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 static void data_abort_decode(unsigned int esr)
  87 {
  88         pr_alert("Data abort info:\n");
  89
  90         if (esr & ESR_ELx_ISV) {
  91                 pr_alert("  Access size = %u byte(s)\n",
  92                          1U << ((esr & ESR_ELx_SAS) >> ESR_ELx_SAS_SHIFT));
  93                 pr_alert("  SSE = %lu, SRT = %lu\n",
  94                          (esr & ESR_ELx_SSE) >> ESR_ELx_SSE_SHIFT,
  95                          (esr & ESR_ELx_SRT_MASK) >> ESR_ELx_SRT_SHIFT);
  96                 pr_alert("  SF = %lu, AR = %lu\n",
  97                          (esr & ESR_ELx_SF) >> ESR_ELx_SF_SHIFT,
  98                          (esr & ESR_ELx_AR) >> ESR_ELx_AR_SHIFT);
  99         } else {
 100                 pr_alert("  ISV = 0, ISS = 0x%08lx\n", esr & ESR_ELx_ISS_MASK);
 101         }
 102
 103         pr_alert("  CM = %lu, WnR = %lu\n",
 104                  (esr & ESR_ELx_CM) >> ESR_ELx_CM_SHIFT,
 105                  (esr & ESR_ELx_WNR) >> ESR_ELx_WNR_SHIFT);
 106 }
 107
 108 static void mem_abort_decode(unsigned int esr)
 109 {
 110         pr_alert("Mem abort info:\n");
 111
 112         pr_alert("  ESR = 0x%08x\n", esr);
 113         pr_alert("  Exception class = %s, IL = %u bits\n",
 114                  esr_get_class_string(esr),
 115                  (esr & ESR_ELx_IL) ? 32 : 16);
 116         pr_alert("  SET = %lu, FnV = %lu\n",
 117                  (esr & ESR_ELx_SET_MASK) >> ESR_ELx_SET_SHIFT,
 118                  (esr & ESR_ELx_FnV) >> ESR_ELx_FnV_SHIFT);
 119         pr_alert("  EA = %lu, S1PTW = %lu\n",
 120                  (esr & ESR_ELx_EA) >> ESR_ELx_EA_SHIFT,
 121                  (esr & ESR_ELx_S1PTW) >> ESR_ELx_S1PTW_SHIFT);
 122
 123         if (esr_is_data_abort(esr))
 124                 data_abort_decode(esr);
 125 }
 126
 127 /*
 128  * Dump out the page tables associated with 'addr' in the currently active mm.
 129  */
 130 void show_pte(unsigned long addr)
 131 {
 132         struct mm_struct *mm;
 133         pgd_t *pgd;
 134
 135         if (addr < TASK_SIZE) {
 136                 /* TTBR0 */
 137                 mm = current->active_mm;
 138                 if (mm == &init_mm) {
 139                         pr_alert("[%016lx] user address but active_mm is swapper\n",
 140                                  addr);
 141                         return;
 142                 }
 143         } else if (addr >= VA_START) {
 144                 /* TTBR1 */
 145                 mm = &init_mm;
 146         } else {
 147                 pr_alert("[%016lx] address between user and kernel address ranges\n",
 148                          addr);
 149                 return;
 150         }
 151
 152         pr_alert("%s pgtable: %luk pages, %u-bit VAs, pgd = %p\n",
 153                  mm == &init_mm ? "swapper" : "user", PAGE_SIZE / SZ_1K,
 154                  VA_BITS, mm->pgd);
 155         pgd = pgd_offset(mm, addr);
 156         pr_alert("[%016lx] *pgd=%016llx", addr, pgd_val(*pgd));
 157
 158         do {
 159                 pud_t *pud;
 160                 pmd_t *pmd;
 161                 pte_t *pte;
 162
 163                 if (pgd_none(*pgd) || pgd_bad(*pgd))
 164                         break;
 165
 166                 pud = pud_offset(pgd, addr);
 167                 pr_cont(", *pud=%016llx", pud_val(*pud));
 168                 if (pud_none(*pud) || pud_bad(*pud))
 169                         break;
 170
 171                 pmd = pmd_offset(pud, addr);
 172                 pr_cont(", *pmd=%016llx", pmd_val(*pmd));
 173                 if (pmd_none(*pmd) || pmd_bad(*pmd))
 174                         break;
 175
 176                 pte = pte_offset_map(pmd, addr);
 177                 pr_cont(", *pte=%016llx", pte_val(*pte));
 178                 pte_unmap(pte);
 179         } while(0);
 180
 181         pr_cont("\n");
 182 }
 183
 184 /*
 185  * This function sets the access flags (dirty, accessed), as well as write
 186  * permission, and only to a more permissive setting.
