[mirror_ubuntu-zesty-kernel.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/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);

		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 AV	20
	21	#include <asm/pgtable.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	*/
	28	struct pg_state {
	29	int level;
	30	pgprot_t current_prot;
	31	unsigned long start_address;
	32	unsigned long current_address;
fe770bf0	33	const struct addr_marker *marker;
3891a04a	34	unsigned long lines;
ef6bea6d	35	bool to_dmesg;
e1a58320 SS	36	bool check_wx;
e1a58320 SS	37	unsigned long wx_pages;
926e5392 AV	38	};
926e5392 AV	39
fe770bf0 PA	40	struct addr_marker {
	41	unsigned long start_address;
	42	const char *name;
3891a04a	43	unsigned long max_lines;
fe770bf0 PA	44	};
fe770bf0 PA	45
92851e2f AS	46	/* indices for address_markers; keep sync'd w/ address_markers below */
	47	enum address_markers_idx {
	48	USER_SPACE_NR = 0,
	49	#ifdef CONFIG_X86_64
	50	KERNEL_SPACE_NR,
	51	LOW_KERNEL_NR,
	52	VMALLOC_START_NR,
	53	VMEMMAP_START_NR,
8a5a5d15	54	# ifdef CONFIG_X86_ESPFIX64
3891a04a	55	ESPFIX_START_NR,
8a5a5d15	56	# endif
92851e2f AS	57	HIGH_KERNEL_NR,
	58	MODULES_VADDR_NR,
	59	MODULES_END_NR,
	60	#else
	61	KERNEL_SPACE_NR,
	62	VMALLOC_START_NR,
	63	VMALLOC_END_NR,
	64	# ifdef CONFIG_HIGHMEM
	65	PKMAP_BASE_NR,
	66	# endif
	67	FIXADDR_START_NR,
	68	#endif
	69	};
	70
fe770bf0 PA	71	/* Address space markers hints */
	72	static struct addr_marker address_markers[] = {
	73	{ 0, "User Space" },
	74	#ifdef CONFIG_X86_64
	75	{ 0x8000000000000000UL, "Kernel Space" },
0483e1fa TG	76	{ 0/* PAGE_OFFSET */, "Low Kernel Mapping" },
	77	{ 0/* VMALLOC_START */, "vmalloc() Area" },
	78	{ 0/* VMEMMAP_START */, "Vmemmap" },
8a5a5d15	79	# ifdef CONFIG_X86_ESPFIX64
3891a04a	80	{ ESPFIX_BASE_ADDR, "ESPfix Area", 16 },
8266e31e MK	81	# endif
	82	# ifdef CONFIG_EFI
	83	{ EFI_VA_END, "EFI Runtime Services" },
8a5a5d15	84	# endif
fe770bf0	85	{ __START_KERNEL_map, "High Kernel Mapping" },
9a79cf9c YL	86	{ MODULES_VADDR, "Modules" },
9a79cf9c YL	87	{ MODULES_END, "End Modules" },
fe770bf0 PA	88	#else
	89	{ PAGE_OFFSET, "Kernel Mapping" },
	90	{ 0/* VMALLOC_START */, "vmalloc() Area" },
	91	{ 0/VMALLOC_END/, "vmalloc() End" },
	92	# ifdef CONFIG_HIGHMEM
173ae9ba	93	{ 0/PKMAP_BASE/, "Persistent kmap() Area" },
fe770bf0 PA	94	# endif
	95	{ 0/FIXADDR_START/, "Fixmap Area" },
	96	#endif
	97	{ -1, NULL } /* End of list */
	98	};
926e5392	99
fe770bf0 PA	100	/* Multipliers for offsets within the PTEs */
	101	#define PTE_LEVEL_MULT (PAGE_SIZE)
	102	#define PMD_LEVEL_MULT (PTRS_PER_PTE * PTE_LEVEL_MULT)
	103	#define PUD_LEVEL_MULT (PTRS_PER_PMD * PMD_LEVEL_MULT)
	104	#define PGD_LEVEL_MULT (PTRS_PER_PUD * PUD_LEVEL_MULT)
926e5392	105
ef6bea6d BP	106	#define pt_dump_seq_printf(m, to_dmesg, fmt, args...) \
	107	({ \
