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/cpufeature.h>
  38 #include <asm/exception.h>
  39 #include <asm/debug-monitors.h>
  40 #include <asm/esr.h>
  41 #include <asm/sysreg.h>
  42 #include <asm/system_misc.h>
  43 #include <asm/pgtable.h>
  44 #include <asm/tlbflush.h>
  45
  46 #include <acpi/ghes.h>
  47
  48 struct fault_info {
  49         int     (*fn)(unsigned long addr, unsigned int esr,
  50                       struct pt_regs *regs);
  51         int     sig;
  52         int     code;
  53         const char *name;
  54 };
  55
  56 static const struct fault_info fault_info[];
  57
  58 static inline const struct fault_info *esr_to_fault_info(unsigned int esr)
  59 {
  60         return fault_info + (esr & 63);
  61 }
  62
  63 #ifdef CONFIG_KPROBES
  64 static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
  65 {
  66         int ret = 0;
  67
  68         /* kprobe_running() needs smp_processor_id() */
  69         if (!user_mode(regs)) {
  70                 preempt_disable();
  71                 if (kprobe_running() && kprobe_fault_handler(regs, esr))
  72                         ret = 1;
  73                 preempt_enable();
  74         }
  75
  76         return ret;
  77 }
  78 #else
  79 static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
  80 {
  81         return 0;
  82 }
  83 #endif
  84
  85 /*
  86  * Dump out the page tables associated with 'addr' in the currently active mm.
  87  */
  88 void show_pte(unsigned long addr)
  89 {
  90         struct mm_struct *mm;
  91         pgd_t *pgd;
  92
  93         if (addr < TASK_SIZE) {
  94                 /* TTBR0 */
  95                 mm = current->active_mm;
  96                 if (mm == &init_mm) {
  97                         pr_alert("[%016lx] user address but active_mm is swapper\n",
  98                                  addr);
  99                         return;
 100                 }
 101         } else if (addr >= VA_START) {
 102                 /* TTBR1 */
 103                 mm = &init_mm;
 104         } else {
 105                 pr_alert("[%016lx] address between user and kernel address ranges\n",
 106                          addr);
 107                 return;
 108         }
 109
 110         pr_alert("%s pgtable: %luk pages, %u-bit VAs, pgd = %p\n",
 111                  mm == &init_mm ? "swapper" : "user", PAGE_SIZE / SZ_1K,
 112                  VA_BITS, mm->pgd);
 113         pgd = pgd_offset(mm, addr);
 114         pr_alert("[%016lx] *pgd=%016llx", addr, pgd_val(*pgd));
 115
 116         do {
 117                 pud_t *pud;
 118                 pmd_t *pmd;
 119                 pte_t *pte;
 120
 121                 if (pgd_none(*pgd) || pgd_bad(*pgd))
 122                         break;
 123
 124                 pud = pud_offset(pgd, addr);
 125                 pr_cont(", *pud=%016llx", pud_val(*pud));
 126                 if (pud_none(*pud) || pud_bad(*pud))
 127                         break;
 128
 129                 pmd = pmd_offset(pud, addr);
 130                 pr_cont(", *pmd=%016llx", pmd_val(*pmd));
 131                 if (pmd_none(*pmd) || pmd_bad(*pmd))
 132                         break;
 133
 134                 pte = pte_offset_map(pmd, addr);
 135                 pr_cont(", *pte=%016llx", pte_val(*pte));
 136                 pte_unmap(pte);
 137         } while(0);
 138
 139         pr_cont("\n");
 140 }
 141
 142 #ifdef CONFIG_ARM64_HW_AFDBM
 143 /*
 144  * This function sets the access flags (dirty, accessed), as well as write
 145  * permission, and only to a more permissive setting.
 146  *
 147  * It needs to cope with hardware update of the accessed/dirty state by other
 148  * agents in the system and can safely skip the __sync_icache_dcache() call as,
 149  * like set_pte_at(), the PTE is never changed from no-exec to exec here.
 150  *
 151  * Returns whether or not the PTE actually changed.
