[linux-2.6-block.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/init.h>
#include <linux/sched.h>
#include <linux/seq_file.h>

#include <asm/kasan.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_KASAN
	KASAN_SHADOW_START_NR,
	KASAN_SHADOW_END_NR,
#endif
# 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_KASAN
	{ KASAN_SHADOW_START,	"KASAN shadow" },
	{ KASAN_SHADOW_END,	"KASAN shadow end" },
#endif
# 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 P4D_LEVEL_MULT (PTRS_PER_PUD * PUD_LEVEL_MULT)
#define PGD_LEVEL_MULT (PTRS_PER_P4D * P4D_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", "p4d", "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 <= 4 && pr & _PAGE_PSE)
			pt_dump_cont_printf(m, dmsg, "PSE ");
		else
			pt_dump_cont_printf(m, dmsg, "    ");
		if ((level == 5 && pr & _PAGE_PAT) ||
		    ((level == 4 || level == 3) && 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)
{
	int shift;
	if (!IS_ENABLED(CONFIG_X86_64))
		return u;

	shift = 64 - (__VIRTUAL_MASK_SHIFT + 1);
	return (signed long)(u << shift) >> shift;
}

/*
 * 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), 5);
		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), 4);
			} else {
				walk_pte_level(m, st, *start,
					       P + i * PMD_LEVEL_MULT);
			}
		} else
			note_page(m, st, __pgprot(0), 4);
		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

/*
 * This is an optimization for CONFIG_DEBUG_WX=y + CONFIG_KASAN=y
 * KASAN fills page tables with the same values. Since there is no
 * point in checking page table more than once we just skip repeated
 * entries. This saves us dozens of seconds during boot.
 */
static bool pud_already_checked(pud_t *prev_pud, pud_t *pud, bool checkwx)
{
	return checkwx && prev_pud && (pud_val(*prev_pud) == pud_val(*pud));
}

static void walk_pud_level(struct seq_file *m, struct pg_state *st, p4d_t addr, unsigned long P)
{
	int i;
	pud_t *start;
	pgprotval_t prot;
	pud_t *prev_pud = NULL;

	start = (pud_t *)p4d_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) &&
		    !pud_already_checked(prev_pud, start, st->check_wx)) {
			if (pud_large(*start) || !pud_present(*start)) {
				prot = pud_flags(*start);
				note_page(m, st, __pgprot(prot), 3);
			} else {
				walk_pmd_level(m, st, *start,
					       P + i * PUD_LEVEL_MULT);
			}
		} else
			note_page(m, st, __pgprot(0), 3);

		prev_pud = start;
		start++;
	}
}

#else
#define walk_pud_level(m,s,a,p) walk_pmd_level(m,s,__pud(p4d_val(a)),p)
#define p4d_large(a) pud_large(__pud(p4d_val(a)))
#define p4d_none(a)  pud_none(__pud(p4d_val(a)))
#endif

#if PTRS_PER_P4D > 1

static void walk_p4d_level(struct seq_file *m, struct pg_state *st, pgd_t addr, unsigned long P)
{
	int i;
	p4d_t *start;
	pgprotval_t prot;

	start = (p4d_t *)pgd_page_vaddr(addr);

	for (i = 0; i < PTRS_PER_P4D; i++) {
		st->current_address = normalize_addr(P + i * P4D_LEVEL_MULT);
		if (!p4d_none(*start)) {
			if (p4d_large(*start) || !p4d_present(*start)) {
				prot = p4d_flags(*start);
				note_page(m, st, __pgprot(prot), 2);
			} else {
				walk_pud_level(m, st, *start,
					       P + i * P4D_LEVEL_MULT);
			}
		} else
			note_page(m, st, __pgprot(0), 2);

		start++;
	}
}

#else
#define walk_p4d_level(m,s,a,p) walk_pud_level(m,s,__p4d(pgd_val(a)),p)
#define pgd_large(a) p4d_large(__p4d(pgd_val(a)))
#define pgd_none(a)  p4d_none(__p4d(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_top_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_p4d_level(m, &st, *start,
					       i * PGD_LEVEL_MULT);
			}
		} else
			note_page(m, &st, __pgprot(0), 1);