	108	if (to_dmesg) \
	109	printk(KERN_INFO fmt, ##args); \
	110	else \
	111	if (m) \
	112	seq_printf(m, fmt, ##args); \
	113	})
	114
	115	#define pt_dump_cont_printf(m, to_dmesg, fmt, args...) \
	116	({ \
	117	if (to_dmesg) \
	118	printk(KERN_CONT fmt, ##args); \
	119	else \
	120	if (m) \
	121	seq_printf(m, fmt, ##args); \
	122	})
	123
926e5392 AV	124	/*
	125	* Print a readable form of a pgprot_t to the seq_file
	126	*/
ef6bea6d	127	static void printk_prot(struct seq_file *m, pgprot_t prot, int level, bool dmsg)
926e5392	128	{
fe770bf0 PA	129	pgprotval_t pr = pgprot_val(prot);
	130	static const char * const level_name[] =
	131	{ "cr3", "pgd", "pud", "pmd", "pte" };
	132
	133	if (!pgprot_val(prot)) {
	134	/* Not present */
f439c429	135	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0 PA	136	} else {
fe770bf0 PA	137	if (pr & _PAGE_USER)
ef6bea6d	138	pt_dump_cont_printf(m, dmsg, "USR ");
926e5392	139	else
ef6bea6d	140	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	141	if (pr & _PAGE_RW)
ef6bea6d	142	pt_dump_cont_printf(m, dmsg, "RW ");
fe770bf0	143	else
ef6bea6d	144	pt_dump_cont_printf(m, dmsg, "ro ");
fe770bf0	145	if (pr & _PAGE_PWT)
ef6bea6d	146	pt_dump_cont_printf(m, dmsg, "PWT ");
fe770bf0	147	else
ef6bea6d	148	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	149	if (pr & _PAGE_PCD)
ef6bea6d	150	pt_dump_cont_printf(m, dmsg, "PCD ");
926e5392	151	else
ef6bea6d	152	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	153
f439c429 JG	154	/* Bit 7 has a different meaning on level 3 vs 4 */
	155	if (level <= 3 && pr & _PAGE_PSE)
	156	pt_dump_cont_printf(m, dmsg, "PSE ");
	157	else
	158	pt_dump_cont_printf(m, dmsg, " ");
	159	if ((level == 4 && pr & _PAGE_PAT) \|\|
	160	((level == 3 \|\| level == 2) && pr & _PAGE_PAT_LARGE))
da25e628	161	pt_dump_cont_printf(m, dmsg, "PAT ");
f439c429 JG	162	else
f439c429 JG	163	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	164	if (pr & _PAGE_GLOBAL)
ef6bea6d	165	pt_dump_cont_printf(m, dmsg, "GLB ");
fe770bf0	166	else
ef6bea6d	167	pt_dump_cont_printf(m, dmsg, " ");
fe770bf0	168	if (pr & _PAGE_NX)
ef6bea6d	169	pt_dump_cont_printf(m, dmsg, "NX ");
fe770bf0	170	else
ef6bea6d	171	pt_dump_cont_printf(m, dmsg, "x ");
926e5392	172	}
ef6bea6d	173	pt_dump_cont_printf(m, dmsg, "%s\n", level_name[level]);
926e5392 AV	174	}
	175
	176	/*
fe770bf0	177	* On 64 bits, sign-extend the 48 bit address to 64 bit
926e5392	178	*/
fe770bf0	179	static unsigned long normalize_addr(unsigned long u)
926e5392	180	{
fe770bf0 PA	181	#ifdef CONFIG_X86_64
	182	return (signed long)(u << 16) >> 16;
	183	#else
926e5392	184	return u;
fe770bf0	185	#endif
926e5392 AV	186	}
	187
	188	/*
	189	* This function gets called on a break in a continuous series
	190	* of PTE entries; the next one is different so we need to
	191	* print what we collected so far.