 152  */
 153 int ptep_set_access_flags(struct vm_area_struct *vma,
 154                           unsigned long address, pte_t *ptep,
 155                           pte_t entry, int dirty)
 156 {
 157         pteval_t old_pteval;
 158         unsigned int tmp;
 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                 pte_val(entry) |= PTE_RDONLY;
 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         asm volatile("//        ptep_set_access_flags\n"
 178         "       prfm    pstl1strm, %2\n"
 179         "1:     ldxr    %0, %2\n"
 180         "       eor     %0, %0, %3              // negate PTE_RDONLY in *ptep\n"
 181         "       orr     %0, %0, %4              // set flags\n"
 182         "       eor     %0, %0, %3              // negate final PTE_RDONLY\n"
 183         "       stxr    %w1, %0, %2\n"
 184         "       cbnz    %w1, 1b\n"
 185         : "=&r" (old_pteval), "=&r" (tmp), "+Q" (pte_val(*ptep))
 186         : "L" (PTE_RDONLY), "r" (pte_val(entry)));
 187
 188         flush_tlb_fix_spurious_fault(vma, address);
 189         return 1;
 190 }
 191 #endif
 192
 193 static bool is_el1_instruction_abort(unsigned int esr)
 194 {
 195         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR;
 196 }
 197
 198 static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs,
 199                                        unsigned long addr)
 200 {
 201         unsigned int ec       = ESR_ELx_EC(esr);
 202         unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE;
 203
 204         if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR)
 205                 return false;
 206
 207         if (fsc_type == ESR_ELx_FSC_PERM)
 208                 return true;
 209
 210         if (addr < USER_DS && system_uses_ttbr0_pan())
 211                 return fsc_type == ESR_ELx_FSC_FAULT &&
 212                         (regs->pstate & PSR_PAN_BIT);
 213
 214         return false;
 215 }
 216
 217 /*
 218  * The kernel tried to access some page that wasn't present.
 219  */
 220 static void __do_kernel_fault(unsigned long addr, unsigned int esr,
 221                               struct pt_regs *regs)
 222 {
 223         const char *msg;
 224
 225         /*
 226          * Are we prepared to handle this kernel fault?
 227          * We are almost certainly not prepared to handle instruction faults.
 228          */
 229         if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
 230                 return;
 231
 232         /*
 233          * No handler, we'll have to terminate things with extreme prejudice.
 234          */
 235         bust_spinlocks(1);
 236
 237         if (is_permission_fault(esr, regs, addr)) {
 238                 if (esr & ESR_ELx_WNR)
 239                         msg = "write to read-only memory";
 240                 else
 241                         msg = "read from unreadable memory";
 242         } else if (addr < PAGE_SIZE) {
 243                 msg = "NULL pointer dereference";
 244         } else {
 245                 msg = "paging request";
 246         }
 247
 248         pr_alert("Unable to handle kernel %s at virtual address %08lx\n", msg,
 249                  addr);
 250
 251         show_pte(addr);
 252         die("Oops", regs, esr);
 253         bust_spinlocks(0);
 254         do_exit(SIGKILL);
 255 }
 256
 257 /*
 258  * Something tried to access memory that isn't in our memory map. User mode
 259  * accesses just cause a SIGSEGV
 260  */
 261 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
 262                             unsigned int esr, unsigned int sig, int code,
 263                             struct pt_regs *regs, int fault)
 264 {
 265         struct siginfo si;
 266         const struct fault_info *inf;
 267         unsigned int lsb = 0;
 268
 269         if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
 270                 inf = esr_to_fault_info(esr);
 271                 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x",
 272                         tsk->comm, task_pid_nr(tsk), inf->name, sig,
 273                         addr, esr);
 274                 print_vma_addr(KERN_CONT ", in ", regs->pc);
 275                 pr_cont("\n");
 276                 __show_regs(regs);
 277         }
 278
 279         tsk->thread.fault_address = addr;
 280         tsk->thread.fault_code = esr;
 281         si.si_signo = sig;
 282         si.si_errno = 0;
 283         si.si_code = code;
 284         si.si_addr = (void __user *)addr;
 285         /*
 286          * Either small page or large page may be poisoned.
