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_AF | PTE_WRITE | PTE_DIRTY;
 165
 166         /* set PTE_RDONLY if actual read-only or clean PTE */
 167         if (!pte_write(entry) || !pte_sw_dirty(entry))
 168                 entry = pte_set_rdonly(entry);
 169
 170         /*
 171          * Setting the flags must be done atomically to avoid racing with the
 172          * hardware update of the access/dirty state. The PTE_RDONLY bit must
 173          * be set to the most permissive (lowest value) of *ptep and entry
 174          * (calculated as: a & b == ~(~a | ~b)).
 175          */
 176         pte_val(entry) ^= PTE_RDONLY;
 177         pteval = READ_ONCE(pte_val(*ptep));
 178         do {
 179                 old_pteval = pteval;
 180                 pteval ^= PTE_RDONLY;
 181                 pteval |= pte_val(entry);
 182                 pteval ^= PTE_RDONLY;
 183                 pteval = cmpxchg_relaxed(&pte_val(*ptep), old_pteval, pteval);
 184         } while (pteval != old_pteval);
 185
 186         flush_tlb_fix_spurious_fault(vma, address);
 187         return 1;
 188 }
 189 #endif
 190
 191 static bool is_el1_instruction_abort(unsigned int esr)
 192 {
 193         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR;
 194 }
 195
 196 static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs,
 197                                        unsigned long addr)
 198 {
 199         unsigned int ec       = ESR_ELx_EC(esr);
 200         unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE;
 201
 202         if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR)
 203                 return false;
 204
 205         if (fsc_type == ESR_ELx_FSC_PERM)
 206                 return true;
 207
 208         if (addr < USER_DS && system_uses_ttbr0_pan())
 209                 return fsc_type == ESR_ELx_FSC_FAULT &&
 210                         (regs->pstate & PSR_PAN_BIT);
 211
 212         return false;
 213 }
 214
 215 /*
 216  * The kernel tried to access some page that wasn't present.
 217  */
 218 static void __do_kernel_fault(unsigned long addr, unsigned int esr,
 219                               struct pt_regs *regs)
 220 {
 221         const char *msg;
 222
 223         /*
 224          * Are we prepared to handle this kernel fault?
 225          * We are almost certainly not prepared to handle instruction faults.
 226          */
 227         if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
 228                 return;
 229
 230         /*
 231          * No handler, we'll have to terminate things with extreme prejudice.
 232          */
 233         bust_spinlocks(1);
 234
 235         if (is_permission_fault(esr, regs, addr)) {
 236                 if (esr & ESR_ELx_WNR)
 237                         msg = "write to read-only memory";
 238                 else
 239                         msg = "read from unreadable memory";
 240         } else if (addr < PAGE_SIZE) {
 241                 msg = "NULL pointer dereference";
 242         } else {
 243                 msg = "paging request";
 244         }
 245
 246         pr_alert("Unable to handle kernel %s at virtual address %08lx\n", msg,
 247                  addr);
 248
 249         show_pte(addr);
 250         die("Oops", regs, esr);
 251         bust_spinlocks(0);
 252         do_exit(SIGKILL);
 253 }
 254
 255 /*
 256  * Something tried to access memory that isn't in our memory map. User mode
 257  * accesses just cause a SIGSEGV
 258  */
 259 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
 260                             unsigned int esr, unsigned int sig, int code,
 261                             struct pt_regs *regs, int fault)
 262 {
 263         struct siginfo si;
 264         const struct fault_info *inf;
 265         unsigned int lsb = 0;
 266
 267         if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
 268                 inf = esr_to_fault_info(esr);
 269                 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x",
 270                         tsk->comm, task_pid_nr(tsk), inf->name, sig,
 271                         addr, esr);
 272                 print_vma_addr(KERN_CONT ", in ", regs->pc);
 273                 pr_cont("\n");
 274                 __show_regs(regs);
 275         }
 276
 277         tsk->thread.fault_address = addr;
 278         tsk->thread.fault_code = esr;
 279         si.si_signo = sig;
 280         si.si_errno = 0;
 281         si.si_code = code;
 282         si.si_addr = (void __user *)addr;
 283         /*
 284          * Either small page or large page may be poisoned.
