Linux Audio

Check our new training course

Loading...
v4.6
 
  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 */
 27struct 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	unsigned long lines;
 34	bool to_dmesg;
 35	bool check_wx;
 36	unsigned long wx_pages;
 
 37};
 38
 39struct addr_marker {
 40	unsigned long start_address;
 41	const char *name;
 42	unsigned long max_lines;
 43};
 44
 45/* indices for address_markers; keep sync'd w/ address_markers below */
 
 
 
 46enum address_markers_idx {
 47	USER_SPACE_NR = 0,
 48#ifdef CONFIG_X86_64
 49	KERNEL_SPACE_NR,
 
 
 
 50	LOW_KERNEL_NR,
 51	VMALLOC_START_NR,
 52	VMEMMAP_START_NR,
 53# ifdef CONFIG_X86_ESPFIX64
 
 
 
 
 
 54	ESPFIX_START_NR,
 55# endif
 
 
 
 56	HIGH_KERNEL_NR,
 57	MODULES_VADDR_NR,
 58	MODULES_END_NR,
 59#else
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 60	KERNEL_SPACE_NR,
 61	VMALLOC_START_NR,
 62	VMALLOC_END_NR,
 63# ifdef CONFIG_HIGHMEM
 64	PKMAP_BASE_NR,
 65# endif
 66	FIXADDR_START_NR,
 67#endif
 
 
 
 
 
 
 68};
 69
 70/* Address space markers hints */
 71static struct addr_marker address_markers[] = {
 72	{ 0, "User Space" },
 73#ifdef CONFIG_X86_64
 74	{ 0x8000000000000000UL, "Kernel Space" },
 75	{ PAGE_OFFSET,		"Low Kernel Mapping" },
 76	{ VMALLOC_START,        "vmalloc() Area" },
 77	{ VMEMMAP_START,        "Vmemmap" },
 78# ifdef CONFIG_X86_ESPFIX64
 79	{ ESPFIX_BASE_ADDR,	"ESPfix Area", 16 },
 80# endif
 81# ifdef CONFIG_EFI
 82	{ EFI_VA_END,		"EFI Runtime Services" },
 83# endif
 84	{ __START_KERNEL_map,   "High Kernel Mapping" },
 85	{ MODULES_VADDR,        "Modules" },
 86	{ MODULES_END,          "End Modules" },
 87#else
 88	{ PAGE_OFFSET,          "Kernel Mapping" },
 89	{ 0/* VMALLOC_START */, "vmalloc() Area" },
 90	{ 0/*VMALLOC_END*/,     "vmalloc() End" },
 91# ifdef CONFIG_HIGHMEM
 92	{ 0/*PKMAP_BASE*/,      "Persistent kmap() Area" },
 93# endif
 94	{ 0/*FIXADDR_START*/,   "Fixmap Area" },
 95#endif
 96	{ -1, NULL }		/* End of list */
 
 
 
 
 
 97};
 98
 
 
 
 
 99/* Multipliers for offsets within the PTEs */
100#define PTE_LEVEL_MULT (PAGE_SIZE)
101#define PMD_LEVEL_MULT (PTRS_PER_PTE * PTE_LEVEL_MULT)
102#define PUD_LEVEL_MULT (PTRS_PER_PMD * PMD_LEVEL_MULT)
103#define PGD_LEVEL_MULT (PTRS_PER_PUD * PUD_LEVEL_MULT)
 
