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         /*
 167          * PTE_RDONLY is cleared by default in the asm below, so set it in
 168          * back if necessary (read-only or clean PTE).
 169          */
 170         if (!pte_write(entry) || !pte_sw_dirty(entry))
 171                 pte_val(entry) |= PTE_RDONLY;
 172
 173         /*
 174          * Setting the flags must be done atomically to avoid racing with the
 175          * hardware update of the access/dirty state.
 176          */
 177         asm volatile("//        ptep_set_access_flags\n"
 178         "       prfm    pstl1strm, %2\n"
 179         "1:     ldxr    %0, %2\n"
 180         "       and     %0, %0, %3              // clear PTE_RDONLY\n"
 181         "       orr     %0, %0, %4              // set flags\n"
 182         "       stxr    %w1, %0, %2\n"
 183         "       cbnz    %w1, 1b\n"
 184         : "=&r" (old_pteval), "=&r" (tmp), "+Q" (pte_val(*ptep))
 185         : "L" (~PTE_RDONLY), "r" (pte_val(entry)));
 186
 187         flush_tlb_fix_spurious_fault(vma, address);
 188         return 1;
 189 }
 190 #endif
 191
 192 static bool is_el1_instruction_abort(unsigned int esr)
 193 {
 194         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR;
 195 }
 196
 197 static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs,
 198                                        unsigned long addr)
 199 {
 200         unsigned int ec       = ESR_ELx_EC(esr);
 201         unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE;
 202
 203         if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR)
 204                 return false;
 205
 206         if (fsc_type == ESR_ELx_FSC_PERM)
 207                 return true;
 208
 209         if (addr < USER_DS && system_uses_ttbr0_pan())
 210                 return fsc_type == ESR_ELx_FSC_FAULT &&
 211                         (regs->pstate & PSR_PAN_BIT);
 212
 213         return false;
 214 }
 215
 216 /*
 217  * The kernel tried to access some page that wasn't present.
 218  */
 219 static void __do_kernel_fault(unsigned long addr, unsigned int esr,
 220                               struct pt_regs *regs)
 221 {
 222         const char *msg;
 223
 224         /*
 225          * Are we prepared to handle this kernel fault?
 226          * We are almost certainly not prepared to handle instruction faults.
 227          */
 228         if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
 229                 return;
 230
 231         /*
 232          * No handler, we'll have to terminate things with extreme prejudice.
 233          */
 234         bust_spinlocks(1);
 235
 236         if (is_permission_fault(esr, regs, addr)) {
 237                 if (esr & ESR_ELx_WNR)
 238                         msg = "write to read-only memory";
 239                 else
 240                         msg = "read from unreadable memory";
 241         } else if (addr < PAGE_SIZE) {
 242                 msg = "NULL pointer dereference";
 243         } else {
 244                 msg = "paging request";
 245         }
 246
 247         pr_alert("Unable to handle kernel %s at virtual address %08lx\n", msg,
 248                  addr);
 249
 250         show_pte(addr);
 251         die("Oops", regs, esr);
 252         bust_spinlocks(0);
 253         do_exit(SIGKILL);
 254 }
 255
 256 /*
 257  * Something tried to access memory that isn't in our memory map. User mode
 258  * accesses just cause a SIGSEGV
 259  */
 260 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
 261                             unsigned int esr, unsigned int sig, int code,
 262                             struct pt_regs *regs, int fault)
 263 {
 264         struct siginfo si;
 265         const struct fault_info *inf;
 266         unsigned int lsb = 0;
 267
 268         if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
 269                 inf = esr_to_fault_info(esr);
 270                 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x",
 271                         tsk->comm, task_pid_nr(tsk), inf->name, sig,
 272                         addr, esr);
 273                 print_vma_addr(KERN_CONT ", in ", regs->pc);
 274                 pr_cont("\n");
 275                 __show_regs(regs);
 276         }
 277
 278         tsk->thread.fault_address = addr;
 279         tsk->thread.fault_code = esr;
 280         si.si_signo = sig;
 281         si.si_errno = 0;
 282         si.si_code = code;
 283         si.si_addr = (void __user *)addr;
 284         /*
 285          * Either small page or large page may be poisoned.