 285          * In other words, VM_FAULT_HWPOISON_LARGE and
 286          * VM_FAULT_HWPOISON are mutually exclusive.
 287          */
 288         if (fault & VM_FAULT_HWPOISON_LARGE)
 289                 lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault));
 290         else if (fault & VM_FAULT_HWPOISON)
 291                 lsb = PAGE_SHIFT;
 292         si.si_addr_lsb = lsb;
 293
 294         force_sig_info(sig, &si, tsk);
 295 }
 296
 297 static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 298 {
 299         struct task_struct *tsk = current;
 300         const struct fault_info *inf;
 301
 302         /*
 303          * If we are in kernel mode at this point, we have no context to
 304          * handle this fault with.
 305          */
 306         if (user_mode(regs)) {
 307                 inf = esr_to_fault_info(esr);
 308                 __do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs, 0);
 309         } else
 310                 __do_kernel_fault(addr, esr, regs);
 311 }
 312
 313 #define VM_FAULT_BADMAP         0x010000
 314 #define VM_FAULT_BADACCESS      0x020000
 315
 316 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
 317                            unsigned int mm_flags, unsigned long vm_flags,
 318                            struct task_struct *tsk)
 319 {
 320         struct vm_area_struct *vma;
 321         int fault;
 322
 323         vma = find_vma(mm, addr);
 324         fault = VM_FAULT_BADMAP;
 325         if (unlikely(!vma))
 326                 goto out;
 327         if (unlikely(vma->vm_start > addr))
 328                 goto check_stack;
 329
 330         /*
 331          * Ok, we have a good vm_area for this memory access, so we can handle
 332          * it.
 333          */
 334 good_area:
 335         /*
 336          * Check that the permissions on the VMA allow for the fault which
 337          * occurred.
 338          */
 339         if (!(vma->vm_flags & vm_flags)) {
 340                 fault = VM_FAULT_BADACCESS;
 341                 goto out;
 342         }
 343
 344         return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags);
 345
 346 check_stack:
 347         if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
 348                 goto good_area;
 349 out:
 350         return fault;
 351 }
 352
 353 static bool is_el0_instruction_abort(unsigned int esr)
 354 {
 355         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW;
 356 }
 357
 358 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
 359                                    struct pt_regs *regs)
 360 {
 361         struct task_struct *tsk;
 362         struct mm_struct *mm;
 363         int fault, sig, code, major = 0;
 364         unsigned long vm_flags = VM_READ | VM_WRITE;
 365         unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
 366
 367         if (notify_page_fault(regs, esr))
 368                 return 0;
 369
 370         tsk = current;
 371         mm  = tsk->mm;
 372
 373         /*
 374          * If we're in an interrupt or have no user context, we must not take
 375          * the fault.
 376          */
 377         if (faulthandler_disabled() || !mm)
 378                 goto no_context;
 379
 380         if (user_mode(regs))
 381                 mm_flags |= FAULT_FLAG_USER;
 382
 383         if (is_el0_instruction_abort(esr)) {
 384                 vm_flags = VM_EXEC;
 385         } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
 386                 vm_flags = VM_WRITE;
 387                 mm_flags |= FAULT_FLAG_WRITE;
 388         }
 389
 390         if (addr < USER_DS && is_permission_fault(esr, regs, addr)) {
 391                 /* regs->orig_addr_limit may be 0 if we entered from EL0 */
 392                 if (regs->orig_addr_limit == KERNEL_DS)
 393                         die("Accessing user space memory with fs=KERNEL_DS", regs, esr);
 394
 395                 if (is_el1_instruction_abort(esr))
 396                         die("Attempting to execute userspace memory", regs, esr);
 397
 398                 if (!search_exception_tables(regs->pc))
 399                         die("Accessing user space memory outside uaccess.h routines", regs, esr);
 400         }
 401
 402         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
 403
 404         /*
 405          * As per x86, we may deadlock here. However, since the kernel only
 406          * validly references user space from well defined areas of the code,
 407          * we can bug out early if this is from code which shouldn't.
