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
  35 #include <asm/bug.h>
  36 #include <asm/cpufeature.h>
  37 #include <asm/exception.h>
  38 #include <asm/debug-monitors.h>
  39 #include <asm/esr.h>
  40 #include <asm/sysreg.h>
  41 #include <asm/system_misc.h>
  42 #include <asm/pgtable.h>
  43 #include <asm/tlbflush.h>
  44
  45 struct fault_info {
  46         int     (*fn)(unsigned long addr, unsigned int esr,
  47                       struct pt_regs *regs);
  48         int     sig;
  49         int     code;
  50         const char *name;
  51 };
  52
  53 static const struct fault_info fault_info[];
  54
  55 static inline const struct fault_info *esr_to_fault_info(unsigned int esr)
  56 {
  57         return fault_info + (esr & 63);
  58 }
  59
  60 #ifdef CONFIG_KPROBES
  61 static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
  62 {
  63         int ret = 0;
  64
  65         /* kprobe_running() needs smp_processor_id() */
  66         if (!user_mode(regs)) {
  67                 preempt_disable();
  68                 if (kprobe_running() && kprobe_fault_handler(regs, esr))
  69                         ret = 1;
  70                 preempt_enable();
  71         }
  72
  73         return ret;
  74 }
  75 #else
  76 static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
  77 {
  78         return 0;
  79 }
  80 #endif
  81
  82 /*
  83  * Dump out the page tables associated with 'addr' in mm 'mm'.
  84  */
  85 void show_pte(struct mm_struct *mm, unsigned long addr)
  86 {
  87         pgd_t *pgd;
  88
  89         if (!mm)
  90                 mm = &init_mm;
  91
  92         pr_alert("pgd = %p\n", mm->pgd);
  93         pgd = pgd_offset(mm, addr);
  94         pr_alert("[%08lx] *pgd=%016llx", addr, pgd_val(*pgd));
  95
  96         do {
  97                 pud_t *pud;
  98                 pmd_t *pmd;
  99                 pte_t *pte;
 100
 101                 if (pgd_none(*pgd) || pgd_bad(*pgd))
 102                         break;
 103
 104                 pud = pud_offset(pgd, addr);
 105                 pr_cont(", *pud=%016llx", pud_val(*pud));
 106                 if (pud_none(*pud) || pud_bad(*pud))
 107                         break;
 108
 109                 pmd = pmd_offset(pud, addr);
 110                 pr_cont(", *pmd=%016llx", pmd_val(*pmd));
 111                 if (pmd_none(*pmd) || pmd_bad(*pmd))
 112                         break;
 113
 114                 pte = pte_offset_map(pmd, addr);
 115                 pr_cont(", *pte=%016llx", pte_val(*pte));
 116                 pte_unmap(pte);
 117         } while(0);
 118
 119         pr_cont("\n");
 120 }
 121
 122 #ifdef CONFIG_ARM64_HW_AFDBM
 123 /*
 124  * This function sets the access flags (dirty, accessed), as well as write
 125  * permission, and only to a more permissive setting.
 126  *
 127  * It needs to cope with hardware update of the accessed/dirty state by other
 128  * agents in the system and can safely skip the __sync_icache_dcache() call as,
 129  * like set_pte_at(), the PTE is never changed from no-exec to exec here.
 130  *
 131  * Returns whether or not the PTE actually changed.
 132  */
 133 int ptep_set_access_flags(struct vm_area_struct *vma,
 134                           unsigned long address, pte_t *ptep,
 135                           pte_t entry, int dirty)
 136 {
 137         pteval_t old_pteval;
 138         unsigned int tmp;
 139
 140         if (pte_same(*ptep, entry))
 141                 return 0;
 142
 143         /* only preserve the access flags and write permission */
 144         pte_val(entry) &= PTE_AF | PTE_WRITE | PTE_DIRTY;
 145
 146         /*
 147          * PTE_RDONLY is cleared by default in the asm below, so set it in
 148          * back if necessary (read-only or clean PTE).
