[linux-2.6-block.git] / arch / powerpc / mm / ptdump / ptdump.c

// SPDX-License-Identifier: GPL-2.0-only
/*
 * Copyright 2016, Rashmica Gupta, IBM Corp.
 *
 * This traverses the kernel pagetables and dumps the
 * information about the used sections of memory to
 * /sys/kernel/debug/kernel_pagetables.
 *
 * Derived from the arm64 implementation:
 * Copyright (c) 2014, The Linux Foundation, Laura Abbott.
 * (C) Copyright 2008 Intel Corporation, Arjan van de Ven.
 */
#include <linux/debugfs.h>
#include <linux/fs.h>
#include <linux/hugetlb.h>
#include <linux/io.h>
#include <linux/mm.h>
#include <linux/highmem.h>
#include <linux/sched.h>
#include <linux/seq_file.h>
#include <asm/fixmap.h>
#include <asm/pgtable.h>
#include <linux/const.h>
#include <asm/page.h>
#include <asm/pgalloc.h>

#include "ptdump.h"

/*
 * To visualise what is happening,
 *
 *  - PTRS_PER_P** = how many entries there are in the corresponding P**
 *  - P**_SHIFT = how many bits of the address we use to index into the
 * corresponding P**
 *  - P**_SIZE is how much memory we can access through the table - not the
 * size of the table itself.
 * P**={PGD, PUD, PMD, PTE}
 *
 *
 * Each entry of the PGD points to a PUD. Each entry of a PUD points to a
 * PMD. Each entry of a PMD points to a PTE. And every PTE entry points to
 * a page.
 *
 * In the case where there are only 3 levels, the PUD is folded into the
 * PGD: every PUD has only one entry which points to the PMD.
 *
 * The page dumper groups page table entries of the same type into a single
 * description. It uses pg_state to track the range information while
 * iterating over the PTE entries. When the continuity is broken it then
 * dumps out a description of the range - ie PTEs that are virtually contiguous
 * with the same PTE flags are chunked together. This is to make it clear how
 * different areas of the kernel virtual memory are used.
 *
 */
struct pg_state {
	struct seq_file *seq;
	const struct addr_marker *marker;
	unsigned long start_address;
	unsigned long start_pa;
	unsigned long last_pa;
	unsigned int level;
	u64 current_flags;
	bool check_wx;
	unsigned long wx_pages;
};

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

static struct addr_marker address_markers[] = {
	{ 0,	"Start of kernel VM" },
	{ 0,	"vmalloc() Area" },
	{ 0,	"vmalloc() End" },
#ifdef CONFIG_PPC64
	{ 0,	"isa I/O start" },
	{ 0,	"isa I/O end" },
	{ 0,	"phb I/O start" },
	{ 0,	"phb I/O end" },
	{ 0,	"I/O remap start" },
	{ 0,	"I/O remap end" },
	{ 0,	"vmemmap start" },
#else
	{ 0,	"Early I/O remap start" },
	{ 0,	"Early I/O remap end" },
#ifdef CONFIG_HIGHMEM
	{ 0,	"Highmem PTEs start" },
	{ 0,	"Highmem PTEs end" },
#endif
	{ 0,	"Fixmap start" },
	{ 0,	"Fixmap end" },
#endif
#ifdef CONFIG_KASAN
	{ 0,	"kasan shadow mem start" },
	{ 0,	"kasan shadow mem end" },
#endif
	{ -1,	NULL },
};

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

#define pt_dump_seq_putc(m, c)		\
({					\
	if (m)				\
		seq_putc(m, c);		\
})

static void dump_flag_info(struct pg_state *st, const struct flag_info
		*flag, u64 pte, int num)
{
	unsigned int i;

	for (i = 0; i < num; i++, flag++) {
		const char *s = NULL;
		u64 val;

