1 /* 2 * Based on arch/arm/mm/fault.c 3 * 4 * Copyright (C) 1995 Linus Torvalds 5 * Copyright (C) 1995-2004 Russell King 6 * Copyright (C) 2012 ARM Ltd. 7 * 8 * This program is free software; you can redistribute it and/or modify 9 * it under the terms of the GNU General Public License version 2 as 10 * published by the Free Software Foundation. 11 * 12 * This program is distributed in the hope that it will be useful, 13 * but WITHOUT ANY WARRANTY; without even the implied warranty of 14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 15 * GNU General Public License for more details. 16 * 17 * You should have received a copy of the GNU General Public License 18 * along with this program. If not, see <http://www.gnu.org/licenses/>. 19 */ 20 21 #include <linux/module.h> 22 #include <linux/signal.h> 23 #include <linux/mm.h> 24 #include <linux/hardirq.h> 25 #include <linux/init.h> 26 #include <linux/kprobes.h> 27 #include <linux/uaccess.h> 28 #include <linux/page-flags.h> 29 #include <linux/sched.h> 30 #include <linux/highmem.h> 31 #include <linux/perf_event.h> 32 33 #include <asm/cpufeature.h> 34 #include <asm/exception.h> 35 #include <asm/debug-monitors.h> 36 #include <asm/esr.h> 37 #include <asm/sysreg.h> 38 #include <asm/system_misc.h> 39 #include <asm/pgtable.h> 40 #include <asm/tlbflush.h> 41 42 static const char *fault_name(unsigned int esr); 43 44 /* 45 * Dump out the page tables associated with 'addr' in mm 'mm'. 46 */ 47 void show_pte(struct mm_struct *mm, unsigned long addr) 48 { 49 pgd_t *pgd; 50 51 if (!mm) 52 mm = &init_mm; 53 54 pr_alert("pgd = %p\n", mm->pgd); 55 pgd = pgd_offset(mm, addr); 56 pr_alert("[%08lx] *pgd=%016llx", addr, pgd_val(*pgd)); 57 58 do { 59 pud_t *pud; 60 pmd_t *pmd; 61 pte_t *pte; 62 63 if (pgd_none(*pgd) || pgd_bad(*pgd)) 64 break; 65 66 pud = pud_offset(pgd, addr); 67 printk(", *pud=%016llx", pud_val(*pud)); 68 if (pud_none(*pud) || pud_bad(*pud)) 69 break; 70 71 pmd = pmd_offset(pud, addr); 72 printk(", *pmd=%016llx", pmd_val(*pmd)); 73 if (pmd_none(*pmd) || pmd_bad(*pmd)) 74 break; 75 76 pte = pte_offset_map(pmd, addr); 77 printk(", *pte=%016llx", pte_val(*pte)); 78 pte_unmap(pte); 79 } while(0); 80 81 printk("\n"); 82 } 83 84 /* 85 * The kernel tried to access some page that wasn't present. 86 */ 87 static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr, 88 unsigned int esr, struct pt_regs *regs) 89 { 90 /* 91 * Are we prepared to handle this kernel fault? 92 */ 93 if (fixup_exception(regs)) 94 return; 95 96 /* 97 * No handler, we'll have to terminate things with extreme prejudice. 98 */ 99 bust_spinlocks(1); 100 pr_alert("Unable to handle kernel %s at virtual address %08lx\n", 101 (addr < PAGE_SIZE) ? "NULL pointer dereference" : 102 "paging request", addr); 103 104 show_pte(mm, addr); 105 die("Oops", regs, esr); 106 bust_spinlocks(0); 107 do_exit(SIGKILL); 108 } 109 110 /* 111 * Something tried to access memory that isn't in our memory map. User mode 112 * accesses just cause a SIGSEGV 113 */ 114 static void __do_user_fault(struct task_struct *tsk, unsigned long addr, 115 unsigned int esr, unsigned int sig, int code, 116 struct pt_regs *regs) 117 { 118 struct siginfo si; 119 120 if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) { 121 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n", 122 tsk->comm, task_pid_nr(tsk), fault_name(esr), sig, 123 addr, esr); 124 show_pte(tsk->mm, addr); 125 show_regs(regs); 126 } 127 128 tsk->thread.fault_address = addr; 129 tsk->thread.fault_code = esr; 130 si.si_signo = sig; 131 si.si_errno = 0; 132 si.si_code = code; 133 si.si_addr = (void __user *)addr; 134 force_sig_info(sig, &si, tsk); 