[deliverable/linux.git] / arch / x86 / mm / dump_pagetables.c

/*
 * Debug helper to dump the current kernel pagetables of the system
 * so that we can see what the various memory ranges are set to.
 *
 * (C) Copyright 2008 Intel Corporation
 *
 * Author: Arjan van de Ven <arjan@linux.intel.com>
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; version 2
 * of the License.
 */

#include <linux/debugfs.h>
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/seq_file.h>

#include <asm/pgtable.h>

/*
 * The dumper groups pagetable entries of the same type into one, and for
 * that it needs to keep some state when walking, and flush this state
 * when a "break" in the continuity is found.
 */
struct pg_state {
	int level;
	pgprot_t current_prot;
	unsigned long start_address;
	unsigned long current_address;
	const struct addr_marker *marker;
	unsigned long lines;
	bool to_dmesg;
	bool check_wx;
	unsigned long wx_pages;
};

struct addr_marker {
	unsigned long start_address;
	const char *name;
	unsigned long max_lines;
};

/* indices for address_markers; keep sync'd w/ address_markers below */
enum address_markers_idx {
	USER_SPACE_NR = 0,
#ifdef CONFIG_X86_64
	KERNEL_SPACE_NR,
	LOW_KERNEL_NR,
	VMALLOC_START_NR,
	VMEMMAP_START_NR,
# ifdef CONFIG_X86_ESPFIX64
	ESPFIX_START_NR,
# endif
	HIGH_KERNEL_NR,
	MODULES_VADDR_NR,
	MODULES_END_NR,
#else
	KERNEL_SPACE_NR,
	VMALLOC_START_NR,
	VMALLOC_END_NR,
# ifdef CONFIG_HIGHMEM
	PKMAP_BASE_NR,
# endif
	FIXADDR_START_NR,
#endif
};

/* Address space markers hints */
static struct addr_marker address_markers[] = {
	{ 0, "User Space" },
#ifdef CONFIG_X86_64
	{ 0x8000000000000000UL, "Kernel Space" },
	{ 0/* PAGE_OFFSET */,   "Low Kernel Mapping" },
	{ 0/* VMALLOC_START */, "vmalloc() Area" },
	{ 0/* VMEMMAP_START */, "Vmemmap" },
# ifdef CONFIG_X86_ESPFIX64
	{ ESPFIX_BASE_ADDR,	"ESPfix Area", 16 },
# endif
# ifdef CONFIG_EFI
	{ EFI_VA_END,		"EFI Runtime Services" },
# endif
	{ __START_KERNEL_map,   "High Kernel Mapping" },
	{ MODULES_VADDR,        "Modules" },
	{ MODULES_END,          "End Modules" },
#else
	{ PAGE_OFFSET,          "Kernel Mapping" },
	{ 0/* VMALLOC_START */, "vmalloc() Area" },
	{ 0/*VMALLOC_END*/,     "vmalloc() End" },
# ifdef CONFIG_HIGHMEM
	{ 0/*PKMAP_BASE*/,      "Persistent kmap() Area" },
# endif
	{ 0/*FIXADDR_START*/,   "Fixmap Area" },
#endif
	{ -1, NULL }		/* End of list */
};

/* Multipliers for offsets within the PTEs */
#define PTE_LEVEL_MULT (PAGE_SIZE)
#define PMD_LEVEL_MULT (PTRS_PER_PTE * PTE_LEVEL_MULT)
#define PUD_LEVEL_MULT (PTRS_PER_PMD * PMD_LEVEL_MULT)
#define PGD_LEVEL_MULT (PTRS_PER_PUD * PUD_LEVEL_MULT)

#define pt_dump_seq_printf(m, to_dmesg, fmt, args...)		\
({								\
	if (to_dmesg)					\
		printk(KERN_INFO fmt, ##args);			\
	else							\
		if (m)						\
			seq_printf(m, fmt, ##args);		\
})

