arch/arm64/mm/fault.c

   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/extable.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/signal.h>
  30 #include <linux/sched/debug.h>
  31 #include <linux/highmem.h>
  32 #include <linux/perf_event.h>
  33 #include <linux/preempt.h>
  34 #include <linux/hugetlb.h>
  35
  36 #include <asm/bug.h>
  37 #include <asm/cmpxchg.h>
  38 #include <asm/cpufeature.h>
  39 #include <asm/exception.h>
  40 #include <asm/debug-monitors.h>
  41 #include <asm/esr.h>
  42 #include <asm/sysreg.h>
  43 #include <asm/system_misc.h>
  44 #include <asm/pgtable.h>
  45 #include <asm/tlbflush.h>
  46
  47 #include <acpi/ghes.h>
  48
  49 struct fault_info {
  50         int     (*fn)(unsigned long addr, unsigned int esr,
  51                       struct pt_regs *regs);
  52         int     sig;
  53         int     code;
  54         const char *name;
  55 };
  56
  57 static const struct fault_info fault_info[];
  58
  59 static inline const struct fault_info *esr_to_fault_info(unsigned int esr)
  60 {
  61         return fault_info + (esr & 63);
  62 }
  63
  64 #ifdef CONFIG_KPROBES
  65 static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
  66 {
  67         int ret = 0;
  68
  69         /* kprobe_running() needs smp_processor_id() */
  70         if (!user_mode(regs)) {
  71                 preempt_disable();
  72                 if (kprobe_running() && kprobe_fault_handler(regs, esr))
  73                         ret = 1;
  74                 preempt_enable();
  75         }
  76
  77         return ret;
  78 }
  79 #else
  80 static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
  81 {
  82         return 0;
  83 }
  84 #endif
  85
  86 /*
  87  * Dump out the page tables associated with 'addr' in the currently active mm.
  88  */
  89 void show_pte(unsigned long addr)
  90 {
  91         struct mm_struct *mm;
  92         pgd_t *pgd;
  93
  94         if (addr < TASK_SIZE) {
  95                 /* TTBR0 */
  96                 mm = current->active_mm;
  97                 if (mm == &init_mm) {
  98                         pr_alert("[%016lx] user address but active_mm is swapper\n",
  99                                  addr);
 100                         return;
 101                 }
 102         } else if (addr >= VA_START) {
 103                 /* TTBR1 */
 104                 mm = &init_mm;
 105         } else {
 106                 pr_alert("[%016lx] address between user and kernel address ranges\n",
 107                          addr);
 108                 return;
 109         }
 110
 111         pr_alert("%s pgtable: %luk pages, %u-bit VAs, pgd = %p\n",
 112                  mm == &init_mm ? "swapper" : "user", PAGE_SIZE / SZ_1K,
 113                  VA_BITS, mm->pgd);
 114         pgd = pgd_offset(mm, addr);
 115         pr_alert("[%016lx] *pgd=%016llx", addr, pgd_val(*pgd));
 116
 117         do {
 118                 pud_t *pud;
 119                 pmd_t *pmd;
 120                 pte_t *pte;
 121
 122                 if (pgd_none(*pgd) || pgd_bad(*pgd))
 123                         break;
 124
 125                 pud = pud_offset(pgd, addr);
 126                 pr_cont(", *pud=%016llx", pud_val(*pud));
 127                 if (pud_none(*pud) || pud_bad(*pud))
 128                         break;
 129
 130                 pmd = pmd_offset(pud, addr);
 131                 pr_cont(", *pmd=%016llx", pmd_val(*pmd));
 132                 if (pmd_none(*pmd) || pmd_bad(*pmd))
 133                         break;
 134
 135                 pte = pte_offset_map(pmd, addr);
 136                 pr_cont(", *pte=%016llx", pte_val(*pte));
 137                 pte_unmap(pte);
 138         } while(0);
 139
 140         pr_cont("\n");
 141 }
 142
 143 #ifdef CONFIG_ARM64_HW_AFDBM
 144 /*
 145  * This function sets the access flags (dirty, accessed), as well as write
 146  * permission, and only to a more permissive setting.
