powerpc/mm/radix: Display if mappings are exec or not
[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>
bfb9956a 19#include <linux/hugetlb.h>
8eb07b18
RG
20#include <linux/io.h>
21#include <linux/mm.h>
22#include <linux/sched.h>
23#include <linux/seq_file.h>
24#include <asm/fixmap.h>
25#include <asm/pgtable.h>
26#include <linux/const.h>
27#include <asm/page.h>
28#include <asm/pgalloc.h>
29
97026b5a
CL
30#include "dump_linuxpagetables.h"
31
6c01bbd2
CL
32#ifdef CONFIG_PPC32
33#define KERN_VIRT_START 0
34#endif
35
8eb07b18
RG
36/*
37 * To visualise what is happening,
38 *
39 * - PTRS_PER_P** = how many entries there are in the corresponding P**
40 * - P**_SHIFT = how many bits of the address we use to index into the
41 * corresponding P**
42 * - P**_SIZE is how much memory we can access through the table - not the
43 * size of the table itself.
44 * P**={PGD, PUD, PMD, PTE}
45 *
46 *
47 * Each entry of the PGD points to a PUD. Each entry of a PUD points to a
48 * PMD. Each entry of a PMD points to a PTE. And every PTE entry points to
49 * a page.
50 *
51 * In the case where there are only 3 levels, the PUD is folded into the
52 * PGD: every PUD has only one entry which points to the PMD.
53 *
54 * The page dumper groups page table entries of the same type into a single
55 * description. It uses pg_state to track the range information while
56 * iterating over the PTE entries. When the continuity is broken it then
57 * dumps out a description of the range - ie PTEs that are virtually contiguous
58 * with the same PTE flags are chunked together. This is to make it clear how
59 * different areas of the kernel virtual memory are used.
60 *
61 */
62struct pg_state {
63 struct seq_file *seq;
64 const struct addr_marker *marker;
65 unsigned long start_address;
aaa22952
OH
66 unsigned long start_pa;
67 unsigned long last_pa;
8eb07b18
RG
68 unsigned int level;
69 u64 current_flags;
70};
71
72struct addr_marker {
73 unsigned long start_address;
74 const char *name;
75};
76
77static struct addr_marker address_markers[] = {
78 { 0, "Start of kernel VM" },
79 { 0, "vmalloc() Area" },
80 { 0, "vmalloc() End" },
6c01bbd2 81#ifdef CONFIG_PPC64
8eb07b18
RG
82 { 0, "isa I/O start" },
83 { 0, "isa I/O end" },
84 { 0, "phb I/O start" },
85 { 0, "phb I/O end" },
86 { 0, "I/O remap start" },
87 { 0, "I/O remap end" },
88 { 0, "vmemmap start" },
6c01bbd2
CL
89#else
90 { 0, "Early I/O remap start" },
91 { 0, "Early I/O remap end" },
92#ifdef CONFIG_NOT_COHERENT_CACHE
93 { 0, "Consistent mem start" },
94 { 0, "Consistent mem end" },
95#endif
96#ifdef CONFIG_HIGHMEM
97 { 0, "Highmem PTEs start" },
98 { 0, "Highmem PTEs end" },
99#endif
100 { 0, "Fixmap start" },
101 { 0, "Fixmap end" },
102#endif
8eb07b18
RG
103 { -1, NULL },
104};
105
8eb07b18
RG
106static void dump_flag_info(struct pg_state *st, const struct flag_info
107 *flag, u64 pte, int num)
108{
109 unsigned int i;
110
111 for (i = 0; i < num; i++, flag++) {
112 const char *s = NULL;
113 u64 val;
114
115 /* flag not defined so don't check it */
116 if (flag->mask == 0)
117 continue;
118 /* Some 'flags' are actually values */
119 if (flag->is_val) {
120 val = pte & flag->val;
121 if (flag->shift)
122 val = val >> flag->shift;
123 seq_printf(st->seq, " %s:%llx", flag->set, val);
124 } else {
125 if ((pte & flag->mask) == flag->val)
126 s = flag->set;
127 else
128 s = flag->clear;
129 if (s)
130 seq_printf(st->seq, " %s", s);
131 }
132 st->current_flags &= ~flag->mask;
133 }
134 if (st->current_flags != 0)
135 seq_printf(st->seq, " unknown flags:%llx", st->current_flags);
136}
137
138static void dump_addr(struct pg_state *st, unsigned long addr)
139{
140 static const char units[] = "KMGTPE";
141 const char *unit = units;
142 unsigned long delta;
143
78a18dbf 144#ifdef CONFIG_PPC64
