1 files changed, 284 insertions, 166 deletions

diff --git a/arch/s390/mm/dump_pagetables.c b/arch/s390/mm/dump_pagetables.c
index 3ad65b04ac15..89badbe72ae7 100644
--- a/arch/s390/mm/dump_pagetables.c
+++ b/arch/s390/mm/dump_pagetables.c
@@ -1,246 +1,364 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/cpufeature.h>
+#include <linux/set_memory.h>
+#include <linux/ptdump.h>
 #include <linux/seq_file.h>
 #include <linux/debugfs.h>
-#include <linux/module.h>
+#include <linux/sort.h>
 #include <linux/mm.h>
+#include <linux/kfence.h>
+#include <linux/kasan.h>
+#include <asm/kasan.h>
+#include <asm/abs_lowcore.h>
+#include <asm/nospec-branch.h>
 #include <asm/sections.h>
-#include <asm/pgtable.h>
+#include <asm/maccess.h>
 
 static unsigned long max_addr;
 
 struct addr_marker {
+	int is_start;
 	unsigned long start_address;
+	unsigned long size;
 	const char *name;
 };
 
-enum address_markers_idx {
-	IDENTITY_NR = 0,
-	KERNEL_START_NR,
-	KERNEL_END_NR,
-	VMEMMAP_NR,
-	VMALLOC_NR,
-#ifdef CONFIG_64BIT
-	MODULES_NR,
-#endif
-};
-
-static struct addr_marker address_markers[] = {
-	[IDENTITY_NR]	  = {0, "Identity Mapping"},
-	[KERNEL_START_NR] = {(unsigned long)&_stext, "Kernel Image Start"},
-	[KERNEL_END_NR]	  = {(unsigned long)&_end, "Kernel Image End"},
-	[VMEMMAP_NR]	  = {0, "vmemmap Area"},
-	[VMALLOC_NR]	  = {0, "vmalloc Area"},
-#ifdef CONFIG_64BIT
-	[MODULES_NR]	  = {0, "Modules Area"},
-#endif
-	{ -1, NULL }
-};
+static struct addr_marker *markers;
+static unsigned int markers_cnt;
 
 struct pg_state {
+	struct ptdump_state ptdump;
+	struct seq_file *seq;
 	int level;
 	unsigned int current_prot;
+	bool check_wx;
+	unsigned long wx_pages;
 	unsigned long start_address;
-	unsigned long current_address;
 	const struct addr_marker *marker;
 };
 
+#define pt_dump_seq_printf(m, fmt, args...)	\
+({						\
+	struct seq_file *__m = (m);		\
+						\
+	if (__m)				\
+		seq_printf(__m, fmt, ##args);	\
+})
+
+#define pt_dump_seq_puts(m, fmt)		\
+({						\
+	struct seq_file *__m = (m);		\
+						\
+	if (__m)				\
+		seq_puts(__m, fmt);		\
+})
+
 static void print_prot(struct seq_file *m, unsigned int pr, int level)
 {
 	static const char * const level_name[] =
 		{ "ASCE", "PGD", "PUD", "PMD", "PTE" };
 
-	seq_printf(m, "%s ", level_name[level]);
+	pt_dump_seq_printf(m, "%s ", level_name[level]);
 	if (pr & _PAGE_INVALID) {
-		seq_printf(m, "I\n");
+		pt_dump_seq_printf(m, "I\n");
 		return;
 	}
-	seq_printf(m, "%s", pr & _PAGE_RO ? "RO " : "RW ");
-	seq_printf(m, "%s", pr & _PAGE_CO ? "CO " : "   ");
-	seq_putc(m, '\n');
+	pt_dump_seq_puts(m, (pr & _PAGE_PROTECT) ? "RO " : "RW ");
+	pt_dump_seq_puts(m, (pr & _PAGE_NOEXEC) ? "NX\n" : "X\n");
 }
 