104
105#define pt_dump_seq_printf(m, to_dmesg, fmt, args...)		\
106({								\
107	if (to_dmesg)					\
108		printk(KERN_INFO fmt, ##args);			\
109	else							\
110		if (m)						\
111			seq_printf(m, fmt, ##args);		\
112})
113
114#define pt_dump_cont_printf(m, to_dmesg, fmt, args...)		\
115({								\
116	if (to_dmesg)					\
117		printk(KERN_CONT fmt, ##args);			\
118	else							\
119		if (m)						\
120			seq_printf(m, fmt, ##args);		\
121})
122
123/*
124 * Print a readable form of a pgprot_t to the seq_file
125 */
126static void printk_prot(struct seq_file *m, pgprot_t prot, int level, bool dmsg)
127{
128	pgprotval_t pr = pgprot_val(prot);
129	static const char * const level_name[] =
130		{ "cr3", "pgd", "pud", "pmd", "pte" };
131
132	if (!pgprot_val(prot)) {
133		/* Not present */
134		pt_dump_cont_printf(m, dmsg, "                              ");
135	} else {
136		if (pr & _PAGE_USER)
137			pt_dump_cont_printf(m, dmsg, "USR ");
138		else
139			pt_dump_cont_printf(m, dmsg, "    ");
140		if (pr & _PAGE_RW)
141			pt_dump_cont_printf(m, dmsg, "RW ");
142		else
143			pt_dump_cont_printf(m, dmsg, "ro ");
144		if (pr & _PAGE_PWT)
145			pt_dump_cont_printf(m, dmsg, "PWT ");
146		else
147			pt_dump_cont_printf(m, dmsg, "    ");
148		if (pr & _PAGE_PCD)
149			pt_dump_cont_printf(m, dmsg, "PCD ");
150		else
151			pt_dump_cont_printf(m, dmsg, "    ");
152
153		/* Bit 7 has a different meaning on level 3 vs 4 */
154		if (level <= 3 && pr & _PAGE_PSE)
155			pt_dump_cont_printf(m, dmsg, "PSE ");
156		else
157			pt_dump_cont_printf(m, dmsg, "    ");
158		if ((level == 4 && pr & _PAGE_PAT) ||
159		    ((level == 3 || level == 2) && pr & _PAGE_PAT_LARGE))
160			pt_dump_cont_printf(m, dmsg, "PAT ");
161		else
162			pt_dump_cont_printf(m, dmsg, "    ");
163		if (pr & _PAGE_GLOBAL)
164			pt_dump_cont_printf(m, dmsg, "GLB ");
165		else
166			pt_dump_cont_printf(m, dmsg, "    ");
167		if (pr & _PAGE_NX)
168			pt_dump_cont_printf(m, dmsg, "NX ");
169		else
170			pt_dump_cont_printf(m, dmsg, "x  ");
171	}
172	pt_dump_cont_printf(m, dmsg, "%s\n", level_name[level]);
173}
174
175/*
176 * On 64 bits, sign-extend the 48 bit address to 64 bit
177 */
178static unsigned long normalize_addr(unsigned long u)
179{
180#ifdef CONFIG_X86_64
181	return (signed long)(u << 16) >> 16;
182#else
183	return u;
 
 
 
 
 
 
 
 
 
 
184#endif
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
185}
186
187/*
188 * This function gets called on a break in a continuous series
189 * of PTE entries; the next one is different so we need to
190 * print what we collected so far.
191 */
192static void note_page(struct seq_file *m, struct pg_state *st,
193		      pgprot_t new_prot, int level)
194{
195	pgprotval_t prot, cur;
 
 
196	static const char units[] = "BKMGTPE";
 
 
 
 
 
 
 
197
198	/*
199	 * If we have a "break" in the series, we need to flush the state that
200	 * we have now. "break" is either changing perms, levels or
201	 * address space marker.
202	 */
203	prot = pgprot_val(new_prot);
204	cur = pgprot_val(st->current_prot);
205
206	if (!st->level) {
207		/* First entry */
208		st->current_prot = new_prot;
 
209		st->level = level;
210		st->marker = address_markers;
211		st->lines = 0;
212		pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
213				   st->marker->name);
214	} else if (prot != cur || level != st->level ||
215		   st->current_address >= st->marker[1].start_address) {
216		const char *unit = units;
217		unsigned long delta;
218		int width = sizeof(unsigned long) * 2;
219		pgprotval_t pr = pgprot_val(st->current_prot);
220
221		if (st->check_wx && (pr & _PAGE_RW) && !(pr & _PAGE_NX)) {
222			WARN_ONCE(1,
223				  "x86/mm: Found insecure W+X mapping at address %p/%pS\n",
224				  (void *)st->start_address,
225				  (void *)st->start_address);
226			st->wx_pages += (st->current_address -
227					 st->start_address) / PAGE_SIZE;
228		}
229
230		/*
231		 * Now print the actual finished series
232		 */
233		if (!st->marker->max_lines ||
234		    st->lines < st->marker->max_lines) {
235			pt_dump_seq_printf(m, st->to_dmesg,
236					   "0x%0*lx-0x%0*lx   ",
237					   width, st->start_address,
238					   width, st->current_address);
239
240			delta = st->current_address - st->start_address;
241			while (!(delta & 1023) && unit[1]) {
242				delta >>= 10;
243				unit++;
244			}
245			pt_dump_cont_printf(m, st->to_dmesg, "%9lu%c ",
246					    delta, *unit);
247			printk_prot(m, st->current_prot, st->level,
248				    st->to_dmesg);
249		}
250		st->lines++;
251
252		/*
253		 * We print markers for special areas of address space,
254		 * such as the start of vmalloc space etc.
255		 * This helps in the interpretation.
256		 */
257		if (st->current_address >= st->marker[1].start_address) {
258			if (st->marker->max_lines &&
259			    st->lines > st->marker->max_lines) {
260				unsigned long nskip =
261					st->lines - st->marker->max_lines;
262				pt_dump_seq_printf(m, st->to_dmesg,
263						   "... %lu entr%s skipped ... \n",
264						   nskip,
265						   nskip == 1 ? "y" : "ies");
266			}
267			st->marker++;
268			st->lines = 0;
269			pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
270					   st->marker->name);
271		}
272
273		st->start_address = st->current_address;
274		st->current_prot = new_prot;
 