 147  *
 148  * It needs to cope with hardware update of the accessed/dirty state by other
 149  * agents in the system and can safely skip the __sync_icache_dcache() call as,
 150  * like set_pte_at(), the PTE is never changed from no-exec to exec here.
 151  *
 152  * Returns whether or not the PTE actually changed.
 153  */
 154 int ptep_set_access_flags(struct vm_area_struct *vma,
 155                           unsigned long address, pte_t *ptep,
 156                           pte_t entry, int dirty)
 157 {
 158         pteval_t old_pteval, pteval;
 159
 160         if (pte_same(*ptep, entry))
 161                 return 0;
 162
 163         /* only preserve the access flags and write permission */
 164         pte_val(entry) &= PTE_RDONLY | PTE_AF | PTE_WRITE | PTE_DIRTY;
 165
 166         /*
 167          * Setting the flags must be done atomically to avoid racing with the
 168          * hardware update of the access/dirty state. The PTE_RDONLY bit must
 169          * be set to the most permissive (lowest value) of *ptep and entry
 170          * (calculated as: a & b == ~(~a | ~b)).
 171          */
 172         pte_val(entry) ^= PTE_RDONLY;
 173         pteval = READ_ONCE(pte_val(*ptep));
 174         do {
 175                 old_pteval = pteval;
 176                 pteval ^= PTE_RDONLY;
 177                 pteval |= pte_val(entry);
 178                 pteval ^= PTE_RDONLY;
 179                 pteval = cmpxchg_relaxed(&pte_val(*ptep), old_pteval, pteval);
 180         } while (pteval != old_pteval);
 181
 182         flush_tlb_fix_spurious_fault(vma, address);
 183         return 1;
 184 }
 185 #endif
 186
 187 static bool is_el1_instruction_abort(unsigned int esr)
 188 {
 189         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR;
 190 }
 191
 192 static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs,
 193                                        unsigned long addr)
 194 {
 195         unsigned int ec       = ESR_ELx_EC(esr);
 196         unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE;
 197
 198         if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR)
 199                 return false;
 200
 201         if (fsc_type == ESR_ELx_FSC_PERM)
 202                 return true;
 203
 204         if (addr < USER_DS && system_uses_ttbr0_pan())
 205                 return fsc_type == ESR_ELx_FSC_FAULT &&
 206                         (regs->pstate & PSR_PAN_BIT);
 207
 208         return false;
 209 }
 210
 211 /*
 212  * The kernel tried to access some page that wasn't present.
 213  */
 214 static void __do_kernel_fault(unsigned long addr, unsigned int esr,
 215                               struct pt_regs *regs)
 216 {
 217         const char *msg;
 218
 219         /*
 220          * Are we prepared to handle this kernel fault?
 221          * We are almost certainly not prepared to handle instruction faults.
 222          */
 223         if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
 224                 return;
 225
 226         /*
 227          * No handler, we'll have to terminate things with extreme prejudice.
 228          */
 229         bust_spinlocks(1);
 230
 231         if (is_permission_fault(esr, regs, addr)) {
 232                 if (esr & ESR_ELx_WNR)
 233                         msg = "write to read-only memory";
 234                 else
 235                         msg = "read from unreadable memory";
 236         } else if (addr < PAGE_SIZE) {
 237                 msg = "NULL pointer dereference";
 238         } else {
 239                 msg = "paging request";
 240         }
 241
 242         pr_alert("Unable to handle kernel %s at virtual address %08lx\n", msg,
 243                  addr);
 244
 245         show_pte(addr);
 246         die("Oops", regs, esr);
 247         bust_spinlocks(0);
 248         do_exit(SIGKILL);
 249 }
 250
 251 /*
 252  * Something tried to access memory that isn't in our memory map. User mode
 253  * accesses just cause a SIGSEGV
 254  */
 255 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
 256                             unsigned int esr, unsigned int sig, int code,
 257                             struct pt_regs *regs, int fault)
 258 {
 259         struct siginfo si;
 260         const struct fault_info *inf;
 261         unsigned int lsb = 0;
 262
 263         if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
 264                 inf = esr_to_fault_info(esr);
 265                 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x",
 266                         tsk->comm, task_pid_nr(tsk), inf->name, sig,
 267                         addr, esr);
 268                 print_vma_addr(KERN_CONT ", in ", regs->pc);
 269                 pr_cont("\n");
 270                 __show_regs(regs);
 271         }
 272
 273         tsk->thread.fault_address = addr;
 274         tsk->thread.fault_code = esr;
 275         si.si_signo = sig;
 276         si.si_errno = 0;
 277         si.si_code = code;
 278         si.si_addr = (void __user *)addr;
 279         /*
 280          * Either small page or large page may be poisoned.