		/* flag not defined so don't check it */
		if (flag->mask == 0)
			continue;
		/* Some 'flags' are actually values */
		if (flag->is_val) {
			val = pte & flag->val;
			if (flag->shift)
				val = val >> flag->shift;
			pt_dump_seq_printf(st->seq, "  %s:%llx", flag->set, val);
		} else {
			if ((pte & flag->mask) == flag->val)
				s = flag->set;
			else
				s = flag->clear;
			if (s)
				pt_dump_seq_printf(st->seq, "  %s", s);
		}
		st->current_flags &= ~flag->mask;
	}
	if (st->current_flags != 0)
		pt_dump_seq_printf(st->seq, "  unknown flags:%llx", st->current_flags);
}

static void dump_addr(struct pg_state *st, unsigned long addr)
{
	static const char units[] = "KMGTPE";
	const char *unit = units;
	unsigned long delta;

#ifdef CONFIG_PPC64
#define REG		"0x%016lx"
#else
#define REG		"0x%08lx"
#endif

	pt_dump_seq_printf(st->seq, REG "-" REG " ", st->start_address, addr - 1);
	if (st->start_pa == st->last_pa && st->start_address + PAGE_SIZE != addr) {
		pt_dump_seq_printf(st->seq, "[" REG "]", st->start_pa);
		delta = PAGE_SIZE >> 10;
	} else {
		pt_dump_seq_printf(st->seq, " " REG " ", st->start_pa);
		delta = (addr - st->start_address) >> 10;
	}
	/* Work out what appropriate unit to use */
	while (!(delta & 1023) && unit[1]) {
		delta >>= 10;
		unit++;
	}
	pt_dump_seq_printf(st->seq, "%9lu%c", delta, *unit);

}

static void note_prot_wx(struct pg_state *st, unsigned long addr)
{
	if (!IS_ENABLED(CONFIG_PPC_DEBUG_WX) || !st->check_wx)
		return;

	if (!((st->current_flags & pgprot_val(PAGE_KERNEL_X)) == pgprot_val(PAGE_KERNEL_X)))
		return;

	WARN_ONCE(1, "powerpc/mm: Found insecure W+X mapping at address %p/%pS\n",
		  (void *)st->start_address, (void *)st->start_address);

	st->wx_pages += (addr - st->start_address) / PAGE_SIZE;
}

static void note_page(struct pg_state *st, unsigned long addr,
	       unsigned int level, u64 val)
{
	u64 flag = val & pg_level[level].mask;
	u64 pa = val & PTE_RPN_MASK;

	/* At first no level is set */
	if (!st->level) {
		st->level = level;
		st->current_flags = flag;
		st->start_address = addr;
		st->start_pa = pa;
		st->last_pa = pa;
		pt_dump_seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
	/*
	 * Dump the section of virtual memory when:
	 *   - the PTE flags from one entry to the next differs.
	 *   - we change levels in the tree.
	 *   - the address is in a different section of memory and is thus
	 *   used for a different purpose, regardless of the flags.
	 *   - the pa of this page is not adjacent to the last inspected page
	 */
	} else if (flag != st->current_flags || level != st->level ||
		   addr >= st->marker[1].start_address ||
		   (pa != st->last_pa + PAGE_SIZE &&
		    (pa != st->start_pa || st->start_pa != st->last_pa))) {

		/* Check the PTE flags */
		if (st->current_flags) {
			note_prot_wx(st, addr);
			dump_addr(st, addr);

			/* Dump all the flags */
			if (pg_level[st->level].flag)
				dump_flag_info(st, pg_level[st->level].flag,
					  st->current_flags,
					  pg_level[st->level].num);

			pt_dump_seq_putc(st->seq, '\n');
		}

		/*
		 * Address indicates we have passed the end of the
		 * current section of virtual memory
		 */
		while (addr >= st->marker[1].start_address) {
			st->marker++;
			pt_dump_seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
		}
		st->start_address = addr;
		st->start_pa = pa;
		st->last_pa = pa;
		st->current_flags = flag;
		st->level = level;
	} else {
		st->last_pa = pa;
	}
}

static void walk_pte(struct pg_state *st, pmd_t *pmd, unsigned long start)
{
	pte_t *pte = pte_offset_kernel(pmd, 0);
	unsigned long addr;
	unsigned int i;

	for (i = 0; i < PTRS_PER_PTE; i++, pte++) {
		addr = start + i * PAGE_SIZE;
		note_page(st, addr, 4, pte_val(*pte));

