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.
5 * (C) Copyright 2008 Intel Corporation
7 * Author: Arjan van de Ven <arjan@linux.intel.com>
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
15 #include <linux/debugfs.h>
17 #include <linux/init.h>
18 #include <linux/sched.h>
19 #include <linux/seq_file.h>
21 #include <asm/pgtable.h>
24 * The dumper groups pagetable entries of the same type into one, and for
25 * that it needs to keep some state when walking, and flush this state
26 * when a "break" in the continuity is found.
30 pgprot_t current_prot;
31 unsigned long start_address;
32 unsigned long current_address;
33 const struct addr_marker *marker;
37 unsigned long wx_pages;
41 unsigned long start_address;
43 unsigned long max_lines;
46 /* indices for address_markers; keep sync'd w/ address_markers below */
47 enum address_markers_idx {
54 # ifdef CONFIG_X86_ESPFIX64
64 # ifdef CONFIG_HIGHMEM
71 /* Address space markers hints */
72 static struct addr_marker address_markers[] = {
75 { 0x8000000000000000UL, "Kernel Space" },
76 { 0/* PAGE_OFFSET */, "Low Kernel Mapping" },
77 { 0/* VMALLOC_START */, "vmalloc() Area" },
78 { 0/* VMEMMAP_START */, "Vmemmap" },
79 # ifdef CONFIG_X86_ESPFIX64
80 { ESPFIX_BASE_ADDR, "ESPfix Area", 16 },
83 { EFI_VA_END, "EFI Runtime Services" },
85 { __START_KERNEL_map, "High Kernel Mapping" },
86 { MODULES_VADDR, "Modules" },
87 { MODULES_END, "End Modules" },
89 { PAGE_OFFSET, "Kernel Mapping" },
90 { 0/* VMALLOC_START */, "vmalloc() Area" },
91 { 0/*VMALLOC_END*/, "vmalloc() End" },
92 # ifdef CONFIG_HIGHMEM
93 { 0/*PKMAP_BASE*/, "Persistent kmap() Area" },
95 { 0/*FIXADDR_START*/, "Fixmap Area" },
97 { -1, NULL } /* End of list */
100 /* Multipliers for offsets within the PTEs */
101 #define PTE_LEVEL_MULT (PAGE_SIZE)
102 #define PMD_LEVEL_MULT (PTRS_PER_PTE * PTE_LEVEL_MULT)
103 #define PUD_LEVEL_MULT (PTRS_PER_PMD * PMD_LEVEL_MULT)
104 #define PGD_LEVEL_MULT (PTRS_PER_PUD * PUD_LEVEL_MULT)
106 #define pt_dump_seq_printf(m, to_dmesg, fmt, args...) \
109 printk(KERN_INFO fmt, ##args); \
112 seq_printf(m, fmt, ##args); \
115 #define pt_dump_cont_printf(m, to_dmesg, fmt, args...) \
118 printk(KERN_CONT fmt, ##args); \
121 seq_printf(m, fmt, ##args); \
125 * Print a readable form of a pgprot_t to the seq_file
127 static void printk_prot(struct seq_file *m, pgprot_t prot, int level, bool dmsg)
129 pgprotval_t pr = pgprot_val(prot);
130 static const char * const level_name[] =
131 { "cr3", "pgd", "pud", "pmd", "pte" };
133 if (!pgprot_val(prot)) {
135 pt_dump_cont_printf(m, dmsg, " ");
138 pt_dump_cont_printf(m, dmsg, "USR ");
140 pt_dump_cont_printf(m, dmsg, " ");
142 pt_dump_cont_printf(m, dmsg, "RW ");
144 pt_dump_cont_printf(m, dmsg, "ro ");
146 pt_dump_cont_printf(m, dmsg, "PWT ");
148 pt_dump_cont_printf(m, dmsg, " ");
150 pt_dump_cont_printf(m, dmsg, "PCD ");
152 pt_dump_cont_printf(m, dmsg, " ");
154 /* Bit 7 has a different meaning on level 3 vs 4 */
155 if (level <= 3 && pr & _PAGE_PSE)
156 pt_dump_cont_printf(m, dmsg, "PSE ");
158 pt_dump_cont_printf(m, dmsg, " ");
159 if ((level == 4 && pr & _PAGE_PAT) ||
160 ((level == 3 || level == 2) && pr & _PAGE_PAT_LARGE))
161 pt_dump_cont_printf(m, dmsg, "PAT ");
163 pt_dump_cont_printf(m, dmsg, " ");
164 if (pr & _PAGE_GLOBAL)
165 pt_dump_cont_printf(m, dmsg, "GLB ");
167 pt_dump_cont_printf(m, dmsg, " ");
169 pt_dump_cont_printf(m, dmsg, "NX ");
171 pt_dump_cont_printf(m, dmsg, "x ");
173 pt_dump_cont_printf(m, dmsg, "%s\n", level_name[level]);
177 * On 64 bits, sign-extend the 48 bit address to 64 bit
179 static unsigned long normalize_addr(unsigned long u)
182 return (signed long)(u << 16) >> 16;
189 * This function gets called on a break in a continuous series
190 * of PTE entries; the next one is different so we need to
191 * print what we collected so far.
