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/module.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.h>
  30 #include <linux/highmem.h>
  31 #include <linux/perf_event.h>
  32
  33 #include <asm/exception.h>
  34 #include <asm/debug-monitors.h>
  35 #include <asm/system_misc.h>
  36 #include <asm/pgtable.h>
  37 #include <asm/tlbflush.h>
  38
  39 static const char *fault_name(unsigned int esr);
  40
  41 /*
  42  * Dump out the page tables associated with 'addr' in mm 'mm'.
  43  */
  44 void show_pte(struct mm_struct *mm, unsigned long addr)
  45 {
  46         pgd_t *pgd;
  47
  48         if (!mm)
  49                 mm = &init_mm;
  50
  51         pr_alert("pgd = %p\n", mm->pgd);
  52         pgd = pgd_offset(mm, addr);
  53         pr_alert("[%08lx] *pgd=%016llx", addr, pgd_val(*pgd));
  54
  55         do {
  56                 pud_t *pud;
  57                 pmd_t *pmd;
  58                 pte_t *pte;
  59
  60                 if (pgd_none(*pgd) || pgd_bad(*pgd))
  61                         break;
  62
  63                 pud = pud_offset(pgd, addr);
  64                 if (pud_none(*pud) || pud_bad(*pud))
  65                         break;
  66
  67                 pmd = pmd_offset(pud, addr);
  68                 printk(", *pmd=%016llx", pmd_val(*pmd));
  69                 if (pmd_none(*pmd) || pmd_bad(*pmd))
  70                         break;
  71
  72                 pte = pte_offset_map(pmd, addr);
  73                 printk(", *pte=%016llx", pte_val(*pte));
  74                 pte_unmap(pte);
  75         } while(0);
  76
  77         printk("\n");
  78 }
  79
  80 /*
  81  * The kernel tried to access some page that wasn't present.
  82  */
  83 static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr,
  84                               unsigned int esr, struct pt_regs *regs)
  85 {
  86         /*
  87          * Are we prepared to handle this kernel fault?
  88          */
  89         if (fixup_exception(regs))
  90                 return;
  91
  92         /*
  93          * No handler, we'll have to terminate things with extreme prejudice.
  94          */
  95         bust_spinlocks(1);
  96         pr_alert("Unable to handle kernel %s at virtual address %08lx\n",
  97                  (addr < PAGE_SIZE) ? "NULL pointer dereference" :
  98                  "paging request", addr);
  99
 100         show_pte(mm, addr);
 101         die("Oops", regs, esr);
 102         bust_spinlocks(0);
 103         do_exit(SIGKILL);
 104 }
 105
 106 /*
 107  * Something tried to access memory that isn't in our memory map. User mode
 108  * accesses just cause a SIGSEGV
 109  */
 110 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
 111                             unsigned int esr, unsigned int sig, int code,
 112                             struct pt_regs *regs)
 113 {
 114         struct siginfo si;
 115
 116         if (show_unhandled_signals) {
 117                 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n",
 118                         tsk->comm, task_pid_nr(tsk), fault_name(esr), sig,
 119                         addr, esr);
 120                 show_pte(tsk->mm, addr);
 121                 show_regs(regs);
 122         }
 123
 124         tsk->thread.fault_address = addr;
 125         si.si_signo = sig;
 126         si.si_errno = 0;
 127         si.si_code = code;
 128         si.si_addr = (void __user *)addr;
 129         force_sig_info(sig, &si, tsk);
 130 }
 131
 132 void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 133 {
 134         struct task_struct *tsk = current;
 135         struct mm_struct *mm = tsk->active_mm;
 136
 137         /*
 138          * If we are in kernel mode at this point, we have no context to
 139          * handle this fault with.