	}
}

static void walk_pmd(struct pg_state *st, pud_t *pud, unsigned long start)
{
	pmd_t *pmd = pmd_offset(pud, 0);
	unsigned long addr;
	unsigned int i;

	for (i = 0; i < PTRS_PER_PMD; i++, pmd++) {
		addr = start + i * PMD_SIZE;
		if (!pmd_none(*pmd) && !pmd_is_leaf(*pmd))
			/* pmd exists */
			walk_pte(st, pmd, addr);
		else
			note_page(st, addr, 3, pmd_val(*pmd));
	}
}

static void walk_pud(struct pg_state *st, pgd_t *pgd, unsigned long start)
{
	pud_t *pud = pud_offset(pgd, 0);
	unsigned long addr;
	unsigned int i;

	for (i = 0; i < PTRS_PER_PUD; i++, pud++) {
		addr = start + i * PUD_SIZE;
		if (!pud_none(*pud) && !pud_is_leaf(*pud))
			/* pud exists */
			walk_pmd(st, pud, addr);
		else
			note_page(st, addr, 2, pud_val(*pud));
	}
}

static void walk_pagetables(struct pg_state *st)
{
	unsigned int i;
	unsigned long addr = st->start_address & PGDIR_MASK;
	pgd_t *pgd = pgd_offset_k(addr);

	/*
	 * Traverse the linux pagetable structure and dump pages that are in
	 * the hash pagetable.
	 */
	for (i = pgd_index(addr); i < PTRS_PER_PGD; i++, pgd++, addr += PGDIR_SIZE) {
		if (!pgd_none(*pgd) && !pgd_is_leaf(*pgd))
			/* pgd exists */
			walk_pud(st, pgd, addr);
		else
			note_page(st, addr, 1, pgd_val(*pgd));
	}
}

static void populate_markers(void)
{
	int i = 0;

	address_markers[i++].start_address = PAGE_OFFSET;
	address_markers[i++].start_address = VMALLOC_START;
	address_markers[i++].start_address = VMALLOC_END;
#ifdef CONFIG_PPC64
	address_markers[i++].start_address = ISA_IO_BASE;
	address_markers[i++].start_address = ISA_IO_END;
	address_markers[i++].start_address = PHB_IO_BASE;
	address_markers[i++].start_address = PHB_IO_END;
	address_markers[i++].start_address = IOREMAP_BASE;
	address_markers[i++].start_address = IOREMAP_END;
	/* What is the ifdef about? */
#ifdef CONFIG_PPC_BOOK3S_64
	address_markers[i++].start_address =  H_VMEMMAP_START;
#else
	address_markers[i++].start_address =  VMEMMAP_BASE;
#endif
#else /* !CONFIG_PPC64 */
	address_markers[i++].start_address = ioremap_bot;
	address_markers[i++].start_address = IOREMAP_TOP;
#ifdef CONFIG_HIGHMEM
	address_markers[i++].start_address = PKMAP_BASE;
	address_markers[i++].start_address = PKMAP_ADDR(LAST_PKMAP);
#endif
	address_markers[i++].start_address = FIXADDR_START;
	address_markers[i++].start_address = FIXADDR_TOP;
#ifdef CONFIG_KASAN
	address_markers[i++].start_address = KASAN_SHADOW_START;
	address_markers[i++].start_address = KASAN_SHADOW_END;
#endif
#endif /* CONFIG_PPC64 */
}

static int ptdump_show(struct seq_file *m, void *v)
{
	struct pg_state st = {
		.seq = m,
		.marker = address_markers,
		.start_address = PAGE_OFFSET,
	};

#ifdef CONFIG_PPC64
	if (!radix_enabled())
		st.start_address = KERN_VIRT_START;
#endif