275		st->level = level;
276	}
277}
278
279static void walk_pte_level(struct seq_file *m, struct pg_state *st, pmd_t addr,
280							unsigned long P)
281{
282	int i;
283	pte_t *start;
284	pgprotval_t prot;
285
286	start = (pte_t *) pmd_page_vaddr(addr);
287	for (i = 0; i < PTRS_PER_PTE; i++) {
288		prot = pte_flags(*start);
289		st->current_address = normalize_addr(P + i * PTE_LEVEL_MULT);
290		note_page(m, st, __pgprot(prot), 4);
291		start++;
292	}
293}
294
295#if PTRS_PER_PMD > 1
296
297static void walk_pmd_level(struct seq_file *m, struct pg_state *st, pud_t addr,
298							unsigned long P)
299{
300	int i;
301	pmd_t *start;
302	pgprotval_t prot;
303
304	start = (pmd_t *) pud_page_vaddr(addr);
305	for (i = 0; i < PTRS_PER_PMD; i++) {
306		st->current_address = normalize_addr(P + i * PMD_LEVEL_MULT);
307		if (!pmd_none(*start)) {
308			if (pmd_large(*start) || !pmd_present(*start)) {
309				prot = pmd_flags(*start);
310				note_page(m, st, __pgprot(prot), 3);
311			} else {
312				walk_pte_level(m, st, *start,
313					       P + i * PMD_LEVEL_MULT);
314			}
315		} else
316			note_page(m, st, __pgprot(0), 3);
317		start++;
318	}
319}
320
321#else
322#define walk_pmd_level(m,s,a,p) walk_pte_level(m,s,__pmd(pud_val(a)),p)
323#define pud_large(a) pmd_large(__pmd(pud_val(a)))
324#define pud_none(a)  pmd_none(__pmd(pud_val(a)))
325#endif
326
327#if PTRS_PER_PUD > 1
328
329static void walk_pud_level(struct seq_file *m, struct pg_state *st, pgd_t addr,
330							unsigned long P)
331{
332	int i;
333	pud_t *start;
334	pgprotval_t prot;
335
336	start = (pud_t *) pgd_page_vaddr(addr);
337
338	for (i = 0; i < PTRS_PER_PUD; i++) {
339		st->current_address = normalize_addr(P + i * PUD_LEVEL_MULT);
340		if (!pud_none(*start)) {
341			if (pud_large(*start) || !pud_present(*start)) {
342				prot = pud_flags(*start);
343				note_page(m, st, __pgprot(prot), 2);
344			} else {
345				walk_pmd_level(m, st, *start,
346					       P + i * PUD_LEVEL_MULT);
347			}
348		} else
349			note_page(m, st, __pgprot(0), 2);
350
351		start++;
352	}
353}
354
355#else
356#define walk_pud_level(m,s,a,p) walk_pmd_level(m,s,__pud(pgd_val(a)),p)
357#define pgd_large(a) pud_large(__pud(pgd_val(a)))
358#define pgd_none(a)  pud_none(__pud(pgd_val(a)))
359#endif
360
361static inline bool is_hypervisor_range(int idx)
362{
363#ifdef CONFIG_X86_64
364	/*
365	 * ffff800000000000 - ffff87ffffffffff is reserved for
366	 * the hypervisor.
367	 */
368	return	(idx >= pgd_index(__PAGE_OFFSET) - 16) &&
369		(idx <  pgd_index(__PAGE_OFFSET));
370#else
371	return false;
372#endif
373}
374
375static void ptdump_walk_pgd_level_core(struct seq_file *m, pgd_t *pgd,
376				       bool checkwx)
377{
 