	192	*/
	193	static void note_page(struct seq_file m, struct pg_state st,
fe770bf0	194	pgprot_t new_prot, int level)
926e5392	195	{
fe770bf0	196	pgprotval_t prot, cur;
3891a04a	197	static const char units[] = "BKMGTPE";
926e5392 AV	198
	199	/*
	200	* If we have a "break" in the series, we need to flush the state that
fe770bf0 PA	201	* we have now. "break" is either changing perms, levels or
fe770bf0 PA	202	* address space marker.
926e5392	203	*/
da25e628 TK	204	prot = pgprot_val(new_prot);
da25e628 TK	205	cur = pgprot_val(st->current_prot);
926e5392	206
fe770bf0 PA	207	if (!st->level) {
	208	/* First entry */
	209	st->current_prot = new_prot;
	210	st->level = level;
	211	st->marker = address_markers;
3891a04a	212	st->lines = 0;
ef6bea6d BP	213	pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
ef6bea6d BP	214	st->marker->name);
fe770bf0 PA	215	} else if (prot != cur \|\| level != st->level \|\|
	216	st->current_address >= st->marker[1].start_address) {
	217	const char *unit = units;
926e5392	218	unsigned long delta;
6424fb38	219	int width = sizeof(unsigned long) * 2;
e1a58320 SS	220	pgprotval_t pr = pgprot_val(st->current_prot);
	221
	222	if (st->check_wx && (pr & _PAGE_RW) && !(pr & _PAGE_NX)) {
	223	WARN_ONCE(1,
	224	"x86/mm: Found insecure W+X mapping at address %p/%pS\n",
	225	(void *)st->start_address,
	226	(void *)st->start_address);
	227	st->wx_pages += (st->current_address -
	228	st->start_address) / PAGE_SIZE;
	229	}
926e5392	230
926e5392 AV	231	/*
	232	* Now print the actual finished series
	233	*/
3891a04a PA	234	if (!st->marker->max_lines \|\|
	235	st->lines < st->marker->max_lines) {
	236	pt_dump_seq_printf(m, st->to_dmesg,
	237	"0x%0lx-0x%0lx ",
	238	width, st->start_address,
	239	width, st->current_address);
926e5392	240
3891a04a PA	241	delta = st->current_address - st->start_address;
	242	while (!(delta & 1023) && unit[1]) {
	243	delta >>= 10;
	244	unit++;
	245	}
	246	pt_dump_cont_printf(m, st->to_dmesg, "%9lu%c ",
	247	delta, *unit);
	248	printk_prot(m, st->current_prot, st->level,
	249	st->to_dmesg);
926e5392	250	}
3891a04a	251	st->lines++;
fe770bf0 PA	252
	253	/*
	254	* We print markers for special areas of address space,
	255	* such as the start of vmalloc space etc.
	256	* This helps in the interpretation.
	257	*/
	258	if (st->current_address >= st->marker[1].start_address) {
3891a04a PA	259	if (st->marker->max_lines &&
	260	st->lines > st->marker->max_lines) {
	261	unsigned long nskip =
	262	st->lines - st->marker->max_lines;
	263	pt_dump_seq_printf(m, st->to_dmesg,
	264	"... %lu entr%s skipped ... \n",
	265	nskip,
	266	nskip == 1 ? "y" : "ies");
	267	}
fe770bf0	268	st->marker++;
3891a04a	269	st->lines = 0;
ef6bea6d BP	270	pt_dump_seq_printf(m, st->to_dmesg, "---[ %s ]---\n",
ef6bea6d BP	271	st->marker->name);
926e5392	272	}
fe770bf0	273
926e5392 AV	274	st->start_address = st->current_address;
	275	st->current_prot = new_prot;