#define pt_dump_cont_printf(m, to_dmesg, fmt, args...)		\
({								\
	if (to_dmesg)					\
		printk(KERN_CONT fmt, ##args);			\
	else							\
		if (m)						\
			seq_printf(m, fmt, ##args);		\
})

/*
 * Print a readable form of a pgprot_t to the seq_file
 */
static void printk_prot(struct seq_file *m, pgprot_t prot, int level, bool dmsg)
{
	pgprotval_t pr = pgprot_val(prot);
	static const char * const level_name[] =
		{ "cr3", "pgd", "pud", "pmd", "pte" };

	if (!pgprot_val(prot)) {
		/* Not present */
		pt_dump_cont_printf(m, dmsg, "                              ");
	} else {
		if (pr & _PAGE_USER)
			pt_dump_cont_printf(m, dmsg, "USR ");
		else
			pt_dump_cont_printf(m, dmsg, "    ");
		if (pr & _PAGE_RW)
			pt_dump_cont_printf(m, dmsg, "RW ");
		else
			pt_dump_cont_printf(m, dmsg, "ro ");
		if (pr & _PAGE_PWT)
			pt_dump_cont_printf(m, dmsg, "PWT ");
		else
			pt_dump_cont_printf(m, dmsg, "    ");
		if (pr & _PAGE_PCD)
			pt_dump_cont_printf(m, dmsg, "PCD ");
		else
			pt_dump_cont_printf(m, dmsg, "    ");

		/* Bit 7 has a different meaning on level 3 vs 4 */
		if (level <= 3 && pr & _PAGE_PSE)
			pt_dump_cont_printf(m, dmsg, "PSE ");
		else
			pt_dump_cont_printf(m, dmsg, "    ");
		if ((level == 4 && pr & _PAGE_PAT) ||
		    ((level == 3 || level == 2) && pr & _PAGE_PAT_LARGE))
			pt_dump_cont_printf(m, dmsg, "PAT ");
		else
			pt_dump_cont_printf(m, dmsg, "    ");
		if (pr & _PAGE_GLOBAL)
			pt_dump_cont_printf(m, dmsg, "GLB ");
		else
			pt_dump_cont_printf(m, dmsg, "    ");
		if (pr & _PAGE_NX)
			pt_dump_cont_printf(m, dmsg, "NX ");
		else
			pt_dump_cont_printf(m, dmsg, "x  ");
	}
	pt_dump_cont_printf(m, dmsg, "%s\n", level_name[level]);
}

/*
 * On 64 bits, sign-extend the 48 bit address to 64 bit
 */
static unsigned long normalize_addr(unsigned long u)
{
#ifdef CONFIG_X86_64
	return (signed long)(u << 16) >> 16;
#else
	return u;
#endif
}

/*
 * This function gets called on a break in a continuous series
 * of PTE entries; the next one is different so we need to
 * print what we collected so far.
 */
static void note_page(struct seq_file *m, struct pg_state *st,
		      pgprot_t new_prot, int level)
{
	pgprotval_t prot, cur;
	static const char units[] = "BKMGTPE";

	/*
	 * If we have a "break" in the series, we need to flush the state that
	 * we have now. "break" is either changing perms, levels or
	 * address space marker.
	 */
	prot = pgprot_val(new_prot);
	cur = pgprot_val(st->current_prot);

	if (!st->level) {
		/* First entry */
		st->current_prot = new_prot;
		st->level = level;
		st->marker = address_markers;
		st->lines = 0;
		pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
				   st->marker->name);
	} else if (prot != cur || level != st->level ||
		   st->current_address >= st->marker[1].start_address) {
		const char *unit = units;
		unsigned long delta;
		int width = sizeof(unsigned long) * 2;
		pgprotval_t pr = pgprot_val(st->current_prot);

		if (st->check_wx && (pr & _PAGE_RW) && !(pr & _PAGE_NX)) {
			WARN_ONCE(1,
				  "x86/mm: Found insecure W+X mapping at address %p/%pS\n",
				  (void *)st->start_address,
				  (void *)st->start_address);
			st->wx_pages += (st->current_address -
					 st->start_address) / PAGE_SIZE;
		}