378#ifdef CONFIG_X86_64
379	pgd_t *start = (pgd_t *) &init_level4_pgt;
 
380#else
381	pgd_t *start = swapper_pg_dir;
382#endif
383	pgprotval_t prot;
384	int i;
385	struct pg_state st = {};
386
387	if (pgd) {
388		start = pgd;
389		st.to_dmesg = true;
390	}
391
392	st.check_wx = checkwx;
393	if (checkwx)
394		st.wx_pages = 0;
395
396	for (i = 0; i < PTRS_PER_PGD; i++) {
397		st.current_address = normalize_addr(i * PGD_LEVEL_MULT);
398		if (!pgd_none(*start) && !is_hypervisor_range(i)) {
399			if (pgd_large(*start) || !pgd_present(*start)) {
400				prot = pgd_flags(*start);
401				note_page(m, &st, __pgprot(prot), 1);
402			} else {
403				walk_pud_level(m, &st, *start,
404					       i * PGD_LEVEL_MULT);
405			}
406		} else
407			note_page(m, &st, __pgprot(0), 1);
408
409		start++;
410	}
411
412	/* Flush out the last page */
413	st.current_address = normalize_addr(PTRS_PER_PGD*PGD_LEVEL_MULT);
414	note_page(m, &st, __pgprot(0), 0);
415	if (!checkwx)
416		return;
417	if (st.wx_pages)
418		pr_info("x86/mm: Checked W+X mappings: FAILED, %lu W+X pages found.\n",
419			st.wx_pages);
420	else
421		pr_info("x86/mm: Checked W+X mappings: passed, no W+X pages found.\n");
422}
423
424void ptdump_walk_pgd_level(struct seq_file *m, pgd_t *pgd)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
425{
426	ptdump_walk_pgd_level_core(m, pgd, false);
 
 
 
 
 
 
 
 
 
 
427}
428EXPORT_SYMBOL_GPL(ptdump_walk_pgd_level);
429
430void ptdump_walk_pgd_level_checkwx(void)
431{
432	ptdump_walk_pgd_level_core(NULL, NULL, true);
433}
434
435static int __init pt_dump_init(void)
436{
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
437#ifdef CONFIG_X86_32
438	/* Not a compile-time constant on x86-32 */
439	address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
440	address_markers[VMALLOC_END_NR].start_address = VMALLOC_END;
441# ifdef CONFIG_HIGHMEM
442	address_markers[PKMAP_BASE_NR].start_address = PKMAP_BASE;
443# endif
444	address_markers[FIXADDR_START_NR].start_address = FIXADDR_START;
 
 
 
 
445#endif
446
447	return 0;
448}
449
450__initcall(pt_dump_init);
451MODULE_LICENSE("GPL");
452MODULE_AUTHOR("Arjan van de Ven <arjan@linux.intel.com>");
453MODULE_DESCRIPTION("Kernel debugging helper that dumps pagetables");
v6.8
  1// SPDX-License-Identifier: GPL-2.0-only
  2/*
  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.
  5 *
  6 * (C) Copyright 2008 Intel Corporation
  7 *
  8 * Author: Arjan van de Ven <arjan@linux.intel.com>
 
 
 
 
 
  9 */
 10
 11#include <linux/debugfs.h>
 12#include <linux/kasan.h>
 13#include <linux/mm.h>
 14#include <linux/init.h>
 15#include <linux/sched.h>
 16#include <linux/seq_file.h>
 17#include <linux/highmem.h>
 18#include <linux/pci.h>
 19#include <linux/ptdump.h>
 20
 21#include <asm/e820/types.h>
 22
 23/*
 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.
 27 */
 28struct pg_state {
 29	struct ptdump_state ptdump;
 30	int level;
 31	pgprotval_t current_prot;
 32	pgprotval_t effective_prot;
 33	pgprotval_t prot_levels[5];
 34	unsigned long start_address;
 
 35	const struct addr_marker *marker;
 36	unsigned long lines;
 37	bool to_dmesg;
 38	bool check_wx;
 39	unsigned long wx_pages;
 40	struct seq_file *seq;
 41};
 42
 43struct addr_marker {
 44	unsigned long start_address;
 45	const char *name;
 46	unsigned long max_lines;
 47};
 48
 49/* Address space markers hints */
 50
 51#ifdef CONFIG_X86_64
 52
 53enum address_markers_idx {
 54	USER_SPACE_NR = 0,
 