 187  *
 188  * It needs to cope with hardware update of the accessed/dirty state by other
 189  * agents in the system and can safely skip the __sync_icache_dcache() call as,
 190  * like set_pte_at(), the PTE is never changed from no-exec to exec here.
 191  *
 192  * Returns whether or not the PTE actually changed.
 193  */
 194 int ptep_set_access_flags(struct vm_area_struct *vma,
 195                           unsigned long address, pte_t *ptep,
 196                           pte_t entry, int dirty)
 197 {
 198         pteval_t old_pteval, pteval;
 199
 200         if (pte_same(*ptep, entry))
 201                 return 0;
 202
 203         /* only preserve the access flags and write permission */
 204         pte_val(entry) &= PTE_RDONLY | PTE_AF | PTE_WRITE | PTE_DIRTY;
 205
 206         /*
 207          * Setting the flags must be done atomically to avoid racing with the
 208          * hardware update of the access/dirty state. The PTE_RDONLY bit must
 209          * be set to the most permissive (lowest value) of *ptep and entry
 210          * (calculated as: a & b == ~(~a | ~b)).
 211          */
 212         pte_val(entry) ^= PTE_RDONLY;
 213         pteval = READ_ONCE(pte_val(*ptep));
 214         do {
 215                 old_pteval = pteval;
 216                 pteval ^= PTE_RDONLY;
 217                 pteval |= pte_val(entry);
 218                 pteval ^= PTE_RDONLY;
 219                 pteval = cmpxchg_relaxed(&pte_val(*ptep), old_pteval, pteval);
 220         } while (pteval != old_pteval);
 221
 222         flush_tlb_fix_spurious_fault(vma, address);
 223         return 1;
 224 }
 225
 226 static bool is_el1_instruction_abort(unsigned int esr)
 227 {
 228         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR;
 229 }
 230
 231 static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs,
 232                                        unsigned long addr)
 233 {
 234         unsigned int ec       = ESR_ELx_EC(esr);
 235         unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE;
 236
 237         if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR)
 238                 return false;
 239
 240         if (fsc_type == ESR_ELx_FSC_PERM)
 241                 return true;
 242
 243         if (addr < USER_DS && system_uses_ttbr0_pan())
 244                 return fsc_type == ESR_ELx_FSC_FAULT &&
 245                         (regs->pstate & PSR_PAN_BIT);
 246
 247         return false;
 248 }
 249
 250 static void __do_kernel_fault(unsigned long addr, unsigned int esr,
 251                               struct pt_regs *regs)
 252 {
 253         const char *msg;
 254
 255         /*
 256          * Are we prepared to handle this kernel fault?
 257          * We are almost certainly not prepared to handle instruction faults.