 408          */
 409         if (!down_read_trylock(&mm->mmap_sem)) {
 410                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 411                         goto no_context;
 412 retry:
 413                 down_read(&mm->mmap_sem);
 414         } else {
 415                 /*
 416                  * The above down_read_trylock() might have succeeded in which
 417                  * case, we'll have missed the might_sleep() from down_read().
 418                  */
 419                 might_sleep();
 420 #ifdef CONFIG_DEBUG_VM
 421                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 422                         goto no_context;
 423 #endif
 424         }
 425
 426         fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
 427         major |= fault & VM_FAULT_MAJOR;
 428
 429         if (fault & VM_FAULT_RETRY) {
 430                 /*
 431                  * If we need to retry but a fatal signal is pending,
 432                  * handle the signal first. We do not need to release
 433                  * the mmap_sem because it would already be released
 434                  * in __lock_page_or_retry in mm/filemap.c.
 435                  */
 436                 if (fatal_signal_pending(current)) {
 437                         if (!user_mode(regs))
 438                                 goto no_context;
 439                         return 0;
 440                 }
 441
 442                 /*
 443                  * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
 444                  * starvation.
 445                  */
 446                 if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
 447                         mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
 448                         mm_flags |= FAULT_FLAG_TRIED;
 449                         goto retry;
 450                 }
 451         }
 452         up_read(&mm->mmap_sem);
 453
 454         /*
 455          * Handle the "normal" (no error) case first.
 456          */
 457         if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
 458                               VM_FAULT_BADACCESS)))) {
 459                 /*
 460                  * Major/minor page fault accounting is only done
 461                  * once. If we go through a retry, it is extremely
 462                  * likely that the page will be found in page cache at
 463                  * that point.
 464                  */
 465                 if (major) {
 466                         tsk->maj_flt++;
 467                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
 468                                       addr);
 469                 } else {
 470                         tsk->min_flt++;
 471                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
 472                                       addr);
 473                 }
 474
 475                 return 0;
 476         }
 477
 478         /*
 479          * If we are in kernel mode at this point, we have no context to
 480          * handle this fault with.
 481          */
 482         if (!user_mode(regs))
 483                 goto no_context;
 484
 485         if (fault & VM_FAULT_OOM) {
 486                 /*
 487                  * We ran out of memory, call the OOM killer, and return to
 488                  * userspace (which will retry the fault, or kill us if we got
 489                  * oom-killed).
 490                  */
 491                 pagefault_out_of_memory();
 492                 return 0;
 493         }
 494
 495         if (fault & VM_FAULT_SIGBUS) {
 496                 /*
 497                  * We had some memory, but were unable to successfully fix up
 498                  * this page fault.
 499                  */
 500                 sig = SIGBUS;
 501                 code = BUS_ADRERR;
 502         } else if (fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) {
 503                 sig = SIGBUS;
 504                 code = BUS_MCEERR_AR;
 505         } else {
 506                 /*
 507                  * Something tried to access memory that isn't in our memory
 508                  * map.
 509                  */
 510                 sig = SIGSEGV;
 511                 code = fault == VM_FAULT_BADACCESS ?
 512                         SEGV_ACCERR : SEGV_MAPERR;
 513         }
 514
 515         __do_user_fault(tsk, addr, esr, sig, code, regs, fault);
 516         return 0;
 517
 518 no_context:
 519         __do_kernel_fault(addr, esr, regs);
 520         return 0;
 521 }
 522
 523 /*
 524  * First Level Translation Fault Handler
 525  *
 526  * We enter here because the first level page table doesn't contain a valid
 527  * entry for the address.
 528  *
 529  * If the address is in kernel space (>= TASK_SIZE), then we are probably
 530  * faulting in the vmalloc() area.
 531  *
 532  * If the init_task's first level page tables contains the relevant entry, we
 533  * copy the it to this task.  If not, we send the process a signal, fixup the
 534  * exception, or oops the kernel.