 149          */
 150         if (!pte_write(entry) || !pte_sw_dirty(entry))
 151                 pte_val(entry) |= PTE_RDONLY;
 152
 153         /*
 154          * Setting the flags must be done atomically to avoid racing with the
 155          * hardware update of the access/dirty state.
 156          */
 157         asm volatile("//        ptep_set_access_flags\n"
 158         "       prfm    pstl1strm, %2\n"
 159         "1:     ldxr    %0, %2\n"
 160         "       and     %0, %0, %3              // clear PTE_RDONLY\n"
 161         "       orr     %0, %0, %4              // set flags\n"
 162         "       stxr    %w1, %0, %2\n"
 163         "       cbnz    %w1, 1b\n"
 164         : "=&r" (old_pteval), "=&r" (tmp), "+Q" (pte_val(*ptep))
 165         : "L" (~PTE_RDONLY), "r" (pte_val(entry)));
 166
 167         flush_tlb_fix_spurious_fault(vma, address);
 168         return 1;
 169 }
 170 #endif
 171
 172 static bool is_el1_instruction_abort(unsigned int esr)
 173 {
 174         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR;
 175 }
 176
 177 /*
 178  * The kernel tried to access some page that wasn't present.
 179  */
 180 static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr,
 181                               unsigned int esr, struct pt_regs *regs)
 182 {
 183         /*
 184          * Are we prepared to handle this kernel fault?
 185          * We are almost certainly not prepared to handle instruction faults.
 186          */
 187         if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
 188                 return;
 189
 190         /*
 191          * No handler, we'll have to terminate things with extreme prejudice.
 192          */
 193         bust_spinlocks(1);
 194         pr_alert("Unable to handle kernel %s at virtual address %08lx\n",
 195                  (addr < PAGE_SIZE) ? "NULL pointer dereference" :
 196                  "paging request", addr);
 197
 198         show_pte(mm, addr);
 199         die("Oops", regs, esr);
 200         bust_spinlocks(0);
 201         do_exit(SIGKILL);
 202 }
 203
 204 /*
 205  * Something tried to access memory that isn't in our memory map. User mode
 206  * accesses just cause a SIGSEGV
 207  */
 208 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
 209                             unsigned int esr, unsigned int sig, int code,
 210                             struct pt_regs *regs)
 211 {
 212         struct siginfo si;
 213         const struct fault_info *inf;
 214
 215         if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
 216                 inf = esr_to_fault_info(esr);
 217                 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n",
 218                         tsk->comm, task_pid_nr(tsk), inf->name, sig,
 219                         addr, esr);
 220                 show_pte(tsk->mm, addr);
 221                 show_regs(regs);
 222         }
 223
 224         tsk->thread.fault_address = addr;
 225         tsk->thread.fault_code = esr;
 226         si.si_signo = sig;
 227         si.si_errno = 0;
 228         si.si_code = code;
 229         si.si_addr = (void __user *)addr;
 230         force_sig_info(sig, &si, tsk);
 231 }
 232
 233 static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 234 {
 235         struct task_struct *tsk = current;
 236         struct mm_struct *mm = tsk->active_mm;
 237         const struct fault_info *inf;
 238
 239         /*
 240          * If we are in kernel mode at this point, we have no context to
 241          * handle this fault with.
 242          */
 243         if (user_mode(regs)) {
 244                 inf = esr_to_fault_info(esr);
 245                 __do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs);
 246         } else
 247                 __do_kernel_fault(mm, addr, esr, regs);
 248 }
 249
 250 #define VM_FAULT_BADMAP         0x010000
 251 #define VM_FAULT_BADACCESS      0x020000
 252
 253 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
 254                            unsigned int mm_flags, unsigned long vm_flags,
 255                            struct task_struct *tsk)
 256 {
 257         struct vm_area_struct *vma;
 258         int fault;
 259
 260         vma = find_vma(mm, addr);
 261         fault = VM_FAULT_BADMAP;
 262         if (unlikely(!vma))
 263                 goto out;
 264         if (unlikely(vma->vm_start > addr))
 265                 goto check_stack;
 266
 267         /*
 268          * Ok, we have a good vm_area for this memory access, so we can handle
 269          * it.
 270          */
 271 good_area:
 272         /*
 273          * Check that the permissions on the VMA allow for the fault which
 274          * occurred.