 258          */
 259         if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
 260                 return;
 261
 262         bust_spinlocks(1);
 263
 264         if (is_permission_fault(esr, regs, addr)) {
 265                 if (esr & ESR_ELx_WNR)
 266                         msg = "write to read-only memory";
 267                 else
 268                         msg = "read from unreadable memory";
 269         } else if (addr < PAGE_SIZE) {
 270                 msg = "NULL pointer dereference";
 271         } else {
 272                 msg = "paging request";
 273         }
 274
 275         pr_alert("Unable to handle kernel %s at virtual address %08lx\n", msg,
 276                  addr);
 277
 278         mem_abort_decode(esr);
 279
 280         show_pte(addr);
 281         die("Oops", regs, esr);
 282         bust_spinlocks(0);
 283         do_exit(SIGKILL);
 284 }
 285
 286 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
 287                             unsigned int esr, unsigned int sig, int code,
 288                             struct pt_regs *regs, int fault)
 289 {
 290         struct siginfo si;
 291         const struct fault_info *inf;
 292         unsigned int lsb = 0;
 293
 294         if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
 295                 inf = esr_to_fault_info(esr);
 296                 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x",
 297                         tsk->comm, task_pid_nr(tsk), inf->name, sig,
 298                         addr, esr);
 299                 print_vma_addr(KERN_CONT ", in ", regs->pc);
 300                 pr_cont("\n");
 301                 __show_regs(regs);
 302         }
 303
 304         tsk->thread.fault_address = addr;
 305         tsk->thread.fault_code = esr;
 306         si.si_signo = sig;
 307         si.si_errno = 0;
 308         si.si_code = code;
 309         si.si_addr = (void __user *)addr;
 310         /*
 311          * Either small page or large page may be poisoned.
 312          * In other words, VM_FAULT_HWPOISON_LARGE and
 313          * VM_FAULT_HWPOISON are mutually exclusive.
 314          */
 315         if (fault & VM_FAULT_HWPOISON_LARGE)
 316                 lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault));
 317         else if (fault & VM_FAULT_HWPOISON)
 318                 lsb = PAGE_SHIFT;
 319         si.si_addr_lsb = lsb;
 320
 321         force_sig_info(sig, &si, tsk);
 322 }
 323
 324 static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 325 {
 326         struct task_struct *tsk = current;
 327         const struct fault_info *inf;
 328
 329         /*
 330          * If we are in kernel mode at this point, we have no context to
 331          * handle this fault with.
 332          */
 333         if (user_mode(regs)) {
 334                 inf = esr_to_fault_info(esr);
 335                 __do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs, 0);
 336         } else
 337                 __do_kernel_fault(addr, esr, regs);
 338 }
 339
 340 #define VM_FAULT_BADMAP         0x010000
 341 #define VM_FAULT_BADACCESS      0x020000
 342
 343 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
 344                            unsigned int mm_flags, unsigned long vm_flags,
 345                            struct task_struct *tsk)
 346 {
 347         struct vm_area_struct *vma;
 348         int fault;
 349
 350         vma = find_vma(mm, addr);
 351         fault = VM_FAULT_BADMAP;
 352         if (unlikely(!vma))
 353                 goto out;
 354         if (unlikely(vma->vm_start > addr))
 355                 goto check_stack;
 356
 357         /*
 358          * Ok, we have a good vm_area for this memory access, so we can handle
 359          * it.
 360          */
 361 good_area:
 362         /*
 363          * Check that the permissions on the VMA allow for the fault which
 364          * occurred.
 365          */
 366         if (!(vma->vm_flags & vm_flags)) {
 367                 fault = VM_FAULT_BADACCESS;
 368                 goto out;
 369         }
 370
 371         return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags);
 372
 373 check_stack:
 374         if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
 375                 goto good_area;
 376 out:
 377         return fault;
 378 }
 379
 380 static bool is_el0_instruction_abort(unsigned int esr)
 381 {
 382         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW;
 383 }
 384
 385 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
 386                                    struct pt_regs *regs)
 387 {
 388         struct task_struct *tsk;
 389         struct mm_struct *mm;
 390         int fault, sig, code, major = 0;
 391         unsigned long vm_flags = VM_READ | VM_WRITE;
 392         unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
 393
 394         if (notify_page_fault(regs, esr))
 395                 return 0;
 396
 397         tsk = current;
 398         mm  = tsk->mm;
 399
 400         /*
 401          * If we're in an interrupt or have no user context, we must not take
 402          * the fault.