 287          * In other words, VM_FAULT_HWPOISON_LARGE and
 288          * VM_FAULT_HWPOISON are mutually exclusive.
 289          */
 290         if (fault & VM_FAULT_HWPOISON_LARGE)
 291                 lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault));
 292         else if (fault & VM_FAULT_HWPOISON)
 293                 lsb = PAGE_SHIFT;
 294         si.si_addr_lsb = lsb;
 295
 296         force_sig_info(sig, &si, tsk);
 297 }
 298
 299 static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 300 {
 301         struct task_struct *tsk = current;
 302         const struct fault_info *inf;
 303
 304         /*
 305          * If we are in kernel mode at this point, we have no context to
 306          * handle this fault with.
 307          */
 308         if (user_mode(regs)) {
 309                 inf = esr_to_fault_info(esr);
 310                 __do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs, 0);
 311         } else
 312                 __do_kernel_fault(addr, esr, regs);
 313 }
 314
 315 #define VM_FAULT_BADMAP         0x010000
 316 #define VM_FAULT_BADACCESS      0x020000
 317
 318 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
 319                            unsigned int mm_flags, unsigned long vm_flags,
 320                            struct task_struct *tsk)
 321 {
 322         struct vm_area_struct *vma;
 323         int fault;
 324
 325         vma = find_vma(mm, addr);
 326         fault = VM_FAULT_BADMAP;
 327         if (unlikely(!vma))
 328                 goto out;
 329         if (unlikely(vma->vm_start > addr))
 330                 goto check_stack;
 331
 332         /*
 333          * Ok, we have a good vm_area for this memory access, so we can handle
 334          * it.
 335          */
 336 good_area:
 337         /*
 338          * Check that the permissions on the VMA allow for the fault which
 339          * occurred.
 340          */
 341         if (!(vma->vm_flags & vm_flags)) {
 342                 fault = VM_FAULT_BADACCESS;
 343                 goto out;
 344         }
 345
 346         return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags);
 347
 348 check_stack:
 349         if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
 350                 goto good_area;
 351 out:
 352         return fault;
 353 }
 354
 355 static bool is_el0_instruction_abort(unsigned int esr)
 356 {
 357         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW;
 358 }
 359
 360 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
 361                                    struct pt_regs *regs)
 362 {
 363         struct task_struct *tsk;
 364         struct mm_struct *mm;
 365         int fault, sig, code, major = 0;
 366         unsigned long vm_flags = VM_READ | VM_WRITE;
 367         unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
 368
 369         if (notify_page_fault(regs, esr))
 370                 return 0;
 371
 372         tsk = current;
 373         mm  = tsk->mm;
 374
 375         /*
 376          * If we're in an interrupt or have no user context, we must not take
 377          * the fault.
 378          */
 379         if (faulthandler_disabled() || !mm)
 380                 goto no_context;
 381
 382         if (user_mode(regs))
 383                 mm_flags |= FAULT_FLAG_USER;
 384
 385         if (is_el0_instruction_abort(esr)) {
 386                 vm_flags = VM_EXEC;
 387         } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
 388                 vm_flags = VM_WRITE;
 389                 mm_flags |= FAULT_FLAG_WRITE;
 390         }
 391
 392         if (addr < USER_DS && is_permission_fault(esr, regs, addr)) {
 393                 /* regs->orig_addr_limit may be 0 if we entered from EL0 */
 394                 if (regs->orig_addr_limit == KERNEL_DS)
 395                         die("Accessing user space memory with fs=KERNEL_DS", regs, esr);
 396
 397                 if (is_el1_instruction_abort(esr))
 398                         die("Attempting to execute userspace memory", regs, esr);
 399
 400                 if (!search_exception_tables(regs->pc))
 401                         die("Accessing user space memory outside uaccess.h routines", regs, esr);
 402         }
 403
 404         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
 405
 406         /*
 407          * As per x86, we may deadlock here. However, since the kernel only
 408          * validly references user space from well defined areas of the code,
 409          * we can bug out early if this is from code which shouldn't.