		cond_resched();
		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);
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>
84e629b6	17	#include <linux/init.h>
146fbb76	18	#include <linux/sched.h>
926e5392	19	#include <linux/seq_file.h>
926e5392	20
025205f8	21	#include <asm/kasan.h>
926e5392 AV	22	#include <asm/pgtable.h>
	23
	24	/*
	25	* The dumper groups pagetable entries of the same type into one, and for
	26	* that it needs to keep some state when walking, and flush this state
	27	* when a "break" in the continuity is found.
	28	*/
	29	struct pg_state {
	30	int level;
	31	pgprot_t current_prot;
	32	unsigned long start_address;
	33	unsigned long current_address;
fe770bf0	34	const struct addr_marker *marker;
3891a04a	35	unsigned long lines;
ef6bea6d	36	bool to_dmesg;
e1a58320 SS	37	bool check_wx;
e1a58320 SS	38	unsigned long wx_pages;
926e5392 AV	39	};
926e5392 AV	40
fe770bf0 PA	41	struct addr_marker {
	42	unsigned long start_address;
	43	const char *name;
3891a04a	44	unsigned long max_lines;
fe770bf0 PA	45	};
fe770bf0 PA	46
92851e2f AS	47	/* indices for address_markers; keep sync'd w/ address_markers below */
	48	enum address_markers_idx {
	49	USER_SPACE_NR = 0,
	50	#ifdef CONFIG_X86_64
	51	KERNEL_SPACE_NR,
	52	LOW_KERNEL_NR,
	53	VMALLOC_START_NR,
	54	VMEMMAP_START_NR,
025205f8 AR	55	#ifdef CONFIG_KASAN
	56	KASAN_SHADOW_START_NR,
	57	KASAN_SHADOW_END_NR,
	58	#endif
8a5a5d15	59	# ifdef CONFIG_X86_ESPFIX64
3891a04a	60	ESPFIX_START_NR,
8a5a5d15	61	# endif
92851e2f AS	62	HIGH_KERNEL_NR,
	63	MODULES_VADDR_NR,
	64	MODULES_END_NR,
	65	#else
	66	KERNEL_SPACE_NR,
	67	VMALLOC_START_NR,
	68	VMALLOC_END_NR,
	69	# ifdef CONFIG_HIGHMEM
	70	PKMAP_BASE_NR,
	71	# endif
	72	FIXADDR_START_NR,
	73	#endif
	74	};
	75
fe770bf0 PA	76	/* Address space markers hints */
	77	static struct addr_marker address_markers[] = {
	78	{ 0, "User Space" },
	79	#ifdef CONFIG_X86_64
	80	{ 0x8000000000000000UL, "Kernel Space" },
0483e1fa TG	81	{ 0/* PAGE_OFFSET */, "Low Kernel Mapping" },
	82	{ 0/* VMALLOC_START */, "vmalloc() Area" },
	83	{ 0/* VMEMMAP_START */, "Vmemmap" },
025205f8 AR	84	#ifdef CONFIG_KASAN
	85	{ KASAN_SHADOW_START, "KASAN shadow" },
	86	{ KASAN_SHADOW_END, "KASAN shadow end" },
	87	#endif
8a5a5d15	88	# ifdef CONFIG_X86_ESPFIX64
3891a04a	89	{ ESPFIX_BASE_ADDR, "ESPfix Area", 16 },
8266e31e MK	90	# endif
	91	# ifdef CONFIG_EFI
	92	{ EFI_VA_END, "EFI Runtime Services" },
8a5a5d15	93	# endif
fe770bf0	94	{ __START_KERNEL_map, "High Kernel Mapping" },
9a79cf9c YL	95	{ MODULES_VADDR, "Modules" },
9a79cf9c YL	96	{ MODULES_END, "End Modules" },
fe770bf0 PA	97	#else
	98	{ PAGE_OFFSET, "Kernel Mapping" },
	99	{ 0/* VMALLOC_START */, "vmalloc() Area" },
	100	{ 0/VMALLOC_END/, "vmalloc() End" },
	101	# ifdef CONFIG_HIGHMEM
173ae9ba	102	{ 0/PKMAP_BASE/, "Persistent kmap() Area" },
