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.h>
  30 #include <linux/highmem.h>
  31 #include <linux/perf_event.h>
  32 #include <linux/preempt.h>
  33
  34 #include <asm/bug.h>
  35 #include <asm/cpufeature.h>
  36 #include <asm/exception.h>
  37 #include <asm/debug-monitors.h>
  38 #include <asm/esr.h>
  39 #include <asm/sysreg.h>
  40 #include <asm/system_misc.h>
  41 #include <asm/pgtable.h>
  42 #include <asm/tlbflush.h>
  43
  44 struct fault_info {
  45         int     (*fn)(unsigned long addr, unsigned int esr,
  46                       struct pt_regs *regs);
  47         int     sig;
  48         int     code;
  49         const char *name;
  50 };
  51
  52 static const struct fault_info fault_info[];
  53
  54 static inline const struct fault_info *esr_to_fault_info(unsigned int esr)
  55 {
  56         return fault_info + (esr & 63);
  57 }
  58
  59 #ifdef CONFIG_KPROBES
  60 static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
  61 {
  62         int ret = 0;
  63
  64         /* kprobe_running() needs smp_processor_id() */
  65         if (!user_mode(regs)) {
  66                 preempt_disable();
  67                 if (kprobe_running() && kprobe_fault_handler(regs, esr))
  68                         ret = 1;
  69                 preempt_enable();
  70         }
  71
  72         return ret;
  73 }
  74 #else
  75 static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
  76 {
  77         return 0;
  78 }
  79 #endif
  80
  81 /*
  82  * Dump out the page tables associated with 'addr' in mm 'mm'.
  83  */
  84 void show_pte(struct mm_struct *mm, unsigned long addr)
  85 {
  86         pgd_t *pgd;
  87
  88         if (!mm)
  89                 mm = &init_mm;
  90
  91         pr_alert("pgd = %p\n", mm->pgd);
  92         pgd = pgd_offset(mm, addr);
  93         pr_alert("[%08lx] *pgd=%016llx", addr, pgd_val(*pgd));
  94
  95         do {
  96                 pud_t *pud;
  97                 pmd_t *pmd;
  98                 pte_t *pte;
  99
 100                 if (pgd_none(*pgd) || pgd_bad(*pgd))
 101                         break;
 102
 103                 pud = pud_offset(pgd, addr);
 104                 printk(", *pud=%016llx", pud_val(*pud));
 105                 if (pud_none(*pud) || pud_bad(*pud))
 106                         break;
 107
 108                 pmd = pmd_offset(pud, addr);
 109                 printk(", *pmd=%016llx", pmd_val(*pmd));
 110                 if (pmd_none(*pmd) || pmd_bad(*pmd))
 111                         break;
 112
 113                 pte = pte_offset_map(pmd, addr);
 114                 printk(", *pte=%016llx", pte_val(*pte));
 115                 pte_unmap(pte);
 116         } while(0);
 117
 118         printk("\n");
 119 }
 120
 121 #ifdef CONFIG_ARM64_HW_AFDBM
 122 /*
 123  * This function sets the access flags (dirty, accessed), as well as write
 124  * permission, and only to a more permissive setting.
 125  *
 126  * It needs to cope with hardware update of the accessed/dirty state by other
 127  * agents in the system and can safely skip the __sync_icache_dcache() call as,
 128  * like set_pte_at(), the PTE is never changed from no-exec to exec here.
 129  *
 130  * Returns whether or not the PTE actually changed.
 131  */
 132 int ptep_set_access_flags(struct vm_area_struct *vma,
 133                           unsigned long address, pte_t *ptep,
 134                           pte_t entry, int dirty)
 135 {
 136         pteval_t old_pteval;
 137         unsigned int tmp;
 138
 139         if (pte_same(*ptep, entry))
 140                 return 0;
 141
 142         /* only preserve the access flags and write permission */
 143         pte_val(entry) &= PTE_AF | PTE_WRITE | PTE_DIRTY;
 144
 145         /*
 146          * PTE_RDONLY is cleared by default in the asm below, so set it in
 147          * back if necessary (read-only or clean PTE).
 148          */
 149         if (!pte_write(entry) || !pte_sw_dirty(entry))
 150                 pte_val(entry) |= PTE_RDONLY;
 151
 152         /*
 153          * Setting the flags must be done atomically to avoid racing with the
 154          * hardware update of the access/dirty state.
 155          */
 156         asm volatile("//        ptep_set_access_flags\n"
 157         "       prfm    pstl1strm, %2\n"
 158         "1:     ldxr    %0, %2\n"
 159         "       and     %0, %0, %3              // clear PTE_RDONLY\n"
 160         "       orr     %0, %0, %4              // set flags\n"
 161         "       stxr    %w1, %0, %2\n"
 162         "       cbnz    %w1, 1b\n"
 163         : "=&r" (old_pteval), "=&r" (tmp), "+Q" (pte_val(*ptep))
 164         : "L" (~PTE_RDONLY), "r" (pte_val(entry)));
 165
 166         flush_tlb_fix_spurious_fault(vma, address);
 167         return 1;
 168 }
 169 #endif
 170
 171 static bool is_el1_instruction_abort(unsigned int esr)
 172 {
 173         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR;
 174 }
 175
 176 /*
 177  * The kernel tried to access some page that wasn't present.
 178  */
 179 static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr,
 180                               unsigned int esr, struct pt_regs *regs)
 181 {
 182         /*
 183          * Are we prepared to handle this kernel fault?
 184          * We are almost certainly not prepared to handle instruction faults.
 185          */
 186         if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
 187                 return;
 188
 189         /*
 190          * No handler, we'll have to terminate things with extreme prejudice.