	/* Traverse kernel page tables */
	walk_pagetables(&st);
	note_page(&st, 0, 0, 0);
	return 0;
}


static int ptdump_open(struct inode *inode, struct file *file)
{
	return single_open(file, ptdump_show, NULL);
}

static const struct file_operations ptdump_fops = {
	.open		= ptdump_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
};

static void build_pgtable_complete_mask(void)
{
	unsigned int i, j;

	for (i = 0; i < ARRAY_SIZE(pg_level); i++)
		if (pg_level[i].flag)
			for (j = 0; j < pg_level[i].num; j++)
				pg_level[i].mask |= pg_level[i].flag[j].mask;
}

#ifdef CONFIG_PPC_DEBUG_WX
void ptdump_check_wx(void)
{
	struct pg_state st = {
		.seq = NULL,
		.marker = address_markers,
		.check_wx = true,
		.start_address = PAGE_OFFSET,
	};

#ifdef CONFIG_PPC64
	if (!radix_enabled())
		st.start_address = KERN_VIRT_START;
#endif

	walk_pagetables(&st);

	if (st.wx_pages)
		pr_warn("Checked W+X mappings: FAILED, %lu W+X pages found\n",
			st.wx_pages);
	else
		pr_info("Checked W+X mappings: passed, no W+X pages found\n");
}
#endif

static int ptdump_init(void)
{
	struct dentry *debugfs_file;