		/*
		 * Now print the actual finished series
		 */
		if (!st->marker->max_lines ||
		    st->lines < st->marker->max_lines) {
			pt_dump_seq_printf(m, st->to_dmesg,
					   "0x%0*lx-0x%0*lx   ",
					   width, st->start_address,
					   width, st->current_address);

			delta = st->current_address - st->start_address;
			while (!(delta & 1023) && unit[1]) {
				delta >>= 10;
				unit++;
			}
			pt_dump_cont_printf(m, st->to_dmesg, "%9lu%c ",
					    delta, *unit);
			printk_prot(m, st->current_prot, st->level,
				    st->to_dmesg);
		}
		st->lines++;

		/*
		 * We print markers for special areas of address space,
		 * such as the start of vmalloc space etc.
		 * This helps in the interpretation.
		 */
		if (st->current_address >= st->marker[1].start_address) {
			if (st->marker->max_lines &&
			    st->lines > st->marker->max_lines) {
				unsigned long nskip =
					st->lines - st->marker->max_lines;
				pt_dump_seq_printf(m, st->to_dmesg,
						   "... %lu entr%s skipped ... \n",
						   nskip,
						   nskip == 1 ? "y" : "ies");
			}
			st->marker++;
			st->lines = 0;
			pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
					   st->marker->name);
		}

		st->start_address = st->current_address;
		st->current_prot = new_prot;
		st->level = level;
	}
}

static void walk_pte_level(struct seq_file *m, struct pg_state *st, pmd_t addr,
							unsigned long P)
{
	int i;
	pte_t *start;
	pgprotval_t prot;

	start = (pte_t *) pmd_page_vaddr(addr);
	for (i = 0; i < PTRS_PER_PTE; i++) {
		prot = pte_flags(*start);
		st->current_address = normalize_addr(P + i * PTE_LEVEL_MULT);
		note_page(m, st, __pgprot(prot), 4);
		start++;
	}
}

#if PTRS_PER_PMD > 1

static void walk_pmd_level(struct seq_file *m, struct pg_state *st, pud_t addr,
							unsigned long P)
{
	int i;
	pmd_t *start;
	pgprotval_t prot;

	start = (pmd_t *) pud_page_vaddr(addr);
	for (i = 0; i < PTRS_PER_PMD; i++) {
		st->current_address = normalize_addr(P + i * PMD_LEVEL_MULT);
		if (!pmd_none(*start)) {
			if (pmd_large(*start) || !pmd_present(*start)) {
				prot = pmd_flags(*start);
				note_page(m, st, __pgprot(prot), 3);
			} else {
				walk_pte_level(m, st, *start,
					       P + i * PMD_LEVEL_MULT);
			}
		} else
			note_page(m, st, __pgprot(0), 3);
		start++;
	}
}

#else
#define walk_pmd_level(m,s,a,p) walk_pte_level(m,s,__pmd(pud_val(a)),p)
#define pud_large(a) pmd_large(__pmd(pud_val(a)))
#define pud_none(a)  pmd_none(__pmd(pud_val(a)))
#endif

