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/module.h>
18 #include <linux/seq_file.h>
20 #include <asm/pgtable.h>
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.
29 pgprot_t current_prot;
30 unsigned long start_address;
31 unsigned long current_address;
32 const struct addr_marker *marker;
36 unsigned long start_address;
40 /* indices for address_markers; keep sync'd w/ address_markers below */
41 enum address_markers_idx {
55 # ifdef CONFIG_HIGHMEM
62 /* Address space markers hints */
63 static struct addr_marker address_markers[] = {
66 { 0x8000000000000000UL, "Kernel Space" },
67 { PAGE_OFFSET, "Low Kernel Mapping" },
68 { VMALLOC_START, "vmalloc() Area" },
69 { VMEMMAP_START, "Vmemmap" },
70 { __START_KERNEL_map, "High Kernel Mapping" },
71 { MODULES_VADDR, "Modules" },
72 { MODULES_END, "End Modules" },
74 { PAGE_OFFSET, "Kernel Mapping" },
75 { 0/* VMALLOC_START */, "vmalloc() Area" },
76 { 0/*VMALLOC_END*/, "vmalloc() End" },
77 # ifdef CONFIG_HIGHMEM
78 { 0/*PKMAP_BASE*/, "Persisent kmap() Area" },
80 { 0/*FIXADDR_START*/, "Fixmap Area" },
82 { -1, NULL } /* End of list */
85 /* Multipliers for offsets within the PTEs */
86 #define PTE_LEVEL_MULT (PAGE_SIZE)
87 #define PMD_LEVEL_MULT (PTRS_PER_PTE * PTE_LEVEL_MULT)
88 #define PUD_LEVEL_MULT (PTRS_PER_PMD * PMD_LEVEL_MULT)
89 #define PGD_LEVEL_MULT (PTRS_PER_PUD * PUD_LEVEL_MULT)
92 * Print a readable form of a pgprot_t to the seq_file
94 static void printk_prot(struct seq_file *m, pgprot_t prot, int level)
96 pgprotval_t pr = pgprot_val(prot);
97 static const char * const level_name[] =
98 { "cr3", "pgd", "pud", "pmd", "pte" };
100 if (!pgprot_val(prot)) {
105 seq_printf(m, "USR ");
109 seq_printf(m, "RW ");
111 seq_printf(m, "ro ");
113 seq_printf(m, "PWT ");
117 seq_printf(m, "PCD ");
121 /* Bit 9 has a different meaning on level 3 vs 4 */
124 seq_printf(m, "PSE ");
129 seq_printf(m, "pat ");
133 if (pr & _PAGE_GLOBAL)
134 seq_printf(m, "GLB ");
138 seq_printf(m, "NX ");
142 seq_printf(m, "%s\n", level_name[level]);
146 * On 64 bits, sign-extend the 48 bit address to 64 bit
148 static unsigned long normalize_addr(unsigned long u)
151 return (signed long)(u << 16) >> 16;
158 * This function gets called on a break in a continuous series
159 * of PTE entries; the next one is different so we need to
160 * print what we collected so far.
162 static void note_page(struct seq_file *m, struct pg_state *st,
163 pgprot_t new_prot, int level)
165 pgprotval_t prot, cur;
166 static const char units[] = "KMGTPE";
169 * If we have a "break" in the series, we need to flush the state that
170 * we have now. "break" is either changing perms, levels or
171 * address space marker.
173 prot = pgprot_val(new_prot) & PTE_FLAGS_MASK;
174 cur = pgprot_val(st->current_prot) & PTE_FLAGS_MASK;
178 st->current_prot = new_prot;
180 st->marker = address_markers;
181 seq_printf(m, "---[ %s ]---\n", st->marker->name);
182 } else if (prot != cur || level != st->level ||
183 st->current_address >= st->marker[1].start_address) {
184 const char *unit = units;
186 int width = sizeof(unsigned long) * 2;
189 * Now print the actual finished series
191 seq_printf(m, "0x%0*lx-0x%0*lx ",
192 width, st->start_address,
193 width, st->current_address);
195 delta = (st->current_address - st->start_address) >> 10;
196 while (!(delta & 1023) && unit[1]) {
200 seq_printf(m, "%9lu%c ", delta, *unit);
201 printk_prot(m, st->current_prot, st->level);
204 * We print markers for special areas of address space,
205 * such as the start of vmalloc space etc.
206 * This helps in the interpretation.