193 static void note_page(struct seq_file *m, struct pg_state *st,
194 pgprot_t new_prot, int level)
196 pgprotval_t prot, cur;
197 static const char units[] = "BKMGTPE";
200 * If we have a "break" in the series, we need to flush the state that
201 * we have now. "break" is either changing perms, levels or
202 * address space marker.
204 prot = pgprot_val(new_prot);
205 cur = pgprot_val(st->current_prot);
209 st->current_prot = new_prot;
211 st->marker = address_markers;
213 pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
215 } else if (prot != cur || level != st->level ||
216 st->current_address >= st->marker[1].start_address) {
217 const char *unit = units;
219 int width = sizeof(unsigned long) * 2;
220 pgprotval_t pr = pgprot_val(st->current_prot);
222 if (st->check_wx && (pr & _PAGE_RW) && !(pr & _PAGE_NX)) {
224 "x86/mm: Found insecure W+X mapping at address %p/%pS\n",
225 (void *)st->start_address,
226 (void *)st->start_address);
227 st->wx_pages += (st->current_address -
228 st->start_address) / PAGE_SIZE;
232 * Now print the actual finished series
234 if (!st->marker->max_lines ||
235 st->lines < st->marker->max_lines) {
236 pt_dump_seq_printf(m, st->to_dmesg,
238 width, st->start_address,
239 width, st->current_address);
241 delta = st->current_address - st->start_address;
242 while (!(delta & 1023) && unit[1]) {
246 pt_dump_cont_printf(m, st->to_dmesg, "%9lu%c ",
248 printk_prot(m, st->current_prot, st->level,
254 * We print markers for special areas of address space,
255 * such as the start of vmalloc space etc.
256 * This helps in the interpretation.