 535  *
 536  * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
 537  * or a critical region, and should only copy the information from the master
 538  * page table, nothing more.
 539  */
 540 static int __kprobes do_translation_fault(unsigned long addr,
 541                                           unsigned int esr,
 542                                           struct pt_regs *regs)
 543 {
 544         if (addr < TASK_SIZE)
 545                 return do_page_fault(addr, esr, regs);
 546
 547         do_bad_area(addr, esr, regs);
 548         return 0;
 549 }
 550
 551 static int do_alignment_fault(unsigned long addr, unsigned int esr,
 552                               struct pt_regs *regs)
 553 {
 554         do_bad_area(addr, esr, regs);
 555         return 0;
 556 }
 557
 558 /*
 559  * This abort handler always returns "fault".
 560  */
 561 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 562 {
 563         return 1;
 564 }
 565
 566 /*
 567  * This abort handler deals with Synchronous External Abort.
 568  * It calls notifiers, and then returns "fault".
 569  */
 570 static int do_sea(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 571 {
 572         struct siginfo info;
 573         const struct fault_info *inf;
 574         int ret = 0;
 575
 576         inf = esr_to_fault_info(esr);
 577         pr_err("Synchronous External Abort: %s (0x%08x) at 0x%016lx\n",
 578                 inf->name, esr, addr);
 579
 580         /*
 581          * Synchronous aborts may interrupt code which had interrupts masked.
 582          * Before calling out into the wider kernel tell the interested
 583          * subsystems.
 584          */
 585         if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) {
 586                 if (interrupts_enabled(regs))
 587                         nmi_enter();
 588
 589                 ret = ghes_notify_sea();
 590
 591                 if (interrupts_enabled(regs))
 592                         nmi_exit();
 593         }
 594
 595         info.si_signo = SIGBUS;
 596         info.si_errno = 0;
 597         info.si_code  = 0;
 598         if (esr & ESR_ELx_FnV)
 599                 info.si_addr = NULL;
 600         else
 601                 info.si_addr  = (void __user *)addr;
 602         arm64_notify_die("", regs, &info, esr);
 603
 604         return ret;
 605 }
 606
 607 static const struct fault_info fault_info[] = {
 608         { do_bad,               SIGBUS,  0,             "ttbr address size fault"       },
 609         { do_bad,               SIGBUS,  0,             "level 1 address size fault"    },
 610         { do_bad,               SIGBUS,  0,             "level 2 address size fault"    },
 611         { do_bad,               SIGBUS,  0,             "level 3 address size fault"    },
 612         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 0 translation fault"     },
 613         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 1 translation fault"     },
 614         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 2 translation fault"     },
 615         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 3 translation fault"     },
 616         { do_bad,               SIGBUS,  0,             "unknown 8"                     },
 617         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 access flag fault"     },
 618         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 access flag fault"     },
 619         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 access flag fault"     },
 620         { do_bad,               SIGBUS,  0,             "unknown 12"                    },
 621         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 permission fault"      },
 622         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 permission fault"      },
 623         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 permission fault"      },
 624         { do_sea,               SIGBUS,  0,             "synchronous external abort"    },
 625         { do_bad,               SIGBUS,  0,             "unknown 17"                    },
 626         { do_bad,               SIGBUS,  0,             "unknown 18"                    },
 627         { do_bad,               SIGBUS,  0,             "unknown 19"                    },
 628         { do_sea,               SIGBUS,  0,             "level 0 (translation table walk)"      },
 629         { do_sea,               SIGBUS,  0,             "level 1 (translation table walk)"      },
 630         { do_sea,               SIGBUS,  0,             "level 2 (translation table walk)"      },
 631         { do_sea,               SIGBUS,  0,             "level 3 (translation table walk)"      },
 632         { do_sea,               SIGBUS,  0,             "synchronous parity or ECC error" },
 633         { do_bad,               SIGBUS,  0,             "unknown 25"                    },
 634         { do_bad,               SIGBUS,  0,             "unknown 26"                    },