 281          * In other words, VM_FAULT_HWPOISON_LARGE and
 282          * VM_FAULT_HWPOISON are mutually exclusive.
 283          */
 284         if (fault & VM_FAULT_HWPOISON_LARGE)
 285                 lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault));
 286         else if (fault & VM_FAULT_HWPOISON)
 287                 lsb = PAGE_SHIFT;
 288         si.si_addr_lsb = lsb;
 289
 290         force_sig_info(sig, &si, tsk);
 291 }
 292
 293 static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 294 {
 295         struct task_struct *tsk = current;
 296         const struct fault_info *inf;
 297
 298         /*
 299          * If we are in kernel mode at this point, we have no context to
 300          * handle this fault with.
 301          */
 302         if (user_mode(regs)) {
 303                 inf = esr_to_fault_info(esr);
 304                 __do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs, 0);
 305         } else
 306                 __do_kernel_fault(addr, esr, regs);
 307 }
 308
 309 #define VM_FAULT_BADMAP         0x010000
 310 #define VM_FAULT_BADACCESS      0x020000
 311
 312 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
 313                            unsigned int mm_flags, unsigned long vm_flags,
 314                            struct task_struct *tsk)
 315 {
 316         struct vm_area_struct *vma;
 317         int fault;
 318
 319         vma = find_vma(mm, addr);
 320         fault = VM_FAULT_BADMAP;
 321         if (unlikely(!vma))
 322                 goto out;
 323         if (unlikely(vma->vm_start > addr))
 324                 goto check_stack;
 325
 326         /*
 327          * Ok, we have a good vm_area for this memory access, so we can handle
 328          * it.
 329          */
 330 good_area:
 331         /*
 332          * Check that the permissions on the VMA allow for the fault which
 333          * occurred.
 334          */
 335         if (!(vma->vm_flags & vm_flags)) {
 336                 fault = VM_FAULT_BADACCESS;
 337                 goto out;
 338         }
 339
 340         return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags);
 341
 342 check_stack:
 343         if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
 344                 goto good_area;
 345 out:
 346         return fault;
 347 }
 348
 349 static bool is_el0_instruction_abort(unsigned int esr)
 350 {
 351         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW;
 352 }
 353
 354 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
 355                                    struct pt_regs *regs)
 356 {
 357         struct task_struct *tsk;
 358         struct mm_struct *mm;
 359         int fault, sig, code, major = 0;
 360         unsigned long vm_flags = VM_READ | VM_WRITE;
 361         unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
 362
 363         if (notify_page_fault(regs, esr))
 364                 return 0;
 365
 366         tsk = current;
 367         mm  = tsk->mm;
 368
 369         /*
 370          * If we're in an interrupt or have no user context, we must not take
 371          * the fault.
 372          */
 373         if (faulthandler_disabled() || !mm)
 374                 goto no_context;
 375
 376         if (user_mode(regs))
 377                 mm_flags |= FAULT_FLAG_USER;
 378
 379         if (is_el0_instruction_abort(esr)) {
 380                 vm_flags = VM_EXEC;
 381         } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
 382                 vm_flags = VM_WRITE;
 383                 mm_flags |= FAULT_FLAG_WRITE;
 384         }
 385
 386         if (addr < USER_DS && is_permission_fault(esr, regs, addr)) {
 387                 /* regs->orig_addr_limit may be 0 if we entered from EL0 */
 388                 if (regs->orig_addr_limit == KERNEL_DS)
 389                         die("Accessing user space memory with fs=KERNEL_DS", regs, esr);
 390
 391                 if (is_el1_instruction_abort(esr))
 392                         die("Attempting to execute userspace memory", regs, esr);
 393
 394                 if (!search_exception_tables(regs->pc))
 395                         die("Accessing user space memory outside uaccess.h routines", regs, esr);
 396         }
 397
 398         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
 399
 400         /*
 401          * As per x86, we may deadlock here. However, since the kernel only
 402          * validly references user space from well defined areas of the code,
 403          * we can bug out early if this is from code which shouldn't.
