Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
[cris-mirror.git] / arch / arm64 / mm / dump.c
blob65dfc8571bf8397c3f2a6297d21b5112794461e1
1 /*
2 * Copyright (c) 2014, The Linux Foundation. All rights reserved.
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.
6 * Derived from x86 and arm implementation:
7 * (C) Copyright 2008 Intel Corporation
9 * Author: Arjan van de Ven <arjan@linux.intel.com>
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public License
13 * as published by the Free Software Foundation; version 2
14 * of the License.
16 #include <linux/debugfs.h>
17 #include <linux/errno.h>
18 #include <linux/fs.h>
19 #include <linux/io.h>
20 #include <linux/init.h>
21 #include <linux/mm.h>
22 #include <linux/sched.h>
23 #include <linux/seq_file.h>
25 #include <asm/fixmap.h>
26 #include <asm/kasan.h>
27 #include <asm/memory.h>
28 #include <asm/pgtable.h>
29 #include <asm/pgtable-hwdef.h>
30 #include <asm/ptdump.h>
32 static const struct addr_marker address_markers[] = {
33 #ifdef CONFIG_KASAN
34 { KASAN_SHADOW_START, "Kasan shadow start" },
35 { KASAN_SHADOW_END, "Kasan shadow end" },
36 #endif
37 { MODULES_VADDR, "Modules start" },
38 { MODULES_END, "Modules end" },
39 { VMALLOC_START, "vmalloc() Area" },
40 { VMALLOC_END, "vmalloc() End" },
41 { FIXADDR_START, "Fixmap start" },
42 { FIXADDR_TOP, "Fixmap end" },
43 { PCI_IO_START, "PCI I/O start" },
44 { PCI_IO_END, "PCI I/O end" },
45 #ifdef CONFIG_SPARSEMEM_VMEMMAP
46 { VMEMMAP_START, "vmemmap start" },
47 { VMEMMAP_START + VMEMMAP_SIZE, "vmemmap end" },
48 #endif
49 { PAGE_OFFSET, "Linear Mapping" },
50 { -1, NULL },
53 #define pt_dump_seq_printf(m, fmt, args...) \
54 ({ \
55 if (m) \
56 seq_printf(m, fmt, ##args); \
59 #define pt_dump_seq_puts(m, fmt) \
60 ({ \
61 if (m) \
62 seq_printf(m, fmt); \
66 * The page dumper groups page table entries of the same type into a single
67 * description. It uses pg_state to track the range information while
68 * iterating over the pte entries. When the continuity is broken it then
69 * dumps out a description of the range.
71 struct pg_state {
72 struct seq_file *seq;
73 const struct addr_marker *marker;
74 unsigned long start_address;
75 unsigned level;
76 u64 current_prot;
77 bool check_wx;
78 unsigned long wx_pages;
79 unsigned long uxn_pages;
82 struct prot_bits {
83 u64 mask;
84 u64 val;
85 const char *set;
86 const char *clear;
89 static const struct prot_bits pte_bits[] = {
91 .mask = PTE_VALID,
92 .val = PTE_VALID,
93 .set = " ",
94 .clear = "F",
95 }, {
96 .mask = PTE_USER,
97 .val = PTE_USER,
98 .set = "USR",
99 .clear = " ",
100 }, {
101 .mask = PTE_RDONLY,
102 .val = PTE_RDONLY,
103 .set = "ro",
104 .clear = "RW",
105 }, {
106 .mask = PTE_PXN,
107 .val = PTE_PXN,
108 .set = "NX",
109 .clear = "x ",
110 }, {
111 .mask = PTE_SHARED,
112 .val = PTE_SHARED,
113 .set = "SHD",
114 .clear = " ",
115 }, {
116 .mask = PTE_AF,
117 .val = PTE_AF,
118 .set = "AF",
119 .clear = " ",
120 }, {
121 .mask = PTE_NG,
122 .val = PTE_NG,
123 .set = "NG",
124 .clear = " ",
125 }, {
126 .mask = PTE_CONT,
127 .val = PTE_CONT,
128 .set = "CON",
129 .clear = " ",
130 }, {
131 .mask = PTE_TABLE_BIT,
132 .val = PTE_TABLE_BIT,
133 .set = " ",
134 .clear = "BLK",
135 }, {
136 .mask = PTE_UXN,