 403          */
 404         if (faulthandler_disabled() || !mm)
 405                 goto no_context;
 406
 407         if (user_mode(regs))
 408                 mm_flags |= FAULT_FLAG_USER;
 409
 410         if (is_el0_instruction_abort(esr)) {
 411                 vm_flags = VM_EXEC;
 412         } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
 413                 vm_flags = VM_WRITE;
 414                 mm_flags |= FAULT_FLAG_WRITE;
 415         }
 416
 417         if (addr < USER_DS && is_permission_fault(esr, regs, addr)) {
 418                 /* regs->orig_addr_limit may be 0 if we entered from EL0 */
 419                 if (regs->orig_addr_limit == KERNEL_DS)
 420                         die("Accessing user space memory with fs=KERNEL_DS", regs, esr);
 421
 422                 if (is_el1_instruction_abort(esr))
 423                         die("Attempting to execute userspace memory", regs, esr);
 424
 425                 if (!search_exception_tables(regs->pc))
 426                         die("Accessing user space memory outside uaccess.h routines", regs, esr);
 427         }
 428
 429         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
 430
 431         /*
 432          * As per x86, we may deadlock here. However, since the kernel only
 433          * validly references user space from well defined areas of the code,
 434          * we can bug out early if this is from code which shouldn't.
 435          */
 436         if (!down_read_trylock(&mm->mmap_sem)) {
 437                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 438                         goto no_context;
 439 retry:
 440                 down_read(&mm->mmap_sem);
 441         } else {
 442                 /*
 443                  * The above down_read_trylock() might have succeeded in which
 444                  * case, we'll have missed the might_sleep() from down_read().
 445                  */
 446                 might_sleep();
 447 #ifdef CONFIG_DEBUG_VM
 448                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 449                         goto no_context;
 450 #endif
 451         }
 452
 453         fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
 454         major |= fault & VM_FAULT_MAJOR;
 455
 456         if (fault & VM_FAULT_RETRY) {
 457                 /*
 458                  * If we need to retry but a fatal signal is pending,
 459                  * handle the signal first. We do not need to release
 460                  * the mmap_sem because it would already be released
 461                  * in __lock_page_or_retry in mm/filemap.c.
 462                  */
 463                 if (fatal_signal_pending(current)) {
 464                         if (!user_mode(regs))
 465                                 goto no_context;
 466                         return 0;
 467                 }
 468
 469                 /*
 470                  * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
 471                  * starvation.
 472                  */
 473                 if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
 474                         mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
 475                         mm_flags |= FAULT_FLAG_TRIED;
 476                         goto retry;
 477                 }
 478         }
 479         up_read(&mm->mmap_sem);
 480
 481         /*
 482          * Handle the "normal" (no error) case first.
 483          */
 484         if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
 485                               VM_FAULT_BADACCESS)))) {
 486                 /*
 487                  * Major/minor page fault accounting is only done
 488                  * once. If we go through a retry, it is extremely
 489                  * likely that the page will be found in page cache at
 490                  * that point.
 491                  */
 492                 if (major) {
 493                         tsk->maj_flt++;
 494                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
 495                                       addr);
 496                 } else {
 497                         tsk->min_flt++;
 498                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
 499                                       addr);
 500                 }
 501
 502                 return 0;
 503         }
 504
 505         /*
 506          * If we are in kernel mode at this point, we have no context to
 507          * handle this fault with.
 508          */
 509         if (!user_mode(regs))
 510                 goto no_context;
 511
 512         if (fault & VM_FAULT_OOM) {
 513                 /*
 514                  * We ran out of memory, call the OOM killer, and return to
 515                  * userspace (which will retry the fault, or kill us if we got
 516                  * oom-killed).
 517                  */
 518                 pagefault_out_of_memory();
 519                 return 0;
 520         }
 521
 522         if (fault & VM_FAULT_SIGBUS) {
 523                 /*
 524                  * We had some memory, but were unable to successfully fix up
 525                  * this page fault.
 526                  */
 527                 sig = SIGBUS;
 528                 code = BUS_ADRERR;
 529         } else if (fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) {
 530                 sig = SIGBUS;
 531                 code = BUS_MCEERR_AR;
 532         } else {
 533                 /*
 534                  * Something tried to access memory that isn't in our memory
 535                  * map.
