Merge branch 'oprofile-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[deliverable/linux.git] / arch / x86 / mm / dump_pagetables.c
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 * (C) Copyright 2008 Intel Corporation
6 *
7 * Author: Arjan van de Ven <arjan@linux.intel.com>
8 *
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; version 2
12 * of the License.
13 */
14
15 #include <linux/debugfs.h>
16 #include <linux/mm.h>
17 #include <linux/module.h>
18 #include <linux/seq_file.h>
19
20 #include <asm/pgtable.h>
21
22 /*
23 * The dumper groups pagetable entries of the same type into one, and for
24 * that it needs to keep some state when walking, and flush this state
25 * when a "break" in the continuity is found.
26 */
27 struct pg_state {
28 int level;
29 pgprot_t current_prot;
30 unsigned long start_address;
31 unsigned long current_address;
32 const struct addr_marker *marker;
33 };
34
35 struct addr_marker {
36 unsigned long start_address;
37 const char *name;
38 };
39
40 /* Address space markers hints */
41 static struct addr_marker address_markers[] = {
42 { 0, "User Space" },
43 #ifdef CONFIG_X86_64
44 { 0x8000000000000000UL, "Kernel Space" },
45 { PAGE_OFFSET, "Low Kernel Mapping" },
46 { VMALLOC_START, "vmalloc() Area" },
47 { VMEMMAP_START, "Vmemmap" },
48 { __START_KERNEL_map, "High Kernel Mapping" },
49 { MODULES_VADDR, "Modules" },
50 { MODULES_END, "End Modules" },
51 #else
52 { PAGE_OFFSET, "Kernel Mapping" },
53 { 0/* VMALLOC_START */, "vmalloc() Area" },
54 { 0/*VMALLOC_END*/, "vmalloc() End" },
55 # ifdef CONFIG_HIGHMEM
56 { 0/*PKMAP_BASE*/, "Persisent kmap() Area" },
57 # endif
58 { 0/*FIXADDR_START*/, "Fixmap Area" },
59 #endif
60 { -1, NULL } /* End of list */
61 };
62
63 /* Multipliers for offsets within the PTEs */
64 #define PTE_LEVEL_MULT (PAGE_SIZE)
65 #define PMD_LEVEL_MULT (PTRS_PER_PTE * PTE_LEVEL_MULT)
66 #define PUD_LEVEL_MULT (PTRS_PER_PMD * PMD_LEVEL_MULT)
67 #define PGD_LEVEL_MULT (PTRS_PER_PUD * PUD_LEVEL_MULT)
68
69 /*
70 * Print a readable form of a pgprot_t to the seq_file
71 */
72 static void printk_prot(struct seq_file *m, pgprot_t prot, int level)
73 {
74 pgprotval_t pr = pgprot_val(prot);
75 static const char * const level_name[] =
76 { "cr3", "pgd", "pud", "pmd", "pte" };
77
78 if (!pgprot_val(prot)) {
79 /* Not present */
80 seq_printf(m, " ");
81 } else {
82 if (pr & _PAGE_USER)
83 seq_printf(m, "USR ");
84 else
85 seq_printf(m, " ");
86 if (pr & _PAGE_RW)
87 seq_printf(m, "RW ");
88 else
89 seq_printf(m, "ro ");
90 if (pr & _PAGE_PWT)
91 seq_printf(m, "PWT ");
92 else
93 seq_printf(m, " ");
94 if (pr & _PAGE_PCD)
95 seq_printf(m, "PCD ");
96 else
97 seq_printf(m, " ");
98
99 /* Bit 9 has a different meaning on level 3 vs 4 */
100 if (level <= 3) {
101 if (pr & _PAGE_PSE)
102 seq_printf(m, "PSE ");
103 else
104 seq_printf(m, " ");
105 } else {
106 if (pr & _PAGE_PAT)
107 seq_printf(m, "pat ");
108 else
109 seq_printf(m, " ");
110 }
111 if (pr & _PAGE_GLOBAL)
112 seq_printf(m, "GLB ");
113 else
114 seq_printf(m, " ");
115 if (pr & _PAGE_NX)
116 seq_printf(m, "NX ");
117 else
118 seq_printf(m, "x ");
119 }
120 seq_printf(m, "%s\n", level_name[level]);
121 }
122
123 /*
124 * On 64 bits, sign-extend the 48 bit address to 64 bit
125 */
126 static unsigned long normalize_addr(unsigned long u)
127 {
128 #ifdef CONFIG_X86_64
129 return (signed long)(u << 16) >> 16;
130 #else
131 return u;
132 #endif
133 }
134
135 /*
136 * This function gets called on a break in a continuous series
137 * of PTE entries; the next one is different so we need to
138 * print what we collected so far.