fe770bf0 PA	103	# endif
	104	{ 0/FIXADDR_START/, "Fixmap Area" },
	105	#endif
	106	{ -1, NULL } /* End of list */
	107	};
926e5392	108
fe770bf0 PA	109	/* Multipliers for offsets within the PTEs */
	110	#define PTE_LEVEL_MULT (PAGE_SIZE)
	111	#define PMD_LEVEL_MULT (PTRS_PER_PTE * PTE_LEVEL_MULT)
	112	#define PUD_LEVEL_MULT (PTRS_PER_PMD * PMD_LEVEL_MULT)
fdd3d8ce	113	#define P4D_LEVEL_MULT (PTRS_PER_PUD * PUD_LEVEL_MULT)
84bbabc3	114	#define PGD_LEVEL_MULT (PTRS_PER_P4D * P4D_LEVEL_MULT)
926e5392	115
ef6bea6d BP	116	#define pt_dump_seq_printf(m, to_dmesg, fmt, args...) \
	117	({ \
	118	if (to_dmesg) \
	119	printk(KERN_INFO fmt, ##args); \
	120	else \
	121	if (m) \
	122	seq_printf(m, fmt, ##args); \
	123	})
	124
	125	#define pt_dump_cont_printf(m, to_dmesg, fmt, args...) \
	126	({ \
	127	if (to_dmesg) \
	128	printk(KERN_CONT fmt, ##args); \
	129	else \
	130	if (m) \
	131	seq_printf(m, fmt, ##args); \
	132	})
	133
926e5392 AV	134	/*
	135	* Print a readable form of a pgprot_t to the seq_file
	136	*/
ef6bea6d	137	static void printk_prot(struct seq_file *m, pgprot_t prot, int level, bool dmsg)
926e5392	138	{
fe770bf0 PA	139	pgprotval_t pr = pgprot_val(prot);
fe770bf0 PA	140	static const char * const level_name[] =
45dcd209	141	{ "cr3", "pgd", "p4d", "pud", "pmd", "pte" };
fe770bf0 PA	142
	143	if (!pgprot_val(prot)) {
	144	/* Not present */
f439c429	145	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0 PA	146	} else {
fe770bf0 PA	147	if (pr & _PAGE_USER)
ef6bea6d	148	pt_dump_cont_printf(m, dmsg, "USR ");
926e5392	149	else
ef6bea6d	150	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	151	if (pr & _PAGE_RW)
ef6bea6d	152	pt_dump_cont_printf(m, dmsg, "RW ");
fe770bf0	153	else
ef6bea6d	154	pt_dump_cont_printf(m, dmsg, "ro ");
fe770bf0	155	if (pr & _PAGE_PWT)
ef6bea6d	156	pt_dump_cont_printf(m, dmsg, "PWT ");
fe770bf0	157	else
ef6bea6d	158	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	159	if (pr & _PAGE_PCD)
ef6bea6d	160	pt_dump_cont_printf(m, dmsg, "PCD ");
926e5392	161	else
ef6bea6d	162	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	163
f439c429	164	/* Bit 7 has a different meaning on level 3 vs 4 */
45dcd209	165	if (level <= 4 && pr & _PAGE_PSE)
f439c429 JG	166	pt_dump_cont_printf(m, dmsg, "PSE ");
	167	else
	168	pt_dump_cont_printf(m, dmsg, " ");
45dcd209 KS	169	if ((level == 5 && pr & _PAGE_PAT) \|\|
45dcd209 KS	170	((level == 4 \|\| level == 3) && pr & _PAGE_PAT_LARGE))
da25e628	171	pt_dump_cont_printf(m, dmsg, "PAT ");
f439c429 JG	172	else
f439c429 JG	173	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	174	if (pr & _PAGE_GLOBAL)
ef6bea6d	175	pt_dump_cont_printf(m, dmsg, "GLB ");
fe770bf0	176	else
ef6bea6d	177	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	178	if (pr & _PAGE_NX)
ef6bea6d	179	pt_dump_cont_printf(m, dmsg, "NX ");