 275          */
 276         if (!(vma->vm_flags & vm_flags)) {
 277                 fault = VM_FAULT_BADACCESS;
 278                 goto out;
 279         }
 280
 281         return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags);
 282
 283 check_stack:
 284         if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
 285                 goto good_area;
 286 out:
 287         return fault;
 288 }
 289
 290 static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs)
 291 {
 292         unsigned int ec       = ESR_ELx_EC(esr);
 293         unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE;
 294
 295         if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR)
 296                 return false;
 297
 298         if (system_uses_ttbr0_pan())
 299                 return fsc_type == ESR_ELx_FSC_FAULT &&
 300                         (regs->pstate & PSR_PAN_BIT);
 301         else
 302                 return fsc_type == ESR_ELx_FSC_PERM;
 303 }
 304
 305 static bool is_el0_instruction_abort(unsigned int esr)
 306 {
 307         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW;
 308 }
 309
 310 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
 311                                    struct pt_regs *regs)
 312 {
 313         struct task_struct *tsk;
 314         struct mm_struct *mm;
 315         int fault, sig, code;
 316         unsigned long vm_flags = VM_READ | VM_WRITE;
 317         unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
 318
 319         if (notify_page_fault(regs, esr))
 320                 return 0;
 321
 322         tsk = current;
 323         mm  = tsk->mm;
 324
 325         /*
 326          * If we're in an interrupt or have no user context, we must not take
 327          * the fault.
 328          */
 329         if (faulthandler_disabled() || !mm)
 330                 goto no_context;
 331
 332         if (user_mode(regs))
 333                 mm_flags |= FAULT_FLAG_USER;
 334
 335         if (is_el0_instruction_abort(esr)) {
 336                 vm_flags = VM_EXEC;
 337         } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
 338                 vm_flags = VM_WRITE;
 339                 mm_flags |= FAULT_FLAG_WRITE;
 340         }
 341
 342         if (addr < USER_DS && is_permission_fault(esr, regs)) {
 343                 /* regs->orig_addr_limit may be 0 if we entered from EL0 */
 344                 if (regs->orig_addr_limit == KERNEL_DS)
 345                         die("Accessing user space memory with fs=KERNEL_DS", regs, esr);
 346
 347                 if (is_el1_instruction_abort(esr))
 348                         die("Attempting to execute userspace memory", regs, esr);
 349
 350                 if (!search_exception_tables(regs->pc))
 351                         die("Accessing user space memory outside uaccess.h routines", regs, esr);
 352         }
 353
 354         /*
 355          * As per x86, we may deadlock here. However, since the kernel only
 356          * validly references user space from well defined areas of the code,
 357          * we can bug out early if this is from code which shouldn't.
 358          */
 359         if (!down_read_trylock(&mm->mmap_sem)) {
 360                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 361                         goto no_context;
 362 retry:
 363                 down_read(&mm->mmap_sem);
 364         } else {
 365                 /*
 366                  * The above down_read_trylock() might have succeeded in which
 367                  * case, we'll have missed the might_sleep() from down_read().
 368                  */
 369                 might_sleep();
 370 #ifdef CONFIG_DEBUG_VM
 371                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 372                         goto no_context;
 373 #endif
 374         }
 375
 376         fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
 377
 378         /*
 379          * If we need to retry but a fatal signal is pending, handle the
 380          * signal first. We do not need to release the mmap_sem because it
 381          * would already be released in __lock_page_or_retry in mm/filemap.c.
 382          */
 383         if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
 384                 return 0;
 385
 386         /*
 387          * Major/minor page fault accounting is only done on the initial
 388          * attempt. If we go through a retry, it is extremely likely that the
 389          * page will be found in page cache at that point.
 390          */
 391
 392         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
 393         if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
 394                 if (fault & VM_FAULT_MAJOR) {
 395                         tsk->maj_flt++;
 396                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
 397                                       addr);
 398                 } else {
 399                         tsk->min_flt++;
 400                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
 401                                       addr);
 402                 }
 403                 if (fault & VM_FAULT_RETRY) {
 404                         /*
 405                          * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
 406                          * starvation.