 55	KERNEL_SPACE_NR,
 56#ifdef CONFIG_MODIFY_LDT_SYSCALL
 57	LDT_NR,
 58#endif
 59	LOW_KERNEL_NR,
 60	VMALLOC_START_NR,
 61	VMEMMAP_START_NR,
 62#ifdef CONFIG_KASAN
 63	KASAN_SHADOW_START_NR,
 64	KASAN_SHADOW_END_NR,
 65#endif
 66	CPU_ENTRY_AREA_NR,
 67#ifdef CONFIG_X86_ESPFIX64
 68	ESPFIX_START_NR,
 69#endif
 70#ifdef CONFIG_EFI
 71	EFI_END_NR,
 72#endif
 73	HIGH_KERNEL_NR,
 74	MODULES_VADDR_NR,
 75	MODULES_END_NR,
 76	FIXADDR_START_NR,
 77	END_OF_SPACE_NR,
 78};
 79
 80static struct addr_marker address_markers[] = {
 81	[USER_SPACE_NR]		= { 0,			"User Space" },
 82	[KERNEL_SPACE_NR]	= { (1UL << 63),	"Kernel Space" },
 83	[LOW_KERNEL_NR]		= { 0UL,		"Low Kernel Mapping" },
 84	[VMALLOC_START_NR]	= { 0UL,		"vmalloc() Area" },
 85	[VMEMMAP_START_NR]	= { 0UL,		"Vmemmap" },
 86#ifdef CONFIG_KASAN
 87	/*
 88	 * These fields get initialized with the (dynamic)
 89	 * KASAN_SHADOW_{START,END} values in pt_dump_init().
 90	 */
 91	[KASAN_SHADOW_START_NR]	= { 0UL,		"KASAN shadow" },
 92	[KASAN_SHADOW_END_NR]	= { 0UL,		"KASAN shadow end" },
 93#endif
 94#ifdef CONFIG_MODIFY_LDT_SYSCALL
 95	[LDT_NR]		= { 0UL,		"LDT remap" },
 96#endif
 97	[CPU_ENTRY_AREA_NR]	= { CPU_ENTRY_AREA_BASE,"CPU entry Area" },
 98#ifdef CONFIG_X86_ESPFIX64
 99	[ESPFIX_START_NR]	= { ESPFIX_BASE_ADDR,	"ESPfix Area", 16 },
100#endif
101#ifdef CONFIG_EFI
102	[EFI_END_NR]		= { EFI_VA_END,		"EFI Runtime Services" },
103#endif
104	[HIGH_KERNEL_NR]	= { __START_KERNEL_map,	"High Kernel Mapping" },
105	[MODULES_VADDR_NR]	= { MODULES_VADDR,	"Modules" },
106	[MODULES_END_NR]	= { MODULES_END,	"End Modules" },
107	[FIXADDR_START_NR]	= { FIXADDR_START,	"Fixmap Area" },
108	[END_OF_SPACE_NR]	= { -1,			NULL }
109};
110
111#define INIT_PGD	((pgd_t *) &init_top_pgt)
112
113#else /* CONFIG_X86_64 */
114
115enum address_markers_idx {
116	USER_SPACE_NR = 0,
117	KERNEL_SPACE_NR,
118	VMALLOC_START_NR,
119	VMALLOC_END_NR,
120#ifdef CONFIG_HIGHMEM
121	PKMAP_BASE_NR,
 
 
122#endif
123#ifdef CONFIG_MODIFY_LDT_SYSCALL
124	LDT_NR,
125#endif
126	CPU_ENTRY_AREA_NR,
127	FIXADDR_START_NR,
128	END_OF_SPACE_NR,
129};
130
 