 536                  */
 537                 sig = SIGSEGV;
 538                 code = fault == VM_FAULT_BADACCESS ?
 539                         SEGV_ACCERR : SEGV_MAPERR;
 540         }
 541
 542         __do_user_fault(tsk, addr, esr, sig, code, regs, fault);
 543         return 0;
 544
 545 no_context:
 546         __do_kernel_fault(addr, esr, regs);
 547         return 0;
 548 }
 549
 550 static int __kprobes do_translation_fault(unsigned long addr,
 551                                           unsigned int esr,
 552                                           struct pt_regs *regs)
 553 {
 554         if (addr < TASK_SIZE)
 555                 return do_page_fault(addr, esr, regs);
 556
 557         do_bad_area(addr, esr, regs);
 558         return 0;
 559 }
 560
 561 static int do_alignment_fault(unsigned long addr, unsigned int esr,
 562                               struct pt_regs *regs)
 563 {
 564         do_bad_area(addr, esr, regs);
 565         return 0;
 566 }
 567
 568 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 569 {
 570         return 1; /* "fault" */
 571 }
 572
 573 static int do_sea(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 574 {
 575         struct siginfo info;
 576         const struct fault_info *inf;
 577         int ret = 0;
 578
 579         inf = esr_to_fault_info(esr);
 580         pr_err("Synchronous External Abort: %s (0x%08x) at 0x%016lx\n",
 581                 inf->name, esr, addr);
 582
 583         /*
 584          * Synchronous aborts may interrupt code which had interrupts masked.
 585          * Before calling out into the wider kernel tell the interested
 586          * subsystems.
 587          */
 588         if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) {
 589                 if (interrupts_enabled(regs))
 590                         nmi_enter();
 591
 592                 ret = ghes_notify_sea();
 593
 594                 if (interrupts_enabled(regs))
 595                         nmi_exit();
 596         }
 597
 598         info.si_signo = SIGBUS;
 599         info.si_errno = 0;
 600         info.si_code  = 0;
 601         if (esr & ESR_ELx_FnV)
 602                 info.si_addr = NULL;
 603         else
 604                 info.si_addr  = (void __user *)addr;
 605         arm64_notify_die("", regs, &info, esr);
 606
 607         return ret;
 608 }
 609
 610 static const struct fault_info fault_info[] = {
 611         { do_bad,               SIGBUS,  0,             "ttbr address size fault"       },
 612         { do_bad,               SIGBUS,  0,             "level 1 address size fault"    },
 613         { do_bad,               SIGBUS,  0,             "level 2 address size fault"    },
 614         { do_bad,               SIGBUS,  0,             "level 3 address size fault"    },
 615         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 0 translation fault"     },
 616         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 1 translation fault"     },
 617         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 2 translation fault"     },
 618         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 3 translation fault"     },
 619         { do_bad,               SIGBUS,  0,             "unknown 8"                     },
 620         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 access flag fault"     },
 621         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 access flag fault"     },
 622         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 access flag fault"     },
 623         { do_bad,               SIGBUS,  0,             "unknown 12"                    },
 624         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 permission fault"      },
 625         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 permission fault"      },
 626         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 permission fault"      },
 627         { do_sea,               SIGBUS,  0,             "synchronous external abort"    },
 628         { do_bad,               SIGBUS,  0,             "unknown 17"                    },
 629         { do_bad,               SIGBUS,  0,             "unknown 18"                    },
 630         { do_bad,               SIGBUS,  0,             "unknown 19"                    },
 631         { do_sea,               SIGBUS,  0,             "level 0 (translation table walk)"      },
 632         { do_sea,               SIGBUS,  0,             "level 1 (translation table walk)"      },
 633         { do_sea,               SIGBUS,  0,             "level 2 (translation table walk)"      },
 634         { do_sea,               SIGBUS,  0,             "level 3 (translation table walk)"      },
 635         { do_sea,               SIGBUS,  0,             "synchronous parity or ECC error" },    // Reserved when RAS is implemented
 636         { do_bad,               SIGBUS,  0,             "unknown 25"                    },
 637         { do_bad,               SIGBUS,  0,             "unknown 26"                    },
 638         { do_bad,               SIGBUS,  0,             "unknown 27"                    },
 639         { do_sea,               SIGBUS,  0,             "level 0 synchronous parity error (translation table walk)"     },      // Reserved when RAS is implemented
 640         { do_sea,               SIGBUS,  0,             "level 1 synchronous parity error (translation table walk)"     },      // Reserved when RAS is implemented