#if PTRS_PER_PUD > 1

static void walk_pud_level(struct seq_file *m, struct pg_state *st, pgd_t addr,
							unsigned long P)
{
	int i;
	pud_t *start;
	pgprotval_t prot;

	start = (pud_t *) pgd_page_vaddr(addr);

	for (i = 0; i < PTRS_PER_PUD; i++) {
		st->current_address = normalize_addr(P + i * PUD_LEVEL_MULT);
		if (!pud_none(*start)) {
			if (pud_large(*start) || !pud_present(*start)) {
				prot = pud_flags(*start);
				note_page(m, st, __pgprot(prot), 2);
			} else {
				walk_pmd_level(m, st, *start,
					       P + i * PUD_LEVEL_MULT);
			}
		} else
			note_page(m, st, __pgprot(0), 2);

		start++;
	}
}

#else
#define walk_pud_level(m,s,a,p) walk_pmd_level(m,s,__pud(pgd_val(a)),p)
#define pgd_large(a) pud_large(__pud(pgd_val(a)))
#define pgd_none(a)  pud_none(__pud(pgd_val(a)))
#endif

static inline bool is_hypervisor_range(int idx)
{
#ifdef CONFIG_X86_64
	/*
	 * ffff800000000000 - ffff87ffffffffff is reserved for
	 * the hypervisor.
	 */
	return	(idx >= pgd_index(__PAGE_OFFSET) - 16) &&
		(idx <  pgd_index(__PAGE_OFFSET));
#else
	return false;
#endif
}

static void ptdump_walk_pgd_level_core(struct seq_file *m, pgd_t *pgd,
				       bool checkwx)
{
#ifdef CONFIG_X86_64
	pgd_t *start = (pgd_t *) &init_level4_pgt;
#else
	pgd_t *start = swapper_pg_dir;
#endif
	pgprotval_t prot;
	int i;
	struct pg_state st = {};

	if (pgd) {
		start = pgd;
		st.to_dmesg = true;
	}

	st.check_wx = checkwx;
	if (checkwx)
		st.wx_pages = 0;

	for (i = 0; i < PTRS_PER_PGD; i++) {
		st.current_address = normalize_addr(i * PGD_LEVEL_MULT);
		if (!pgd_none(*start) && !is_hypervisor_range(i)) {
			if (pgd_large(*start) || !pgd_present(*start)) {
				prot = pgd_flags(*start);
				note_page(m, &st, __pgprot(prot), 1);
			} else {
				walk_pud_level(m, &st, *start,
					       i * PGD_LEVEL_MULT);
			}
		} else
			note_page(m, &st, __pgprot(0), 1);

		start++;
	}

	/* Flush out the last page */
	st.current_address = normalize_addr(PTRS_PER_PGD*PGD_LEVEL_MULT);
	note_page(m, &st, __pgprot(0), 0);
	if (!checkwx)
		return;
	if (st.wx_pages)
		pr_info("x86/mm: Checked W+X mappings: FAILED, %lu W+X pages found.\n",
			st.wx_pages);
	else
		pr_info("x86/mm: Checked W+X mappings: passed, no W+X pages found.\n");
}

void ptdump_walk_pgd_level(struct seq_file *m, pgd_t *pgd)
{
	ptdump_walk_pgd_level_core(m, pgd, false);
}
EXPORT_SYMBOL_GPL(ptdump_walk_pgd_level);

void ptdump_walk_pgd_level_checkwx(void)
{
	ptdump_walk_pgd_level_core(NULL, NULL, true);
}

static int __init pt_dump_init(void)
{
	/*
	 * Various markers are not compile-time constants, so assign them
	 * here.
	 */
#ifdef CONFIG_X86_64
	address_markers[LOW_KERNEL_NR].start_address = PAGE_OFFSET;
	address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
	address_markers[VMEMMAP_START_NR].start_address = VMEMMAP_START;
#endif
#ifdef CONFIG_X86_32
	address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
	address_markers[VMALLOC_END_NR].start_address = VMALLOC_END;
# ifdef CONFIG_HIGHMEM
	address_markers[PKMAP_BASE_NR].start_address = PKMAP_BASE;
# endif
	address_markers[FIXADDR_START_NR].start_address = FIXADDR_START;
#endif