 404          */
 405         if (!down_read_trylock(&mm->mmap_sem)) {
 406                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 407                         goto no_context;
 408 retry:
 409                 down_read(&mm->mmap_sem);
 410         } else {
 411                 /*
 412                  * The above down_read_trylock() might have succeeded in which
 413                  * case, we'll have missed the might_sleep() from down_read().
 414                  */
 415                 might_sleep();
 416 #ifdef CONFIG_DEBUG_VM
 417                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 418                         goto no_context;
 419 #endif
 420         }
 421
 422         fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
 423         major |= fault & VM_FAULT_MAJOR;
 424
 425         if (fault & VM_FAULT_RETRY) {
 426                 /*
 427                  * If we need to retry but a fatal signal is pending,
 428                  * handle the signal first. We do not need to release
 429                  * the mmap_sem because it would already be released
 430                  * in __lock_page_or_retry in mm/filemap.c.
 431                  */
 432                 if (fatal_signal_pending(current)) {
 433                         if (!user_mode(regs))
 434                                 goto no_context;
 435                         return 0;
 436                 }
 437
 438                 /*
 439                  * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
 440                  * starvation.
 441                  */
 442                 if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
 443                         mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
 444                         mm_flags |= FAULT_FLAG_TRIED;
 445                         goto retry;
 446                 }
 447         }
 448         up_read(&mm->mmap_sem);
 449
 450         /*
 451          * Handle the "normal" (no error) case first.
 452          */
 453         if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
 454                               VM_FAULT_BADACCESS)))) {
 455                 /*
 456                  * Major/minor page fault accounting is only done
 457                  * once. If we go through a retry, it is extremely
 458                  * likely that the page will be found in page cache at
 459                  * that point.
 460                  */
 461                 if (major) {
 462                         tsk->maj_flt++;
 463                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
 464                                       addr);
 465                 } else {
 466                         tsk->min_flt++;
 467                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
 468                                       addr);
 469                 }
 470
 471                 return 0;
 472         }
 473
 474         /*
 475          * If we are in kernel mode at this point, we have no context to
 476          * handle this fault with.
 477          */
 478         if (!user_mode(regs))
 479                 goto no_context;
 480
 481         if (fault & VM_FAULT_OOM) {
 482                 /*
 483                  * We ran out of memory, call the OOM killer, and return to
 484                  * userspace (which will retry the fault, or kill us if we got
 485                  * oom-killed).
 486                  */
 487                 pagefault_out_of_memory();
 488                 return 0;
 489         }
 490
 491         if (fault & VM_FAULT_SIGBUS) {
 492                 /*
 493                  * We had some memory, but were unable to successfully fix up
 494                  * this page fault.
 495                  */
 496                 sig = SIGBUS;
 497                 code = BUS_ADRERR;
 498         } else if (fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) {
 499                 sig = SIGBUS;
 500                 code = BUS_MCEERR_AR;
 501         } else {
 502                 /*
 503                  * Something tried to access memory that isn't in our memory
 504                  * map.
 505                  */
 506                 sig = SIGSEGV;
 507                 code = fault == VM_FAULT_BADACCESS ?
 508                         SEGV_ACCERR : SEGV_MAPERR;
 509         }
 510
 511         __do_user_fault(tsk, addr, esr, sig, code, regs, fault);
 512         return 0;
 513
 514 no_context:
 515         __do_kernel_fault(addr, esr, regs);
 516         return 0;
 517 }
 518
 519 /*
 520  * First Level Translation Fault Handler
 521  *
 522  * We enter here because the first level page table doesn't contain a valid
 523  * entry for the address.
 524  *
 525  * If the address is in kernel space (>= TASK_SIZE), then we are probably
 526  * faulting in the vmalloc() area.