fe770bf0	180	else
ef6bea6d	181	pt_dump_cont_printf(m, dmsg, "x ");
926e5392	182	}
ef6bea6d	183	pt_dump_cont_printf(m, dmsg, "%s\n", level_name[level]);
926e5392 AV	184	}
	185
	186	/*
fe770bf0	187	* On 64 bits, sign-extend the 48 bit address to 64 bit
926e5392	188	*/
fe770bf0	189	static unsigned long normalize_addr(unsigned long u)
926e5392	190	{
3a366f79 KS	191	int shift;
	192	if (!IS_ENABLED(CONFIG_X86_64))
	193	return u;
	194
	195	shift = 64 - (__VIRTUAL_MASK_SHIFT + 1);
	196	return (signed long)(u << shift) >> shift;
926e5392 AV	197	}
	198
	199	/*
	200	* This function gets called on a break in a continuous series
	201	* of PTE entries; the next one is different so we need to
	202	* print what we collected so far.
	203	*/
	204	static void note_page(struct seq_file m, struct pg_state st,
fe770bf0	205	pgprot_t new_prot, int level)
926e5392	206	{
fe770bf0	207	pgprotval_t prot, cur;
3891a04a	208	static const char units[] = "BKMGTPE";
926e5392 AV	209
	210	/*
	211	* If we have a "break" in the series, we need to flush the state that
fe770bf0 PA	212	* we have now. "break" is either changing perms, levels or
fe770bf0 PA	213	* address space marker.
926e5392	214	*/
da25e628 TK	215	prot = pgprot_val(new_prot);
da25e628 TK	216	cur = pgprot_val(st->current_prot);
926e5392	217
fe770bf0 PA	218	if (!st->level) {
	219	/* First entry */
	220	st->current_prot = new_prot;
	221	st->level = level;
	222	st->marker = address_markers;
3891a04a	223	st->lines = 0;
ef6bea6d BP	224	pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
ef6bea6d BP	225	st->marker->name);
fe770bf0 PA	226	} else if (prot != cur \|\| level != st->level \|\|
	227	st->current_address >= st->marker[1].start_address) {
	228	const char *unit = units;
926e5392	229	unsigned long delta;
6424fb38	230	int width = sizeof(unsigned long) * 2;
e1a58320 SS	231	pgprotval_t pr = pgprot_val(st->current_prot);
	232
	233	if (st->check_wx && (pr & _PAGE_RW) && !(pr & _PAGE_NX)) {
	234	WARN_ONCE(1,
	235	"x86/mm: Found insecure W+X mapping at address %p/%pS\n",
	236	(void *)st->start_address,
	237	(void *)st->start_address);
	238	st->wx_pages += (st->current_address -
	239	st->start_address) / PAGE_SIZE;
	240	}
926e5392	241
926e5392 AV	242	/*
	243	* Now print the actual finished series
	244	*/
3891a04a PA	245	if (!st->marker->max_lines \|\|
	246	st->lines < st->marker->max_lines) {
	247	pt_dump_seq_printf(m, st->to_dmesg,
	248	"0x%0lx-0x%0lx ",
	249	width, st->start_address,
	250	width, st->current_address);
926e5392	251
3891a04a PA	252	delta = st->current_address - st->start_address;
	253	while (!(delta & 1023) && unit[1]) {
	254	delta >>= 10;
	255	unit++;
	256	}
	257	pt_dump_cont_printf(m, st->to_dmesg, "%9lu%c ",
	258	delta, *unit);
	259	printk_prot(m, st->current_prot, st->level,
	260	st->to_dmesg);
926e5392	261	}
3891a04a	262	st->lines++;