131static struct addr_marker address_markers[] = {
132	[USER_SPACE_NR]		= { 0,			"User Space" },
133	[KERNEL_SPACE_NR]	= { PAGE_OFFSET,	"Kernel Mapping" },
134	[VMALLOC_START_NR]	= { 0UL,		"vmalloc() Area" },
135	[VMALLOC_END_NR]	= { 0UL,		"vmalloc() End" },
136#ifdef CONFIG_HIGHMEM
137	[PKMAP_BASE_NR]		= { 0UL,		"Persistent kmap() Area" },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
138#endif
139#ifdef CONFIG_MODIFY_LDT_SYSCALL
140	[LDT_NR]		= { 0UL,		"LDT remap" },
141#endif
142	[CPU_ENTRY_AREA_NR]	= { 0UL,		"CPU entry area" },
143	[FIXADDR_START_NR]	= { 0UL,		"Fixmap area" },
144	[END_OF_SPACE_NR]	= { -1,			NULL }
145};
146
147#define INIT_PGD	(swapper_pg_dir)
148
149#endif /* !CONFIG_X86_64 */
150
151/* Multipliers for offsets within the PTEs */
152#define PTE_LEVEL_MULT (PAGE_SIZE)
153#define PMD_LEVEL_MULT (PTRS_PER_PTE * PTE_LEVEL_MULT)
154#define PUD_LEVEL_MULT (PTRS_PER_PMD * PMD_LEVEL_MULT)
155#define P4D_LEVEL_MULT (PTRS_PER_PUD * PUD_LEVEL_MULT)
156#define PGD_LEVEL_MULT (PTRS_PER_P4D * P4D_LEVEL_MULT)
157
158#define pt_dump_seq_printf(m, to_dmesg, fmt, args...)		\
159({								\
160	if (to_dmesg)					\
161		printk(KERN_INFO fmt, ##args);			\
162	else							\
163		if (m)						\
164			seq_printf(m, fmt, ##args);		\
165})
166
167#define pt_dump_cont_printf(m, to_dmesg, fmt, args...)		\
168({								\
169	if (to_dmesg)					\
170		printk(KERN_CONT fmt, ##args);			\
171	else							\
172		if (m)						\
173			seq_printf(m, fmt, ##args);		\
174})
175
176/*
177 * Print a readable form of a pgprot_t to the seq_file
178 */
179static void printk_prot(struct seq_file *m, pgprotval_t pr, int level, bool dmsg)
180{
 
181	static const char * const level_name[] =
182		{ "pgd", "p4d", "pud", "pmd", "pte" };
183
184	if (!(pr & _PAGE_PRESENT)) {
185		/* Not present */
186		pt_dump_cont_printf(m, dmsg, "                              ");
187	} else {
188		if (pr & _PAGE_USER)
189			pt_dump_cont_printf(m, dmsg, "USR ");
190		else
191			pt_dump_cont_printf(m, dmsg, "    ");
192		if (pr & _PAGE_RW)
193			pt_dump_cont_printf(m, dmsg, "RW ");
194		else
195			pt_dump_cont_printf(m, dmsg, "ro ");
196		if (pr & _PAGE_PWT)
197			pt_dump_cont_printf(m, dmsg, "PWT ");
198		else
199			pt_dump_cont_printf(m, dmsg, "    ");
200		if (pr & _PAGE_PCD)
201			pt_dump_cont_printf(m, dmsg, "PCD ");
202		else
203			pt_dump_cont_printf(m, dmsg, "    ");
204
205		/* Bit 7 has a different meaning on level 3 vs 4 */
206		if (level <= 3 && pr & _PAGE_PSE)
207			pt_dump_cont_printf(m, dmsg, "PSE ");
208		else
209			pt_dump_cont_printf(m, dmsg, "    ");
210		if ((level == 4 && pr & _PAGE_PAT) ||
211		    ((level == 3 || level == 2) && pr & _PAGE_PAT_LARGE))
212			pt_dump_cont_printf(m, dmsg, "PAT ");
213		else
214			pt_dump_cont_printf(m, dmsg, "    ");
215		if (pr & _PAGE_GLOBAL)
216			pt_dump_cont_printf(m, dmsg, "GLB ");
217		else
218			pt_dump_cont_printf(m, dmsg, "    ");
219		if (pr & _PAGE_NX)
220			pt_dump_cont_printf(m, dmsg, "NX ");
221		else
222			pt_dump_cont_printf(m, dmsg, "x  ");
223	}
224	pt_dump_cont_printf(m, dmsg, "%s\n", level_name[level]);
225}
226
227static void note_wx(struct pg_state *st, unsigned long addr)
 
 
 