 410          */
 411         if (!down_read_trylock(&mm->mmap_sem)) {
 412                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 413                         goto no_context;
 414 retry:
 415                 down_read(&mm->mmap_sem);
 416         } else {
 417                 /*
 418                  * The above down_read_trylock() might have succeeded in which
 419                  * case, we'll have missed the might_sleep() from down_read().
 420                  */
 421                 might_sleep();
 422 #ifdef CONFIG_DEBUG_VM
 423                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 424                         goto no_context;
 425 #endif
 426         }
 427
 428         fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
 429         major |= fault & VM_FAULT_MAJOR;
 430
 431         if (fault & VM_FAULT_RETRY) {
 432                 /*
 433                  * If we need to retry but a fatal signal is pending,
 434                  * handle the signal first. We do not need to release
 435                  * the mmap_sem because it would already be released
 436                  * in __lock_page_or_retry in mm/filemap.c.
 437                  */
 438                 if (fatal_signal_pending(current)) {
 439                         if (!user_mode(regs))
 440                                 goto no_context;
 441                         return 0;
 442                 }
 443
 444                 /*
 445                  * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
 446                  * starvation.
 447                  */
 448                 if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
 449                         mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
 450                         mm_flags |= FAULT_FLAG_TRIED;
 451                         goto retry;
 452                 }
 453         }
 454         up_read(&mm->mmap_sem);
 455
 456         /*
 457          * Handle the "normal" (no error) case first.
 458          */
 459         if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
 460                               VM_FAULT_BADACCESS)))) {
 461                 /*
 462                  * Major/minor page fault accounting is only done
 463                  * once. If we go through a retry, it is extremely
 464                  * likely that the page will be found in page cache at
 465                  * that point.
 466                  */
 467                 if (major) {
 468                         tsk->maj_flt++;
 469                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
 470                                       addr);
 471                 } else {
 472                         tsk->min_flt++;
 473                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
 474                                       addr);
 475                 }
 476
 477                 return 0;
 478         }
 479
 480         /*
 481          * If we are in kernel mode at this point, we have no context to
 482          * handle this fault with.
 483          */
 484         if (!user_mode(regs))
 485                 goto no_context;
 486
 487         if (fault & VM_FAULT_OOM) {
 488                 /*
 489                  * We ran out of memory, call the OOM killer, and return to
 490                  * userspace (which will retry the fault, or kill us if we got
 491                  * oom-killed).
 492                  */
 493                 pagefault_out_of_memory();
 494                 return 0;
 495         }
 496
 497         if (fault & VM_FAULT_SIGBUS) {
 498                 /*
 499                  * We had some memory, but were unable to successfully fix up
 500                  * this page fault.
 501                  */
 502                 sig = SIGBUS;
 503                 code = BUS_ADRERR;
 504         } else if (fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) {
 505                 sig = SIGBUS;
 506                 code = BUS_MCEERR_AR;
 507         } else {
 508                 /*
 509                  * Something tried to access memory that isn't in our memory
 510                  * map.
 511                  */
 512                 sig = SIGSEGV;
 513                 code = fault == VM_FAULT_BADACCESS ?
 514                         SEGV_ACCERR : SEGV_MAPERR;
 515         }
 516
 517         __do_user_fault(tsk, addr, esr, sig, code, regs, fault);
 518         return 0;
 519
 520 no_context:
 521         __do_kernel_fault(addr, esr, regs);
 522         return 0;
 523 }
 524
 525 /*
 526  * First Level Translation Fault Handler
 527  *
 528  * We enter here because the first level page table doesn't contain a valid
 529  * entry for the address.
 530  *
 531  * If the address is in kernel space (>= TASK_SIZE), then we are probably
 532  * faulting in the vmalloc() area.
 533  *
 534  * If the init_task's first level page tables contains the relevant entry, we
 535  * copy the it to this task.  If not, we send the process a signal, fixup the
 536  * exception, or oops the kernel.
 537  *
 538  * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
 539  * or a critical region, and should only copy the information from the master
 540  * page table, nothing more.