-static void note_page(struct seq_file *m, struct pg_state *st,
-		     unsigned int new_prot, int level)
+static void note_prot_wx(struct pg_state *st, unsigned long addr)
+{
+	if (!st->check_wx)
+		return;
+	if (st->current_prot & _PAGE_INVALID)
+		return;
+	if (st->current_prot & _PAGE_PROTECT)
+		return;
+	if (st->current_prot & _PAGE_NOEXEC)
+		return;
+	/*
+	 * The first lowcore page is W+X if spectre mitigations are using
+	 * trampolines or the BEAR enhancements facility is not installed,
+	 * in which case we have two lpswe instructions in lowcore that need
+	 * to be executable.
+	 */
+	if (addr == PAGE_SIZE && (nospec_uses_trampoline() || !cpu_has_bear()))
+		return;
+	WARN_ONCE(IS_ENABLED(CONFIG_DEBUG_WX),
+		  "s390/mm: Found insecure W+X mapping at address %pS\n",
+		  (void *)st->start_address);
+	st->wx_pages += (addr - st->start_address) / PAGE_SIZE;
+}
+
+static void note_page_update_state(struct pg_state *st, unsigned long addr, unsigned int prot, int level)
+{
+	struct seq_file *m = st->seq;
+
+	while (addr >= st->marker[1].start_address) {
+		st->marker++;
+		pt_dump_seq_printf(m, "---[ %s %s ]---\n", st->marker->name,
+				   st->marker->is_start ? "Start" : "End");
+	}
+	st->start_address = addr;
+	st->current_prot = prot;
+	st->level = level;
+}
+
+static void note_page(struct ptdump_state *pt_st, unsigned long addr, int level, u64 val)
 {
-	static const char units[] = "KMGTPE";
 	int width = sizeof(unsigned long) * 2;
+	static const char units[] = "KMGTPE";
 	const char *unit = units;
-	unsigned int prot, cur;
 	unsigned long delta;
+	struct pg_state *st;
+	struct seq_file *m;
+	unsigned int prot;
 
-	/*
-	 * 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 = new_prot;
-	cur = st->current_prot;
-
-	if (!st->level) {
-		/* First entry */
-		st->current_prot = new_prot;
-		st->level = level;
-		st->marker = address_markers;
-		seq_printf(m, "---[ %s ]---\n", st->marker->name);
-	} else if (prot != cur || level != st->level ||
-		   st->current_address >= st->marker[1].start_address) {
-		/* Print the actual finished series */
-		seq_printf(m, "0x%0*lx-0x%0*lx",
-			   width, st->start_address,
-			   width, st->current_address);
-		delta = (st->current_address - st->start_address) >> 10;
+	st = container_of(pt_st, struct pg_state, ptdump);
+	m = st->seq;
+	prot = val & (_PAGE_PROTECT | _PAGE_NOEXEC);
+	if (level == 4 && (val & _PAGE_INVALID))
+		prot = _PAGE_INVALID;
+	/* For pmd_none() & friends val gets passed as zero. */
+	if (level != 4 && !val)
+		prot = _PAGE_INVALID;
+	/* Final flush from generic code. */
+	if (level == -1)
+		addr = max_addr;
+	if (st->level == -1) {
+		pt_dump_seq_puts(m, "---[ Kernel Virtual Address Space ]---\n");
+		note_page_update_state(st, addr, prot, level);
+	} else if (prot != st->current_prot || level != st->level ||
+		   addr >= st->marker[1].start_address) {
+		note_prot_wx(st, addr);
+		pt_dump_seq_printf(m, "0x%0*lx-0x%0*lx ",
+				   width, st->start_address,
+				   width, addr);
+		delta = (addr - st->start_address) >> 10;
 		while (!(delta & 0x3ff) && unit[1]) {
 			delta >>= 10;
 			unit++;
 		}
-		seq_printf(m, "%9lu%c ", delta, *unit);
+		pt_dump_seq_printf(m, "%9lu%c ", delta, *unit);
 		print_prot(m, st->current_prot, st->level);
-		if (st->current_address >= st->marker[1].start_address) {
-			st->marker++;
-			seq_printf(m, "---[ %s ]---\n", st->marker->name);
-		}
-		st->start_address = st->current_address;
-		st->current_prot = new_prot;
-		st->level = level;
+		note_page_update_state(st, addr, prot, level);
 	}
 }
 