aaa22952
OH
145 seq_printf(st->seq, "0x%016lx-0x%016lx ", st->start_address, addr-1);
146 seq_printf(st->seq, "0x%016lx ", st->start_pa);
78a18dbf
CL
147#else
148 seq_printf(st->seq, "0x%08lx-0x%08lx ", st->start_address, addr - 1);
149 seq_printf(st->seq, "0x%08lx ", st->start_pa);
150#endif
aaa22952 151
8eb07b18
RG
152 delta = (addr - st->start_address) >> 10;
153 /* Work out what appropriate unit to use */
154 while (!(delta & 1023) && unit[1]) {
155 delta >>= 10;
156 unit++;
157 }
158 seq_printf(st->seq, "%9lu%c", delta, *unit);
159
160}
161
162static void note_page(struct pg_state *st, unsigned long addr,
163 unsigned int level, u64 val)
164{
165 u64 flag = val & pg_level[level].mask;
aaa22952
OH
166 u64 pa = val & PTE_RPN_MASK;
167
8eb07b18
RG
168 /* At first no level is set */
169 if (!st->level) {
170 st->level = level;
171 st->current_flags = flag;
172 st->start_address = addr;
aaa22952
OH
173 st->start_pa = pa;
174 st->last_pa = pa;
8eb07b18
RG
175 seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
176 /*
177 * Dump the section of virtual memory when:
178 * - the PTE flags from one entry to the next differs.
179 * - we change levels in the tree.
180 * - the address is in a different section of memory and is thus
181 * used for a different purpose, regardless of the flags.
aaa22952 182 * - the pa of this page is not adjacent to the last inspected page
8eb07b18
RG
183 */
184 } else if (flag != st->current_flags || level != st->level ||
aaa22952
OH
185 addr >= st->marker[1].start_address ||
186 pa != st->last_pa + PAGE_SIZE) {
8eb07b18
RG
187
188 /* Check the PTE flags */
189 if (st->current_flags) {
190 dump_addr(st, addr);
191
192 /* Dump all the flags */
193 if (pg_level[st->level].flag)
194 dump_flag_info(st, pg_level[st->level].flag,
195 st->current_flags,
196 pg_level[st->level].num);
197
aae85e3c 198 seq_putc(st->seq, '\n');
8eb07b18
RG
199 }
200
201 /*
202 * Address indicates we have passed the end of the
203 * current section of virtual memory
204 */
205 while (addr >= st->marker[1].start_address) {
206 st->marker++;
207 seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
208 }
209 st->start_address = addr;
aaa22952
OH
210 st->start_pa = pa;
211 st->last_pa = pa;
8eb07b18
RG
212 st->current_flags = flag;
213 st->level = level;
aaa22952
OH
214 } else {
215 st->last_pa = pa;
8eb07b18
RG
216 }
217}
218
219static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start)
220{
221 pte_t *pte = pte_offset_kernel(pmd, 0);
222 unsigned long addr;
223 unsigned int i;
224
225 for (i = 0; i < PTRS_PER_PTE; i++, pte++) {
226 addr = start + i * PAGE_SIZE;
227 note_page(st, addr, 4, pte_val(*pte));
228
229 }
230}
231
232static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
233{
234 pmd_t *pmd = pmd_offset(pud, 0);
235 unsigned long addr;
236 unsigned int i;
237
238 for (i = 0; i < PTRS_PER_PMD; i++, pmd++) {
239 addr = start + i * PMD_SIZE;
bfb9956a 240 if (!pmd_none(*pmd) && !pmd_huge(*pmd))
8eb07b18
RG
241 /* pmd exists */
242 walk_pte(st, pmd, addr);
243 else
244 note_page(st, addr, 3, pmd_val(*pmd));
245 }
246}
247
248static void walk_pud(struct pg_state *st, pgd_t *pgd, unsigned long start)
249{
250 pud_t *pud = pud_offset(pgd, 0);
251 unsigned long addr;
252 unsigned int i;
253
254 for (i = 0; i < PTRS_PER_PUD; i++, pud++) {
255 addr = start + i * PUD_SIZE;
bfb9956a 256 if (!pud_none(*pud) && !pud_huge(*pud))
8eb07b18
RG
257 /* pud exists */
258 walk_pmd(st, pud, addr);
259 else
260 note_page(st, addr, 2, pud_val(*pud));
261 }
262}
263
264static void walk_pagetables(struct pg_state *st)
265{
266 pgd_t *pgd = pgd_offset_k(0UL);
267 unsigned int i;
268 unsigned long addr;
269
270 /*
271 * Traverse the linux pagetable structure and dump pages that are in
272 * the hash pagetable.