135 } 136 137 static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs) 138 { 139 struct task_struct *tsk = current; 140 struct mm_struct *mm = tsk->active_mm; 141 142 /* 143 * If we are in kernel mode at this point, we have no context to 144 * handle this fault with. 145 */ 146 if (user_mode(regs)) 147 __do_user_fault(tsk, addr, esr, SIGSEGV, SEGV_MAPERR, regs); 148 else 149 __do_kernel_fault(mm, addr, esr, regs); 150 } 151 152 #define VM_FAULT_BADMAP 0x010000 153 #define VM_FAULT_BADACCESS 0x020000 154 155 #define ESR_LNX_EXEC (1 << 24) 156 157 static int __do_page_fault(struct mm_struct *mm, unsigned long addr, 158 unsigned int mm_flags, unsigned long vm_flags, 159 struct task_struct *tsk) 160 { 161 struct vm_area_struct *vma; 162 int fault; 163 164 vma = find_vma(mm, addr); 165 fault = VM_FAULT_BADMAP; 166 if (unlikely(!vma)) 167 goto out; 168 if (unlikely(vma->vm_start > addr)) 169 goto check_stack; 170 171 /* 172 * Ok, we have a good vm_area for this memory access, so we can handle 173 * it. 174 */ 175 good_area: 176 /* 177 * Check that the permissions on the VMA allow for the fault which 178 * occurred. If we encountered a write or exec fault, we must have 179 * appropriate permissions, otherwise we allow any permission. 180 */ 181 if (!(vma->vm_flags & vm_flags)) { 182 fault = VM_FAULT_BADACCESS; 183 goto out; 184 } 185 186 return handle_mm_fault(mm, vma, addr & PAGE_MASK, mm_flags); 187 188 check_stack: 189 if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr)) 190 goto good_area; 191 out: 192 return fault; 193 } 194 195 static inline int permission_fault(unsigned int esr) 196 { 197 unsigned int ec = (esr & ESR_ELx_EC_MASK) >> ESR_ELx_EC_SHIFT; 198 unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE; 199 200 return (ec == ESR_ELx_EC_DABT_CUR && fsc_type == ESR_ELx_FSC_PERM); 201 } 202 203 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr, 204 struct pt_regs *regs) 205 { 206 struct task_struct *tsk; 207 struct mm_struct *mm; 208 int fault, sig, code; 209 unsigned long vm_flags = VM_READ | VM_WRITE | VM_EXEC; 210 unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE; 211 212 tsk = current; 213 mm = tsk->mm; 214 215 /* Enable interrupts if they were enabled in the parent context. */ 216 if (interrupts_enabled(regs)) 217 local_irq_enable(); 218 219 /* 220 * If we're in an interrupt or have no user context, we must not take 221 * the fault. 222 */ 223 if (faulthandler_disabled() || !mm) 224 goto no_context; 225 226 if (user_mode(regs)) 227 mm_flags |= FAULT_FLAG_USER; 228 229 if (esr & ESR_LNX_EXEC) { 230 vm_flags = VM_EXEC; 231 } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) { 232 vm_flags = VM_WRITE; 233 mm_flags |= FAULT_FLAG_WRITE; 234 } 235 236 if (permission_fault(esr) && (addr < USER_DS)) { 237 if (get_fs() == KERNEL_DS) 238 die("Accessing user space memory with fs=KERNEL_DS", regs, esr); 239 240 if (!search_exception_tables(regs->pc)) 241 die("Accessing user space memory outside uaccess.h routines", regs, esr); 242 } 243 244 /* 245 * As per x86, we may deadlock here. However, since the kernel only 246 * validly references user space from well defined areas of the code, 247 * we can bug out early if this is from code which shouldn't. 248 */ 249 if (!down_read_trylock(&mm->mmap_sem)) { 250 if (!user_mode(regs) && !search_exception_tables(regs->pc)) 251 goto no_context; 252 retry: 253 down_read(&mm->mmap_sem); 254 } else { 255 /* 256 * The above down_read_trylock() might have succeeded in which 257 * case, we'll have missed the might_sleep() from down_read(). 