258 if (st->current_address >= st->marker[1].start_address) {
259 if (st->marker->max_lines &&
260 st->lines > st->marker->max_lines) {
261 unsigned long nskip =
262 st->lines - st->marker->max_lines;
263 pt_dump_seq_printf(m, st->to_dmesg,
264 "... %lu entr%s skipped ... \n",
266 nskip == 1 ? "y" : "ies");
270 pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
274 st->start_address = st->current_address;
275 st->current_prot = new_prot;
280 static void walk_pte_level(struct seq_file *m, struct pg_state *st, pmd_t addr,
287 start = (pte_t *) pmd_page_vaddr(addr);
288 for (i = 0; i < PTRS_PER_PTE; i++) {
289 prot = pte_flags(*start);
290 st->current_address = normalize_addr(P + i * PTE_LEVEL_MULT);
291 note_page(m, st, __pgprot(prot), 4);
298 static void walk_pmd_level(struct seq_file *m, struct pg_state *st, pud_t addr,
305 start = (pmd_t *) pud_page_vaddr(addr);
306 for (i = 0; i < PTRS_PER_PMD; i++) {
307 st->current_address = normalize_addr(P + i * PMD_LEVEL_MULT);
308 if (!pmd_none(*start)) {
309 if (pmd_large(*start) || !pmd_present(*start)) {
310 prot = pmd_flags(*start);
311 note_page(m, st, __pgprot(prot), 3);
313 walk_pte_level(m, st, *start,
314 P + i * PMD_LEVEL_MULT);
317 note_page(m, st, __pgprot(0), 3);
323 #define walk_pmd_level(m,s,a,p) walk_pte_level(m,s,__pmd(pud_val(a)),p)
324 #define pud_large(a) pmd_large(__pmd(pud_val(a)))
325 #define pud_none(a) pmd_none(__pmd(pud_val(a)))
330 static void walk_pud_level(struct seq_file *m, struct pg_state *st, pgd_t addr,
337 start = (pud_t *) pgd_page_vaddr(addr);
339 for (i = 0; i < PTRS_PER_PUD; i++) {
340 st->current_address = normalize_addr(P + i * PUD_LEVEL_MULT);
341 if (!pud_none(*start)) {
342 if (pud_large(*start) || !pud_present(*start)) {
343 prot = pud_flags(*start);
344 note_page(m, st, __pgprot(prot), 2);
346 walk_pmd_level(m, st, *start,
347 P + i * PUD_LEVEL_MULT);
350 note_page(m, st, __pgprot(0), 2);
357 #define walk_pud_level(m,s,a,p) walk_pmd_level(m,s,__pud(pgd_val(a)),p)
358 #define pgd_large(a) pud_large(__pud(pgd_val(a)))
359 #define pgd_none(a) pud_none(__pud(pgd_val(a)))
362 static inline bool is_hypervisor_range(int idx)
366 * ffff800000000000 - ffff87ffffffffff is reserved for
369 return (idx >= pgd_index(__PAGE_OFFSET) - 16) &&
370 (idx < pgd_index(__PAGE_OFFSET));
376 static void ptdump_walk_pgd_level_core(struct seq_file *m, pgd_t *pgd,
380 pgd_t *start = (pgd_t *) &init_level4_pgt;
382 pgd_t *start = swapper_pg_dir;
386 struct pg_state st = {};
393 st.check_wx = checkwx;
397 for (i = 0; i < PTRS_PER_PGD; i++) {
398 st.current_address = normalize_addr(i * PGD_LEVEL_MULT);
399 if (!pgd_none(*start) && !is_hypervisor_range(i)) {
400 if (pgd_large(*start) || !pgd_present(*start)) {
401 prot = pgd_flags(*start);
402 note_page(m, &st, __pgprot(prot), 1);
404 walk_pud_level(m, &st, *start,
408 note_page(m, &st, __pgprot(0), 1);
414 /* Flush out the last page */
415 st.current_address = normalize_addr(PTRS_PER_PGD*PGD_LEVEL_MULT);
416 note_page(m, &st, __pgprot(0), 0);
420 pr_info("x86/mm: Checked W+X mappings: FAILED, %lu W+X pages found.\n",
423 pr_info("x86/mm: Checked W+X mappings: passed, no W+X pages found.\n");
426 void ptdump_walk_pgd_level(struct seq_file *m, pgd_t *pgd)
428 ptdump_walk_pgd_level_core(m, pgd, false);
430 EXPORT_SYMBOL_GPL(ptdump_walk_pgd_level);
432 void ptdump_walk_pgd_level_checkwx(void)
434 ptdump_walk_pgd_level_core(NULL, NULL, true);
437 static int __init pt_dump_init(void)
440 * Various markers are not compile-time constants, so assign them
444 address_markers[LOW_KERNEL_NR].start_address = PAGE_OFFSET;
445 address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
446 address_markers[VMEMMAP_START_NR].start_address = VMEMMAP_START;
449 address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
450 address_markers[VMALLOC_END_NR].start_address = VMALLOC_END;
451 # ifdef CONFIG_HIGHMEM
452 address_markers[PKMAP_BASE_NR].start_address = PKMAP_BASE;
454 address_markers[FIXADDR_START_NR].start_address = FIXADDR_START;
459 __initcall(pt_dump_init);