 641         { do_sea,               SIGBUS,  0,             "level 2 synchronous parity error (translation table walk)"     },      // Reserved when RAS is implemented
 642         { do_sea,               SIGBUS,  0,             "level 3 synchronous parity error (translation table walk)"     },      // Reserved when RAS is implemented
 643         { do_bad,               SIGBUS,  0,             "unknown 32"                    },
 644         { do_alignment_fault,   SIGBUS,  BUS_ADRALN,    "alignment fault"               },
 645         { do_bad,               SIGBUS,  0,             "unknown 34"                    },
 646         { do_bad,               SIGBUS,  0,             "unknown 35"                    },
 647         { do_bad,               SIGBUS,  0,             "unknown 36"                    },
 648         { do_bad,               SIGBUS,  0,             "unknown 37"                    },
 649         { do_bad,               SIGBUS,  0,             "unknown 38"                    },
 650         { do_bad,               SIGBUS,  0,             "unknown 39"                    },
 651         { do_bad,               SIGBUS,  0,             "unknown 40"                    },
 652         { do_bad,               SIGBUS,  0,             "unknown 41"                    },
 653         { do_bad,               SIGBUS,  0,             "unknown 42"                    },
 654         { do_bad,               SIGBUS,  0,             "unknown 43"                    },
 655         { do_bad,               SIGBUS,  0,             "unknown 44"                    },
 656         { do_bad,               SIGBUS,  0,             "unknown 45"                    },
 657         { do_bad,               SIGBUS,  0,             "unknown 46"                    },
 658         { do_bad,               SIGBUS,  0,             "unknown 47"                    },
 659         { do_bad,               SIGBUS,  0,             "TLB conflict abort"            },
 660         { do_bad,               SIGBUS,  0,             "Unsupported atomic hardware update fault"      },
 661         { do_bad,               SIGBUS,  0,             "unknown 50"                    },
 662         { do_bad,               SIGBUS,  0,             "unknown 51"                    },
 663         { do_bad,               SIGBUS,  0,             "implementation fault (lockdown abort)" },
 664         { do_bad,               SIGBUS,  0,             "implementation fault (unsupported exclusive)" },
 665         { do_bad,               SIGBUS,  0,             "unknown 54"                    },
 666         { do_bad,               SIGBUS,  0,             "unknown 55"                    },
 667         { do_bad,               SIGBUS,  0,             "unknown 56"                    },
 668         { do_bad,               SIGBUS,  0,             "unknown 57"                    },
 669         { do_bad,               SIGBUS,  0,             "unknown 58"                    },
 670         { do_bad,               SIGBUS,  0,             "unknown 59"                    },
 671         { do_bad,               SIGBUS,  0,             "unknown 60"                    },
 672         { do_bad,               SIGBUS,  0,             "section domain fault"          },
 673         { do_bad,               SIGBUS,  0,             "page domain fault"             },
 674         { do_bad,               SIGBUS,  0,             "unknown 63"                    },
 675 };
 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 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
 688                                          struct pt_regs *regs)
 689 {
 690         const struct fault_info *inf = esr_to_fault_info(esr);
 691         struct siginfo info;
 692
 693         if (!inf->fn(addr, esr, regs))
 694                 return;
 695
 696         pr_alert("Unhandled fault: %s at 0x%016lx\n",
 697                  inf->name, addr);
 698
 699         mem_abort_decode(esr);
 700
 701         if (!user_mode(regs))
 702                 show_pte(addr);
 703
 704         info.si_signo = inf->sig;
 705         info.si_errno = 0;
 706         info.si_code  = inf->code;
 707         info.si_addr  = (void __user *)addr;
 708         arm64_notify_die("", regs, &info, esr);
 709 }
 710
 711 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
 712                                            unsigned int esr,
 713                                            struct pt_regs *regs)
 714 {
 715         struct siginfo info;
 716         struct task_struct *tsk = current;
 717
 718         if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
 719                 pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
 720                                     tsk->comm, task_pid_nr(tsk),
 721                                     esr_get_class_string(esr), (void *)regs->pc,
 722                                     (void *)regs->sp);
 723
 724         info.si_signo = SIGBUS;
 725         info.si_errno = 0;
 726         info.si_code  = BUS_ADRALN;
 727         info.si_addr  = (void __user *)addr;
 728         arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
 729 }
 730
 731 int __init early_brk64(unsigned long addr, unsigned int esr,
 732                        struct pt_regs *regs);
 733
 734 /*
 735  * __refdata because early_brk64 is __init, but the reference to it is
 736  * clobbered at arch_initcall time.