	populate_markers();
	build_pgtable_complete_mask();
	debugfs_file = debugfs_create_file("kernel_page_tables", 0400, NULL,
			NULL, &ptdump_fops);
	return debugfs_file ? 0 : -ENOMEM;
}
device_initcall(ptdump_init);
Commit	Line	Data
b886d83c	1	// SPDX-License-Identifier: GPL-2.0-only
8eb07b18 RG	2	/*
	3	* Copyright 2016, Rashmica Gupta, IBM Corp.
	4	*
	5	* This traverses the kernel pagetables and dumps the
	6	* information about the used sections of memory to
	7	* /sys/kernel/debug/kernel_pagetables.
	8	*
	9	* Derived from the arm64 implementation:
	10	* Copyright (c) 2014, The Linux Foundation, Laura Abbott.
	11	* (C) Copyright 2008 Intel Corporation, Arjan van de Ven.
8eb07b18 RG	12	*/
	13	#include <linux/debugfs.h>
	14	#include <linux/fs.h>
bfb9956a	15	#include <linux/hugetlb.h>
8eb07b18 RG	16	#include <linux/io.h>
8eb07b18 RG	17	#include <linux/mm.h>
462951cd	18	#include <linux/highmem.h>
8eb07b18 RG	19	#include <linux/sched.h>
	20	#include <linux/seq_file.h>
	21	#include <asm/fixmap.h>
	22	#include <asm/pgtable.h>
	23	#include <linux/const.h>
	24	#include <asm/page.h>
	25	#include <asm/pgalloc.h>
	26
e66c3209	27	#include "ptdump.h"
97026b5a	28
8eb07b18 RG	29	/*
	30	* To visualise what is happening,
	31	*
	32	* - PTRS_PER_P = how many entries there are in the corresponding P
	33	* - P**_SHIFT = how many bits of the address we use to index into the
	34	* corresponding P**
	35	* - P**_SIZE is how much memory we can access through the table - not the
	36	* size of the table itself.
	37	* P**={PGD, PUD, PMD, PTE}
	38	*
	39	*
	40	* Each entry of the PGD points to a PUD. Each entry of a PUD points to a
	41	* PMD. Each entry of a PMD points to a PTE. And every PTE entry points to
	42	* a page.
	43	*
	44	* In the case where there are only 3 levels, the PUD is folded into the
	45	* PGD: every PUD has only one entry which points to the PMD.
	46	*
	47	* The page dumper groups page table entries of the same type into a single
	48	* description. It uses pg_state to track the range information while
	49	* iterating over the PTE entries. When the continuity is broken it then
	50	* dumps out a description of the range - ie PTEs that are virtually contiguous
	51	* with the same PTE flags are chunked together. This is to make it clear how
	52	* different areas of the kernel virtual memory are used.
	53	*
	54	*/
	55	struct pg_state {
	56	struct seq_file *seq;
	57	const struct addr_marker *marker;
	58	unsigned long start_address;
aaa22952 OH	59	unsigned long start_pa;
aaa22952 OH	60	unsigned long last_pa;
8eb07b18 RG	61	unsigned int level;
8eb07b18 RG	62	u64 current_flags;
453d87f6 RC	63	bool check_wx;
453d87f6 RC	64	unsigned long wx_pages;
8eb07b18 RG	65	};
	66
	67	struct addr_marker {
	68	unsigned long start_address;
	69	const char *name;
	70	};
	71
	72	static struct addr_marker address_markers[] = {
	73	{ 0, "Start of kernel VM" },
	74	{ 0, "vmalloc() Area" },
	75	{ 0, "vmalloc() End" },
6c01bbd2	76	#ifdef CONFIG_PPC64
8eb07b18 RG	77	{ 0, "isa I/O start" },
	78	{ 0, "isa I/O end" },
	79	{ 0, "phb I/O start" },
	80	{ 0, "phb I/O end" },
	81	{ 0, "I/O remap start" },
	82	{ 0, "I/O remap end" },
	83	{ 0, "vmemmap start" },
6c01bbd2 CL	84	#else
	85	{ 0, "Early I/O remap start" },
	86	{ 0, "Early I/O remap end" },
6c01bbd2 CL	87	#ifdef CONFIG_HIGHMEM
	88	{ 0, "Highmem PTEs start" },
	89	{ 0, "Highmem PTEs end" },
	90	#endif
	91	{ 0, "Fixmap start" },
	92	{ 0, "Fixmap end" },
b4abe38f CL	93	#endif
	94	#ifdef CONFIG_KASAN
	95	{ 0, "kasan shadow mem start" },