 541  */
 542 static int __kprobes do_translation_fault(unsigned long addr,
 543                                           unsigned int esr,
 544                                           struct pt_regs *regs)
 545 {
 546         if (addr < TASK_SIZE)
 547                 return do_page_fault(addr, esr, regs);
 548
 549         do_bad_area(addr, esr, regs);
 550         return 0;
 551 }
 552
 553 static int do_alignment_fault(unsigned long addr, unsigned int esr,
 554                               struct pt_regs *regs)
 555 {
 556         do_bad_area(addr, esr, regs);
 557         return 0;
 558 }
 559
 560 /*
 561  * This abort handler always returns "fault".
 562  */
 563 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 564 {
 565         return 1;
 566 }
 567
 568 /*
 569  * This abort handler deals with Synchronous External Abort.
 570  * It calls notifiers, and then returns "fault".
 571  */
 572 static int do_sea(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 573 {
 574         struct siginfo info;
 575         const struct fault_info *inf;
 576         int ret = 0;
 577
 578         inf = esr_to_fault_info(esr);
 579         pr_err("Synchronous External Abort: %s (0x%08x) at 0x%016lx\n",
 580                 inf->name, esr, addr);
 581
 582         /*
 583          * Synchronous aborts may interrupt code which had interrupts masked.
 584          * Before calling out into the wider kernel tell the interested
 585          * subsystems.
 586          */
 587         if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) {
 588                 if (interrupts_enabled(regs))
 589                         nmi_enter();
 590
 591                 ret = ghes_notify_sea();
 592
 593                 if (interrupts_enabled(regs))
 594                         nmi_exit();
 595         }
 596
 597         info.si_signo = SIGBUS;
 598         info.si_errno = 0;
 599         info.si_code  = 0;
 600         if (esr & ESR_ELx_FnV)
 601                 info.si_addr = NULL;
 602         else
 603                 info.si_addr  = (void __user *)addr;
 604         arm64_notify_die("", regs, &info, esr);
 605
 606         return ret;
 607 }
 608
 609 static const struct fault_info fault_info[] = {
 610         { do_bad,               SIGBUS,  0,             "ttbr address size fault"       },
 611         { do_bad,               SIGBUS,  0,             "level 1 address size fault"    },
 612         { do_bad,               SIGBUS,  0,             "level 2 address size fault"    },
 613         { do_bad,               SIGBUS,  0,             "level 3 address size fault"    },
 614         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 0 translation fault"     },
 615         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 1 translation fault"     },
 616         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 2 translation fault"     },
 617         { do_page_fault,        SIGSEGV, SEGV_MAPERR,   "level 3 translation fault"     },
 618         { do_bad,               SIGBUS,  0,             "unknown 8"                     },
 619         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 access flag fault"     },
 620         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 access flag fault"     },
 621         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 access flag fault"     },
 622         { do_bad,               SIGBUS,  0,             "unknown 12"                    },
 623         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 permission fault"      },
 624         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 permission fault"      },
 625         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 permission fault"      },
 626         { do_sea,               SIGBUS,  0,             "synchronous external abort"    },
 627         { do_bad,               SIGBUS,  0,             "unknown 17"                    },
 628         { do_bad,               SIGBUS,  0,             "unknown 18"                    },
 629         { do_bad,               SIGBUS,  0,             "unknown 19"                    },
 630         { do_sea,               SIGBUS,  0,             "level 0 (translation table walk)"      },
 631         { do_sea,               SIGBUS,  0,             "level 1 (translation table walk)"      },
 632         { do_sea,               SIGBUS,  0,             "level 2 (translation table walk)"      },
 633         { do_sea,               SIGBUS,  0,             "level 3 (translation table walk)"      },
 634         { do_sea,               SIGBUS,  0,             "synchronous parity or ECC error" },
 635         { do_bad,               SIGBUS,  0,             "unknown 25"                    },
 636         { do_bad,               SIGBUS,  0,             "unknown 26"                    },
 637         { do_bad,               SIGBUS,  0,             "unknown 27"                    },