fe770bf0 PA	263
	264	/*
	265	* We print markers for special areas of address space,
	266	* such as the start of vmalloc space etc.
	267	* This helps in the interpretation.
	268	*/
	269	if (st->current_address >= st->marker[1].start_address) {
3891a04a PA	270	if (st->marker->max_lines &&
	271	st->lines > st->marker->max_lines) {
	272	unsigned long nskip =
	273	st->lines - st->marker->max_lines;
	274	pt_dump_seq_printf(m, st->to_dmesg,
	275	"... %lu entr%s skipped ... \n",
	276	nskip,
	277	nskip == 1 ? "y" : "ies");
	278	}
fe770bf0	279	st->marker++;
3891a04a	280	st->lines = 0;
ef6bea6d BP	281	pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
ef6bea6d BP	282	st->marker->name);
926e5392	283	}
fe770bf0	284
926e5392 AV	285	st->start_address = st->current_address;
	286	st->current_prot = new_prot;
	287	st->level = level;
fe770bf0	288	}
926e5392 AV	289	}
926e5392 AV	290
fdd3d8ce	291	static void walk_pte_level(struct seq_file m, struct pg_state st, pmd_t addr, unsigned long P)
926e5392 AV	292	{
	293	int i;
	294	pte_t *start;
da25e628	295	pgprotval_t prot;
926e5392	296
fdd3d8ce	297	start = (pte_t *)pmd_page_vaddr(addr);
926e5392	298	for (i = 0; i < PTRS_PER_PTE; i++) {
da25e628	299	prot = pte_flags(*start);
fe770bf0	300	st->current_address = normalize_addr(P + i * PTE_LEVEL_MULT);
45dcd209	301	note_page(m, st, __pgprot(prot), 5);
926e5392 AV	302	start++;
	303	}
	304	}
	305
fe770bf0	306	#if PTRS_PER_PMD > 1
926e5392	307
fdd3d8ce	308	static void walk_pmd_level(struct seq_file m, struct pg_state st, pud_t addr, unsigned long P)
926e5392 AV	309	{
	310	int i;
	311	pmd_t *start;
da25e628	312	pgprotval_t prot;
926e5392	313
fdd3d8ce	314	start = (pmd_t *)pud_page_vaddr(addr);
926e5392	315	for (i = 0; i < PTRS_PER_PMD; i++) {
fe770bf0	316	st->current_address = normalize_addr(P + i * PMD_LEVEL_MULT);
926e5392	317	if (!pmd_none(*start)) {
da25e628 TK	318	if (pmd_large(start) \|\| !pmd_present(start)) {
da25e628 TK	319	prot = pmd_flags(*start);
45dcd209	320	note_page(m, st, __pgprot(prot), 4);
da25e628	321	} else {
fe770bf0 PA	322	walk_pte_level(m, st, *start,
fe770bf0 PA	323	P + i * PMD_LEVEL_MULT);
da25e628	324	}
926e5392	325	} else
45dcd209	326	note_page(m, st, __pgprot(0), 4);
926e5392 AV	327	start++;
	328	}
	329	}
	330
fe770bf0 PA	331	#else
	332	#define walk_pmd_level(m,s,a,p) walk_pte_level(m,s,__pmd(pud_val(a)),p)
	333	#define pud_large(a) pmd_large(__pmd(pud_val(a)))
	334	#define pud_none(a) pmd_none(__pmd(pud_val(a)))
	335	#endif
926e5392	336
fe770bf0 PA	337	#if PTRS_PER_PUD > 1
fe770bf0 PA	338
243b72aa AR	339	/*
	340	* This is an optimization for CONFIG_DEBUG_WX=y + CONFIG_KASAN=y
	341	* KASAN fills page tables with the same values. Since there is no
	342	* point in checking page table more than once we just skip repeated
	343	* entries. This saves us dozens of seconds during boot.