 286          * In other words, VM_FAULT_HWPOISON_LARGE and
 287          * VM_FAULT_HWPOISON are mutually exclusive.
 288          */
 289         if (fault & VM_FAULT_HWPOISON_LARGE)
 290                 lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault));
 291         else if (fault & VM_FAULT_HWPOISON)
 292                 lsb = PAGE_SHIFT;
 293         si.si_addr_lsb = lsb;
 294
 295         force_sig_info(sig, &si, tsk);
 296 }
 297
 298 static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 299 {
 300         struct task_struct *tsk = current;
 301         const struct fault_info *inf;
 302
 303         /*
 304          * If we are in kernel mode at this point, we have no context to
 305          * handle this fault with.
 306          */
 307         if (user_mode(regs)) {
 308                 inf = esr_to_fault_info(esr);
 309                 __do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs, 0);
 310         } else
 311                 __do_kernel_fault(addr, esr, regs);
 312 }
 313
 314 #define VM_FAULT_BADMAP         0x010000
 315 #define VM_FAULT_BADACCESS      0x020000
 316
 317 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
 318                            unsigned int mm_flags, unsigned long vm_flags,
 319                            struct task_struct *tsk)
 320 {
 321         struct vm_area_struct *vma;
 322         int fault;
 323
 324         vma = find_vma(mm, addr);
 325         fault = VM_FAULT_BADMAP;
 326         if (unlikely(!vma))
 327                 goto out;
 328         if (unlikely(vma->vm_start > addr))
 329                 goto check_stack;
 330
 331         /*
 332          * Ok, we have a good vm_area for this memory access, so we can handle
 333          * it.
 334          */
 335 good_area:
 336         /*
 337          * Check that the permissions on the VMA allow for the fault which
 338          * occurred.
 339          */
 340         if (!(vma->vm_flags & vm_flags)) {
 341                 fault = VM_FAULT_BADACCESS;
 342                 goto out;
 343         }
 344
 345         return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags);
 346
 347 check_stack:
 348         if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
 349                 goto good_area;
 350 out:
 351         return fault;
 352 }
 353
 354 static bool is_el0_instruction_abort(unsigned int esr)
 355 {
 356         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW;
 357 }
 358
 359 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
 360                                    struct pt_regs *regs)
 361 {
 362         struct task_struct *tsk;
 363         struct mm_struct *mm;
 364         int fault, sig, code, major = 0;
 365         unsigned long vm_flags = VM_READ | VM_WRITE;
 366         unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
 367
 368         if (notify_page_fault(regs, esr))
 369                 return 0;
 370
 371         tsk = current;
 372         mm  = tsk->mm;
 373
 374         /*
 375          * If we're in an interrupt or have no user context, we must not take
 376          * the fault.
 377          */
 378         if (faulthandler_disabled() || !mm)
 379                 goto no_context;
 380
 381         if (user_mode(regs))
 382                 mm_flags |= FAULT_FLAG_USER;
 383
 384         if (is_el0_instruction_abort(esr)) {
 385                 vm_flags = VM_EXEC;
 386         } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
 387                 vm_flags = VM_WRITE;
 388                 mm_flags |= FAULT_FLAG_WRITE;
 389         }
 390
 391         if (addr < USER_DS && is_permission_fault(esr, regs, addr)) {
 392                 /* regs->orig_addr_limit may be 0 if we entered from EL0 */
 393                 if (regs->orig_addr_limit == KERNEL_DS)
 394                         die("Accessing user space memory with fs=KERNEL_DS", regs, esr);
 395
 396                 if (is_el1_instruction_abort(esr))
 397                         die("Attempting to execute userspace memory", regs, esr);
 398
 399                 if (!search_exception_tables(regs->pc))
 400                         die("Accessing user space memory outside uaccess.h routines", regs, esr);
 401         }
 402
 403         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
 404
 405         /*
 406          * As per x86, we may deadlock here. However, since the kernel only
 407          * validly references user space from well defined areas of the code,
 408          * we can bug out early if this is from code which shouldn't.