	return 0;
}

__initcall(pt_dump_init);
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Arjan van de Ven <arjan@linux.intel.com>");
MODULE_DESCRIPTION("Kernel debugging helper that dumps pagetables");
Commit	Line	Data
926e5392 AV	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
fe770bf0 PA	15	#include <linux/debugfs.h>
fe770bf0 PA	16	#include <linux/mm.h>
926e5392 AV	17	#include <linux/module.h>
926e5392 AV	18	#include <linux/seq_file.h>
926e5392 AV	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	*/
	27	struct pg_state {
	28	int level;
	29	pgprot_t current_prot;
	30	unsigned long start_address;
	31	unsigned long current_address;
fe770bf0	32	const struct addr_marker *marker;
3891a04a	33	unsigned long lines;
ef6bea6d	34	bool to_dmesg;
e1a58320 SS	35	bool check_wx;
e1a58320 SS	36	unsigned long wx_pages;
926e5392 AV	37	};
926e5392 AV	38
fe770bf0 PA	39	struct addr_marker {
	40	unsigned long start_address;
	41	const char *name;
3891a04a	42	unsigned long max_lines;
fe770bf0 PA	43	};
fe770bf0 PA	44
92851e2f AS	45	/* indices for address_markers; keep sync'd w/ address_markers below */
	46	enum 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,
8a5a5d15	53	# ifdef CONFIG_X86_ESPFIX64
3891a04a	54	ESPFIX_START_NR,
8a5a5d15	55	# endif
92851e2f AS	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
fe770bf0 PA	70	/* Address space markers hints */
	71	static struct addr_marker address_markers[] = {
	72	{ 0, "User Space" },
	73	#ifdef CONFIG_X86_64
	74	{ 0x8000000000000000UL, "Kernel Space" },
0483e1fa TG	75	{ 0/* PAGE_OFFSET */, "Low Kernel Mapping" },
	76	{ 0/* VMALLOC_START */, "vmalloc() Area" },
	77	{ 0/* VMEMMAP_START */, "Vmemmap" },
8a5a5d15	78	# ifdef CONFIG_X86_ESPFIX64
3891a04a	79	{ ESPFIX_BASE_ADDR, "ESPfix Area", 16 },
8266e31e MK	80	# endif
	81	# ifdef CONFIG_EFI
	82	{ EFI_VA_END, "EFI Runtime Services" },
8a5a5d15	83	# endif
fe770bf0	84	{ __START_KERNEL_map, "High Kernel Mapping" },
9a79cf9c YL	85	{ MODULES_VADDR, "Modules" },
9a79cf9c YL	86	{ MODULES_END, "End Modules" },
fe770bf0 PA	87	#else
	88	{ PAGE_OFFSET, "Kernel Mapping" },
	89	{ 0/* VMALLOC_START */, "vmalloc() Area" },
	90	{ 0/VMALLOC_END/, "vmalloc() End" },
	91	# ifdef CONFIG_HIGHMEM
173ae9ba	92	{ 0/PKMAP_BASE/, "Persistent kmap() Area" },
fe770bf0 PA	93	# endif
	94	{ 0/FIXADDR_START/, "Fixmap Area" },
	95	#endif
	96	{ -1, NULL } /* End of list */
	97	};
926e5392	98
fe770bf0 PA	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)
926e5392	104
ef6bea6d BP	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
926e5392 AV	123	/*
	124	* Print a readable form of a pgprot_t to the seq_file
	125	*/
ef6bea6d	126	static void printk_prot(struct seq_file *m, pgprot_t prot, int level, bool dmsg)
926e5392	127	{
fe770bf0 PA	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 */
f439c429	134	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0 PA	135	} else {
fe770bf0 PA	136	if (pr & _PAGE_USER)
ef6bea6d	137	pt_dump_cont_printf(m, dmsg, "USR ");
926e5392	138	else