	96	{ 0, "kasan shadow mem end" },
6c01bbd2	97	#endif
8eb07b18 RG	98	{ -1, NULL },
	99	};
	100
5f18cbdb RC	101	#define pt_dump_seq_printf(m, fmt, args...) \
	102	({ \
	103	if (m) \
	104	seq_printf(m, fmt, ##args); \
	105	})
	106
	107	#define pt_dump_seq_putc(m, c) \
	108	({ \
	109	if (m) \
	110	seq_putc(m, c); \
	111	})
	112
8eb07b18 RG	113	static void dump_flag_info(struct pg_state *st, const struct flag_info
	114	*flag, u64 pte, int num)
	115	{
	116	unsigned int i;
	117
	118	for (i = 0; i < num; i++, flag++) {
	119	const char *s = NULL;
	120	u64 val;
	121
	122	/* flag not defined so don't check it */
	123	if (flag->mask == 0)
	124	continue;
	125	/* Some 'flags' are actually values */
	126	if (flag->is_val) {
	127	val = pte & flag->val;
	128	if (flag->shift)
	129	val = val >> flag->shift;
5f18cbdb	130	pt_dump_seq_printf(st->seq, " %s:%llx", flag->set, val);
8eb07b18 RG	131	} else {
	132	if ((pte & flag->mask) == flag->val)
	133	s = flag->set;
	134	else
	135	s = flag->clear;
	136	if (s)
5f18cbdb	137	pt_dump_seq_printf(st->seq, " %s", s);
8eb07b18 RG	138	}
	139	st->current_flags &= ~flag->mask;
	140	}
	141	if (st->current_flags != 0)
5f18cbdb	142	pt_dump_seq_printf(st->seq, " unknown flags:%llx", st->current_flags);
8eb07b18 RG	143	}
	144
	145	static void dump_addr(struct pg_state *st, unsigned long addr)
	146	{
	147	static const char units[] = "KMGTPE";
	148	const char *unit = units;
	149	unsigned long delta;
	150
78a18dbf	151	#ifdef CONFIG_PPC64
cabe8138	152	#define REG "0x%016lx"
78a18dbf	153	#else
cabe8138	154	#define REG "0x%08lx"
78a18dbf	155	#endif
aaa22952	156
5f18cbdb	157	pt_dump_seq_printf(st->seq, REG "-" REG " ", st->start_address, addr - 1);
cabe8138	158	if (st->start_pa == st->last_pa && st->start_address + PAGE_SIZE != addr) {
5f18cbdb	159	pt_dump_seq_printf(st->seq, "[" REG "]", st->start_pa);
cabe8138 CL	160	delta = PAGE_SIZE >> 10;
cabe8138 CL	161	} else {
5f18cbdb	162	pt_dump_seq_printf(st->seq, " " REG " ", st->start_pa);
cabe8138 CL	163	delta = (addr - st->start_address) >> 10;
cabe8138 CL	164	}
8eb07b18 RG	165	/* Work out what appropriate unit to use */
	166	while (!(delta & 1023) && unit[1]) {
	167	delta >>= 10;
	168	unit++;
	169	}
5f18cbdb	170	pt_dump_seq_printf(st->seq, "%9lu%c", delta, *unit);
8eb07b18 RG	171
	172	}
	173
453d87f6 RC	174	static void note_prot_wx(struct pg_state *st, unsigned long addr)
453d87f6 RC	175	{
f3a2ac05	176	if (!IS_ENABLED(CONFIG_PPC_DEBUG_WX) \|\| !st->check_wx)
453d87f6 RC	177	return;
	178
	179	if (!((st->current_flags & pgprot_val(PAGE_KERNEL_X)) == pgprot_val(PAGE_KERNEL_X)))
	180	return;
	181
	182	WARN_ONCE(1, "powerpc/mm: Found insecure W+X mapping at address %p/%pS\n",
	183	(void )st->start_address, (void )st->start_address);
	184
	185	st->wx_pages += (addr - st->start_address) / PAGE_SIZE;
	186	}
	187
8eb07b18 RG	188	static void note_page(struct pg_state *st, unsigned long addr,
	189	unsigned int level, u64 val)
	190	{
	191	u64 flag = val & pg_level[level].mask;
aaa22952 OH	192	u64 pa = val & PTE_RPN_MASK;
aaa22952 OH	193
8eb07b18 RG	194	/* At first no level is set */
	195	if (!st->level) {
	196	st->level = level;
	197	st->current_flags = flag;
	198	st->start_address = addr;
aaa22952 OH	199	st->start_pa = pa;
aaa22952 OH	200	st->last_pa = pa;
5f18cbdb	201	pt_dump_seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
8eb07b18 RG	202	/*
	203	* Dump the section of virtual memory when:
	204	* - the PTE flags from one entry to the next differs.