-/*
- * The actual page table walker functions. In order to keep the implementation
- * of print_prot() short, we only check and pass _PAGE_INVALID and _PAGE_RO
- * flags to note_page() if a region, segment or page table entry is invalid or
- * read-only.
- * After all it's just a hint that the current level being walked contains an
- * invalid or read-only entry.
- */
-static void walk_pte_level(struct seq_file *m, struct pg_state *st,
-			   pmd_t *pmd, unsigned long addr)
+static void note_page_pte(struct ptdump_state *pt_st, unsigned long addr, pte_t pte)
 {
-	unsigned int prot;
-	pte_t *pte;
-	int i;
-
-	for (i = 0; i < PTRS_PER_PTE && addr < max_addr; i++) {
-		st->current_address = addr;
-		pte = pte_offset_kernel(pmd, addr);
-		prot = pte_val(*pte) & (_PAGE_RO | _PAGE_INVALID);
-		note_page(m, st, prot, 4);
-		addr += PAGE_SIZE;
-	}
+	note_page(pt_st, addr, 4, pte_val(pte));
 }
 
-#ifdef CONFIG_64BIT
-#define _PMD_PROT_MASK (_SEGMENT_ENTRY_RO | _SEGMENT_ENTRY_CO)
-#else
-#define _PMD_PROT_MASK 0
-#endif
+static void note_page_pmd(struct ptdump_state *pt_st, unsigned long addr, pmd_t pmd)
+{
+	note_page(pt_st, addr, 3, pmd_val(pmd));
+}
 
-static void walk_pmd_level(struct seq_file *m, struct pg_state *st,
-			   pud_t *pud, unsigned long addr)
+static void note_page_pud(struct ptdump_state *pt_st, unsigned long addr, pud_t pud)
 {
-	unsigned int prot;
-	pmd_t *pmd;
-	int i;
-
-	for (i = 0; i < PTRS_PER_PMD && addr < max_addr; i++) {
-		st->current_address = addr;
-		pmd = pmd_offset(pud, addr);
-		if (!pmd_none(*pmd)) {
-			if (pmd_large(*pmd)) {
-				prot = pmd_val(*pmd) & _PMD_PROT_MASK;
-				note_page(m, st, prot, 3);
-			} else
-				walk_pte_level(m, st, pmd, addr);
-		} else
-			note_page(m, st, _PAGE_INVALID, 3);
-		addr += PMD_SIZE;
-	}
+	note_page(pt_st, addr, 2, pud_val(pud));
 }
 
-#ifdef CONFIG_64BIT
-#define _PUD_PROT_MASK (_REGION3_ENTRY_RO | _REGION3_ENTRY_CO)
-#else
-#define _PUD_PROT_MASK 0
-#endif
+static void note_page_p4d(struct ptdump_state *pt_st, unsigned long addr, p4d_t p4d)
+{
+	note_page(pt_st, addr, 1, p4d_val(p4d));
+}
 
-static void walk_pud_level(struct seq_file *m, struct pg_state *st,
-			   pgd_t *pgd, unsigned long addr)
+static void note_page_pgd(struct ptdump_state *pt_st, unsigned long addr, pgd_t pgd)
 {
-	unsigned int prot;
-	pud_t *pud;
-	int i;
-
-	for (i = 0; i < PTRS_PER_PUD && addr < max_addr; i++) {
-		st->current_address = addr;
-		pud = pud_offset(pgd, addr);
-		if (!pud_none(*pud))
-			if (pud_large(*pud)) {
-				prot = pud_val(*pud) & _PUD_PROT_MASK;
-				note_page(m, st, prot, 2);
-			} else
-				walk_pmd_level(m, st, pud, addr);
-		else
-			note_page(m, st, _PAGE_INVALID, 2);
-		addr += PUD_SIZE;
-	}
+	note_page(pt_st, addr, 0, pgd_val(pgd));
+}
+
+static void note_page_flush(struct ptdump_state *pt_st)
+{
+	pte_t pte_zero = {0};
+
+	note_page(pt_st, 0, -1, pte_val(pte_zero));
 }
 