ef6bea6d	139	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	140	if (pr & _PAGE_RW)
ef6bea6d	141	pt_dump_cont_printf(m, dmsg, "RW ");
fe770bf0	142	else
ef6bea6d	143	pt_dump_cont_printf(m, dmsg, "ro ");
fe770bf0	144	if (pr & _PAGE_PWT)
ef6bea6d	145	pt_dump_cont_printf(m, dmsg, "PWT ");
fe770bf0	146	else
ef6bea6d	147	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	148	if (pr & _PAGE_PCD)
ef6bea6d	149	pt_dump_cont_printf(m, dmsg, "PCD ");
926e5392	150	else
ef6bea6d	151	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	152
f439c429 JG	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))
da25e628	160	pt_dump_cont_printf(m, dmsg, "PAT ");
f439c429 JG	161	else
f439c429 JG	162	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	163	if (pr & _PAGE_GLOBAL)
ef6bea6d	164	pt_dump_cont_printf(m, dmsg, "GLB ");
fe770bf0	165	else
ef6bea6d	166	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	167	if (pr & _PAGE_NX)
ef6bea6d	168	pt_dump_cont_printf(m, dmsg, "NX ");
fe770bf0	169	else
ef6bea6d	170	pt_dump_cont_printf(m, dmsg, "x ");
926e5392	171	}
ef6bea6d	172	pt_dump_cont_printf(m, dmsg, "%s\n", level_name[level]);
926e5392 AV	173	}
	174
	175	/*
fe770bf0	176	* On 64 bits, sign-extend the 48 bit address to 64 bit
926e5392	177	*/
fe770bf0	178	static unsigned long normalize_addr(unsigned long u)
926e5392	179	{
fe770bf0 PA	180	#ifdef CONFIG_X86_64
	181	return (signed long)(u << 16) >> 16;
	182	#else
926e5392	183	return u;
fe770bf0	184	#endif
926e5392 AV	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	*/
	192	static void note_page(struct seq_file m, struct pg_state st,
fe770bf0	193	pgprot_t new_prot, int level)
926e5392	194	{
fe770bf0	195	pgprotval_t prot, cur;
3891a04a	196	static const char units[] = "BKMGTPE";
926e5392 AV	197
	198	/*
	199	* If we have a "break" in the series, we need to flush the state that
fe770bf0 PA	200	* we have now. "break" is either changing perms, levels or
fe770bf0 PA	201	* address space marker.
926e5392	202	*/
da25e628 TK	203	prot = pgprot_val(new_prot);
da25e628 TK	204	cur = pgprot_val(st->current_prot);
926e5392	205
fe770bf0 PA	206	if (!st->level) {
	207	/* First entry */
	208	st->current_prot = new_prot;
	209	st->level = level;
	210	st->marker = address_markers;
3891a04a	211	st->lines = 0;
ef6bea6d BP	212	pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
ef6bea6d BP	213	st->marker->name);
fe770bf0 PA	214	} else if (prot != cur \|\| level != st->level \|\|
	215	st->current_address >= st->marker[1].start_address) {
	216	const char *unit = units;
926e5392	217	unsigned long delta;
6424fb38	218	int width = sizeof(unsigned long) * 2;
e1a58320 SS	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	}
926e5392	229
926e5392 AV	230	/*
	231	* Now print the actual finished series
	232	*/
3891a04a PA	233	if (!st->marker->max_lines \|\|
	234	st->lines < st->marker->max_lines) {
	235	pt_dump_seq_printf(m, st->to_dmesg,
	236	"0x%0lx-0x%0lx ",
	237	width, st->start_address,
	238	width, st->current_address);
926e5392	239