	205	* - we change levels in the tree.
	206	* - the address is in a different section of memory and is thus
	207	* used for a different purpose, regardless of the flags.
aaa22952	208	* - the pa of this page is not adjacent to the last inspected page
8eb07b18 RG	209	*/
8eb07b18 RG	210	} else if (flag != st->current_flags \|\| level != st->level \|\|
aaa22952	211	addr >= st->marker[1].start_address \|\|
cabe8138 CL	212	(pa != st->last_pa + PAGE_SIZE &&
cabe8138 CL	213	(pa != st->start_pa \|\| st->start_pa != st->last_pa))) {
8eb07b18 RG	214
	215	/* Check the PTE flags */
	216	if (st->current_flags) {
453d87f6	217	note_prot_wx(st, addr);
8eb07b18 RG	218	dump_addr(st, addr);
	219
	220	/* Dump all the flags */
	221	if (pg_level[st->level].flag)
	222	dump_flag_info(st, pg_level[st->level].flag,
	223	st->current_flags,
	224	pg_level[st->level].num);
	225
5f18cbdb	226	pt_dump_seq_putc(st->seq, '\n');
8eb07b18 RG	227	}
	228
	229	/*
	230	* Address indicates we have passed the end of the
	231	* current section of virtual memory
	232	*/
	233	while (addr >= st->marker[1].start_address) {
	234	st->marker++;
5f18cbdb	235	pt_dump_seq_printf(st->seq, "---[ %s ]---\n", st->marker->name);
8eb07b18 RG	236	}
8eb07b18 RG	237	st->start_address = addr;
aaa22952 OH	238	st->start_pa = pa;
aaa22952 OH	239	st->last_pa = pa;
8eb07b18 RG	240	st->current_flags = flag;
8eb07b18 RG	241	st->level = level;
aaa22952 OH	242	} else {
aaa22952 OH	243	st->last_pa = pa;
8eb07b18 RG	244	}
	245	}
	246
	247	static void walk_pte(struct pg_state st, pmd_t pmd, unsigned long start)
	248	{
	249	pte_t *pte = pte_offset_kernel(pmd, 0);
	250	unsigned long addr;
	251	unsigned int i;
	252
	253	for (i = 0; i < PTRS_PER_PTE; i++, pte++) {
	254	addr = start + i * PAGE_SIZE;
	255	note_page(st, addr, 4, pte_val(*pte));
	256
	257	}
	258	}
	259
	260	static void walk_pmd(struct pg_state st, pud_t pud, unsigned long start)
	261	{
	262	pmd_t *pmd = pmd_offset(pud, 0);
	263	unsigned long addr;
	264	unsigned int i;
	265
	266	for (i = 0; i < PTRS_PER_PMD; i++, pmd++) {
	267	addr = start + i * PMD_SIZE;
d6eacedd	268	if (!pmd_none(pmd) && !pmd_is_leaf(pmd))
8eb07b18 RG	269	/* pmd exists */
	270	walk_pte(st, pmd, addr);
	271	else
	272	note_page(st, addr, 3, pmd_val(*pmd));
	273	}
	274	}
	275
	276	static void walk_pud(struct pg_state st, pgd_t pgd, unsigned long start)
	277	{
	278	pud_t *pud = pud_offset(pgd, 0);
	279	unsigned long addr;
	280	unsigned int i;
	281
	282	for (i = 0; i < PTRS_PER_PUD; i++, pud++) {
	283	addr = start + i * PUD_SIZE;
d6eacedd	284	if (!pud_none(pud) && !pud_is_leaf(pud))
8eb07b18 RG	285	/* pud exists */
	286	walk_pmd(st, pud, addr);
	287	else
	288	note_page(st, addr, 2, pud_val(*pud));
	289	}
	290	}
	291
	292	static void walk_pagetables(struct pg_state *st)
	293	{
8eb07b18	294	unsigned int i;
e033829d CL	295	unsigned long addr = st->start_address & PGDIR_MASK;
e033829d CL	296	pgd_t *pgd = pgd_offset_k(addr);
0d923962	297
8eb07b18 RG	298	/*
	299	* Traverse the linux pagetable structure and dump pages that are in
	300	* the hash pagetable.
	301	*/
e033829d	302	for (i = pgd_index(addr); i < PTRS_PER_PGD; i++, pgd++, addr += PGDIR_SIZE) {
d6eacedd	303	if (!pgd_none(pgd) && !pgd_is_leaf(pgd))
8eb07b18 RG	304	/* pgd exists */
	305	walk_pud(st, pgd, addr);
	306	else
	307	note_page(st, addr, 1, pgd_val(*pgd));
	308	}
	309	}
	310
	311	static void populate_markers(void)
	312	{
6c01bbd2 CL	313	int i = 0;
	314
	315	address_markers[i++].start_address = PAGE_OFFSET;