 407                          */
 408                         mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
 409                         mm_flags |= FAULT_FLAG_TRIED;
 410                         goto retry;
 411                 }
 412         }
 413
 414         up_read(&mm->mmap_sem);
 415
 416         /*
 417          * Handle the "normal" case first - VM_FAULT_MAJOR
 418          */
 419         if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
 420                               VM_FAULT_BADACCESS))))
 421                 return 0;
 422
 423         /*
 424          * If we are in kernel mode at this point, we have no context to
 425          * handle this fault with.
 426          */
 427         if (!user_mode(regs))
 428                 goto no_context;
 429
 430         if (fault & VM_FAULT_OOM) {
 431                 /*
 432                  * We ran out of memory, call the OOM killer, and return to
 433                  * userspace (which will retry the fault, or kill us if we got
 434                  * oom-killed).
 435                  */
 436                 pagefault_out_of_memory();
 437                 return 0;
 438         }
 439
 440         if (fault & VM_FAULT_SIGBUS) {
 441                 /*
 442                  * We had some memory, but were unable to successfully fix up
 443                  * this page fault.
 444                  */
 445                 sig = SIGBUS;
 446                 code = BUS_ADRERR;
 447         } else {
 448                 /*
 449                  * Something tried to access memory that isn't in our memory
 450                  * map.
 451                  */
 452                 sig = SIGSEGV;
 453                 code = fault == VM_FAULT_BADACCESS ?
 454                         SEGV_ACCERR : SEGV_MAPERR;
 455         }
 456
 457         __do_user_fault(tsk, addr, esr, sig, code, regs);
 458         return 0;
 459
 460 no_context:
 461         __do_kernel_fault(mm, addr, esr, regs);
 462         return 0;
 463 }
 464
 465 /*
 466  * First Level Translation Fault Handler
 467  *
 468  * We enter here because the first level page table doesn't contain a valid
 469  * entry for the address.
 470  *
 471  * If the address is in kernel space (>= TASK_SIZE), then we are probably
 472  * faulting in the vmalloc() area.
 473  *
 474  * If the init_task's first level page tables contains the relevant entry, we
 475  * copy the it to this task.  If not, we send the process a signal, fixup the
 476  * exception, or oops the kernel.
 477  *
 478  * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
 479  * or a critical region, and should only copy the information from the master
 480  * page table, nothing more.
 481  */
 482 static int __kprobes do_translation_fault(unsigned long addr,
 483                                           unsigned int esr,
 484                                           struct pt_regs *regs)
 485 {
 486         if (addr < TASK_SIZE)
 487                 return do_page_fault(addr, esr, regs);
 488
 489         do_bad_area(addr, esr, regs);
 490         return 0;
 491 }
 492
 493 static int do_alignment_fault(unsigned long addr, unsigned int esr,
 494                               struct pt_regs *regs)
 495 {
 496         do_bad_area(addr, esr, regs);
 497         return 0;
 498 }
 499
 500 /*
 501  * This abort handler always returns "fault".