 409          */
 410         if (!down_read_trylock(&mm->mmap_sem)) {
 411                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 412                         goto no_context;
 413 retry:
 414                 down_read(&mm->mmap_sem);
 415         } else {
 416                 /*
 417                  * The above down_read_trylock() might have succeeded in which
 418                  * case, we'll have missed the might_sleep() from down_read().
 419                  */
 420                 might_sleep();
 421 #ifdef CONFIG_DEBUG_VM
 422                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 423                         goto no_context;
 424 #endif
 425         }
 426
 427         fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
 428         major |= fault & VM_FAULT_MAJOR;
 429
 430         if (fault & VM_FAULT_RETRY) {
 431                 /*
 432                  * If we need to retry but a fatal signal is pending,
 433                  * handle the signal first. We do not need to release
 434                  * the mmap_sem because it would already be released
 435                  * in __lock_page_or_retry in mm/filemap.c.
 436                  */
 437                 if (fatal_signal_pending(current))
 438                         return 0;
 439
 440                 /*
 441                  * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
 442                  * starvation.
 443                  */
 444                 if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
 445                         mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
 446                         mm_flags |= FAULT_FLAG_TRIED;
 447                         goto retry;
 448                 }
 449         }
 450         up_read(&mm->mmap_sem);
 451
 452         /*
 453          * Handle the "normal" (no error) case first.
 454          */
 455         if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
 456                               VM_FAULT_BADACCESS)))) {
 457                 /*
 458                  * Major/minor page fault accounting is only done
 459                  * once. If we go through a retry, it is extremely
 460                  * likely that the page will be found in page cache at
 461                  * that point.
 462                  */
 463                 if (major) {
 464                         tsk->maj_flt++;
 465                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
 466                                       addr);
 467                 } else {
 468                         tsk->min_flt++;
 469                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
 470                                       addr);
 471                 }
 472
 473                 return 0;
 474         }
 475
 476         /*
 477          * If we are in kernel mode at this point, we have no context to
 478          * handle this fault with.
 479          */
 480         if (!user_mode(regs))
 481                 goto no_context;
 482
 483         if (fault & VM_FAULT_OOM) {
 484                 /*
 485                  * We ran out of memory, call the OOM killer, and return to
 486                  * userspace (which will retry the fault, or kill us if we got
 487                  * oom-killed).
 488                  */
 489                 pagefault_out_of_memory();
 490                 return 0;
 491         }
 492
 493         if (fault & VM_FAULT_SIGBUS) {
 494                 /*
 495                  * We had some memory, but were unable to successfully fix up
 496                  * this page fault.
 497                  */
 498                 sig = SIGBUS;
 499                 code = BUS_ADRERR;
 500         } else if (fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) {
 501                 sig = SIGBUS;
 502                 code = BUS_MCEERR_AR;
 503         } else {
 504                 /*
 505                  * Something tried to access memory that isn't in our memory
 506                  * map.
 507                  */
 508                 sig = SIGSEGV;
 509                 code = fault == VM_FAULT_BADACCESS ?
 510                         SEGV_ACCERR : SEGV_MAPERR;
 511         }
 512
 513         __do_user_fault(tsk, addr, esr, sig, code, regs, fault);
 514         return 0;
 515
 516 no_context:
 517         __do_kernel_fault(addr, esr, regs);
 518         return 0;
 519 }
 520
 521 /*
 522  * First Level Translation Fault Handler
 523  *
 524  * We enter here because the first level page table doesn't contain a valid
 525  * entry for the address.
 526  *
 527  * If the address is in kernel space (>= TASK_SIZE), then we are probably
 528  * faulting in the vmalloc() area.
 529  *
 530  * If the init_task's first level page tables contains the relevant entry, we
 531  * copy the it to this task.  If not, we send the process a signal, fixup the
 532  * exception, or oops the kernel.