	316	address_markers[i++].start_address = VMALLOC_START;
	317	address_markers[i++].start_address = VMALLOC_END;
	318	#ifdef CONFIG_PPC64
	319	address_markers[i++].start_address = ISA_IO_BASE;
	320	address_markers[i++].start_address = ISA_IO_END;
	321	address_markers[i++].start_address = PHB_IO_BASE;
	322	address_markers[i++].start_address = PHB_IO_END;
	323	address_markers[i++].start_address = IOREMAP_BASE;
	324	address_markers[i++].start_address = IOREMAP_END;
0034d395	325	/* What is the ifdef about? */
4e003747	326	#ifdef CONFIG_PPC_BOOK3S_64
0034d395	327	address_markers[i++].start_address = H_VMEMMAP_START;
8eb07b18	328	#else
6c01bbd2 CL	329	address_markers[i++].start_address = VMEMMAP_BASE;
	330	#endif
	331	#else /* !CONFIG_PPC64 */
	332	address_markers[i++].start_address = ioremap_bot;
	333	address_markers[i++].start_address = IOREMAP_TOP;
6c01bbd2 CL	334	#ifdef CONFIG_HIGHMEM
	335	address_markers[i++].start_address = PKMAP_BASE;
	336	address_markers[i++].start_address = PKMAP_ADDR(LAST_PKMAP);
8eb07b18	337	#endif
6c01bbd2 CL	338	address_markers[i++].start_address = FIXADDR_START;
6c01bbd2 CL	339	address_markers[i++].start_address = FIXADDR_TOP;
b4abe38f CL	340	#ifdef CONFIG_KASAN
	341	address_markers[i++].start_address = KASAN_SHADOW_START;
	342	address_markers[i++].start_address = KASAN_SHADOW_END;
	343	#endif
6c01bbd2	344	#endif /* CONFIG_PPC64 */
8eb07b18 RG	345	}
	346
	347	static int ptdump_show(struct seq_file m, void v)
	348	{
	349	struct pg_state st = {
	350	.seq = m,
8eb07b18	351	.marker = address_markers,
82242352	352	.start_address = PAGE_OFFSET,
8eb07b18	353	};
0d923962	354
82242352 CL	355	#ifdef CONFIG_PPC64
82242352 CL	356	if (!radix_enabled())
0d923962	357	st.start_address = KERN_VIRT_START;
82242352	358	#endif
0d923962	359
8eb07b18 RG	360	/* Traverse kernel page tables */
	361	walk_pagetables(&st);
	362	note_page(&st, 0, 0, 0);
	363	return 0;
	364	}
	365
	366
	367	static int ptdump_open(struct inode inode, struct file file)
	368	{
	369	return single_open(file, ptdump_show, NULL);
	370	}
	371
	372	static const struct file_operations ptdump_fops = {
	373	.open = ptdump_open,
	374	.read = seq_read,
	375	.llseek = seq_lseek,
	376	.release = single_release,
	377	};
	378
	379	static void build_pgtable_complete_mask(void)
	380	{
	381	unsigned int i, j;
	382
	383	for (i = 0; i < ARRAY_SIZE(pg_level); i++)
	384	if (pg_level[i].flag)
	385	for (j = 0; j < pg_level[i].num; j++)
	386	pg_level[i].mask \|= pg_level[i].flag[j].mask;
	387	}
	388
453d87f6 RC	389	#ifdef CONFIG_PPC_DEBUG_WX
	390	void ptdump_check_wx(void)
	391	{
	392	struct pg_state st = {
	393	.seq = NULL,
	394	.marker = address_markers,
	395	.check_wx = true,
82242352	396	.start_address = PAGE_OFFSET,
453d87f6 RC	397	};
453d87f6 RC	398
82242352 CL	399	#ifdef CONFIG_PPC64
82242352 CL	400	if (!radix_enabled())
453d87f6	401	st.start_address = KERN_VIRT_START;
82242352	402	#endif
453d87f6 RC	403
	404	walk_pagetables(&st);
	405
	406	if (st.wx_pages)
	407	pr_warn("Checked W+X mappings: FAILED, %lu W+X pages found\n",
	408	st.wx_pages);
	409	else
	410	pr_info("Checked W+X mappings: passed, no W+X pages found\n");
	411	}
	412	#endif
	413
8eb07b18 RG	414	static int ptdump_init(void)
	415	{
	416	struct dentry *debugfs_file;
	417
	418	populate_markers();
	419	build_pgtable_complete_mask();
2505820f	420	debugfs_file = debugfs_create_file("kernel_page_tables", 0400, NULL,
8eb07b18 RG	421	NULL, &ptdump_fops);
	422	return debugfs_file ? 0 : -ENOMEM;
	423	}
	424	device_initcall(ptdump_init);