 737  * See traps.c and debug-monitors.c:debug_traps_init().
 738  */
 739 static struct fault_info __refdata debug_fault_info[] = {
 740         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware breakpoint"   },
 741         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware single-step"  },
 742         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware watchpoint"   },
 743         { do_bad,       SIGBUS,         0,              "unknown 3"             },
 744         { do_bad,       SIGTRAP,        TRAP_BRKPT,     "aarch32 BKPT"          },
 745         { do_bad,       SIGTRAP,        0,              "aarch32 vector catch"  },
 746         { early_brk64,  SIGTRAP,        TRAP_BRKPT,     "aarch64 BRK"           },
 747         { do_bad,       SIGBUS,         0,              "unknown 7"             },
 748 };
 749
 750 void __init hook_debug_fault_code(int nr,
 751                                   int (*fn)(unsigned long, unsigned int, struct pt_regs *),
 752                                   int sig, int code, const char *name)
 753 {
 754         BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
 755
 756         debug_fault_info[nr].fn         = fn;
 757         debug_fault_info[nr].sig        = sig;
 758         debug_fault_info[nr].code       = code;
 759         debug_fault_info[nr].name       = name;
 760 }
 761
 762 asmlinkage int __exception do_debug_exception(unsigned long addr,
 763                                               unsigned int esr,
 764                                               struct pt_regs *regs)
 765 {
 766         const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
 767         struct siginfo info;
 768         int rv;
 769
 770         /*
 771          * Tell lockdep we disabled irqs in entry.S. Do nothing if they were
 772          * already disabled to preserve the last enabled/disabled addresses.
 773          */
 774         if (interrupts_enabled(regs))
 775                 trace_hardirqs_off();
 776
 777         if (!inf->fn(addr, esr, regs)) {
 778                 rv = 1;
 779         } else {
 780                 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
 781                          inf->name, esr, addr);
 782
 783                 info.si_signo = inf->sig;
 784                 info.si_errno = 0;
 785                 info.si_code  = inf->code;
 786                 info.si_addr  = (void __user *)addr;
 787                 arm64_notify_die("", regs, &info, 0);
 788                 rv = 0;
 789         }
 790
 791         if (interrupts_enabled(regs))
 792                 trace_hardirqs_on();
 793
 794         return rv;
 795 }
 796 NOKPROBE_SYMBOL(do_debug_exception);
 797
 798 #ifdef CONFIG_ARM64_PAN
 799 int cpu_enable_pan(void *__unused)
 800 {
 801         /*
 802          * We modify PSTATE. This won't work from irq context as the PSTATE
 803          * is discarded once we return from the exception.
 804          */
 805         WARN_ON_ONCE(in_interrupt());
 806
 807         config_sctlr_el1(SCTLR_EL1_SPAN, 0);
 808         asm(SET_PSTATE_PAN(1));
 809         return 0;
 810 }
 811 #endif /* CONFIG_ARM64_PAN */