258 */ 259 might_sleep(); 260 #ifdef CONFIG_DEBUG_VM 261 if (!user_mode(regs) && !search_exception_tables(regs->pc)) 262 goto no_context; 263 #endif 264 } 265 266 fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk); 267 268 /* 269 * If we need to retry but a fatal signal is pending, handle the 270 * signal first. We do not need to release the mmap_sem because it 271 * would already be released in __lock_page_or_retry in mm/filemap.c. 272 */ 273 if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current)) 274 return 0; 275 276 /* 277 * Major/minor page fault accounting is only done on the initial 278 * attempt. If we go through a retry, it is extremely likely that the 279 * page will be found in page cache at that point. 280 */ 281 282 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr); 283 if (mm_flags & FAULT_FLAG_ALLOW_RETRY) { 284 if (fault & VM_FAULT_MAJOR) { 285 tsk->maj_flt++; 286 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, 287 addr); 288 } else { 289 tsk->min_flt++; 290 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, 291 addr); 292 } 293 if (fault & VM_FAULT_RETRY) { 294 /* 295 * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of 296 * starvation. 297 */ 298 mm_flags &= ~FAULT_FLAG_ALLOW_RETRY; 299 mm_flags |= FAULT_FLAG_TRIED; 300 goto retry; 301 } 302 } 303 304 up_read(&mm->mmap_sem); 305 306 /* 307 * Handle the "normal" case first - VM_FAULT_MAJOR 308 */ 309 if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP | 310 VM_FAULT_BADACCESS)))) 311 return 0; 312 313 /* 314 * If we are in kernel mode at this point, we have no context to 315 * handle this fault with. 316 */ 317 if (!user_mode(regs)) 318 goto no_context; 319 320 if (fault & VM_FAULT_OOM) { 321 /* 322 * We ran out of memory, call the OOM killer, and return to 323 * userspace (which will retry the fault, or kill us if we got 324 * oom-killed). 325 */ 326 pagefault_out_of_memory(); 327 return 0; 328 } 329 330 if (fault & VM_FAULT_SIGBUS) { 331 /* 332 * We had some memory, but were unable to successfully fix up 333 * this page fault. 334 */ 335 sig = SIGBUS; 336 code = BUS_ADRERR; 337 } else { 338 /* 339 * Something tried to access memory that isn't in our memory 340 * map. 341 */ 342 sig = SIGSEGV; 343 code = fault == VM_FAULT_BADACCESS ? 344 SEGV_ACCERR : SEGV_MAPERR; 345 } 346 347 __do_user_fault(tsk, addr, esr, sig, code, regs); 348 return 0; 349 350 no_context: 351 __do_kernel_fault(mm, addr, esr, regs); 352 return 0; 353 } 354 355 /* 356 * First Level Translation Fault Handler 357 * 358 * We enter here because the first level page table doesn't contain a valid 359 * entry for the address. 360 * 361 * If the address is in kernel space (>= TASK_SIZE), then we are probably 362 * faulting in the vmalloc() area. 363 * 364 * If the init_task's first level page tables contains the relevant entry, we 365 * copy the it to this task. If not, we send the process a signal, fixup the 366 * exception, or oops the kernel. 367 * 368 * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt 369 * or a critical region, and should only copy the information from the master 370 * page table, nothing more. 371 */ 372 static int __kprobes do_translation_fault(unsigned long addr, 373 unsigned int esr, 374 struct pt_regs *regs) 375 { 376 if (addr < TASK_SIZE) 377 return do_page_fault(addr, esr, regs); 378 379 do_bad_area(addr, esr, regs); 380 return 0; 381 } 382 383 static int do_alignment_fault(unsigned long addr, unsigned int esr, 384 struct pt_regs *regs) 385 { 386 do_bad_area(addr, esr, regs); 387 return 0; 388 } 389 390 /* 391 * This abort handler always returns "fault". 