228{
229	unsigned long npages;
230
231	npages = (addr - st->start_address) / PAGE_SIZE;
232
233#ifdef CONFIG_PCI_BIOS
234	/*
235	 * If PCI BIOS is enabled, the PCI BIOS area is forced to WX.
236	 * Inform about it, but avoid the warning.
237	 */
238	if (pcibios_enabled && st->start_address >= PAGE_OFFSET + BIOS_BEGIN &&
239	    addr <= PAGE_OFFSET + BIOS_END) {
240		pr_warn_once("x86/mm: PCI BIOS W+X mapping %lu pages\n", npages);
241		return;
242	}
243#endif
244	/* Account the WX pages */
245	st->wx_pages += npages;
246	WARN_ONCE(__supported_pte_mask & _PAGE_NX,
247		  "x86/mm: Found insecure W+X mapping at address %pS\n",
248		  (void *)st->start_address);
249}
250
251static void effective_prot(struct ptdump_state *pt_st, int level, u64 val)
252{
253	struct pg_state *st = container_of(pt_st, struct pg_state, ptdump);
254	pgprotval_t prot = val & PTE_FLAGS_MASK;
255	pgprotval_t effective;
256
257	if (level > 0) {
258		pgprotval_t higher_prot = st->prot_levels[level - 1];
259
260		effective = (higher_prot & prot & (_PAGE_USER | _PAGE_RW)) |
261			    ((higher_prot | prot) & _PAGE_NX);
262	} else {
263		effective = prot;
264	}
265
266	st->prot_levels[level] = effective;
267}
268
269/*
270 * This function gets called on a break in a continuous series
271 * of PTE entries; the next one is different so we need to
272 * print what we collected so far.
273 */
274static void note_page(struct ptdump_state *pt_st, unsigned long addr, int level,
275		      u64 val)
276{
277	struct pg_state *st = container_of(pt_st, struct pg_state, ptdump);
278	pgprotval_t new_prot, new_eff;
279	pgprotval_t cur, eff;
280	static const char units[] = "BKMGTPE";
281	struct seq_file *m = st->seq;
282
283	new_prot = val & PTE_FLAGS_MASK;
284	if (!val)
285		new_eff = 0;
286	else
287		new_eff = st->prot_levels[level];
288
289	/*
290	 * If we have a "break" in the series, we need to flush the state that
291	 * we have now. "break" is either changing perms, levels or
292	 * address space marker.
293	 */
294	cur = st->current_prot;
295	eff = st->effective_prot;
296
297	if (st->level == -1) {
298		/* First entry */
299		st->current_prot = new_prot;
300		st->effective_prot = new_eff;
301		st->level = level;
302		st->marker = address_markers;
303		st->lines = 0;
304		pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
305				   st->marker->name);
306	} else if (new_prot != cur || new_eff != eff || level != st->level ||
307		   addr >= st->marker[1].start_address) {
308		const char *unit = units;
309		unsigned long delta;
310		int width = sizeof(unsigned long) * 2;
 
311
312		if (st->check_wx && (eff & _PAGE_RW) && !(eff & _PAGE_NX))
313			note_wx(st, addr);
 
 
 
 
 
 
314
315		/*
316		 * Now print the actual finished series
317		 */
318		if (!st->marker->max_lines ||
319		    st->lines < st->marker->max_lines) {
320			pt_dump_seq_printf(m, st->to_dmesg,
321					   "0x%0*lx-0x%0*lx   ",
322					   width, st->start_address,
323					   width, addr);
324
325			delta = addr - st->start_address;
326			while (!(delta & 1023) && unit[1]) {
327				delta >>= 10;
328				unit++;
329			}
330			pt_dump_cont_printf(m, st->to_dmesg, "%9lu%c ",
331					    delta, *unit);
332			printk_prot(m, st->current_prot, st->level,
333				    st->to_dmesg);
334		}
335		st->lines++;
336
337		/*
338		 * We print markers for special areas of address space,
339		 * such as the start of vmalloc space etc.
340		 * This helps in the interpretation.
341		 */
342		if (addr >= st->marker[1].start_address) {
343			if (st->marker->max_lines &&
344			    st->lines > st->marker->max_lines) {
345				unsigned long nskip =
346					st->lines - st->marker->max_lines;
347				pt_dump_seq_printf(m, st->to_dmesg,
348						   "... %lu entr%s skipped ... \n",
349						   nskip,
350						   nskip == 1 ? "y" : "ies");
351			}
352			st->marker++;
353			st->lines = 0;
354			pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
355					   st->marker->name);
356		}
357
358		st->start_address = addr;
359		st->current_prot = new_prot;
360		st->effective_prot = new_eff;
361		st->level = level;
362	}
363}
364
365static void ptdump_walk_pgd_level_core(struct seq_file *m,
366				       struct mm_struct *mm, pgd_t *pgd,
367				       bool checkwx, bool dmesg)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
368{
369	const struct ptdump_range ptdump_ranges[] = {
370#ifdef CONFIG_X86_64
371	{0, PTRS_PER_PGD * PGD_LEVEL_MULT / 2},
372	{GUARD_HOLE_END_ADDR, ~0UL},
373#else
374	{0, ~0UL},
375#endif
376	{0, 0}
377};
 