 140          */
 141         if (user_mode(regs))
 142                 __do_user_fault(tsk, addr, esr, SIGSEGV, SEGV_MAPERR, regs);
 143         else
 144                 __do_kernel_fault(mm, addr, esr, regs);
 145 }
 146
 147 #define VM_FAULT_BADMAP         0x010000
 148 #define VM_FAULT_BADACCESS      0x020000
 149
 150 #define ESR_WRITE               (1 << 6)
 151 #define ESR_CM                  (1 << 8)
 152 #define ESR_LNX_EXEC            (1 << 24)
 153
 154 /*
 155  * Check that the permissions on the VMA allow for the fault which occurred.
 156  * If we encountered a write fault, we must have write permission, otherwise
 157  * we allow any permission.
 158  */
 159 static inline bool access_error(unsigned int esr, struct vm_area_struct *vma)
 160 {
 161         unsigned int mask = VM_READ | VM_WRITE | VM_EXEC;
 162
 163         if (esr & ESR_WRITE)
 164                 mask = VM_WRITE;
 165         if (esr & ESR_LNX_EXEC)
 166                 mask = VM_EXEC;
 167
 168         return vma->vm_flags & mask ? false : true;
 169 }
 170
 171 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
 172                            unsigned int esr, unsigned int flags,
 173                            struct task_struct *tsk)
 174 {
 175         struct vm_area_struct *vma;
 176         int fault;
 177
 178         vma = find_vma(mm, addr);
 179         fault = VM_FAULT_BADMAP;
 180         if (unlikely(!vma))
 181                 goto out;
 182         if (unlikely(vma->vm_start > addr))
 183                 goto check_stack;
 184
 185         /*
 186          * Ok, we have a good vm_area for this memory access, so we can handle
 187          * it.
 188          */
 189 good_area:
 190         if (access_error(esr, vma)) {
 191                 fault = VM_FAULT_BADACCESS;
 192                 goto out;
 193         }
 194
 195         return handle_mm_fault(mm, vma, addr & PAGE_MASK, flags);
 196
 197 check_stack:
 198         if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
 199                 goto good_area;
 200 out:
 201         return fault;
 202 }
 203
 204 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
 205                                    struct pt_regs *regs)
 206 {
 207         struct task_struct *tsk;
 208         struct mm_struct *mm;
 209         int fault, sig, code;
 210         bool write = (esr & ESR_WRITE) && !(esr & ESR_CM);
 211         unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE |
 212                 (write ? FAULT_FLAG_WRITE : 0);
 213
 214         tsk = current;
 215         mm  = tsk->mm;
 216
 217         /* Enable interrupts if they were enabled in the parent context. */
 218         if (interrupts_enabled(regs))
 219                 local_irq_enable();
 220
 221         /*
 222          * If we're in an interrupt or have no user context, we must not take
 223          * the fault.
 224          */
 225         if (in_atomic() || !mm)
 226                 goto no_context;
 227
 228         /*
 229          * As per x86, we may deadlock here. However, since the kernel only
 230          * validly references user space from well defined areas of the code,
 231          * we can bug out early if this is from code which shouldn't.
 232          */
 233         if (!down_read_trylock(&mm->mmap_sem)) {
 234                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 235                         goto no_context;
 236 retry:
 237                 down_read(&mm->mmap_sem);
 238         } else {
 239                 /*
 240                  * The above down_read_trylock() might have succeeded in which
 241                  * case, we'll have missed the might_sleep() from down_read().
 242                  */
 243                 might_sleep();
 244 #ifdef CONFIG_DEBUG_VM
 245                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 246                         goto no_context;
 247 #endif
 248         }
 249
 250         fault = __do_page_fault(mm, addr, esr, flags, tsk);
 251
 252         /*
 253          * If we need to retry but a fatal signal is pending, handle the
 254          * signal first. We do not need to release the mmap_sem because it
 255          * would already be released in __lock_page_or_retry in mm/filemap.c.
 256          */
 257         if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
 258                 return 0;
 259
 260         /*
 261          * Major/minor page fault accounting is only done on the initial
 262          * attempt. If we go through a retry, it is extremely likely that the
 263          * page will be found in page cache at that point.