 502  */
 503 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 504 {
 505         return 1;
 506 }
 507
 508 static const struct fault_info fault_info[] = {
 509         { do_bad,               SIGBUS,  0,             "ttbr address size fault"       },
 510         { do_bad,               SIGBUS,  0,             "level 1 address size fault"    },
 511         { do_bad,               SIGBUS,  0,             "level 2 address size fault"    },
 512         { do_bad,               SIGBUS,  0,             "level 3 address size fault"    },
 513         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 0 translation fault"     },
 514         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 1 translation fault"     },
 515         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 2 translation fault"     },
 516         { do_page_fault,        SIGSEGV, SEGV_MAPERR,   "level 3 translation fault"     },
 517         { do_bad,               SIGBUS,  0,             "unknown 8"                     },
 518         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 access flag fault"     },
 519         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 access flag fault"     },
 520         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 access flag fault"     },
 521         { do_bad,               SIGBUS,  0,             "unknown 12"                    },
 522         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 permission fault"      },
 523         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 permission fault"      },
 524         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 permission fault"      },
 525         { do_bad,               SIGBUS,  0,             "synchronous external abort"    },
 526         { do_bad,               SIGBUS,  0,             "unknown 17"                    },
 527         { do_bad,               SIGBUS,  0,             "unknown 18"                    },
 528         { do_bad,               SIGBUS,  0,             "unknown 19"                    },
 529         { do_bad,               SIGBUS,  0,             "synchronous external abort (translation table walk)" },
 530         { do_bad,               SIGBUS,  0,             "synchronous external abort (translation table walk)" },
 531         { do_bad,               SIGBUS,  0,             "synchronous external abort (translation table walk)" },
 532         { do_bad,               SIGBUS,  0,             "synchronous external abort (translation table walk)" },
 533         { do_bad,               SIGBUS,  0,             "synchronous parity error"      },
 534         { do_bad,               SIGBUS,  0,             "unknown 25"                    },
 535         { do_bad,               SIGBUS,  0,             "unknown 26"                    },
 536         { do_bad,               SIGBUS,  0,             "unknown 27"                    },
 537         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk)" },
 538         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk)" },
 539         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk)" },
 540         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk)" },
 541         { do_bad,               SIGBUS,  0,             "unknown 32"                    },
 542         { do_alignment_fault,   SIGBUS,  BUS_ADRALN,    "alignment fault"               },
 543         { do_bad,               SIGBUS,  0,             "unknown 34"                    },
 544         { do_bad,               SIGBUS,  0,             "unknown 35"                    },
 545         { do_bad,               SIGBUS,  0,             "unknown 36"                    },
 546         { do_bad,               SIGBUS,  0,             "unknown 37"                    },
 547         { do_bad,               SIGBUS,  0,             "unknown 38"                    },
 548         { do_bad,               SIGBUS,  0,             "unknown 39"                    },
 549         { do_bad,               SIGBUS,  0,             "unknown 40"                    },
 550         { do_bad,               SIGBUS,  0,             "unknown 41"                    },
 551         { do_bad,               SIGBUS,  0,             "unknown 42"                    },
 552         { do_bad,               SIGBUS,  0,             "unknown 43"                    },
 553         { do_bad,               SIGBUS,  0,             "unknown 44"                    },
 554         { do_bad,               SIGBUS,  0,             "unknown 45"                    },
 555         { do_bad,               SIGBUS,  0,             "unknown 46"                    },
 556         { do_bad,               SIGBUS,  0,             "unknown 47"                    },
 557         { do_bad,               SIGBUS,  0,             "TLB conflict abort"            },
 558         { do_bad,               SIGBUS,  0,             "unknown 49"                    },
 559         { do_bad,               SIGBUS,  0,             "unknown 50"                    },
 560         { do_bad,               SIGBUS,  0,             "unknown 51"                    },
 561         { do_bad,               SIGBUS,  0,             "implementation fault (lockdown abort)" },
 562         { do_bad,               SIGBUS,  0,             "implementation fault (unsupported exclusive)" },
 563         { do_bad,               SIGBUS,  0,             "unknown 54"                    },
 564         { do_bad,               SIGBUS,  0,             "unknown 55"                    },
 565         { do_bad,               SIGBUS,  0,             "unknown 56"                    },
 566         { do_bad,               SIGBUS,  0,             "unknown 57"                    },
 567         { do_bad,               SIGBUS,  0,             "unknown 58"                    },
 568         { do_bad,               SIGBUS,  0,             "unknown 59"                    },
 569         { do_bad,               SIGBUS,  0,             "unknown 60"                    },
 570         { do_bad,               SIGBUS,  0,             "section domain fault"          },
 571         { do_bad,               SIGBUS,  0,             "page domain fault"             },
 572         { do_bad,               SIGBUS,  0,             "unknown 63"                    },
 573 };
 574
 575 /*
 576  * Dispatch a data abort to the relevant handler.