 635         { do_bad,               SIGBUS,  0,             "unknown 27"                    },
 636         { do_sea,               SIGBUS,  0,             "level 0 synchronous parity error (translation table walk)"     },
 637         { do_sea,               SIGBUS,  0,             "level 1 synchronous parity error (translation table walk)"     },
 638         { do_sea,               SIGBUS,  0,             "level 2 synchronous parity error (translation table walk)"     },
 639         { do_sea,               SIGBUS,  0,             "level 3 synchronous parity error (translation table walk)"     },
 640         { do_bad,               SIGBUS,  0,             "unknown 32"                    },
 641         { do_alignment_fault,   SIGBUS,  BUS_ADRALN,    "alignment fault"               },
 642         { do_bad,               SIGBUS,  0,             "unknown 34"                    },
 643         { do_bad,               SIGBUS,  0,             "unknown 35"                    },
 644         { do_bad,               SIGBUS,  0,             "unknown 36"                    },
 645         { do_bad,               SIGBUS,  0,             "unknown 37"                    },
 646         { do_bad,               SIGBUS,  0,             "unknown 38"                    },
 647         { do_bad,               SIGBUS,  0,             "unknown 39"                    },
 648         { do_bad,               SIGBUS,  0,             "unknown 40"                    },
 649         { do_bad,               SIGBUS,  0,             "unknown 41"                    },
 650         { do_bad,               SIGBUS,  0,             "unknown 42"                    },
 651         { do_bad,               SIGBUS,  0,             "unknown 43"                    },
 652         { do_bad,               SIGBUS,  0,             "unknown 44"                    },
 653         { do_bad,               SIGBUS,  0,             "unknown 45"                    },
 654         { do_bad,               SIGBUS,  0,             "unknown 46"                    },
 655         { do_bad,               SIGBUS,  0,             "unknown 47"                    },
 656         { do_bad,               SIGBUS,  0,             "TLB conflict abort"            },
 657         { do_bad,               SIGBUS,  0,             "unknown 49"                    },
 658         { do_bad,               SIGBUS,  0,             "unknown 50"                    },
 659         { do_bad,               SIGBUS,  0,             "unknown 51"                    },
 660         { do_bad,               SIGBUS,  0,             "implementation fault (lockdown abort)" },
 661         { do_bad,               SIGBUS,  0,             "implementation fault (unsupported exclusive)" },
 662         { do_bad,               SIGBUS,  0,             "unknown 54"                    },
 663         { do_bad,               SIGBUS,  0,             "unknown 55"                    },
 664         { do_bad,               SIGBUS,  0,             "unknown 56"                    },
 665         { do_bad,               SIGBUS,  0,             "unknown 57"                    },
 666         { do_bad,               SIGBUS,  0,             "unknown 58"                    },
 667         { do_bad,               SIGBUS,  0,             "unknown 59"                    },
 668         { do_bad,               SIGBUS,  0,             "unknown 60"                    },
 669         { do_bad,               SIGBUS,  0,             "section domain fault"          },
 670         { do_bad,               SIGBUS,  0,             "page domain fault"             },
 671         { do_bad,               SIGBUS,  0,             "unknown 63"                    },
 672 };
 673
 674 /*
 675  * Handle Synchronous External Aborts that occur in a guest kernel.
 676  *
 677  * The return value will be zero if the SEA was successfully handled
 678  * and non-zero if there was an error processing the error or there was
 679  * no error to process.
 680  */
 681 int handle_guest_sea(phys_addr_t addr, unsigned int esr)
 682 {
 683         int ret = -ENOENT;
 684
 685         if (IS_ENABLED(CONFIG_ACPI_APEI_SEA))
 686                 ret = ghes_notify_sea();
 687
 688         return ret;
 689 }
 690
 691 /*
 692  * Dispatch a data abort to the relevant handler.
 693  */
 694 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
 695                                          struct pt_regs *regs)
 696 {
 697         const struct fault_info *inf = esr_to_fault_info(esr);
 698         struct siginfo info;
 699
 700         if (!inf->fn(addr, esr, regs))
 701                 return;
 702
 703         pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
 704                  inf->name, esr, addr);
 705
 706         info.si_signo = inf->sig;
 707         info.si_errno = 0;
 708         info.si_code  = inf->code;
 709         info.si_addr  = (void __user *)addr;
 710         arm64_notify_die("", regs, &info, esr);
 711 }
 712
 713 /*
 714  * Handle stack alignment exceptions.