-static void walk_pgd_level(struct seq_file *m)
+bool ptdump_check_wx(void)
 {
-	unsigned long addr = 0;
-	struct pg_state st;
-	pgd_t *pgd;
-	int i;
-
-	memset(&st, 0, sizeof(st));
-	for (i = 0; i < PTRS_PER_PGD && addr < max_addr; i++) {
-		st.current_address = addr;
-		pgd = pgd_offset_k(addr);
-		if (!pgd_none(*pgd))
-			walk_pud_level(m, &st, pgd, addr);
-		else
-			note_page(m, &st, _PAGE_INVALID, 1);
-		addr += PGDIR_SIZE;
+	struct pg_state st = {
+		.ptdump = {
+			.note_page_pte = note_page_pte,
+			.note_page_pmd = note_page_pmd,
+			.note_page_pud = note_page_pud,
+			.note_page_p4d = note_page_p4d,
+			.note_page_pgd = note_page_pgd,
+			.note_page_flush = note_page_flush,
+			.range = (struct ptdump_range[]) {
+				{.start = 0, .end = max_addr},
+				{.start = 0, .end = 0},
+			}
+		},
+		.seq = NULL,
+		.level = -1,
+		.current_prot = 0,
+		.check_wx = true,
+		.wx_pages = 0,
+		.start_address = 0,
+		.marker = (struct addr_marker[]) {
+			{ .start_address =  0, .name = NULL},
+			{ .start_address = -1, .name = NULL},
+		},
+	};
+
+	if (!cpu_has_nx())
+		return true;
+	ptdump_walk_pgd(&st.ptdump, &init_mm, NULL);
+	if (st.wx_pages) {
+		pr_warn("Checked W+X mappings: FAILED, %lu W+X pages found\n", st.wx_pages);
+
+		return false;
+	} else {
+		pr_info("Checked W+X mappings: passed, no %sW+X pages found\n",
+			(nospec_uses_trampoline() || !cpu_has_bear()) ?
+			"unexpected " : "");
+
+		return true;
 	}
-	/* Flush out the last page */
-	st.current_address = max_addr;
-	note_page(m, &st, 0, 0);
 }
 
+#ifdef CONFIG_PTDUMP_DEBUGFS
 static int ptdump_show(struct seq_file *m, void *v)
 {
-	walk_pgd_level(m);
+	struct pg_state st = {
+		.ptdump = {
+			.note_page_pte = note_page_pte,
+			.note_page_pmd = note_page_pmd,
+			.note_page_pud = note_page_pud,
+			.note_page_p4d = note_page_p4d,
+			.note_page_pgd = note_page_pgd,
+			.note_page_flush = note_page_flush,
+			.range = (struct ptdump_range[]) {
+				{.start = 0, .end = max_addr},
+				{.start = 0, .end = 0},
+			}
+		},
+		.seq = m,
+		.level = -1,
+		.current_prot = 0,
+		.check_wx = false,
+		.wx_pages = 0,
+		.start_address = 0,
+		.marker = markers,
+	};
+
+	mutex_lock(&cpa_mutex);
+	ptdump_walk_pgd(&st.ptdump, &init_mm, NULL);
+	mutex_unlock(&cpa_mutex);
 	return 0;
 }
+DEFINE_SHOW_ATTRIBUTE(ptdump);
+#endif /* CONFIG_PTDUMP_DEBUGFS */
 
-static int ptdump_open(struct inode *inode, struct file *filp)
+static int ptdump_cmp(const void *a, const void *b)
 {
-	return single_open(filp, ptdump_show, NULL);
+	const struct addr_marker *ama = a;
+	const struct addr_marker *amb = b;
+
+	if (ama->start_address > amb->start_address)
+		return 1;
+	if (ama->start_address < amb->start_address)
+		return -1;
+	/*
+	 * If the start addresses of two markers are identical sort markers in an
+	 * order that considers areas contained within other areas correctly.
+	 */
+	if (ama->is_start && amb->is_start) {
+		if (ama->size > amb->size)
+			return -1;
+		if (ama->size < amb->size)
+			return 1;
+		return 0;
+	}
+	if (!ama->is_start && !amb->is_start) {
+		if (ama->size > amb->size)
+			return 1;
+		if (ama->size < amb->size)
+			return -1;
+		return 0;
+	}
+	if (ama->is_start)
+		return 1;
+	if (amb->is_start)
+		return -1;
+	return 0;
 }
 