 533  *
 534  * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
 535  * or a critical region, and should only copy the information from the master
 536  * page table, nothing more.
 537  */
 538 static int __kprobes do_translation_fault(unsigned long addr,
 539                                           unsigned int esr,
 540                                           struct pt_regs *regs)
 541 {
 542         if (addr < TASK_SIZE)
 543                 return do_page_fault(addr, esr, regs);
 544
 545         do_bad_area(addr, esr, regs);
 546         return 0;
 547 }
 548
 549 static int do_alignment_fault(unsigned long addr, unsigned int esr,
 550                               struct pt_regs *regs)
 551 {
 552         do_bad_area(addr, esr, regs);
 553         return 0;
 554 }
 555
 556 /*
 557  * This abort handler always returns "fault".
 558  */
 559 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 560 {
 561         return 1;
 562 }
 563
 564 /*
 565  * This abort handler deals with Synchronous External Abort.
 566  * It calls notifiers, and then returns "fault".
 567  */
 568 static int do_sea(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 569 {
 570         struct siginfo info;
 571         const struct fault_info *inf;
 572         int ret = 0;
 573
 574         inf = esr_to_fault_info(esr);
 575         pr_err("Synchronous External Abort: %s (0x%08x) at 0x%016lx\n",
 576                 inf->name, esr, addr);
 577
 578         /*
 579          * Synchronous aborts may interrupt code which had interrupts masked.
 580          * Before calling out into the wider kernel tell the interested
 581          * subsystems.
 582          */
 583         if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) {
 584                 if (interrupts_enabled(regs))
 585                         nmi_enter();
 586
 587                 ret = ghes_notify_sea();
 588
 589                 if (interrupts_enabled(regs))
 590                         nmi_exit();
 591         }
 592
 593         info.si_signo = SIGBUS;
 594         info.si_errno = 0;
 595         info.si_code  = 0;
 596         if (esr & ESR_ELx_FnV)
 597                 info.si_addr = NULL;
 598         else
 599                 info.si_addr  = (void __user *)addr;
 600         arm64_notify_die("", regs, &info, esr);
 601
 602         return ret;
 603 }
 604
 605 static const struct fault_info fault_info[] = {
 606         { do_bad,               SIGBUS,  0,             "ttbr address size fault"       },
 607         { do_bad,               SIGBUS,  0,             "level 1 address size fault"    },
 608         { do_bad,               SIGBUS,  0,             "level 2 address size fault"    },
 609         { do_bad,               SIGBUS,  0,             "level 3 address size fault"    },
 610         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 0 translation fault"     },
 611         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 1 translation fault"     },
 612         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 2 translation fault"     },
 613         { do_page_fault,        SIGSEGV, SEGV_MAPERR,   "level 3 translation fault"     },
 614         { do_bad,               SIGBUS,  0,             "unknown 8"                     },
 615         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 access flag fault"     },
 616         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 access flag fault"     },
 617         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 access flag fault"     },
 618         { do_bad,               SIGBUS,  0,             "unknown 12"                    },
 619         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 permission fault"      },
 620         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 permission fault"      },
 621         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 permission fault"      },
 622         { do_sea,               SIGBUS,  0,             "synchronous external abort"    },
 623         { do_bad,               SIGBUS,  0,             "unknown 17"                    },
 624         { do_bad,               SIGBUS,  0,             "unknown 18"                    },
 625         { do_bad,               SIGBUS,  0,             "unknown 19"                    },
 626         { do_sea,               SIGBUS,  0,             "level 0 (translation table walk)"      },
 627         { do_sea,               SIGBUS,  0,             "level 1 (translation table walk)"      },
 628         { do_sea,               SIGBUS,  0,             "level 2 (translation table walk)"      },