	276	st->level = level;
fe770bf0	277	}
926e5392 AV	278	}
926e5392 AV	279
fe770bf0	280	static void walk_pte_level(struct seq_file m, struct pg_state st, pmd_t addr,
926e5392 AV	281	unsigned long P)
	282	{
	283	int i;
	284	pte_t *start;
da25e628	285	pgprotval_t prot;
926e5392 AV	286
	287	start = (pte_t *) pmd_page_vaddr(addr);
	288	for (i = 0; i < PTRS_PER_PTE; i++) {
da25e628	289	prot = pte_flags(*start);
fe770bf0	290	st->current_address = normalize_addr(P + i * PTE_LEVEL_MULT);
da25e628	291	note_page(m, st, __pgprot(prot), 4);
926e5392 AV	292	start++;
	293	}
	294	}
	295
fe770bf0	296	#if PTRS_PER_PMD > 1
926e5392	297
fe770bf0	298	static void walk_pmd_level(struct seq_file m, struct pg_state st, pud_t addr,
926e5392 AV	299	unsigned long P)
	300	{
	301	int i;
	302	pmd_t *start;
da25e628	303	pgprotval_t prot;
926e5392 AV	304
	305	start = (pmd_t *) pud_page_vaddr(addr);
	306	for (i = 0; i < PTRS_PER_PMD; i++) {
fe770bf0	307	st->current_address = normalize_addr(P + i * PMD_LEVEL_MULT);
926e5392	308	if (!pmd_none(*start)) {
da25e628 TK	309	if (pmd_large(start) \|\| !pmd_present(start)) {
da25e628 TK	310	prot = pmd_flags(*start);
926e5392	311	note_page(m, st, __pgprot(prot), 3);
da25e628	312	} else {
fe770bf0 PA	313	walk_pte_level(m, st, *start,
fe770bf0 PA	314	P + i * PMD_LEVEL_MULT);
da25e628	315	}
926e5392 AV	316	} else
	317	note_page(m, st, __pgprot(0), 3);
	318	start++;
	319	}
	320	}
	321
fe770bf0 PA	322	#else
	323	#define walk_pmd_level(m,s,a,p) walk_pte_level(m,s,__pmd(pud_val(a)),p)
	324	#define pud_large(a) pmd_large(__pmd(pud_val(a)))
	325	#define pud_none(a) pmd_none(__pmd(pud_val(a)))
	326	#endif
926e5392	327
fe770bf0 PA	328	#if PTRS_PER_PUD > 1
	329
	330	static void walk_pud_level(struct seq_file m, struct pg_state st, pgd_t addr,
926e5392 AV	331	unsigned long P)
	332	{
	333	int i;
	334	pud_t *start;
da25e628	335	pgprotval_t prot;
926e5392 AV	336
	337	start = (pud_t *) pgd_page_vaddr(addr);
	338
	339	for (i = 0; i < PTRS_PER_PUD; i++) {
fe770bf0	340	st->current_address = normalize_addr(P + i * PUD_LEVEL_MULT);
926e5392	341	if (!pud_none(*start)) {
da25e628 TK	342	if (pud_large(start) \|\| !pud_present(start)) {
da25e628 TK	343	prot = pud_flags(*start);
926e5392	344	note_page(m, st, __pgprot(prot), 2);
da25e628	345	} else {
fe770bf0 PA	346	walk_pmd_level(m, st, *start,
fe770bf0 PA	347	P + i * PUD_LEVEL_MULT);
da25e628	348	}
926e5392 AV	349	} else
	350	note_page(m, st, __pgprot(0), 2);
	351
	352	start++;
	353	}
	354	}
	355
fe770bf0 PA	356	#else
	357	#define walk_pud_level(m,s,a,p) walk_pmd_level(m,s,__pud(pgd_val(a)),p)
	358	#define pgd_large(a) pud_large(__pud(pgd_val(a)))
	359	#define pgd_none(a) pud_none(__pud(pgd_val(a)))
	360	#endif
	361
f4e342c8 BO	362	static inline bool is_hypervisor_range(int idx)
f4e342c8 BO	363	{
b176862f	364	#ifdef CONFIG_X86_64
f4e342c8 BO	365	/*
	366	* ffff800000000000 - ffff87ffffffffff is reserved for
	367	* the hypervisor.