-static const struct file_operations ptdump_fops = {
-	.open		= ptdump_open,
-	.read		= seq_read,
-	.llseek		= seq_lseek,
-	.release	= single_release,
-};
+static int add_marker(unsigned long start, unsigned long end, const char *name)
+{
+	struct addr_marker *new;
+	size_t newsize;
+
+	newsize = (markers_cnt + 2) * sizeof(*markers);
+	new = kvrealloc(markers, newsize, GFP_KERNEL);
+	if (!new)
+		return -ENOMEM;
+	markers = new;
+	markers[markers_cnt].is_start = 1;
+	markers[markers_cnt].start_address = start;
+	markers[markers_cnt].size = end - start;
+	markers[markers_cnt].name = name;
+	markers_cnt++;
+	markers[markers_cnt].is_start = 0;
+	markers[markers_cnt].start_address = end;
+	markers[markers_cnt].size = end - start;
+	markers[markers_cnt].name = name;
+	markers_cnt++;
+	return 0;
+}
 
 static int pt_dump_init(void)
 {
+#ifdef CONFIG_KFENCE
+	unsigned long kfence_start = (unsigned long)__kfence_pool;
+#endif
+	unsigned long lowcore = (unsigned long)get_lowcore();
+	int rc;
+
 	/*
 	 * Figure out the maximum virtual address being accessible with the
 	 * kernel ASCE. We need this to keep the page table walker functions
 	 * from accessing non-existent entries.
 	 */
-#ifdef CONFIG_32BIT
-	max_addr = 1UL << 31;
-#else
-	max_addr = (S390_lowcore.kernel_asce & _REGION_ENTRY_TYPE_MASK) >> 2;
+	max_addr = (get_lowcore()->kernel_asce.val & _REGION_ENTRY_TYPE_MASK) >> 2;
 	max_addr = 1UL << (max_addr * 11 + 31);
-	address_markers[MODULES_NR].start_address = MODULES_VADDR;
+	/* start + end markers - must be added first */
+	rc = add_marker(0, -1UL, NULL);
+	rc |= add_marker((unsigned long)_stext, (unsigned long)_end, "Kernel Image");
+	rc |= add_marker(lowcore, lowcore + sizeof(struct lowcore), "Lowcore");
+	rc |= add_marker(__identity_base, __identity_base + ident_map_size, "Identity Mapping");
+	rc |= add_marker((unsigned long)__samode31, (unsigned long)__eamode31, "Amode31 Area");
+	rc |= add_marker(MODULES_VADDR, MODULES_END, "Modules Area");
+	rc |= add_marker(__abs_lowcore, __abs_lowcore + ABS_LOWCORE_MAP_SIZE, "Lowcore Area");
+	rc |= add_marker(__memcpy_real_area, __memcpy_real_area + MEMCPY_REAL_SIZE, "Real Memory Copy Area");
+	rc |= add_marker((unsigned long)vmemmap, (unsigned long)vmemmap + vmemmap_size, "vmemmap Area");
+	rc |= add_marker(VMALLOC_START, VMALLOC_END, "vmalloc Area");
+#ifdef CONFIG_KFENCE
+	rc |= add_marker(kfence_start, kfence_start + KFENCE_POOL_SIZE, "KFence Pool");
+#endif
+#ifdef CONFIG_KMSAN
+	rc |= add_marker(KMSAN_VMALLOC_SHADOW_START, KMSAN_VMALLOC_SHADOW_END, "Kmsan vmalloc Shadow");
+	rc |= add_marker(KMSAN_VMALLOC_ORIGIN_START, KMSAN_VMALLOC_ORIGIN_END, "Kmsan vmalloc Origins");
+	rc |= add_marker(KMSAN_MODULES_SHADOW_START, KMSAN_MODULES_SHADOW_END, "Kmsan Modules Shadow");
+	rc |= add_marker(KMSAN_MODULES_ORIGIN_START, KMSAN_MODULES_ORIGIN_END, "Kmsan Modules Origins");
+#endif
+#ifdef CONFIG_KASAN
+	rc |= add_marker(KASAN_SHADOW_START, KASAN_SHADOW_END, "Kasan Shadow");
 #endif
-	address_markers[VMEMMAP_NR].start_address = (unsigned long) vmemmap;
-	address_markers[VMALLOC_NR].start_address = VMALLOC_START;
+	if (rc)
+		goto error;
+	sort(&markers[1], markers_cnt - 1, sizeof(*markers), ptdump_cmp, NULL);
+#ifdef CONFIG_PTDUMP_DEBUGFS
 	debugfs_create_file("kernel_page_tables", 0400, NULL, NULL, &ptdump_fops);
+#endif /* CONFIG_PTDUMP_DEBUGFS */
 	return 0;
+error:
+	kvfree(markers);
+	return -ENOMEM;
 }
 device_initcall(pt_dump_init);