 527  *
 528  * If the init_task's first level page tables contains the relevant entry, we
 529  * copy the it to this task.  If not, we send the process a signal, fixup the
 530  * exception, or oops the kernel.
 531  *
 532  * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
 533  * or a critical region, and should only copy the information from the master
 534  * page table, nothing more.
 535  */
 536 static int __kprobes do_translation_fault(unsigned long addr,
 537                                           unsigned int esr,
 538                                           struct pt_regs *regs)
 539 {
 540         if (addr < TASK_SIZE)
 541                 return do_page_fault(addr, esr, regs);
 542
 543         do_bad_area(addr, esr, regs);
 544         return 0;
 545 }
 546
 547 static int do_alignment_fault(unsigned long addr, unsigned int esr,
 548                               struct pt_regs *regs)
 549 {
 550         do_bad_area(addr, esr, regs);
 551         return 0;
 552 }
 553
 554 /*
 555  * This abort handler always returns "fault".
 556  */
 557 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 558 {
 559         return 1;
 560 }
 561
 562 /*
 563  * This abort handler deals with Synchronous External Abort.
 564  * It calls notifiers, and then returns "fault".
 565  */
 566 static int do_sea(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 567 {
 568         struct siginfo info;
 569         const struct fault_info *inf;
 570         int ret = 0;
 571
 572         inf = esr_to_fault_info(esr);
 573         pr_err("Synchronous External Abort: %s (0x%08x) at 0x%016lx\n",
 574                 inf->name, esr, addr);
 575
 576         /*
 577          * Synchronous aborts may interrupt code which had interrupts masked.
 578          * Before calling out into the wider kernel tell the interested
 579          * subsystems.
 580          */
 581         if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) {
 582                 if (interrupts_enabled(regs))
 583                         nmi_enter();
 584
 585                 ret = ghes_notify_sea();
 586
 587                 if (interrupts_enabled(regs))
 588                         nmi_exit();
 589         }
 590
 591         info.si_signo = SIGBUS;
 592         info.si_errno = 0;
 593         info.si_code  = 0;
 594         if (esr & ESR_ELx_FnV)
 595                 info.si_addr = NULL;
 596         else
 597                 info.si_addr  = (void __user *)addr;
 598         arm64_notify_die("", regs, &info, esr);
 599
 600         return ret;
 601 }
 602
 603 static const struct fault_info fault_info[] = {
 604         { do_bad,               SIGBUS,  0,             "ttbr address size fault"       },
 605         { do_bad,               SIGBUS,  0,             "level 1 address size fault"    },
 606         { do_bad,               SIGBUS,  0,             "level 2 address size fault"    },
 607         { do_bad,               SIGBUS,  0,             "level 3 address size fault"    },
 608         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 0 translation fault"     },
 609         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 1 translation fault"     },
 610         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 2 translation fault"     },
 611         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 3 translation fault"     },
 612         { do_bad,               SIGBUS,  0,             "unknown 8"                     },
 613         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 access flag fault"     },
 614         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 access flag fault"     },
 615         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 access flag fault"     },
 616         { do_bad,               SIGBUS,  0,             "unknown 12"                    },
 617         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 permission fault"      },
 618         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 permission fault"      },
 619         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 permission fault"      },
 620         { do_sea,               SIGBUS,  0,             "synchronous external abort"    },
 621         { do_bad,               SIGBUS,  0,             "unknown 17"                    },
 622         { do_bad,               SIGBUS,  0,             "unknown 18"                    },
 623         { do_bad,               SIGBUS,  0,             "unknown 19"                    },
 624         { do_sea,               SIGBUS,  0,             "level 0 (translation table walk)"      },
 625         { do_sea,               SIGBUS,  0,             "level 1 (translation table walk)"      },