	368	*/
b176862f BP	369	return (idx >= pgd_index(__PAGE_OFFSET) - 16) &&
b176862f BP	370	(idx < pgd_index(__PAGE_OFFSET));
f4e342c8	371	#else
b176862f	372	return false;
f4e342c8	373	#endif
b176862f	374	}
f4e342c8	375
e1a58320 SS	376	static void ptdump_walk_pgd_level_core(struct seq_file m, pgd_t pgd,
e1a58320 SS	377	bool checkwx)
926e5392	378	{
fe770bf0	379	#ifdef CONFIG_X86_64
926e5392	380	pgd_t start = (pgd_t ) &init_level4_pgt;
fe770bf0 PA	381	#else
	382	pgd_t *start = swapper_pg_dir;
	383	#endif
da25e628	384	pgprotval_t prot;
926e5392	385	int i;
ef6bea6d	386	struct pg_state st = {};
926e5392	387
ef6bea6d BP	388	if (pgd) {
	389	start = pgd;
	390	st.to_dmesg = true;
	391	}
926e5392	392
e1a58320 SS	393	st.check_wx = checkwx;
	394	if (checkwx)
	395	st.wx_pages = 0;
	396
926e5392	397	for (i = 0; i < PTRS_PER_PGD; i++) {
fe770bf0	398	st.current_address = normalize_addr(i * PGD_LEVEL_MULT);
f4e342c8	399	if (!pgd_none(*start) && !is_hypervisor_range(i)) {
da25e628 TK	400	if (pgd_large(start) \|\| !pgd_present(start)) {
da25e628 TK	401	prot = pgd_flags(*start);
fe770bf0	402	note_page(m, &st, __pgprot(prot), 1);
da25e628	403	} else {
fe770bf0 PA	404	walk_pud_level(m, &st, *start,
fe770bf0 PA	405	i * PGD_LEVEL_MULT);
da25e628	406	}
fe770bf0	407	} else
926e5392	408	note_page(m, &st, __pgprot(0), 1);
fe770bf0	409
146fbb76	410	cond_resched();
926e5392 AV	411	start++;
926e5392 AV	412	}
fe770bf0 PA	413
	414	/* Flush out the last page */
	415	st.current_address = normalize_addr(PTRS_PER_PGD*PGD_LEVEL_MULT);
	416	note_page(m, &st, __pgprot(0), 0);
e1a58320 SS	417	if (!checkwx)
	418	return;
	419	if (st.wx_pages)
	420	pr_info("x86/mm: Checked W+X mappings: FAILED, %lu W+X pages found.\n",
	421	st.wx_pages);
	422	else
	423	pr_info("x86/mm: Checked W+X mappings: passed, no W+X pages found.\n");
	424	}
	425
	426	void ptdump_walk_pgd_level(struct seq_file m, pgd_t pgd)
	427	{
	428	ptdump_walk_pgd_level_core(m, pgd, false);
926e5392	429	}
8609d1b5	430	EXPORT_SYMBOL_GPL(ptdump_walk_pgd_level);
926e5392	431
e1a58320 SS	432	void ptdump_walk_pgd_level_checkwx(void)
	433	{
	434	ptdump_walk_pgd_level_core(NULL, NULL, true);
	435	}
	436
8609d1b5	437	static int __init pt_dump_init(void)
926e5392	438	{
0483e1fa TG	439	/*
	440	* Various markers are not compile-time constants, so assign them
	441	* here.
	442	*/
	443	#ifdef CONFIG_X86_64
	444	address_markers[LOW_KERNEL_NR].start_address = PAGE_OFFSET;
	445	address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
	446	address_markers[VMEMMAP_START_NR].start_address = VMEMMAP_START;
	447	#endif
fe770bf0	448	#ifdef CONFIG_X86_32
92851e2f AS	449	address_markers[VMALLOC_START_NR].start_address = VMALLOC_START;
92851e2f AS	450	address_markers[VMALLOC_END_NR].start_address = VMALLOC_END;
fe770bf0	451	# ifdef CONFIG_HIGHMEM
92851e2f	452	address_markers[PKMAP_BASE_NR].start_address = PKMAP_BASE;
fe770bf0	453	# endif
92851e2f	454	address_markers[FIXADDR_START_NR].start_address = FIXADDR_START;
fe770bf0 PA	455	#endif
fe770bf0 PA	456
926e5392 AV	457	return 0;
926e5392 AV	458	}
926e5392	459	__initcall(pt_dump_init);