3891a04a PA	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);
926e5392	249	}
3891a04a	250	st->lines++;
fe770bf0 PA	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) {
3891a04a PA	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	}
fe770bf0	267	st->marker++;
3891a04a	268	st->lines = 0;
ef6bea6d BP	269	pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
ef6bea6d BP	270	st->marker->name);
926e5392	271	}
fe770bf0	272
926e5392 AV	273	st->start_address = st->current_address;
	274	st->current_prot = new_prot;
	275	st->level = level;
fe770bf0	276	}
926e5392 AV	277	}
926e5392 AV	278
fe770bf0	279	static void walk_pte_level(struct seq_file m, struct pg_state st, pmd_t addr,
926e5392 AV	280	unsigned long P)
	281	{
	282	int i;
	283	pte_t *start;
da25e628	284	pgprotval_t prot;
926e5392 AV	285
	286	start = (pte_t *) pmd_page_vaddr(addr);
	287	for (i = 0; i < PTRS_PER_PTE; i++) {
da25e628	288	prot = pte_flags(*start);
fe770bf0	289	st->current_address = normalize_addr(P + i * PTE_LEVEL_MULT);
da25e628	290	note_page(m, st, __pgprot(prot), 4);
926e5392 AV	291	start++;
	292	}
	293	}
	294
fe770bf0	295	#if PTRS_PER_PMD > 1
926e5392	296
fe770bf0	297	static void walk_pmd_level(struct seq_file m, struct pg_state st, pud_t addr,
926e5392 AV	298	unsigned long P)
	299	{
	300	int i;
	301	pmd_t *start;
da25e628	302	pgprotval_t prot;
926e5392 AV	303
	304	start = (pmd_t *) pud_page_vaddr(addr);
	305	for (i = 0; i < PTRS_PER_PMD; i++) {
fe770bf0	306	st->current_address = normalize_addr(P + i * PMD_LEVEL_MULT);
926e5392	307	if (!pmd_none(*start)) {
da25e628 TK	308	if (pmd_large(start) \|\| !pmd_present(start)) {
da25e628 TK	309	prot = pmd_flags(*start);
926e5392	310	note_page(m, st, __pgprot(prot), 3);
da25e628	311	} else {
fe770bf0 PA	312	walk_pte_level(m, st, *start,
fe770bf0 PA	313	P + i * PMD_LEVEL_MULT);
da25e628	314	}
926e5392 AV	315	} else
	316	note_page(m, st, __pgprot(0), 3);
	317	start++;
	318	}
	319	}
	320
fe770bf0 PA	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
926e5392	326
fe770bf0 PA	327	#if PTRS_PER_PUD > 1
	328
	329	static void walk_pud_level(struct seq_file m, struct pg_state st, pgd_t addr,
926e5392 AV	330	unsigned long P)
	331	{
	332	int i;
	333	pud_t *start;
da25e628	334	pgprotval_t prot;
926e5392 AV	335
	336	start = (pud_t *) pgd_page_vaddr(addr);
	337
	338	for (i = 0; i < PTRS_PER_PUD; i++) {
fe770bf0	339	st->current_address = normalize_addr(P + i * PUD_LEVEL_MULT);
926e5392	340	if (!pud_none(*start)) {
da25e628 TK	341	if (pud_large(start) \|\| !pud_present(start)) {
da25e628 TK	342	prot = pud_flags(*start);
926e5392	343	note_page(m, st, __pgprot(prot), 2);
da25e628	344	} else {
fe770bf0 PA	345	walk_pmd_level(m, st, *start,
fe770bf0 PA	346	P + i * PUD_LEVEL_MULT);
da25e628	347	}
926e5392 AV	348	} else
	349	note_page(m, st, __pgprot(0), 2);
	350
	351	start++;
	352	}
	353	}
	354
fe770bf0 PA	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
f4e342c8 BO	361	static inline bool is_hypervisor_range(int idx)
f4e342c8 BO	362	{
b176862f	363	#ifdef CONFIG_X86_64