 626         { do_sea,               SIGBUS,  0,             "level 2 (translation table walk)"      },
 627         { do_sea,               SIGBUS,  0,             "level 3 (translation table walk)"      },
 628         { do_sea,               SIGBUS,  0,             "synchronous parity or ECC error" },
 629         { do_bad,               SIGBUS,  0,             "unknown 25"                    },
 630         { do_bad,               SIGBUS,  0,             "unknown 26"                    },
 631         { do_bad,               SIGBUS,  0,             "unknown 27"                    },
 632         { do_sea,               SIGBUS,  0,             "level 0 synchronous parity error (translation table walk)"     },
 633         { do_sea,               SIGBUS,  0,             "level 1 synchronous parity error (translation table walk)"     },
 634         { do_sea,               SIGBUS,  0,             "level 2 synchronous parity error (translation table walk)"     },
 635         { do_sea,               SIGBUS,  0,             "level 3 synchronous parity error (translation table walk)"     },
 636         { do_bad,               SIGBUS,  0,             "unknown 32"                    },
 637         { do_alignment_fault,   SIGBUS,  BUS_ADRALN,    "alignment fault"               },
 638         { do_bad,               SIGBUS,  0,             "unknown 34"                    },
 639         { do_bad,               SIGBUS,  0,             "unknown 35"                    },
 640         { do_bad,               SIGBUS,  0,             "unknown 36"                    },
 641         { do_bad,               SIGBUS,  0,             "unknown 37"                    },
 642         { do_bad,               SIGBUS,  0,             "unknown 38"                    },
 643         { do_bad,               SIGBUS,  0,             "unknown 39"                    },
 644         { do_bad,               SIGBUS,  0,             "unknown 40"                    },
 645         { do_bad,               SIGBUS,  0,             "unknown 41"                    },
 646         { do_bad,               SIGBUS,  0,             "unknown 42"                    },
 647         { do_bad,               SIGBUS,  0,             "unknown 43"                    },
 648         { do_bad,               SIGBUS,  0,             "unknown 44"                    },
 649         { do_bad,               SIGBUS,  0,             "unknown 45"                    },
 650         { do_bad,               SIGBUS,  0,             "unknown 46"                    },
 651         { do_bad,               SIGBUS,  0,             "unknown 47"                    },
 652         { do_bad,               SIGBUS,  0,             "TLB conflict abort"            },
 653         { do_bad,               SIGBUS,  0,             "unknown 49"                    },
 654         { do_bad,               SIGBUS,  0,             "unknown 50"                    },
 655         { do_bad,               SIGBUS,  0,             "unknown 51"                    },
 656         { do_bad,               SIGBUS,  0,             "implementation fault (lockdown abort)" },
 657         { do_bad,               SIGBUS,  0,             "implementation fault (unsupported exclusive)" },
 658         { do_bad,               SIGBUS,  0,             "unknown 54"                    },
 659         { do_bad,               SIGBUS,  0,             "unknown 55"                    },
 660         { do_bad,               SIGBUS,  0,             "unknown 56"                    },
 661         { do_bad,               SIGBUS,  0,             "unknown 57"                    },
 662         { do_bad,               SIGBUS,  0,             "unknown 58"                    },
 663         { do_bad,               SIGBUS,  0,             "unknown 59"                    },
 664         { do_bad,               SIGBUS,  0,             "unknown 60"                    },
 665         { do_bad,               SIGBUS,  0,             "section domain fault"          },
 666         { do_bad,               SIGBUS,  0,             "page domain fault"             },
 667         { do_bad,               SIGBUS,  0,             "unknown 63"                    },
 668 };
 669
 670 /*
 671  * Handle Synchronous External Aborts that occur in a guest kernel.
 672  *
 673  * The return value will be zero if the SEA was successfully handled
 674  * and non-zero if there was an error processing the error or there was
 675  * no error to process.
 676  */
 677 int handle_guest_sea(phys_addr_t addr, unsigned int esr)
 678 {
 679         int ret = -ENOENT;
 680
 681         if (IS_ENABLED(CONFIG_ACPI_APEI_SEA))
 682                 ret = ghes_notify_sea();
 683
 684         return ret;
 685 }
 686
 687 /*
 688  * Dispatch a data abort to the relevant handler.