 264          */
 265
 266         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
 267         if (flags & FAULT_FLAG_ALLOW_RETRY) {
 268                 if (fault & VM_FAULT_MAJOR) {
 269                         tsk->maj_flt++;
 270                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
 271                                       addr);
 272                 } else {
 273                         tsk->min_flt++;
 274                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
 275                                       addr);
 276                 }
 277                 if (fault & VM_FAULT_RETRY) {
 278                         /*
 279                          * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
 280                          * starvation.
 281                          */
 282                         flags &= ~FAULT_FLAG_ALLOW_RETRY;
 283                         goto retry;
 284                 }
 285         }
 286
 287         up_read(&mm->mmap_sem);
 288
 289         /*
 290          * Handle the "normal" case first - VM_FAULT_MAJOR / VM_FAULT_MINOR
 291          */
 292         if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
 293                               VM_FAULT_BADACCESS))))
 294                 return 0;
 295
 296         if (fault & VM_FAULT_OOM) {
 297                 /*
 298                  * We ran out of memory, call the OOM killer, and return to
 299                  * userspace (which will retry the fault, or kill us if we got
 300                  * oom-killed).
 301                  */
 302                 pagefault_out_of_memory();
 303                 return 0;
 304         }
 305
 306         /*
 307          * If we are in kernel mode at this point, we have no context to
 308          * handle this fault with.
 309          */
 310         if (!user_mode(regs))
 311                 goto no_context;
 312
 313         if (fault & VM_FAULT_SIGBUS) {
 314                 /*
 315                  * We had some memory, but were unable to successfully fix up
 316                  * this page fault.
 317                  */
 318                 sig = SIGBUS;
 319                 code = BUS_ADRERR;
 320         } else {
 321                 /*
 322                  * Something tried to access memory that isn't in our memory
 323                  * map.
 324                  */
 325                 sig = SIGSEGV;
 326                 code = fault == VM_FAULT_BADACCESS ?
 327                         SEGV_ACCERR : SEGV_MAPERR;
 328         }
 329
 330         __do_user_fault(tsk, addr, esr, sig, code, regs);
 331         return 0;
 332
 333 no_context:
 334         __do_kernel_fault(mm, addr, esr, regs);
 335         return 0;
 336 }
 337
 338 /*
 339  * First Level Translation Fault Handler
 340  *
 341  * We enter here because the first level page table doesn't contain a valid
 342  * entry for the address.
 343  *
 344  * If the address is in kernel space (>= TASK_SIZE), then we are probably
 345  * faulting in the vmalloc() area.
 346  *
 347  * If the init_task's first level page tables contains the relevant entry, we
 348  * copy the it to this task.  If not, we send the process a signal, fixup the
 349  * exception, or oops the kernel.
 350  *
 351  * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
 352  * or a critical region, and should only copy the information from the master
 353  * page table, nothing more.
 354  */
 355 static int __kprobes do_translation_fault(unsigned long addr,
 356                                           unsigned int esr,
 357                                           struct pt_regs *regs)
 358 {
 359         if (addr < TASK_SIZE)
 360                 return do_page_fault(addr, esr, regs);
 361
 362         do_bad_area(addr, esr, regs);
 363         return 0;
 364 }
 365
 366 /*
 367  * Some section permission faults need to be handled gracefully.  They can
 368  * happen due to a __{get,put}_user during an oops.
 369  */
 370 static int do_sect_fault(unsigned long addr, unsigned int esr,
 371                          struct pt_regs *regs)
 372 {
 373         do_bad_area(addr, esr, regs);
 374         return 0;
 375 }
 376
 377 /*
 378  * This abort handler always returns "fault".