	344	*/
	345	static bool pud_already_checked(pud_t prev_pud, pud_t pud, bool checkwx)
	346	{
	347	return checkwx && prev_pud && (pud_val(prev_pud) == pud_val(pud));
	348	}
	349
fdd3d8ce	350	static void walk_pud_level(struct seq_file m, struct pg_state st, p4d_t addr, unsigned long P)
926e5392 AV	351	{
	352	int i;
	353	pud_t *start;
da25e628	354	pgprotval_t prot;
243b72aa	355	pud_t *prev_pud = NULL;
926e5392	356
fdd3d8ce	357	start = (pud_t *)p4d_page_vaddr(addr);
926e5392 AV	358
926e5392 AV	359	for (i = 0; i < PTRS_PER_PUD; i++) {
fe770bf0	360	st->current_address = normalize_addr(P + i * PUD_LEVEL_MULT);
243b72aa AR	361	if (!pud_none(*start) &&
243b72aa AR	362	!pud_already_checked(prev_pud, start, st->check_wx)) {
da25e628 TK	363	if (pud_large(start) \|\| !pud_present(start)) {
da25e628 TK	364	prot = pud_flags(*start);
45dcd209	365	note_page(m, st, __pgprot(prot), 3);
da25e628	366	} else {
fe770bf0 PA	367	walk_pmd_level(m, st, *start,
fe770bf0 PA	368	P + i * PUD_LEVEL_MULT);
da25e628	369	}
926e5392	370	} else
45dcd209	371	note_page(m, st, __pgprot(0), 3);
926e5392	372
243b72aa	373	prev_pud = start;
926e5392 AV	374	start++;
	375	}
	376	}
	377
fe770bf0	378	#else
fdd3d8ce KS	379	#define walk_pud_level(m,s,a,p) walk_pmd_level(m,s,__pud(p4d_val(a)),p)
	380	#define p4d_large(a) pud_large(__pud(p4d_val(a)))
	381	#define p4d_none(a) pud_none(__pud(p4d_val(a)))
	382	#endif
	383
	384	#if PTRS_PER_P4D > 1
	385
	386	static void walk_p4d_level(struct seq_file m, struct pg_state st, pgd_t addr, unsigned long P)
	387	{
	388	int i;
	389	p4d_t *start;
	390	pgprotval_t prot;
	391
	392	start = (p4d_t *)pgd_page_vaddr(addr);
	393
	394	for (i = 0; i < PTRS_PER_P4D; i++) {
	395	st->current_address = normalize_addr(P + i * P4D_LEVEL_MULT);
	396	if (!p4d_none(*start)) {
	397	if (p4d_large(start) \|\| !p4d_present(start)) {
	398	prot = p4d_flags(*start);
	399	note_page(m, st, __pgprot(prot), 2);
	400	} else {
	401	walk_pud_level(m, st, *start,
	402	P + i * P4D_LEVEL_MULT);
	403	}
	404	} else
	405	note_page(m, st, __pgprot(0), 2);
	406
	407	start++;
	408	}
	409	}
	410
	411	#else
	412	#define walk_p4d_level(m,s,a,p) walk_pud_level(m,s,__p4d(pgd_val(a)),p)
	413	#define pgd_large(a) p4d_large(__p4d(pgd_val(a)))
	414	#define pgd_none(a) p4d_none(__p4d(pgd_val(a)))
fe770bf0 PA	415	#endif
fe770bf0 PA	416
f4e342c8 BO	417	static inline bool is_hypervisor_range(int idx)
f4e342c8 BO	418	{
b176862f	419	#ifdef CONFIG_X86_64
f4e342c8 BO	420	/*
	421	* ffff800000000000 - ffff87ffffffffff is reserved for
	422	* the hypervisor.
	423	*/
b176862f BP	424	return (idx >= pgd_index(__PAGE_OFFSET) - 16) &&
b176862f BP	425	(idx < pgd_index(__PAGE_OFFSET));
f4e342c8	426	#else
b176862f	427	return false;
f4e342c8	428	#endif
b176862f	429	}
f4e342c8	430