f4e342c8 BO	364	/*
	365	* ffff800000000000 - ffff87ffffffffff is reserved for
	366	* the hypervisor.
	367	*/
b176862f BP	368	return (idx >= pgd_index(__PAGE_OFFSET) - 16) &&
b176862f BP	369	(idx < pgd_index(__PAGE_OFFSET));
f4e342c8	370	#else
b176862f	371	return false;
f4e342c8	372	#endif
b176862f	373	}
f4e342c8	374
e1a58320 SS	375	static void ptdump_walk_pgd_level_core(struct seq_file m, pgd_t pgd,
e1a58320 SS	376	bool checkwx)
926e5392	377	{
fe770bf0	378	#ifdef CONFIG_X86_64
926e5392	379	pgd_t start = (pgd_t ) &init_level4_pgt;
fe770bf0 PA	380	#else
	381	pgd_t *start = swapper_pg_dir;
	382	#endif
da25e628	383	pgprotval_t prot;
926e5392	384	int i;
ef6bea6d	385	struct pg_state st = {};
926e5392	386
ef6bea6d BP	387	if (pgd) {
	388	start = pgd;
	389	st.to_dmesg = true;
	390	}
926e5392	391
e1a58320 SS	392	st.check_wx = checkwx;
	393	if (checkwx)
	394	st.wx_pages = 0;
	395
926e5392	396	for (i = 0; i < PTRS_PER_PGD; i++) {
fe770bf0	397	st.current_address = normalize_addr(i * PGD_LEVEL_MULT);
f4e342c8	398	if (!pgd_none(*start) && !is_hypervisor_range(i)) {
da25e628 TK	399	if (pgd_large(start) \|\| !pgd_present(start)) {
da25e628 TK	400	prot = pgd_flags(*start);
fe770bf0	401	note_page(m, &st, __pgprot(prot), 1);
da25e628	402	} else {
fe770bf0 PA	403	walk_pud_level(m, &st, *start,
fe770bf0 PA	404	i * PGD_LEVEL_MULT);
da25e628	405	}
fe770bf0	406	} else
926e5392	407	note_page(m, &st, __pgprot(0), 1);
fe770bf0	408
926e5392 AV	409	start++;
926e5392 AV	410	}
fe770bf0 PA	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);
e1a58320 SS	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
	424	void ptdump_walk_pgd_level(struct seq_file m, pgd_t pgd)
	425	{
	426	ptdump_walk_pgd_level_core(m, pgd, false);
926e5392	427	}
8609d1b5	428	EXPORT_SYMBOL_GPL(ptdump_walk_pgd_level);
926e5392	429
e1a58320 SS	430	void ptdump_walk_pgd_level_checkwx(void)
	431	{
	432	ptdump_walk_pgd_level_core(NULL, NULL, true);
	433	}
	434
8609d1b5	435	static int __init pt_dump_init(void)
926e5392	436	{
0483e1fa TG	437	/*
	438	* Various markers are not compile-time constants, so assign them
	439	* here.
	440	*/
	441	#ifdef CONFIG_X86_64
	442	address_markers[LOW_KERNEL_NR].start_address = PAGE_OFFSET;
	443	address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
	444	address_markers[VMEMMAP_START_NR].start_address = VMEMMAP_START;
	445	#endif
fe770bf0	446	#ifdef CONFIG_X86_32
92851e2f AS	447	address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
92851e2f AS	448	address_markers[VMALLOC_END_NR].start_address = VMALLOC_END;
fe770bf0	449	# ifdef CONFIG_HIGHMEM
92851e2f	450	address_markers[PKMAP_BASE_NR].start_address = PKMAP_BASE;
fe770bf0	451	# endif
92851e2f	452	address_markers[FIXADDR_START_NR].start_address = FIXADDR_START;
fe770bf0 PA	453	#endif
fe770bf0 PA	454
926e5392 AV	455	return 0;
	456	}
	457
	458	__initcall(pt_dump_init);
	459	MODULE_LICENSE("GPL");
	460	MODULE_AUTHOR("Arjan van de Ven <arjan@linux.intel.com>");
	461	MODULE_DESCRIPTION("Kernel debugging helper that dumps pagetables");