137 .val = PTE_UXN,
138 .set = "UXN",
139 }, {
140 .mask = PTE_ATTRINDX_MASK,
141 .val = PTE_ATTRINDX(MT_DEVICE_nGnRnE),
142 .set = "DEVICE/nGnRnE",
143 }, {
144 .mask = PTE_ATTRINDX_MASK,
145 .val = PTE_ATTRINDX(MT_DEVICE_nGnRE),
146 .set = "DEVICE/nGnRE",
147 }, {
148 .mask = PTE_ATTRINDX_MASK,
149 .val = PTE_ATTRINDX(MT_DEVICE_GRE),
150 .set = "DEVICE/GRE",
151 }, {
152 .mask = PTE_ATTRINDX_MASK,
153 .val = PTE_ATTRINDX(MT_NORMAL_NC),
154 .set = "MEM/NORMAL-NC",
155 }, {
156 .mask = PTE_ATTRINDX_MASK,
157 .val = PTE_ATTRINDX(MT_NORMAL),
158 .set = "MEM/NORMAL",
162 struct pg_level {
163 const struct prot_bits *bits;
164 const char *name;
165 size_t num;
166 u64 mask;
169 static struct pg_level pg_level[] = {
171 }, { /* pgd */
172 .name = "PGD",
173 .bits = pte_bits,
174 .num = ARRAY_SIZE(pte_bits),
175 }, { /* pud */
176 .name = (CONFIG_PGTABLE_LEVELS > 3) ? "PUD" : "PGD",
177 .bits = pte_bits,
178 .num = ARRAY_SIZE(pte_bits),
179 }, { /* pmd */
180 .name = (CONFIG_PGTABLE_LEVELS > 2) ? "PMD" : "PGD",
181 .bits = pte_bits,
182 .num = ARRAY_SIZE(pte_bits),
183 }, { /* pte */
184 .name = "PTE",
185 .bits = pte_bits,
186 .num = ARRAY_SIZE(pte_bits),
190 static void dump_prot(struct pg_state *st, const struct prot_bits *bits,
191 size_t num)
193 unsigned i;
195 for (i = 0; i < num; i++, bits++) {
196 const char *s;
198 if ((st->current_prot & bits->mask) == bits->val)
199 s = bits->set;
200 else
201 s = bits->clear;
203 if (s)
204 pt_dump_seq_printf(st->seq, " %s", s);
208 static void note_prot_uxn(struct pg_state *st, unsigned long addr)
210 if (!st->check_wx)
211 return;
213 if ((st->current_prot & PTE_UXN) == PTE_UXN)
214 return;
216 WARN_ONCE(1, "arm64/mm: Found non-UXN mapping at address %p/%pS\n",
217 (void *)st->start_address, (void *)st->start_address);
219 st->uxn_pages += (addr - st->start_address) / PAGE_SIZE;
222 static void note_prot_wx(struct pg_state *st, unsigned long addr)
224 if (!st->check_wx)
225 return;
226 if ((st->current_prot & PTE_RDONLY) == PTE_RDONLY)
227 return;
228 if ((st->current_prot & PTE_PXN) == PTE_PXN)
229 return;
231 WARN_ONCE(1, "arm64/mm: Found insecure W+X mapping at address %p/%pS\n",
232 (void *)st->start_address, (void *)st->start_address);
234 st->wx_pages += (addr - st->start_address) / PAGE_SIZE;
237 static void note_page(struct pg_state *st, unsigned long addr, unsigned level,
238 u64 val)
240 static const char units[] = "KMGTPE";
241 u64 prot = val & pg_level[level].mask;
243 if (!st->level) {
244 st->level = level;
245 st->current_prot = prot;
246 st->start_address = addr;
247 pt_dump_seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
248 } else if (prot != st->current_prot || level != st->level ||
249 addr >= st->marker[1].start_address) {
250 const char *unit = units;
251 unsigned long delta;
253 if (st->current_prot) {
254 note_prot_uxn(st, addr);
255 note_prot_wx(st, addr);
256 pt_dump_seq_printf(st->seq, "0x%016lx-0x%016lx ",
257 st->start_address, addr);
259 delta = (addr - st->start_address) >> 10;
260 while (!(delta & 1023) && unit[1]) {
261 delta >>= 10;
262 unit++;
264 pt_dump_seq_printf(st->seq, "%9lu%c %s", delta, *unit,
265 pg_level[st->level].name);
266 if (pg_level[st->level].bits)
267 dump_prot(st, pg_level[st->level].bits,