 638         { do_sea,               SIGBUS,  0,             "level 0 synchronous parity error (translation table walk)"     },
 639         { do_sea,               SIGBUS,  0,             "level 1 synchronous parity error (translation table walk)"     },
 640         { do_sea,               SIGBUS,  0,             "level 2 synchronous parity error (translation table walk)"     },
 641         { do_sea,               SIGBUS,  0,             "level 3 synchronous parity error (translation table walk)"     },
 642         { do_bad,               SIGBUS,  0,             "unknown 32"                    },
 643         { do_alignment_fault,   SIGBUS,  BUS_ADRALN,    "alignment fault"               },
 644         { do_bad,               SIGBUS,  0,             "unknown 34"                    },
 645         { do_bad,               SIGBUS,  0,             "unknown 35"                    },
 646         { do_bad,               SIGBUS,  0,             "unknown 36"                    },
 647         { do_bad,               SIGBUS,  0,             "unknown 37"                    },
 648         { do_bad,               SIGBUS,  0,             "unknown 38"                    },
 649         { do_bad,               SIGBUS,  0,             "unknown 39"                    },
 650         { do_bad,               SIGBUS,  0,             "unknown 40"                    },
 651         { do_bad,               SIGBUS,  0,             "unknown 41"                    },
 652         { do_bad,               SIGBUS,  0,             "unknown 42"                    },
 653         { do_bad,               SIGBUS,  0,             "unknown 43"                    },
 654         { do_bad,               SIGBUS,  0,             "unknown 44"                    },
 655         { do_bad,               SIGBUS,  0,             "unknown 45"                    },
 656         { do_bad,               SIGBUS,  0,             "unknown 46"                    },
 657         { do_bad,               SIGBUS,  0,             "unknown 47"                    },
 658         { do_bad,               SIGBUS,  0,             "TLB conflict abort"            },
 659         { do_bad,               SIGBUS,  0,             "unknown 49"                    },
 660         { do_bad,               SIGBUS,  0,             "unknown 50"                    },
 661         { do_bad,               SIGBUS,  0,             "unknown 51"                    },
 662         { do_bad,               SIGBUS,  0,             "implementation fault (lockdown abort)" },
 663         { do_bad,               SIGBUS,  0,             "implementation fault (unsupported exclusive)" },
 664         { do_bad,               SIGBUS,  0,             "unknown 54"                    },
 665         { do_bad,               SIGBUS,  0,             "unknown 55"                    },
 666         { do_bad,               SIGBUS,  0,             "unknown 56"                    },
 667         { do_bad,               SIGBUS,  0,             "unknown 57"                    },
 668         { do_bad,               SIGBUS,  0,             "unknown 58"                    },
 669         { do_bad,               SIGBUS,  0,             "unknown 59"                    },
 670         { do_bad,               SIGBUS,  0,             "unknown 60"                    },
 671         { do_bad,               SIGBUS,  0,             "section domain fault"          },
 672         { do_bad,               SIGBUS,  0,             "page domain fault"             },
 673         { do_bad,               SIGBUS,  0,             "unknown 63"                    },
 674 };
 675
 676 /*
 677  * Handle Synchronous External Aborts that occur in a guest kernel.
 678  *
 679  * The return value will be zero if the SEA was successfully handled
 680  * and non-zero if there was an error processing the error or there was
 681  * no error to process.
 682  */
 683 int handle_guest_sea(phys_addr_t addr, unsigned int esr)
 684 {
 685         int ret = -ENOENT;
 686
 687         if (IS_ENABLED(CONFIG_ACPI_APEI_SEA))
 688                 ret = ghes_notify_sea();
 689
 690         return ret;
 691 }
 692
 693 /*
 694  * Dispatch a data abort to the relevant handler.
 695  */
 696 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
 697                                          struct pt_regs *regs)
 698 {
 699         const struct fault_info *inf = esr_to_fault_info(esr);
 700         struct siginfo info;
 701
 702         if (!inf->fn(addr, esr, regs))
 703                 return;
 704
 705         pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
 706                  inf->name, esr, addr);
 707
 708         info.si_signo = inf->sig;
 709         info.si_errno = 0;
 710         info.si_code  = inf->code;
 711         info.si_addr  = (void __user *)addr;
 712         arm64_notify_die("", regs, &info, esr);
 713 }
 714
 715 /*
 716  * Handle stack alignment exceptions.