139 */
140 static void note_page(struct seq_file *m, struct pg_state *st,
141 pgprot_t new_prot, int level)
142 {
143 pgprotval_t prot, cur;
144 static const char units[] = "KMGTPE";
145
146 /*
147 * If we have a "break" in the series, we need to flush the state that
148 * we have now. "break" is either changing perms, levels or
149 * address space marker.
150 */
151 prot = pgprot_val(new_prot) & PTE_FLAGS_MASK;
152 cur = pgprot_val(st->current_prot) & PTE_FLAGS_MASK;
153
154 if (!st->level) {
155 /* First entry */
156 st->current_prot = new_prot;
157 st->level = level;
158 st->marker = address_markers;
159 seq_printf(m, "---[ %s ]---\n", st->marker->name);
160 } else if (prot != cur || level != st->level ||
161 st->current_address >= st->marker[1].start_address) {
162 const char *unit = units;
163 unsigned long delta;
164 int width = sizeof(unsigned long) * 2;
165
166 /*
167 * Now print the actual finished series
168 */
169 seq_printf(m, "0x%0*lx-0x%0*lx ",
170 width, st->start_address,
171 width, st->current_address);
172
173 delta = (st->current_address - st->start_address) >> 10;
174 while (!(delta & 1023) && unit[1]) {
175 delta >>= 10;
176 unit++;
177 }
178 seq_printf(m, "%9lu%c ", delta, *unit);
179 printk_prot(m, st->current_prot, st->level);
180
181 /*
182 * We print markers for special areas of address space,
183 * such as the start of vmalloc space etc.
184 * This helps in the interpretation.
185 */
186 if (st->current_address >= st->marker[1].start_address) {
187 st->marker++;
188 seq_printf(m, "---[ %s ]---\n", st->marker->name);
189 }
190
191 st->start_address = st->current_address;
192 st->current_prot = new_prot;
193 st->level = level;
194 }
195 }
196
197 static void walk_pte_level(struct seq_file *m, struct pg_state *st, pmd_t addr,
198 unsigned long P)
199 {
200 int i;
201 pte_t *start;
202
203 start = (pte_t *) pmd_page_vaddr(addr);
204 for (i = 0; i < PTRS_PER_PTE; i++) {
205 pgprot_t prot = pte_pgprot(*start);
206
207 st->current_address = normalize_addr(P + i * PTE_LEVEL_MULT);
208 note_page(m, st, prot, 4);
209 start++;
210 }
211 }
212
213 #if PTRS_PER_PMD > 1
214
215 static void walk_pmd_level(struct seq_file *m, struct pg_state *st, pud_t addr,
216 unsigned long P)
217 {
218 int i;
219 pmd_t *start;
220
221 start = (pmd_t *) pud_page_vaddr(addr);
222 for (i = 0; i < PTRS_PER_PMD; i++) {
223 st->current_address = normalize_addr(P + i * PMD_LEVEL_MULT);
224 if (!pmd_none(*start)) {
225 pgprotval_t prot = pmd_val(*start) & PTE_FLAGS_MASK;
226
227 if (pmd_large(*start) || !pmd_present(*start))
228 note_page(m, st, __pgprot(prot), 3);
229 else
230 walk_pte_level(m, st, *start,
231 P + i * PMD_LEVEL_MULT);
232 } else
233 note_page(m, st, __pgprot(0), 3);
234 start++;
235 }
236 }
237
238 #else
239 #define walk_pmd_level(m,s,a,p) walk_pte_level(m,s,__pmd(pud_val(a)),p)
240 #define pud_large(a) pmd_large(__pmd(pud_val(a)))
241 #define pud_none(a) pmd_none(__pmd(pud_val(a)))
242 #endif
243
244 #if PTRS_PER_PUD > 1