 629         { do_sea,               SIGBUS,  0,             "level 3 (translation table walk)"      },
 630         { do_sea,               SIGBUS,  0,             "synchronous parity or ECC error" },
 631         { do_bad,               SIGBUS,  0,             "unknown 25"                    },
 632         { do_bad,               SIGBUS,  0,             "unknown 26"                    },
 633         { do_bad,               SIGBUS,  0,             "unknown 27"                    },
 634         { do_sea,               SIGBUS,  0,             "level 0 synchronous parity error (translation table walk)"     },
 635         { do_sea,               SIGBUS,  0,             "level 1 synchronous parity error (translation table walk)"     },
 636         { do_sea,               SIGBUS,  0,             "level 2 synchronous parity error (translation table walk)"     },
 637         { do_sea,               SIGBUS,  0,             "level 3 synchronous parity error (translation table walk)"     },
 638         { do_bad,               SIGBUS,  0,             "unknown 32"                    },
 639         { do_alignment_fault,   SIGBUS,  BUS_ADRALN,    "alignment fault"               },
 640         { do_bad,               SIGBUS,  0,             "unknown 34"                    },
 641         { do_bad,               SIGBUS,  0,             "unknown 35"                    },
 642         { do_bad,               SIGBUS,  0,             "unknown 36"                    },
 643         { do_bad,               SIGBUS,  0,             "unknown 37"                    },
 644         { do_bad,               SIGBUS,  0,             "unknown 38"                    },
 645         { do_bad,               SIGBUS,  0,             "unknown 39"                    },
 646         { do_bad,               SIGBUS,  0,             "unknown 40"                    },
 647         { do_bad,               SIGBUS,  0,             "unknown 41"                    },
 648         { do_bad,               SIGBUS,  0,             "unknown 42"                    },
 649         { do_bad,               SIGBUS,  0,             "unknown 43"                    },
 650         { do_bad,               SIGBUS,  0,             "unknown 44"                    },
 651         { do_bad,               SIGBUS,  0,             "unknown 45"                    },
 652         { do_bad,               SIGBUS,  0,             "unknown 46"                    },
 653         { do_bad,               SIGBUS,  0,             "unknown 47"                    },
 654         { do_bad,               SIGBUS,  0,             "TLB conflict abort"            },
 655         { do_bad,               SIGBUS,  0,             "unknown 49"                    },
 656         { do_bad,               SIGBUS,  0,             "unknown 50"                    },
 657         { do_bad,               SIGBUS,  0,             "unknown 51"                    },
 658         { do_bad,               SIGBUS,  0,             "implementation fault (lockdown abort)" },
 659         { do_bad,               SIGBUS,  0,             "implementation fault (unsupported exclusive)" },
 660         { do_bad,               SIGBUS,  0,             "unknown 54"                    },
 661         { do_bad,               SIGBUS,  0,             "unknown 55"                    },
 662         { do_bad,               SIGBUS,  0,             "unknown 56"                    },
 663         { do_bad,               SIGBUS,  0,             "unknown 57"                    },
 664         { do_bad,               SIGBUS,  0,             "unknown 58"                    },
 665         { do_bad,               SIGBUS,  0,             "unknown 59"                    },
 666         { do_bad,               SIGBUS,  0,             "unknown 60"                    },
 667         { do_bad,               SIGBUS,  0,             "section domain fault"          },
 668         { do_bad,               SIGBUS,  0,             "page domain fault"             },
 669         { do_bad,               SIGBUS,  0,             "unknown 63"                    },
 670 };
 671
 672 /*
 673  * Handle Synchronous External Aborts that occur in a guest kernel.
 674  *
 675  * The return value will be zero if the SEA was successfully handled
 676  * and non-zero if there was an error processing the error or there was
 677  * no error to process.
 678  */
 679 int handle_guest_sea(phys_addr_t addr, unsigned int esr)
 680 {
 681         int ret = -ENOENT;
 682
 683         if (IS_ENABLED(CONFIG_ACPI_APEI_SEA))
 684                 ret = ghes_notify_sea();
 685
 686         return ret;
 687 }
 688
 689 /*
 690  * Dispatch a data abort to the relevant handler.
