
185 u64 mask;
186};
187
188static struct pg_level pg_level[] = {
189 {
190 }, { /* pgd */
191 }, { /* pud */
192 }, { /* pmd */
193 .bits = section_bits,
194 .num = ARRAY_SIZE(section_bits),
195 }, { /* pte */
196 .bits = pte_bits,
197 .num = ARRAY_SIZE(pte_bits),
198 },
199};
200
201static void dump_prot(struct pg_state *st, const struct prot_bits *bits, size_t num)
202{
203 unsigned i;
204
205 for (i = 0; i < num; i++, bits++) {
206 const char *s;
207
208 if ((st->current_prot & bits->mask) == bits->val)
209 s = bits->set;
210 else
211 s = bits->clear;
212
213 if (s)
214 seq_printf(st->seq, " %s", s);
215 }
216}
217
218static void note_page(struct pg_state *st, unsigned long addr, unsigned level, u64 val)
219{
220 static const char units[] = "KMGTPE";
221 u64 prot = val & pg_level[level].mask;
222
223 if (!st->level) {
224 st->level = level;
225 st->current_prot = prot;
226 seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
227 } else if (prot != st->current_prot || level != st->level ||
228 addr >= st->marker[1].start_address) {
229 const char *unit = units;
230 unsigned long delta;
231
232 if (st->current_prot) {
233 seq_printf(st->seq, "0x%08lx-0x%08lx ",
234 st->start_address, addr);
235
236 delta = (addr - st->start_address) >> 10;
237 while (!(delta & 1023) && unit[1]) {
238 delta >>= 10;
239 unit++;
240 }
241 seq_printf(st->seq, "%9lu%c", delta, *unit);
242 if (pg_level[st->level].bits)
243 dump_prot(st, pg_level[st->level].bits, pg_level[st->level].num);
244 seq_printf(st->seq, "\n");
245 }
246
247 if (addr >= st->marker[1].start_address) {
248 st->marker++;
249 seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
250 }
251 st->start_address = addr;
252 st->current_prot = prot;
253 st->level = level;
254 }
255}
256
257static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start)
258{
259 pte_t *pte = pte_offset_kernel(pmd, 0);
260 unsigned long addr;
261 unsigned i;
262
263 for (i = 0; i < PTRS_PER_PTE; i++, pte++) {
264 addr = start + i * PAGE_SIZE;
265 note_page(st, addr, 4, pte_val(*pte));
266 }
267}
268
269static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
270{
271 pmd_t *pmd = pmd_offset(pud, 0);
272 unsigned long addr;
273 unsigned i;
274
275 for (i = 0; i < PTRS_PER_PMD; i++, pmd++) {
276 addr = start + i * PMD_SIZE;
277 if (pmd_none(*pmd) || pmd_large(*pmd) || !pmd_present(*pmd))
278 note_page(st, addr, 3, pmd_val(*pmd));
279 else
280 walk_pte(st, pmd, addr);
281
282 if (SECTION_SIZE < PMD_SIZE && pmd_large(pmd[1]))
283 note_page(st, addr + SECTION_SIZE, 3, pmd_val(pmd[1]));
284 }
285}
286
287static void walk_pud(struct pg_state *st, pgd_t *pgd, unsigned long start)
288{
289 pud_t *pud = pud_offset(pgd, 0);
290 unsigned long addr;
291 unsigned i;
292
293 for (i = 0; i < PTRS_PER_PUD; i++, pud++) {
294 addr = start + i * PUD_SIZE;
295 if (!pud_none(*pud)) {
296 walk_pmd(st, pud, addr);
297 } else {
298 note_page(st, addr, 2, pud_val(*pud));
299 }
300 }
301}
302
303static void walk_pgd(struct seq_file *m)
304{
305 pgd_t *pgd = swapper_pg_dir;
306 struct pg_state st;
307 unsigned long addr;
308 unsigned i;
309
310 memset(&st, 0, sizeof(st));
311 st.seq = m;
312 st.marker = address_markers;
313
314 for (i = 0; i < PTRS_PER_PGD; i++, pgd++) {
315 addr = i * PGDIR_SIZE;
316 if (!pgd_none(*pgd)) {
317 walk_pud(&st, pgd, addr);
318 } else {
319 note_page(&st, addr, 1, pgd_val(*pgd));
320 }
321 }
322
323 note_page(&st, 0, 0, 0);
324}
325
326static int ptdump_show(struct seq_file *m, void *v)
327{
328 walk_pgd(m);
329 return 0;
330}
331
332static int ptdump_open(struct inode *inode, struct file *file)
333{
334 return single_open(file, ptdump_show, NULL);
335}
336
337static const struct file_operations ptdump_fops = {
338 .open = ptdump_open,
339 .read = seq_read,
340 .llseek = seq_lseek,
341 .release = single_release,
342};
343
344static int ptdump_init(void)
345{
346 struct dentry *pe;
347 unsigned i, j;
348
349 for (i = 0; i < ARRAY_SIZE(pg_level); i++)
350 if (pg_level[i].bits)
351 for (j = 0; j < pg_level[i].num; j++)
352 pg_level[i].mask |= pg_level[i].bits[j].mask;
353
354 address_markers[2].start_address = VMALLOC_START;
355
356 pe = debugfs_create_file("kernel_page_tables", 0400, NULL, NULL,
357 &ptdump_fops);
358 return pe ? 0 : -ENOMEM;
359}
360__initcall(ptdump_init);