245
246 static void walk_pud_level(struct seq_file *m, struct pg_state *st, pgd_t addr,
247 unsigned long P)
248 {
249 int i;
250 pud_t *start;
251
252 start = (pud_t *) pgd_page_vaddr(addr);
253
254 for (i = 0; i < PTRS_PER_PUD; i++) {
255 st->current_address = normalize_addr(P + i * PUD_LEVEL_MULT);
256 if (!pud_none(*start)) {
257 pgprotval_t prot = pud_val(*start) & PTE_FLAGS_MASK;
258
259 if (pud_large(*start) || !pud_present(*start))
260 note_page(m, st, __pgprot(prot), 2);
261 else
262 walk_pmd_level(m, st, *start,
263 P + i * PUD_LEVEL_MULT);
264 } else
265 note_page(m, st, __pgprot(0), 2);
266
267 start++;
268 }
269 }
270
271 #else
272 #define walk_pud_level(m,s,a,p) walk_pmd_level(m,s,__pud(pgd_val(a)),p)
273 #define pgd_large(a) pud_large(__pud(pgd_val(a)))
274 #define pgd_none(a) pud_none(__pud(pgd_val(a)))
275 #endif
276
277 static void walk_pgd_level(struct seq_file *m)
278 {
279 #ifdef CONFIG_X86_64
280 pgd_t *start = (pgd_t *) &init_level4_pgt;
281 #else
282 pgd_t *start = swapper_pg_dir;
283 #endif
284 int i;
285 struct pg_state st;
286
287 memset(&st, 0, sizeof(st));
288
289 for (i = 0; i < PTRS_PER_PGD; i++) {
290 st.current_address = normalize_addr(i * PGD_LEVEL_MULT);
291 if (!pgd_none(*start)) {
292 pgprotval_t prot = pgd_val(*start) & PTE_FLAGS_MASK;
293
294 if (pgd_large(*start) || !pgd_present(*start))
295 note_page(m, &st, __pgprot(prot), 1);
296 else
297 walk_pud_level(m, &st, *start,
298 i * PGD_LEVEL_MULT);
299 } else
300 note_page(m, &st, __pgprot(0), 1);
301
302 start++;
303 }
304
305 /* Flush out the last page */
306 st.current_address = normalize_addr(PTRS_PER_PGD*PGD_LEVEL_MULT);
307 note_page(m, &st, __pgprot(0), 0);
308 }
309
310 static int ptdump_show(struct seq_file *m, void *v)
311 {
312 walk_pgd_level(m);
313 return 0;
314 }
315
316 static int ptdump_open(struct inode *inode, struct file *filp)
317 {
318 return single_open(filp, ptdump_show, NULL);
319 }
320
321 static const struct file_operations ptdump_fops = {
322 .open = ptdump_open,
323 .read = seq_read,
324 .llseek = seq_lseek,
325 .release = single_release,
326 };
327
328 static int pt_dump_init(void)
329 {
330 struct dentry *pe;
331
332 #ifdef CONFIG_X86_32
333 /* Not a compile-time constant on x86-32 */
334 address_markers[2].start_address = VMALLOC_START;
335 address_markers[3].start_address = VMALLOC_END;
336 # ifdef CONFIG_HIGHMEM
337 address_markers[4].start_address = PKMAP_BASE;
338 address_markers[5].start_address = FIXADDR_START;
339 # else
340 address_markers[4].start_address = FIXADDR_START;
341 # endif
342 #endif
343
344 pe = debugfs_create_file("kernel_page_tables", 0600, NULL, NULL,
345 &ptdump_fops);
346 if (!pe)
347 return -ENOMEM;
348
349 return 0;
350 }
351
352 __initcall(pt_dump_init);
353 MODULE_LICENSE("GPL");
354 MODULE_AUTHOR("Arjan van de Ven <arjan@linux.intel.com>");
355 MODULE_DESCRIPTION("Kernel debugging helper that dumps pagetables");
This page took 0.045931 seconds and 6 git commands to generate.