 715  */
 716 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
 717                                            unsigned int esr,
 718                                            struct pt_regs *regs)
 719 {
 720         struct siginfo info;
 721         struct task_struct *tsk = current;
 722
 723         if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
 724                 pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
 725                                     tsk->comm, task_pid_nr(tsk),
 726                                     esr_get_class_string(esr), (void *)regs->pc,
 727                                     (void *)regs->sp);
 728
 729         info.si_signo = SIGBUS;
 730         info.si_errno = 0;
 731         info.si_code  = BUS_ADRALN;
 732         info.si_addr  = (void __user *)addr;
 733         arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
 734 }
 735
 736 int __init early_brk64(unsigned long addr, unsigned int esr,
 737                        struct pt_regs *regs);
 738
 739 /*
 740  * __refdata because early_brk64 is __init, but the reference to it is
 741  * clobbered at arch_initcall time.
 742  * See traps.c and debug-monitors.c:debug_traps_init().
 743  */
 744 static struct fault_info __refdata debug_fault_info[] = {
 745         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware breakpoint"   },
 746         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware single-step"  },
 747         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware watchpoint"   },
 748         { do_bad,       SIGBUS,         0,              "unknown 3"             },
 749         { do_bad,       SIGTRAP,        TRAP_BRKPT,     "aarch32 BKPT"          },
 750         { do_bad,       SIGTRAP,        0,              "aarch32 vector catch"  },
 751         { early_brk64,  SIGTRAP,        TRAP_BRKPT,     "aarch64 BRK"           },
 752         { do_bad,       SIGBUS,         0,              "unknown 7"             },
 753 };
 754
 755 void __init hook_debug_fault_code(int nr,
 756                                   int (*fn)(unsigned long, unsigned int, struct pt_regs *),
 757                                   int sig, int code, const char *name)
 758 {
 759         BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
 760
 761         debug_fault_info[nr].fn         = fn;
 762         debug_fault_info[nr].sig        = sig;
 763         debug_fault_info[nr].code       = code;
 764         debug_fault_info[nr].name       = name;
 765 }
 766
 767 asmlinkage int __exception do_debug_exception(unsigned long addr,
 768                                               unsigned int esr,
 769                                               struct pt_regs *regs)
 770 {
 771         const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
 772         struct siginfo info;
 773         int rv;
 774
 775         /*
 776          * Tell lockdep we disabled irqs in entry.S. Do nothing if they were
 777          * already disabled to preserve the last enabled/disabled addresses.
 778          */
 779         if (interrupts_enabled(regs))
 780                 trace_hardirqs_off();
 781
 782         if (!inf->fn(addr, esr, regs)) {
 783                 rv = 1;
 784         } else {
 785                 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
 786                          inf->name, esr, addr);
 787
 788                 info.si_signo = inf->sig;
 789                 info.si_errno = 0;
 790                 info.si_code  = inf->code;
 791                 info.si_addr  = (void __user *)addr;
 792                 arm64_notify_die("", regs, &info, 0);
 793                 rv = 0;
 794         }
 795
 796         if (interrupts_enabled(regs))
 797                 trace_hardirqs_on();
 798
 799         return rv;
 800 }
 801 NOKPROBE_SYMBOL(do_debug_exception);
 802
 803 #ifdef CONFIG_ARM64_PAN
 804 int cpu_enable_pan(void *__unused)
 805 {
 806         /*
 807          * We modify PSTATE. This won't work from irq context as the PSTATE
 808          * is discarded once we return from the exception.
 809          */
 810         WARN_ON_ONCE(in_interrupt());
 811
 812         config_sctlr_el1(SCTLR_EL1_SPAN, 0);
 813         asm(SET_PSTATE_PAN(1));
 814         return 0;
 815 }
 816 #endif /* CONFIG_ARM64_PAN */