392 */ 393 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs) 394 { 395 return 1; 396 } 397 398 static struct fault_info { 399 int (*fn)(unsigned long addr, unsigned int esr, struct pt_regs *regs); 400 int sig; 401 int code; 402 const char *name; 403 } fault_info[] = { 404 { do_bad, SIGBUS, 0, "ttbr address size fault" }, 405 { do_bad, SIGBUS, 0, "level 1 address size fault" }, 406 { do_bad, SIGBUS, 0, "level 2 address size fault" }, 407 { do_bad, SIGBUS, 0, "level 3 address size fault" }, 408 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 0 translation fault" }, 409 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 1 translation fault" }, 410 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 2 translation fault" }, 411 { do_page_fault, SIGSEGV, SEGV_MAPERR, "level 3 translation fault" }, 412 { do_bad, SIGBUS, 0, "unknown 8" }, 413 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 access flag fault" }, 414 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 access flag fault" }, 415 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 access flag fault" }, 416 { do_bad, SIGBUS, 0, "unknown 12" }, 417 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 permission fault" }, 418 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 permission fault" }, 419 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 permission fault" }, 420 { do_bad, SIGBUS, 0, "synchronous external abort" }, 421 { do_bad, SIGBUS, 0, "unknown 17" }, 422 { do_bad, SIGBUS, 0, "unknown 18" }, 423 { do_bad, SIGBUS, 0, "unknown 19" }, 424 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" }, 425 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" }, 426 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" }, 427 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" }, 428 { do_bad, SIGBUS, 0, "synchronous parity error" }, 429 { do_bad, SIGBUS, 0, "unknown 25" }, 430 { do_bad, SIGBUS, 0, "unknown 26" }, 431 { do_bad, SIGBUS, 0, "unknown 27" }, 432 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" }, 433 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" }, 434 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" }, 435 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" }, 436 { do_bad, SIGBUS, 0, "unknown 32" }, 437 { do_alignment_fault, SIGBUS, BUS_ADRALN, "alignment fault" }, 438 { do_bad, SIGBUS, 0, "unknown 34" }, 439 { do_bad, SIGBUS, 0, "unknown 35" }, 440 { do_bad, SIGBUS, 0, "unknown 36" }, 441 { do_bad, SIGBUS, 0, "unknown 37" }, 442 { do_bad, SIGBUS, 0, "unknown 38" }, 443 { do_bad, SIGBUS, 0, "unknown 39" }, 444 { do_bad, SIGBUS, 0, "unknown 40" }, 445 { do_bad, SIGBUS, 0, "unknown 41" }, 446 { do_bad, SIGBUS, 0, "unknown 42" }, 447 { do_bad, SIGBUS, 0, "unknown 43" }, 448 { do_bad, SIGBUS, 0, "unknown 44" }, 449 { do_bad, SIGBUS, 0, "unknown 45" }, 450 { do_bad, SIGBUS, 0, "unknown 46" }, 451 { do_bad, SIGBUS, 0, "unknown 47" }, 452 { do_bad, SIGBUS, 0, "TLB conflict abort" }, 453 { do_bad, SIGBUS, 0, "unknown 49" }, 454 { do_bad, SIGBUS, 0, "unknown 50" }, 455 { do_bad, SIGBUS, 0, "unknown 51" }, 456 { do_bad, SIGBUS, 0, "implementation fault (lockdown abort)" }, 457 { do_bad, SIGBUS, 0, "implementation fault (unsupported exclusive)" }, 458 { do_bad, SIGBUS, 0, "unknown 54" }, 459 { do_bad, SIGBUS, 0, "unknown 55" }, 460 { do_bad, SIGBUS, 0, "unknown 56" }, 461 { do_bad, SIGBUS, 0, "unknown 57" }, 462 { do_bad, SIGBUS, 0, "unknown 58" }, 463 { do_bad, SIGBUS, 0, "unknown 59" }, 464 { do_bad, SIGBUS, 0, "unknown 60" }, 465 { do_bad, SIGBUS, 0, "section domain fault" }, 466 { do_bad, SIGBUS, 0, "page domain fault" }, 467 { do_bad, SIGBUS, 0, "unknown 63" }, 468 }; 469 470 static const char *fault_name(unsigned int esr) 471 { 472 const struct fault_info *inf = fault_info + (esr & 63); 473 return inf->name; 474 } 475 476 /* 477 * Dispatch a data abort to the relevant handler. 