e1a58320 SS	431	static void ptdump_walk_pgd_level_core(struct seq_file m, pgd_t pgd,
e1a58320 SS	432	bool checkwx)
926e5392	433	{
fe770bf0	434	#ifdef CONFIG_X86_64
65ade2f8	435	pgd_t start = (pgd_t ) &init_top_pgt;
fe770bf0 PA	436	#else
	437	pgd_t *start = swapper_pg_dir;
	438	#endif
da25e628	439	pgprotval_t prot;
926e5392	440	int i;
ef6bea6d	441	struct pg_state st = {};
926e5392	442
ef6bea6d BP	443	if (pgd) {
	444	start = pgd;
	445	st.to_dmesg = true;
	446	}
926e5392	447
e1a58320 SS	448	st.check_wx = checkwx;
	449	if (checkwx)
	450	st.wx_pages = 0;
	451
926e5392	452	for (i = 0; i < PTRS_PER_PGD; i++) {
fe770bf0	453	st.current_address = normalize_addr(i * PGD_LEVEL_MULT);
f4e342c8	454	if (!pgd_none(*start) && !is_hypervisor_range(i)) {
da25e628 TK	455	if (pgd_large(start) \|\| !pgd_present(start)) {
da25e628 TK	456	prot = pgd_flags(*start);
fe770bf0	457	note_page(m, &st, __pgprot(prot), 1);
da25e628	458	} else {
fdd3d8ce	459	walk_p4d_level(m, &st, *start,
fe770bf0	460	i * PGD_LEVEL_MULT);
da25e628	461	}
fe770bf0	462	} else
926e5392	463	note_page(m, &st, __pgprot(0), 1);
fe770bf0	464
146fbb76	465	cond_resched();
926e5392 AV	466	start++;
926e5392 AV	467	}
fe770bf0 PA	468
	469	/* Flush out the last page */
	470	st.current_address = normalize_addr(PTRS_PER_PGD*PGD_LEVEL_MULT);
	471	note_page(m, &st, __pgprot(0), 0);
e1a58320 SS	472	if (!checkwx)
	473	return;
	474	if (st.wx_pages)
	475	pr_info("x86/mm: Checked W+X mappings: FAILED, %lu W+X pages found.\n",
	476	st.wx_pages);
	477	else
	478	pr_info("x86/mm: Checked W+X mappings: passed, no W+X pages found.\n");
	479	}
	480
	481	void ptdump_walk_pgd_level(struct seq_file m, pgd_t pgd)
	482	{
	483	ptdump_walk_pgd_level_core(m, pgd, false);
926e5392	484	}
8609d1b5	485	EXPORT_SYMBOL_GPL(ptdump_walk_pgd_level);
926e5392	486
e1a58320 SS	487	void ptdump_walk_pgd_level_checkwx(void)
	488	{
	489	ptdump_walk_pgd_level_core(NULL, NULL, true);
	490	}
	491
8609d1b5	492	static int __init pt_dump_init(void)
926e5392	493	{
0483e1fa TG	494	/*
	495	* Various markers are not compile-time constants, so assign them
	496	* here.
	497	*/
	498	#ifdef CONFIG_X86_64
	499	address_markers[LOW_KERNEL_NR].start_address = PAGE_OFFSET;
	500	address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
	501	address_markers[VMEMMAP_START_NR].start_address = VMEMMAP_START;
	502	#endif
fe770bf0	503	#ifdef CONFIG_X86_32
92851e2f AS	504	address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
92851e2f AS	505	address_markers[VMALLOC_END_NR].start_address = VMALLOC_END;
fe770bf0	506	# ifdef CONFIG_HIGHMEM
92851e2f	507	address_markers[PKMAP_BASE_NR].start_address = PKMAP_BASE;
fe770bf0	508	# endif
92851e2f	509	address_markers[FIXADDR_START_NR].start_address = FIXADDR_START;
fe770bf0 PA	510	#endif
fe770bf0 PA	511
926e5392 AV	512	return 0;
926e5392 AV	513	}
926e5392	514	__initcall(pt_dump_init);