 717  */
 718 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
 719                                            unsigned int esr,
 720                                            struct pt_regs *regs)
 721 {
 722         struct siginfo info;
 723         struct task_struct *tsk = current;
 724
 725         if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
 726                 pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
 727                                     tsk->comm, task_pid_nr(tsk),
 728                                     esr_get_class_string(esr), (void *)regs->pc,
 729                                     (void *)regs->sp);
 730
 731         info.si_signo = SIGBUS;
 732         info.si_errno = 0;
 733         info.si_code  = BUS_ADRALN;
 734         info.si_addr  = (void __user *)addr;
 735         arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
 736 }
 737
 738 int __init early_brk64(unsigned long addr, unsigned int esr,
 739                        struct pt_regs *regs);
 740
 741 /*
 742  * __refdata because early_brk64 is __init, but the reference to it is
 743  * clobbered at arch_initcall time.
 744  * See traps.c and debug-monitors.c:debug_traps_init().
 745  */
 746 static struct fault_info __refdata debug_fault_info[] = {
 747         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware breakpoint"   },
 748         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware single-step"  },
 749         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware watchpoint"   },
 750         { do_bad,       SIGBUS,         0,              "unknown 3"             },
 751         { do_bad,       SIGTRAP,        TRAP_BRKPT,     "aarch32 BKPT"          },
 752         { do_bad,       SIGTRAP,        0,              "aarch32 vector catch"  },
 753         { early_brk64,  SIGTRAP,        TRAP_BRKPT,     "aarch64 BRK"           },
 754         { do_bad,       SIGBUS,         0,              "unknown 7"             },
 755 };
 756
 757 void __init hook_debug_fault_code(int nr,
 758                                   int (*fn)(unsigned long, unsigned int, struct pt_regs *),
 759                                   int sig, int code, const char *name)
 760 {
 761         BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
 762
 763         debug_fault_info[nr].fn         = fn;
 764         debug_fault_info[nr].sig        = sig;
 765         debug_fault_info[nr].code       = code;
 766         debug_fault_info[nr].name       = name;
 767 }
 768
 769 asmlinkage int __exception do_debug_exception(unsigned long addr,
 770                                               unsigned int esr,
 771                                               struct pt_regs *regs)
 772 {
 773         const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
 774         struct siginfo info;
 775         int rv;
 776
 777         /*
 778          * Tell lockdep we disabled irqs in entry.S. Do nothing if they were
 779          * already disabled to preserve the last enabled/disabled addresses.
 780          */
 781         if (interrupts_enabled(regs))
 782                 trace_hardirqs_off();
 783
 784         if (!inf->fn(addr, esr, regs)) {
 785                 rv = 1;
 786         } else {
 787                 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
 788                          inf->name, esr, addr);
 789
 790                 info.si_signo = inf->sig;
 791                 info.si_errno = 0;
 792                 info.si_code  = inf->code;
 793                 info.si_addr  = (void __user *)addr;
 794                 arm64_notify_die("", regs, &info, 0);
 795                 rv = 0;
 796         }
 797
 798         if (interrupts_enabled(regs))
 799                 trace_hardirqs_on();
 800
 801         return rv;
 802 }
 803 NOKPROBE_SYMBOL(do_debug_exception);
 804
 805 #ifdef CONFIG_ARM64_PAN
 806 int cpu_enable_pan(void *__unused)
 807 {
 808         /*
 809          * We modify PSTATE. This won't work from irq context as the PSTATE
 810          * is discarded once we return from the exception.
 811          */
 812         WARN_ON_ONCE(in_interrupt());
 813
 814         config_sctlr_el1(SCTLR_EL1_SPAN, 0);
 815         asm(SET_PSTATE_PAN(1));
 816         return 0;
 817 }
 818 #endif /* CONFIG_ARM64_PAN */