 379  */
 380 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 381 {
 382         return 1;
 383 }
 384
 385 static struct fault_info {
 386         int     (*fn)(unsigned long addr, unsigned int esr, struct pt_regs *regs);
 387         int     sig;
 388         int     code;
 389         const char *name;
 390 } fault_info[] = {
 391         { do_bad,               SIGBUS,  0,             "ttbr address size fault"       },
 392         { do_bad,               SIGBUS,  0,             "level 1 address size fault"    },
 393         { do_bad,               SIGBUS,  0,             "level 2 address size fault"    },
 394         { do_bad,               SIGBUS,  0,             "level 3 address size fault"    },
 395         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "input address range fault"     },
 396         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 1 translation fault"     },
 397         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 2 translation fault"     },
 398         { do_page_fault,        SIGSEGV, SEGV_MAPERR,   "level 3 translation fault"     },
 399         { do_bad,               SIGBUS,  0,             "reserved access flag fault"    },
 400         { do_bad,               SIGSEGV, SEGV_ACCERR,   "level 1 access flag fault"     },
 401         { do_bad,               SIGSEGV, SEGV_ACCERR,   "level 2 access flag fault"     },
 402         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 access flag fault"     },
 403         { do_bad,               SIGBUS,  0,             "reserved permission fault"     },
 404         { do_bad,               SIGSEGV, SEGV_ACCERR,   "level 1 permission fault"      },
 405         { do_sect_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 permission fault"      },
 406         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 permission fault"      },
 407         { do_bad,               SIGBUS,  0,             "synchronous external abort"    },
 408         { do_bad,               SIGBUS,  0,             "asynchronous external abort"   },
 409         { do_bad,               SIGBUS,  0,             "unknown 18"                    },
 410         { do_bad,               SIGBUS,  0,             "unknown 19"                    },
 411         { do_bad,               SIGBUS,  0,             "synchronous abort (translation table walk)" },
 412         { do_bad,               SIGBUS,  0,             "synchronous abort (translation table walk)" },
 413         { do_bad,               SIGBUS,  0,             "synchronous abort (translation table walk)" },
 414         { do_bad,               SIGBUS,  0,             "synchronous abort (translation table walk)" },
 415         { do_bad,               SIGBUS,  0,             "synchronous parity error"      },
 416         { do_bad,               SIGBUS,  0,             "asynchronous parity error"     },
 417         { do_bad,               SIGBUS,  0,             "unknown 26"                    },
 418         { do_bad,               SIGBUS,  0,             "unknown 27"                    },
 419         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk" },
 420         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk" },
 421         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk" },
 422         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk" },
 423         { do_bad,               SIGBUS,  0,             "unknown 32"                    },
 424         { do_bad,               SIGBUS,  BUS_ADRALN,    "alignment fault"               },
 425         { do_bad,               SIGBUS,  0,             "debug event"                   },
 426         { do_bad,               SIGBUS,  0,             "unknown 35"                    },
 427         { do_bad,               SIGBUS,  0,             "unknown 36"                    },
 428         { do_bad,               SIGBUS,  0,             "unknown 37"                    },
 429         { do_bad,               SIGBUS,  0,             "unknown 38"                    },
 430         { do_bad,               SIGBUS,  0,             "unknown 39"                    },
 431         { do_bad,               SIGBUS,  0,             "unknown 40"                    },
 432         { do_bad,               SIGBUS,  0,             "unknown 41"                    },
 433         { do_bad,               SIGBUS,  0,             "unknown 42"                    },
 434         { do_bad,               SIGBUS,  0,             "unknown 43"                    },
 435         { do_bad,               SIGBUS,  0,             "unknown 44"                    },
 436         { do_bad,               SIGBUS,  0,             "unknown 45"                    },
 437         { do_bad,               SIGBUS,  0,             "unknown 46"                    },
 438         { do_bad,               SIGBUS,  0,             "unknown 47"                    },
 439         { do_bad,               SIGBUS,  0,             "unknown 48"                    },
 440         { do_bad,               SIGBUS,  0,             "unknown 49"                    },
 441         { do_bad,               SIGBUS,  0,             "unknown 50"                    },
 442         { do_bad,               SIGBUS,  0,             "unknown 51"                    },
 443         { do_bad,               SIGBUS,  0,             "implementation fault (lockdown abort)" },
 444         { do_bad,               SIGBUS,  0,             "unknown 53"                    },
 445         { do_bad,               SIGBUS,  0,             "unknown 54"                    },
 446         { do_bad,               SIGBUS,  0,             "unknown 55"                    },
 447         { do_bad,               SIGBUS,  0,             "unknown 56"                    },
 448         { do_bad,               SIGBUS,  0,             "unknown 57"                    },
 449         { do_bad,               SIGBUS,  0,             "implementation fault (coprocessor abort)" },
 450         { do_bad,               SIGBUS,  0,             "unknown 59"                    },
 451         { do_bad,               SIGBUS,  0,             "unknown 60"                    },
 452         { do_bad,               SIGBUS,  0,             "unknown 61"                    },
 453         { do_bad,               SIGBUS,  0,             "unknown 62"                    },
 454         { do_bad,               SIGBUS,  0,             "unknown 63"                    },
 455 };
 456
 457 static const char *fault_name(unsigned int esr)
 458 {
 459         const struct fault_info *inf = fault_info + (esr & 63);
 460         return inf->name;
 461 }
 462
 463 /*
 464  * Dispatch a data abort to the relevant handler.