273 */
274 for (i = 0; i < PTRS_PER_PGD; i++, pgd++) {
275 addr = KERN_VIRT_START + i * PGDIR_SIZE;
bfb9956a 276 if (!pgd_none(*pgd) && !pgd_huge(*pgd))
8eb07b18
RG
277 /* pgd exists */
278 walk_pud(st, pgd, addr);
279 else
280 note_page(st, addr, 1, pgd_val(*pgd));
281 }
282}
283
284static void populate_markers(void)
285{
6c01bbd2
CL
286 int i = 0;
287
288 address_markers[i++].start_address = PAGE_OFFSET;
289 address_markers[i++].start_address = VMALLOC_START;
290 address_markers[i++].start_address = VMALLOC_END;
291#ifdef CONFIG_PPC64
292 address_markers[i++].start_address = ISA_IO_BASE;
293 address_markers[i++].start_address = ISA_IO_END;
294 address_markers[i++].start_address = PHB_IO_BASE;
295 address_markers[i++].start_address = PHB_IO_END;
296 address_markers[i++].start_address = IOREMAP_BASE;
297 address_markers[i++].start_address = IOREMAP_END;
4e003747 298#ifdef CONFIG_PPC_BOOK3S_64
6c01bbd2 299 address_markers[i++].start_address = H_VMEMMAP_BASE;
8eb07b18 300#else
6c01bbd2
CL
301 address_markers[i++].start_address = VMEMMAP_BASE;
302#endif
303#else /* !CONFIG_PPC64 */
304 address_markers[i++].start_address = ioremap_bot;
305 address_markers[i++].start_address = IOREMAP_TOP;
306#ifdef CONFIG_NOT_COHERENT_CACHE
307 address_markers[i++].start_address = IOREMAP_TOP;
308 address_markers[i++].start_address = IOREMAP_TOP +
309 CONFIG_CONSISTENT_SIZE;
310#endif
311#ifdef CONFIG_HIGHMEM
312 address_markers[i++].start_address = PKMAP_BASE;
313 address_markers[i++].start_address = PKMAP_ADDR(LAST_PKMAP);
8eb07b18 314#endif
6c01bbd2
CL
315 address_markers[i++].start_address = FIXADDR_START;
316 address_markers[i++].start_address = FIXADDR_TOP;
317#endif /* CONFIG_PPC64 */
8eb07b18
RG
318}
319
320static int ptdump_show(struct seq_file *m, void *v)
321{
322 struct pg_state st = {
323 .seq = m,
324 .start_address = KERN_VIRT_START,
325 .marker = address_markers,
326 };
327 /* Traverse kernel page tables */
328 walk_pagetables(&st);
329 note_page(&st, 0, 0, 0);
330 return 0;
331}
332
333
334static int ptdump_open(struct inode *inode, struct file *file)
335{
336 return single_open(file, ptdump_show, NULL);
337}
338
339static const struct file_operations ptdump_fops = {
340 .open = ptdump_open,
341 .read = seq_read,
342 .llseek = seq_lseek,
343 .release = single_release,
344};
345
346static void build_pgtable_complete_mask(void)
347{
348 unsigned int i, j;
349
350 for (i = 0; i < ARRAY_SIZE(pg_level); i++)
351 if (pg_level[i].flag)
352 for (j = 0; j < pg_level[i].num; j++)
353 pg_level[i].mask |= pg_level[i].flag[j].mask;
354}
355
356static int ptdump_init(void)
357{
358 struct dentry *debugfs_file;
359
360 populate_markers();
361 build_pgtable_complete_mask();
2505820f 362 debugfs_file = debugfs_create_file("kernel_page_tables", 0400, NULL,
8eb07b18
RG
363 NULL, &ptdump_fops);
364 return debugfs_file ? 0 : -ENOMEM;
365}
366device_initcall(ptdump_init);