478 */ 479 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr, 480 struct pt_regs *regs) 481 { 482 const struct fault_info *inf = fault_info + (esr & 63); 483 struct siginfo info; 484 485 if (!inf->fn(addr, esr, regs)) 486 return; 487 488 pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n", 489 inf->name, esr, addr); 490 491 info.si_signo = inf->sig; 492 info.si_errno = 0; 493 info.si_code = inf->code; 494 info.si_addr = (void __user *)addr; 495 arm64_notify_die("", regs, &info, esr); 496 } 497 498 /* 499 * Handle stack alignment exceptions. 500 */ 501 asmlinkage void __exception do_sp_pc_abort(unsigned long addr, 502 unsigned int esr, 503 struct pt_regs *regs) 504 { 505 struct siginfo info; 506 struct task_struct *tsk = current; 507 508 if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS)) 509 pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n", 510 tsk->comm, task_pid_nr(tsk), 511 esr_get_class_string(esr), (void *)regs->pc, 512 (void *)regs->sp); 513 514 info.si_signo = SIGBUS; 515 info.si_errno = 0; 516 info.si_code = BUS_ADRALN; 517 info.si_addr = (void __user *)addr; 518 arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr); 519 } 520 521 int __init early_brk64(unsigned long addr, unsigned int esr, 522 struct pt_regs *regs); 523 524 /* 525 * __refdata because early_brk64 is __init, but the reference to it is 526 * clobbered at arch_initcall time. 527 * See traps.c and debug-monitors.c:debug_traps_init(). 528 */ 529 static struct fault_info __refdata debug_fault_info[] = { 530 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware breakpoint" }, 531 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware single-step" }, 532 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware watchpoint" }, 533 { do_bad, SIGBUS, 0, "unknown 3" }, 534 { do_bad, SIGTRAP, TRAP_BRKPT, "aarch32 BKPT" }, 535 { do_bad, SIGTRAP, 0, "aarch32 vector catch" }, 536 { early_brk64, SIGTRAP, TRAP_BRKPT, "aarch64 BRK" }, 537 { do_bad, SIGBUS, 0, "unknown 7" }, 538 }; 539 540 void __init hook_debug_fault_code(int nr, 541 int (*fn)(unsigned long, unsigned int, struct pt_regs *), 542 int sig, int code, const char *name) 543 { 544 BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info)); 545 546 debug_fault_info[nr].fn = fn; 547 debug_fault_info[nr].sig = sig; 548 debug_fault_info[nr].code = code; 549 debug_fault_info[nr].name = name; 550 } 551 552 asmlinkage int __exception do_debug_exception(unsigned long addr, 553 unsigned int esr, 554 struct pt_regs *regs) 555 { 556 const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr); 557 struct siginfo info; 558 559 if (!inf->fn(addr, esr, regs)) 560 return 1; 561 562 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n", 563 inf->name, esr, addr); 564 565 info.si_signo = inf->sig; 566 info.si_errno = 0; 567 info.si_code = inf->code; 568 info.si_addr = (void __user *)addr; 569 arm64_notify_die("", regs, &info, 0); 570 571 return 0; 572 } 573 574 #ifdef CONFIG_ARM64_PAN 575 void cpu_enable_pan(void *__unused) 576 { 577 config_sctlr_el1(SCTLR_EL1_SPAN, 0); 578 } 579 #endif /* CONFIG_ARM64_PAN */ 580 581 #ifdef CONFIG_ARM64_UAO 582 /* 583 * Kernel threads have fs=KERNEL_DS by default, and don't need to call 584 * set_fs(), devtmpfs in particular relies on this behaviour. 585 * We need to enable the feature at runtime (instead of adding it to 586 * PSR_MODE_EL1h) as the feature may not be implemented by the cpu. 587 */ 588 void cpu_enable_uao(void *__unused) 589 { 590 asm(SET_PSTATE_UAO(1)); 591 } 592 #endif /* CONFIG_ARM64_UAO */ 593