 
 
 
 
 
378
379	struct pg_state st = {
380		.ptdump = {
381			.note_page	= note_page,
382			.effective_prot = effective_prot,
383			.range		= ptdump_ranges
384		},
385		.level = -1,
386		.to_dmesg	= dmesg,
387		.check_wx	= checkwx,
388		.seq		= m
389	};
 
 
 
 
 
390
391	ptdump_walk_pgd(&st.ptdump, mm, pgd);
 
392
 
 
 
393	if (!checkwx)
394		return;
395	if (st.wx_pages)
396		pr_info("x86/mm: Checked W+X mappings: FAILED, %lu W+X pages found.\n",
397			st.wx_pages);
398	else
399		pr_info("x86/mm: Checked W+X mappings: passed, no W+X pages found.\n");
400}
401
402void ptdump_walk_pgd_level(struct seq_file *m, struct mm_struct *mm)
403{
404	ptdump_walk_pgd_level_core(m, mm, mm->pgd, false, true);
405}
406
407void ptdump_walk_pgd_level_debugfs(struct seq_file *m, struct mm_struct *mm,
408				   bool user)
409{
410	pgd_t *pgd = mm->pgd;
411#ifdef CONFIG_PAGE_TABLE_ISOLATION
412	if (user && boot_cpu_has(X86_FEATURE_PTI))
413		pgd = kernel_to_user_pgdp(pgd);
414#endif
415	ptdump_walk_pgd_level_core(m, mm, pgd, false, false);
416}
417EXPORT_SYMBOL_GPL(ptdump_walk_pgd_level_debugfs);
418
419void ptdump_walk_user_pgd_level_checkwx(void)
420{
421#ifdef CONFIG_PAGE_TABLE_ISOLATION
422	pgd_t *pgd = INIT_PGD;
423
424	if (!(__supported_pte_mask & _PAGE_NX) ||
425	    !boot_cpu_has(X86_FEATURE_PTI))
426		return;
427
428	pr_info("x86/mm: Checking user space page tables\n");
429	pgd = kernel_to_user_pgdp(pgd);
430	ptdump_walk_pgd_level_core(NULL, &init_mm, pgd, true, false);
431#endif
432}
 
433
434void ptdump_walk_pgd_level_checkwx(void)
435{
436	ptdump_walk_pgd_level_core(NULL, &init_mm, INIT_PGD, true, false);
437}
438
439static int __init pt_dump_init(void)
440{
441	/*
442	 * Various markers are not compile-time constants, so assign them
443	 * here.
444	 */
445#ifdef CONFIG_X86_64
446	address_markers[LOW_KERNEL_NR].start_address = PAGE_OFFSET;
447	address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
448	address_markers[VMEMMAP_START_NR].start_address = VMEMMAP_START;
449#ifdef CONFIG_MODIFY_LDT_SYSCALL
450	address_markers[LDT_NR].start_address = LDT_BASE_ADDR;
451#endif
452#ifdef CONFIG_KASAN
453	address_markers[KASAN_SHADOW_START_NR].start_address = KASAN_SHADOW_START;
454	address_markers[KASAN_SHADOW_END_NR].start_address = KASAN_SHADOW_END;
455#endif
456#endif
457#ifdef CONFIG_X86_32
 
458	address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
459	address_markers[VMALLOC_END_NR].start_address = VMALLOC_END;
460# ifdef CONFIG_HIGHMEM
461	address_markers[PKMAP_BASE_NR].start_address = PKMAP_BASE;
462# endif
463	address_markers[FIXADDR_START_NR].start_address = FIXADDR_START;
464	address_markers[CPU_ENTRY_AREA_NR].start_address = CPU_ENTRY_AREA_BASE;
465# ifdef CONFIG_MODIFY_LDT_SYSCALL
466	address_markers[LDT_NR].start_address = LDT_BASE_ADDR;
467# endif
468#endif
 
469	return 0;
470}
 
471__initcall(pt_dump_init);