208 if (st->current_address >= st->marker[1].start_address) {
210 seq_printf(m, "---[ %s ]---\n", st->marker->name);
213 st->start_address = st->current_address;
214 st->current_prot = new_prot;
219 static void walk_pte_level(struct seq_file *m, struct pg_state *st, pmd_t addr,
225 start = (pte_t *) pmd_page_vaddr(addr);
226 for (i = 0; i < PTRS_PER_PTE; i++) {
227 pgprot_t prot = pte_pgprot(*start);
229 st->current_address = normalize_addr(P + i * PTE_LEVEL_MULT);
230 note_page(m, st, prot, 4);
237 static void walk_pmd_level(struct seq_file *m, struct pg_state *st, pud_t addr,
243 start = (pmd_t *) pud_page_vaddr(addr);
244 for (i = 0; i < PTRS_PER_PMD; i++) {
245 st->current_address = normalize_addr(P + i * PMD_LEVEL_MULT);
246 if (!pmd_none(*start)) {
247 pgprotval_t prot = pmd_val(*start) & PTE_FLAGS_MASK;
249 if (pmd_large(*start) || !pmd_present(*start))
250 note_page(m, st, __pgprot(prot), 3);
252 walk_pte_level(m, st, *start,
253 P + i * PMD_LEVEL_MULT);
255 note_page(m, st, __pgprot(0), 3);
261 #define walk_pmd_level(m,s,a,p) walk_pte_level(m,s,__pmd(pud_val(a)),p)
262 #define pud_large(a) pmd_large(__pmd(pud_val(a)))
263 #define pud_none(a) pmd_none(__pmd(pud_val(a)))
268 static void walk_pud_level(struct seq_file *m, struct pg_state *st, pgd_t addr,
274 start = (pud_t *) pgd_page_vaddr(addr);
276 for (i = 0; i < PTRS_PER_PUD; i++) {
277 st->current_address = normalize_addr(P + i * PUD_LEVEL_MULT);
278 if (!pud_none(*start)) {
279 pgprotval_t prot = pud_val(*start) & PTE_FLAGS_MASK;
281 if (pud_large(*start) || !pud_present(*start))
282 note_page(m, st, __pgprot(prot), 2);
284 walk_pmd_level(m, st, *start,
285 P + i * PUD_LEVEL_MULT);
287 note_page(m, st, __pgprot(0), 2);
294 #define walk_pud_level(m,s,a,p) walk_pmd_level(m,s,__pud(pgd_val(a)),p)
295 #define pgd_large(a) pud_large(__pud(pgd_val(a)))
296 #define pgd_none(a) pud_none(__pud(pgd_val(a)))
299 static void walk_pgd_level(struct seq_file *m)
302 pgd_t *start = (pgd_t *) &init_level4_pgt;
304 pgd_t *start = swapper_pg_dir;
309 memset(&st, 0, sizeof(st));
311 for (i = 0; i < PTRS_PER_PGD; i++) {
312 st.current_address = normalize_addr(i * PGD_LEVEL_MULT);
313 if (!pgd_none(*start)) {
314 pgprotval_t prot = pgd_val(*start) & PTE_FLAGS_MASK;
316 if (pgd_large(*start) || !pgd_present(*start))
317 note_page(m, &st, __pgprot(prot), 1);
319 walk_pud_level(m, &st, *start,
322 note_page(m, &st, __pgprot(0), 1);
327 /* Flush out the last page */
328 st.current_address = normalize_addr(PTRS_PER_PGD*PGD_LEVEL_MULT);
329 note_page(m, &st, __pgprot(0), 0);
332 static int ptdump_show(struct seq_file *m, void *v)
338 static int ptdump_open(struct inode *inode, struct file *filp)
340 return single_open(filp, ptdump_show, NULL);
343 static const struct file_operations ptdump_fops = {
347 .release = single_release,
350 static int pt_dump_init(void)
355 /* Not a compile-time constant on x86-32 */
356 address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
357 address_markers[VMALLOC_END_NR].start_address = VMALLOC_END;
358 # ifdef CONFIG_HIGHMEM
359 address_markers[PKMAP_BASE_NR].start_address = PKMAP_BASE;
361 address_markers[FIXADDR_START_NR].start_address = FIXADDR_START;
364 pe = debugfs_create_file("kernel_page_tables", 0600, NULL, NULL,
372 __initcall(pt_dump_init);
373 MODULE_LICENSE("GPL");
374 MODULE_AUTHOR("Arjan van de Ven <arjan@linux.intel.com>");
375 MODULE_DESCRIPTION("Kernel debugging helper that dumps pagetables");