 691  */
 692 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
 693                                          struct pt_regs *regs)
 694 {
 695         const struct fault_info *inf = esr_to_fault_info(esr);
 696         struct siginfo info;
 697
 698         if (!inf->fn(addr, esr, regs))
 699                 return;
 700
 701         pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
 702                  inf->name, esr, 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 /*
 712  * Handle stack alignment exceptions.
 713  */
 714 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
 715                                            unsigned int esr,
 716                                            struct pt_regs *regs)
 717 {
 718         struct siginfo info;
 719         struct task_struct *tsk = current;
 720
 721         if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
 722                 pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
 723                                     tsk->comm, task_pid_nr(tsk),
 724                                     esr_get_class_string(esr), (void *)regs->pc,
 725                                     (void *)regs->sp);
 726
 727         info.si_signo = SIGBUS;
 728         info.si_errno = 0;
 729         info.si_code  = BUS_ADRALN;
 730         info.si_addr  = (void __user *)addr;
 731         arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
 732 }
 733
 734 int __init early_brk64(unsigned long addr, unsigned int esr,
 735                        struct pt_regs *regs);
 736
 737 /*
 738  * __refdata because early_brk64 is __init, but the reference to it is
 739  * clobbered at arch_initcall time.
 740  * See traps.c and debug-monitors.c:debug_traps_init().
 741  */
 742 static struct fault_info __refdata debug_fault_info[] = {
 743         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware breakpoint"   },
 744         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware single-step"  },
 745         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware watchpoint"   },
 746         { do_bad,       SIGBUS,         0,              "unknown 3"             },
 747         { do_bad,       SIGTRAP,        TRAP_BRKPT,     "aarch32 BKPT"          },
 748         { do_bad,       SIGTRAP,        0,              "aarch32 vector catch"  },
 749         { early_brk64,  SIGTRAP,        TRAP_BRKPT,     "aarch64 BRK"           },
 750         { do_bad,       SIGBUS,         0,              "unknown 7"             },
 751 };
 752
 753 void __init hook_debug_fault_code(int nr,
 754                                   int (*fn)(unsigned long, unsigned int, struct pt_regs *),
 755                                   int sig, int code, const char *name)
 756 {
 757         BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
 758
 759         debug_fault_info[nr].fn         = fn;
 760         debug_fault_info[nr].sig        = sig;
 761         debug_fault_info[nr].code       = code;
 762         debug_fault_info[nr].name       = name;
 763 }
 764
 765 asmlinkage int __exception do_debug_exception(unsigned long addr,
 766                                               unsigned int esr,
 767                                               struct pt_regs *regs)
 768 {
 769         const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
 770         struct siginfo info;
 771         int rv;
 772
 773         /*
 774          * Tell lockdep we disabled irqs in entry.S. Do nothing if they were
 775          * already disabled to preserve the last enabled/disabled addresses.
 776          */
 777         if (interrupts_enabled(regs))
 778                 trace_hardirqs_off();
 779
 780         if (!inf->fn(addr, esr, regs)) {
 781                 rv = 1;
 782         } else {
 783                 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
 784                          inf->name, esr, addr);
 785
 786                 info.si_signo = inf->sig;
 787                 info.si_errno = 0;
 788                 info.si_code  = inf->code;
 789                 info.si_addr  = (void __user *)addr;
 790                 arm64_notify_die("", regs, &info, 0);
 791                 rv = 0;
 792         }
 793
 794         if (interrupts_enabled(regs))
 795                 trace_hardirqs_on();
 796
 797         return rv;
 798 }
 799 NOKPROBE_SYMBOL(do_debug_exception);
 800
 801 #ifdef CONFIG_ARM64_PAN
 802 int cpu_enable_pan(void *__unused)
 803 {
 804         /*
 805          * We modify PSTATE. This won't work from irq context as the PSTATE
 806          * is discarded once we return from the exception.
 807          */
 808         WARN_ON_ONCE(in_interrupt());
 809
 810         config_sctlr_el1(SCTLR_EL1_SPAN, 0);
 811         asm(SET_PSTATE_PAN(1));
 812         return 0;
 813 }
 814 #endif /* CONFIG_ARM64_PAN */