268 pg_level[st->level].num);
269 pt_dump_seq_puts(st->seq, "\n");
272 if (addr >= st->marker[1].start_address) {
273 st->marker++;
274 pt_dump_seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
277 st->start_address = addr;
278 st->current_prot = prot;
279 st->level = level;
282 if (addr >= st->marker[1].start_address) {
283 st->marker++;
284 pt_dump_seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
289 static void walk_pte(struct pg_state *st, pmd_t *pmdp, unsigned long start)
291 pte_t *ptep = pte_offset_kernel(pmdp, 0UL);
292 unsigned long addr;
293 unsigned i;
295 for (i = 0; i < PTRS_PER_PTE; i++, ptep++) {
296 addr = start + i * PAGE_SIZE;
297 note_page(st, addr, 4, READ_ONCE(pte_val(*ptep)));
301 static void walk_pmd(struct pg_state *st, pud_t *pudp, unsigned long start)
303 pmd_t *pmdp = pmd_offset(pudp, 0UL);
304 unsigned long addr;
305 unsigned i;
307 for (i = 0; i < PTRS_PER_PMD; i++, pmdp++) {
308 pmd_t pmd = READ_ONCE(*pmdp);
310 addr = start + i * PMD_SIZE;
311 if (pmd_none(pmd) || pmd_sect(pmd)) {
312 note_page(st, addr, 3, pmd_val(pmd));
313 } else {
314 BUG_ON(pmd_bad(pmd));
315 walk_pte(st, pmdp, addr);
320 static void walk_pud(struct pg_state *st, pgd_t *pgdp, unsigned long start)
322 pud_t *pudp = pud_offset(pgdp, 0UL);
323 unsigned long addr;
324 unsigned i;
326 for (i = 0; i < PTRS_PER_PUD; i++, pudp++) {
327 pud_t pud = READ_ONCE(*pudp);
329 addr = start + i * PUD_SIZE;
330 if (pud_none(pud) || pud_sect(pud)) {
331 note_page(st, addr, 2, pud_val(pud));
332 } else {
333 BUG_ON(pud_bad(pud));
334 walk_pmd(st, pudp, addr);
339 static void walk_pgd(struct pg_state *st, struct mm_struct *mm,
340 unsigned long start)
342 pgd_t *pgdp = pgd_offset(mm, 0UL);
343 unsigned i;
344 unsigned long addr;
346 for (i = 0; i < PTRS_PER_PGD; i++, pgdp++) {
347 pgd_t pgd = READ_ONCE(*pgdp);
349 addr = start + i * PGDIR_SIZE;
350 if (pgd_none(pgd)) {
351 note_page(st, addr, 1, pgd_val(pgd));
352 } else {
353 BUG_ON(pgd_bad(pgd));
354 walk_pud(st, pgdp, addr);
359 void ptdump_walk_pgd(struct seq_file *m, struct ptdump_info *info)
361 struct pg_state st = {
362 .seq = m,
363 .marker = info->markers,
366 walk_pgd(&st, info->mm, info->base_addr);
368 note_page(&st, 0, 0, 0);
371 static void ptdump_initialize(void)
373 unsigned i, j;
375 for (i = 0; i < ARRAY_SIZE(pg_level); i++)
376 if (pg_level[i].bits)
377 for (j = 0; j < pg_level[i].num; j++)
378 pg_level[i].mask |= pg_level[i].bits[j].mask;
381 static struct ptdump_info kernel_ptdump_info = {
382 .mm = &init_mm,
383 .markers = address_markers,
384 .base_addr = VA_START,
387 void ptdump_check_wx(void)
389 struct pg_state st = {
390 .seq = NULL,
391 .marker = (struct addr_marker[]) {
392 { 0, NULL},
393 { -1, NULL},
395 .check_wx = true,
398 walk_pgd(&st, &init_mm, VA_START);
399 note_page(&st, 0, 0, 0);
400 if (st.wx_pages || st.uxn_pages)
401 pr_warn("Checked W+X mappings: FAILED, %lu W+X pages found, %lu non-UXN pages found\n",
402 st.wx_pages, st.uxn_pages);
403 else
404 pr_info("Checked W+X mappings: passed, no W+X pages found\n");
407 static int ptdump_init(void)
409 ptdump_initialize();
410 return ptdump_debugfs_register(&kernel_ptdump_info,
411 "kernel_page_tables");
413 device_initcall(ptdump_init);