 577  */
 578 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
 579                                          struct pt_regs *regs)
 580 {
 581         const struct fault_info *inf = esr_to_fault_info(esr);
 582         struct siginfo info;
 583
 584         if (!inf->fn(addr, esr, regs))
 585                 return;
 586
 587         pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
 588                  inf->name, esr, addr);
 589
 590         info.si_signo = inf->sig;
 591         info.si_errno = 0;
 592         info.si_code  = inf->code;
 593         info.si_addr  = (void __user *)addr;
 594         arm64_notify_die("", regs, &info, esr);
 595 }
 596
 597 /*
 598  * Handle stack alignment exceptions.
 599  */
 600 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
 601                                            unsigned int esr,
 602                                            struct pt_regs *regs)
 603 {
 604         struct siginfo info;
 605         struct task_struct *tsk = current;
 606
 607         if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
 608                 pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
 609                                     tsk->comm, task_pid_nr(tsk),
 610                                     esr_get_class_string(esr), (void *)regs->pc,
 611                                     (void *)regs->sp);
 612
 613         info.si_signo = SIGBUS;
 614         info.si_errno = 0;
 615         info.si_code  = BUS_ADRALN;
 616         info.si_addr  = (void __user *)addr;
 617         arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
 618 }
 619
 620 int __init early_brk64(unsigned long addr, unsigned int esr,
 621                        struct pt_regs *regs);
 622
 623 /*
 624  * __refdata because early_brk64 is __init, but the reference to it is
 625  * clobbered at arch_initcall time.
 626  * See traps.c and debug-monitors.c:debug_traps_init().
 627  */
 628 static struct fault_info __refdata debug_fault_info[] = {
 629         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware breakpoint"   },
 630         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware single-step"  },
 631         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware watchpoint"   },
 632         { do_bad,       SIGBUS,         0,              "unknown 3"             },
 633         { do_bad,       SIGTRAP,        TRAP_BRKPT,     "aarch32 BKPT"          },
 634         { do_bad,       SIGTRAP,        0,              "aarch32 vector catch"  },
 635         { early_brk64,  SIGTRAP,        TRAP_BRKPT,     "aarch64 BRK"           },
 636         { do_bad,       SIGBUS,         0,              "unknown 7"             },
 637 };
 638
 639 void __init hook_debug_fault_code(int nr,
 640                                   int (*fn)(unsigned long, unsigned int, struct pt_regs *),
 641                                   int sig, int code, const char *name)
 642 {
 643         BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
 644
 645         debug_fault_info[nr].fn         = fn;
 646         debug_fault_info[nr].sig        = sig;
 647         debug_fault_info[nr].code       = code;
 648         debug_fault_info[nr].name       = name;
 649 }
 650
 651 asmlinkage int __exception do_debug_exception(unsigned long addr,
 652                                               unsigned int esr,
 653                                               struct pt_regs *regs)
 654 {
 655         const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
 656         struct siginfo info;
 657         int rv;
 658
 659         /*
 660          * Tell lockdep we disabled irqs in entry.S. Do nothing if they were
 661          * already disabled to preserve the last enabled/disabled addresses.
 662          */
 663         if (interrupts_enabled(regs))
 664                 trace_hardirqs_off();
 665
 666         if (!inf->fn(addr, esr, regs)) {
 667                 rv = 1;
 668         } else {
 669                 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
 670                          inf->name, esr, addr);
 671
 672                 info.si_signo = inf->sig;
 673                 info.si_errno = 0;
 674                 info.si_code  = inf->code;
 675                 info.si_addr  = (void __user *)addr;
 676                 arm64_notify_die("", regs, &info, 0);
 677                 rv = 0;
 678         }
 679
 680         if (interrupts_enabled(regs))
 681                 trace_hardirqs_on();
 682
 683         return rv;
 684 }
 685 NOKPROBE_SYMBOL(do_debug_exception);
 686
 687 #ifdef CONFIG_ARM64_PAN
 688 int cpu_enable_pan(void *__unused)
 689 {
 690         /*
 691          * We modify PSTATE. This won't work from irq context as the PSTATE
 692          * is discarded once we return from the exception.
 693          */
 694         WARN_ON_ONCE(in_interrupt());
 695
 696         config_sctlr_el1(SCTLR_EL1_SPAN, 0);
 697         asm(SET_PSTATE_PAN(1));
 698         return 0;
 699 }
 700 #endif /* CONFIG_ARM64_PAN */