 465  */
 466 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
 467                                          struct pt_regs *regs)
 468 {
 469         const struct fault_info *inf = fault_info + (esr & 63);
 470         struct siginfo info;
 471
 472         if (!inf->fn(addr, esr, regs))
 473                 return;
 474
 475         pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
 476                  inf->name, esr, addr);
 477
 478         info.si_signo = inf->sig;
 479         info.si_errno = 0;
 480         info.si_code  = inf->code;
 481         info.si_addr  = (void __user *)addr;
 482         arm64_notify_die("", regs, &info, esr);
 483 }
 484
 485 /*
 486  * Handle stack alignment exceptions.
 487  */
 488 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
 489                                            unsigned int esr,
 490                                            struct pt_regs *regs)
 491 {
 492         struct siginfo info;
 493
 494         info.si_signo = SIGBUS;
 495         info.si_errno = 0;
 496         info.si_code  = BUS_ADRALN;
 497         info.si_addr  = (void __user *)addr;
 498         arm64_notify_die("", regs, &info, esr);
 499 }
 500
 501 static struct fault_info debug_fault_info[] = {
 502         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware breakpoint"   },
 503         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware single-step"  },
 504         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware watchpoint"   },
 505         { do_bad,       SIGBUS,         0,              "unknown 3"             },
 506         { do_bad,       SIGTRAP,        TRAP_BRKPT,     "aarch32 BKPT"          },
 507         { do_bad,       SIGTRAP,        0,              "aarch32 vector catch"  },
 508         { do_bad,       SIGTRAP,        TRAP_BRKPT,     "aarch64 BRK"           },
 509         { do_bad,       SIGBUS,         0,              "unknown 7"             },
 510 };
 511
 512 void __init hook_debug_fault_code(int nr,
 513                                   int (*fn)(unsigned long, unsigned int, struct pt_regs *),
 514                                   int sig, int code, const char *name)
 515 {
 516         BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
 517
 518         debug_fault_info[nr].fn         = fn;
 519         debug_fault_info[nr].sig        = sig;
 520         debug_fault_info[nr].code       = code;
 521         debug_fault_info[nr].name       = name;
 522 }
 523
 524 asmlinkage int __exception do_debug_exception(unsigned long addr,
 525                                               unsigned int esr,
 526                                               struct pt_regs *regs)
 527 {
 528         const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
 529         struct siginfo info;
 530
 531         if (!inf->fn(addr, esr, regs))
 532                 return 1;
 533
 534         pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
 535                  inf->name, esr, addr);
 536
 537         info.si_signo = inf->sig;
 538         info.si_errno = 0;
 539         info.si_code  = inf->code;
 540         info.si_addr  = (void __user *)addr;
 541         arm64_notify_die("", regs, &info, esr);
 542
 543         return 0;
 544 }