 689  */
 690 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
 691                                          struct pt_regs *regs)
 692 {
 693         const struct fault_info *inf = esr_to_fault_info(esr);
 694         struct siginfo info;
 695
 696         if (!inf->fn(addr, esr, regs))
 697                 return;
 698
 699         pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
 700                  inf->name, esr, addr);
 701
 702         info.si_signo = inf->sig;
 703         info.si_errno = 0;
 704         info.si_code  = inf->code;
 705         info.si_addr  = (void __user *)addr;
 706         arm64_notify_die("", regs, &info, esr);
 707 }
 708
 709 /*
 710  * Handle stack alignment exceptions.
 711  */
 712 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
 713                                            unsigned int esr,
 714                                            struct pt_regs *regs)
 715 {
 716         struct siginfo info;
 717         struct task_struct *tsk = current;
 718
 719         if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
 720                 pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
 721                                     tsk->comm, task_pid_nr(tsk),
 722                                     esr_get_class_string(esr), (void *)regs->pc,
 723                                     (void *)regs->sp);
 724
 725         info.si_signo = SIGBUS;
 726         info.si_errno = 0;
 727         info.si_code  = BUS_ADRALN;
 728         info.si_addr  = (void __user *)addr;
 729         arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
 730 }
 731
 732 int __init early_brk64(unsigned long addr, unsigned int esr,
 733                        struct pt_regs *regs);
 734
 735 /*
 736  * __refdata because early_brk64 is __init, but the reference to it is
 737  * clobbered at arch_initcall time.
 738  * See traps.c and debug-monitors.c:debug_traps_init().
 739  */
 740 static struct fault_info __refdata debug_fault_info[] = {
 741         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware breakpoint"   },
 742         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware single-step"  },
 743         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware watchpoint"   },
 744         { do_bad,       SIGBUS,         0,              "unknown 3"             },
 745         { do_bad,       SIGTRAP,        TRAP_BRKPT,     "aarch32 BKPT"          },
 746         { do_bad,       SIGTRAP,        0,              "aarch32 vector catch"  },
 747         { early_brk64,  SIGTRAP,        TRAP_BRKPT,     "aarch64 BRK"           },
 748         { do_bad,       SIGBUS,         0,              "unknown 7"             },
 749 };
 750
 751 void __init hook_debug_fault_code(int nr,
 752                                   int (*fn)(unsigned long, unsigned int, struct pt_regs *),
 753                                   int sig, int code, const char *name)
 754 {
 755         BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
 756
 757         debug_fault_info[nr].fn         = fn;
 758         debug_fault_info[nr].sig        = sig;
 759         debug_fault_info[nr].code       = code;
 760         debug_fault_info[nr].name       = name;
 761 }
 762
 763 asmlinkage int __exception do_debug_exception(unsigned long addr,
 764                                               unsigned int esr,
 765                                               struct pt_regs *regs)
 766 {
 767         const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
 768         struct siginfo info;
 769         int rv;
 770
 771         /*
 772          * Tell lockdep we disabled irqs in entry.S. Do nothing if they were
 773          * already disabled to preserve the last enabled/disabled addresses.
 774          */
 775         if (interrupts_enabled(regs))
 776                 trace_hardirqs_off();
 777
 778         if (!inf->fn(addr, esr, regs)) {
 779                 rv = 1;
 780         } else {
 781                 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
 782                          inf->name, esr, addr);
 783
 784                 info.si_signo = inf->sig;
 785                 info.si_errno = 0;
 786                 info.si_code  = inf->code;
 787                 info.si_addr  = (void __user *)addr;
 788                 arm64_notify_die("", regs, &info, 0);
 789                 rv = 0;
 790         }
 791
 792         if (interrupts_enabled(regs))
 793                 trace_hardirqs_on();
 794
 795         return rv;
 796 }
 797 NOKPROBE_SYMBOL(do_debug_exception);
 798
 799 #ifdef CONFIG_ARM64_PAN
 800 int cpu_enable_pan(void *__unused)
 801 {
 802         /*
 803          * We modify PSTATE. This won't work from irq context as the PSTATE
 804          * is discarded once we return from the exception.
 805          */
 806         WARN_ON_ONCE(in_interrupt());
 807
 808         config_sctlr_el1(SCTLR_EL1_SPAN, 0);
 809         asm(SET_PSTATE_PAN(1));
 810         return 0;
 811 }
 812 #endif /* CONFIG_ARM64_PAN */