1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * Copyright (c) 2014, The Linux Foundation. All rights reserved. 4 * Debug helper to dump the current kernel pagetables of the system 5 * so that we can see what the various memory ranges are set to. 6 * 7 * Derived from x86 and arm implementation: 8 * (C) Copyright 2008 Intel Corporation 9 * 10 * Author: Arjan van de Ven <arjan@linux.intel.com> 11 */ 12 #include <linux/debugfs.h> 13 #include <linux/errno.h> 14 #include <linux/fs.h> 15 #include <linux/io.h> 16 #include <linux/init.h> 17 #include <linux/mm.h> 18 #include <linux/ptdump.h> 19 #include <linux/sched.h> 20 #include <linux/seq_file.h> 21 22 #include <asm/fixmap.h> 23 #include <asm/kasan.h> 24 #include <asm/memory.h> 25 #include <asm/pgtable-hwdef.h> 26 #include <asm/ptdump.h> 27 28 29 enum address_markers_idx { 30 PAGE_OFFSET_NR = 0, 31 PAGE_END_NR, 32 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 33 KASAN_START_NR, 34 #endif 35 }; 36 37 static struct addr_marker address_markers[] = { 38 { PAGE_OFFSET, "Linear Mapping start" }, 39 { 0 /* PAGE_END */, "Linear Mapping end" }, 40 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 41 { 0 /* KASAN_SHADOW_START */, "Kasan shadow start" }, 42 { KASAN_SHADOW_END, "Kasan shadow end" }, 43 #endif 44 { BPF_JIT_REGION_START, "BPF start" }, 45 { BPF_JIT_REGION_END, "BPF end" }, 46 { MODULES_VADDR, "Modules start" }, 47 { MODULES_END, "Modules end" }, 48 { VMALLOC_START, "vmalloc() area" }, 49 { VMALLOC_END, "vmalloc() end" }, 50 { FIXADDR_START, "Fixmap start" }, 51 { FIXADDR_TOP, "Fixmap end" }, 52 { PCI_IO_START, "PCI I/O start" }, 53 { PCI_IO_END, "PCI I/O end" }, 54 { VMEMMAP_START, "vmemmap start" }, 55 { VMEMMAP_START + VMEMMAP_SIZE, "vmemmap end" }, 56 { -1, NULL }, 57 }; 58 59 #define pt_dump_seq_printf(m, fmt, args...) \ 60 ({ \ 61 if (m) \ 62 seq_printf(m, fmt, ##args); \ 63 }) 64 65 #define pt_dump_seq_puts(m, fmt) \ 66 ({ \ 67 if (m) \ 68 seq_printf(m, fmt); \ 69 }) 70 71 /* 72 * The page dumper groups page table entries of the same type into a single 73 * description. It uses pg_state to track the range information while 74 * iterating over the pte entries. When the continuity is broken it then 75 * dumps out a description of the range. 76 */ 77 struct pg_state { 78 struct ptdump_state ptdump; 79 struct seq_file *seq; 80 const struct addr_marker *marker; 81 unsigned long start_address; 82 int level; 83 u64 current_prot; 84 bool check_wx; 85 unsigned long wx_pages; 86 unsigned long uxn_pages; 87 }; 88 89 struct prot_bits { 90 u64 mask; 91 u64 val; 92 const char *set; 93 const char *clear; 94 }; 95 96 static const struct prot_bits pte_bits[] = { 97 { 98 .mask = PTE_VALID, 99 .val = PTE_VALID, 100 .set = " ", 101 .clear = "F", 102 }, { 103 .mask = PTE_USER, 104 .val = PTE_USER, 105 .set = "USR", 106 .clear = " ", 107 }, { 108 .mask = PTE_RDONLY, 109 .val = PTE_RDONLY, 110 .set = "ro", 111 .clear = "RW", 112 }, { 113 .mask = PTE_PXN, 114 .val = PTE_PXN, 115 .set = "NX", 116 .clear = "x ", 117 }, { 118 .mask = PTE_SHARED, 119 .val = PTE_SHARED, 120 .set = "SHD", 121 .clear = " ", 122 }, { 123 .mask = PTE_AF, 124 .val = PTE_AF, 125 .set = "AF", 126 .clear = " ", 127 }, { 128 .mask = PTE_NG, 129 .val = PTE_NG, 130 .set = "NG", 131 .clear = " ", 132 }, { 133 .mask = PTE_CONT, 134 .val = PTE_CONT, 135 .set = "CON", 136 .clear = " ", 137 }, { 138 .mask = PTE_TABLE_BIT, 139 .val = PTE_TABLE_BIT, 140 .set = " ", 141 .clear = "BLK", 142 }, { 143 .mask = PTE_UXN, 144 .val = PTE_UXN, 145 .set = "UXN", 146 .clear = " ", 147 }, { 148 .mask = PTE_GP, 149 .val = PTE_GP, 150 .set = "GP", 151 .clear = " ", 152 }, { 153 .mask = PTE_ATTRINDX_MASK, 154 .val = PTE_ATTRINDX(MT_DEVICE_nGnRnE), 155 .set = "DEVICE/nGnRnE", 156 }, { 157 .mask = PTE_ATTRINDX_MASK, 158 .val = PTE_ATTRINDX(MT_DEVICE_nGnRE), 159 .set = "DEVICE/nGnRE", 160 }, { 161 .mask = PTE_ATTRINDX_MASK, 162 .val = PTE_ATTRINDX(MT_NORMAL_NC), 163 .set = "MEM/NORMAL-NC", 164 }, { 165 .mask = PTE_ATTRINDX_MASK, 166 .val = PTE_ATTRINDX(MT_NORMAL), 167 .set = "MEM/NORMAL", 168 }, { 169 .mask = PTE_ATTRINDX_MASK, 170 .val = PTE_ATTRINDX(MT_NORMAL_TAGGED), 171 .set = "MEM/NORMAL-TAGGED", 172 } 173 }; 174 175 struct pg_level { 176 const struct prot_bits *bits; 177 const char *name; 178 size_t num; 179 u64 mask; 180 }; 181 182 static struct pg_level pg_level[] = { 183 { /* pgd */ 184 .name = "PGD", 185 .bits = pte_bits, 186 .num = ARRAY_SIZE(pte_bits), 187 }, { /* p4d */ 188 .name = "P4D", 189 .bits = pte_bits, 190 .num = ARRAY_SIZE(pte_bits), 191 }, { /* pud */ 192 .name = (CONFIG_PGTABLE_LEVELS > 3) ? "PUD" : "PGD", 193 .bits = pte_bits, 194 .num = ARRAY_SIZE(pte_bits), 195 }, { /* pmd */ 196 .name = (CONFIG_PGTABLE_LEVELS > 2) ? "PMD" : "PGD", 197 .bits = pte_bits, 198 .num = ARRAY_SIZE(pte_bits), 199 }, { /* pte */ 200 .name = "PTE", 201 .bits = pte_bits, 202 .num = ARRAY_SIZE(pte_bits), 203 }, 204 }; 205 206 static void dump_prot(struct pg_state *st, const struct prot_bits *bits, 207 size_t num) 208 { 209 unsigned i; 210 211 for (i = 0; i < num; i++, bits++) { 212 const char *s; 213 214 if ((st->current_prot & bits->mask) == bits->val) 215 s = bits->set; 216 else 217 s = bits->clear; 218 219 if (s) 220 pt_dump_seq_printf(st->seq, " %s", s); 221 } 222 } 223 224 static void note_prot_uxn(struct pg_state *st, unsigned long addr) 225 { 226 if (!st->check_wx) 227 return; 228 229 if ((st->current_prot & PTE_UXN) == PTE_UXN) 230 return; 231 232 WARN_ONCE(1, "arm64/mm: Found non-UXN mapping at address %p/%pS\n", 233 (void *)st->start_address, (void *)st->start_address); 234 235 st->uxn_pages += (addr - st->start_address) / PAGE_SIZE; 236 } 237 238 static void note_prot_wx(struct pg_state *st, unsigned long addr) 239 { 240 if (!st->check_wx) 241 return; 242 if ((st->current_prot & PTE_RDONLY) == PTE_RDONLY) 243 return; 244 if ((st->current_prot & PTE_PXN) == PTE_PXN) 245 return; 246 247 WARN_ONCE(1, "arm64/mm: Found insecure W+X mapping at address %p/%pS\n", 248 (void *)st->start_address, (void *)st->start_address); 249 250 st->wx_pages += (addr - st->start_address) / PAGE_SIZE; 251 } 252 253 static void note_page(struct ptdump_state *pt_st, unsigned long addr, int level, 254 u64 val) 255 { 256 struct pg_state *st = container_of(pt_st, struct pg_state, ptdump); 257 static const char units[] = "KMGTPE"; 258 u64 prot = 0; 259 260 if (level >= 0) 261 prot = val & pg_level[level].mask; 262 263 if (st->level == -1) { 264 st->level = level; 265 st->current_prot = prot; 266 st->start_address = addr; 267 pt_dump_seq_printf(st->seq, "---[ %s ]---\n", st->marker->name); 268 } else if (prot != st->current_prot || level != st->level || 269 addr >= st->marker[1].start_address) { 270 const char *unit = units; 271 unsigned long delta; 272 273 if (st->current_prot) { 274 note_prot_uxn(st, addr); 275 note_prot_wx(st, addr); 276 } 277 278 pt_dump_seq_printf(st->seq, "0x%016lx-0x%016lx ", 279 st->start_address, addr); 280 281 delta = (addr - st->start_address) >> 10; 282 while (!(delta & 1023) && unit[1]) { 283 delta >>= 10; 284 unit++; 285 } 286 pt_dump_seq_printf(st->seq, "%9lu%c %s", delta, *unit, 287 pg_level[st->level].name); 288 if (st->current_prot && pg_level[st->level].bits) 289 dump_prot(st, pg_level[st->level].bits, 290 pg_level[st->level].num); 291 pt_dump_seq_puts(st->seq, "\n"); 292 293 if (addr >= st->marker[1].start_address) { 294 st->marker++; 295 pt_dump_seq_printf(st->seq, "---[ %s ]---\n", st->marker->name); 296 } 297 298 st->start_address = addr; 299 st->current_prot = prot; 300 st->level = level; 301 } 302 303 if (addr >= st->marker[1].start_address) { 304 st->marker++; 305 pt_dump_seq_printf(st->seq, "---[ %s ]---\n", st->marker->name); 306 } 307 308 } 309 310 void ptdump_walk(struct seq_file *s, struct ptdump_info *info) 311 { 312 unsigned long end = ~0UL; 313 struct pg_state st; 314 315 if (info->base_addr < TASK_SIZE_64) 316 end = TASK_SIZE_64; 317 318 st = (struct pg_state){ 319 .seq = s, 320 .marker = info->markers, 321 .level = -1, 322 .ptdump = { 323 .note_page = note_page, 324 .range = (struct ptdump_range[]){ 325 {info->base_addr, end}, 326 {0, 0} 327 } 328 } 329 }; 330 331 ptdump_walk_pgd(&st.ptdump, info->mm, NULL); 332 } 333 334 static void __init ptdump_initialize(void) 335 { 336 unsigned i, j; 337 338 for (i = 0; i < ARRAY_SIZE(pg_level); i++) 339 if (pg_level[i].bits) 340 for (j = 0; j < pg_level[i].num; j++) 341 pg_level[i].mask |= pg_level[i].bits[j].mask; 342 } 343 344 static struct ptdump_info kernel_ptdump_info = { 345 .mm = &init_mm, 346 .markers = address_markers, 347 .base_addr = PAGE_OFFSET, 348 }; 349 350 void ptdump_check_wx(void) 351 { 352 struct pg_state st = { 353 .seq = NULL, 354 .marker = (struct addr_marker[]) { 355 { 0, NULL}, 356 { -1, NULL}, 357 }, 358 .level = -1, 359 .check_wx = true, 360 .ptdump = { 361 .note_page = note_page, 362 .range = (struct ptdump_range[]) { 363 {PAGE_OFFSET, ~0UL}, 364 {0, 0} 365 } 366 } 367 }; 368 369 ptdump_walk_pgd(&st.ptdump, &init_mm, NULL); 370 371 if (st.wx_pages || st.uxn_pages) 372 pr_warn("Checked W+X mappings: FAILED, %lu W+X pages found, %lu non-UXN pages found\n", 373 st.wx_pages, st.uxn_pages); 374 else 375 pr_info("Checked W+X mappings: passed, no W+X pages found\n"); 376 } 377 378 static int __init ptdump_init(void) 379 { 380 address_markers[PAGE_END_NR].start_address = PAGE_END; 381 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS) 382 address_markers[KASAN_START_NR].start_address = KASAN_SHADOW_START; 383 #endif 384 ptdump_initialize(); 385 ptdump_debugfs_register(&kernel_ptdump_info, "kernel_page_tables"); 386 return 0; 387 } 388 device_initcall(ptdump_init); 389