 191          */
 192         bust_spinlocks(1);
 193         pr_alert("Unable to handle kernel %s at virtual address %08lx\n",
 194                  (addr < PAGE_SIZE) ? "NULL pointer dereference" :
 195                  "paging request", addr);
 196
 197         show_pte(mm, addr);
 198         die("Oops", regs, esr);
 199         bust_spinlocks(0);
 200         do_exit(SIGKILL);
 201 }
 202
 203 /*
 204  * Something tried to access memory that isn't in our memory map. User mode
 205  * accesses just cause a SIGSEGV
 206  */
 207 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
 208                             unsigned int esr, unsigned int sig, int code,
 209                             struct pt_regs *regs)
 210 {
 211         struct siginfo si;
 212         const struct fault_info *inf;
 213
 214         if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
 215                 inf = esr_to_fault_info(esr);
 216                 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n",
 217                         tsk->comm, task_pid_nr(tsk), inf->name, sig,
 218                         addr, esr);
 219                 show_pte(tsk->mm, addr);
 220                 show_regs(regs);
 221         }
 222
 223         tsk->thread.fault_address = addr;
 224         tsk->thread.fault_code = esr;
 225         si.si_signo = sig;
 226         si.si_errno = 0;
 227         si.si_code = code;
 228         si.si_addr = (void __user *)addr;
 229         force_sig_info(sig, &si, tsk);
 230 }
 231
 232 static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 233 {
 234         struct task_struct *tsk = current;
 235         struct mm_struct *mm = tsk->active_mm;
 236         const struct fault_info *inf;
 237
 238         /*
 239          * If we are in kernel mode at this point, we have no context to
 240          * handle this fault with.
 241          */
 242         if (user_mode(regs)) {
 243                 inf = esr_to_fault_info(esr);
 244                 __do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs);
 245         } else
 246                 __do_kernel_fault(mm, addr, esr, regs);
 247 }
 248
 249 #define VM_FAULT_BADMAP         0x010000
 250 #define VM_FAULT_BADACCESS      0x020000
 251
 252 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
 253                            unsigned int mm_flags, unsigned long vm_flags,
 254                            struct task_struct *tsk)
 255 {
 256         struct vm_area_struct *vma;
 257         int fault;
 258
 259         vma = find_vma(mm, addr);
 260         fault = VM_FAULT_BADMAP;
 261         if (unlikely(!vma))
 262                 goto out;
 263         if (unlikely(vma->vm_start > addr))
 264                 goto check_stack;
 265
 266         /*
 267          * Ok, we have a good vm_area for this memory access, so we can handle
 268          * it.
 269          */
 270 good_area:
 271         /*
 272          * Check that the permissions on the VMA allow for the fault which
 273          * occurred.
 274          */
 275         if (!(vma->vm_flags & vm_flags)) {
 276                 fault = VM_FAULT_BADACCESS;
 277                 goto out;
 278         }
 279
 280         return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags);
 281
 282 check_stack:
 283         if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
 284                 goto good_area;
 285 out:
 286         return fault;
 287 }
 288
 289 static inline bool is_permission_fault(unsigned int esr)
 290 {
 291         unsigned int ec       = ESR_ELx_EC(esr);
 292         unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE;
 293
 294         return (ec == ESR_ELx_EC_DABT_CUR && fsc_type == ESR_ELx_FSC_PERM) ||
 295                (ec == ESR_ELx_EC_IABT_CUR && fsc_type == ESR_ELx_FSC_PERM);
 296 }
 297
 298 static bool is_el0_instruction_abort(unsigned int esr)
 299 {
 300         return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW;
 301 }
 302
 303 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
 304                                    struct pt_regs *regs)
 305 {
 306         struct task_struct *tsk;
 307         struct mm_struct *mm;
 308         int fault, sig, code;
 309         unsigned long vm_flags = VM_READ | VM_WRITE;
 310         unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
 311
 312         if (notify_page_fault(regs, esr))
 313                 return 0;
 314
 315         tsk = current;
 316         mm  = tsk->mm;
 317
 318         /*
 319          * If we're in an interrupt or have no user context, we must not take
 320          * the fault.
 321          */
 322         if (faulthandler_disabled() || !mm)
 323                 goto no_context;
 324
 325         if (user_mode(regs))
 326                 mm_flags |= FAULT_FLAG_USER;
 327
 328         if (is_el0_instruction_abort(esr)) {
 329                 vm_flags = VM_EXEC;
 330         } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
 331                 vm_flags = VM_WRITE;
 332                 mm_flags |= FAULT_FLAG_WRITE;
 333         }
 334
 335         if (is_permission_fault(esr) && (addr < USER_DS)) {
 336                 /* regs->orig_addr_limit may be 0 if we entered from EL0 */
 337                 if (regs->orig_addr_limit == KERNEL_DS)
 338                         die("Accessing user space memory with fs=KERNEL_DS", regs, esr);
 339
 340                 if (is_el1_instruction_abort(esr))
 341                         die("Attempting to execute userspace memory", regs, esr);
 342
 343                 if (!search_exception_tables(regs->pc))
 344                         die("Accessing user space memory outside uaccess.h routines", regs, esr);
 345         }
 346
 347         /*
 348          * As per x86, we may deadlock here. However, since the kernel only
 349          * validly references user space from well defined areas of the code,
 350          * we can bug out early if this is from code which shouldn't.
 351          */
 352         if (!down_read_trylock(&mm->mmap_sem)) {
 353                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 354                         goto no_context;
 355 retry:
 356                 down_read(&mm->mmap_sem);
 357         } else {
 358                 /*
 359                  * The above down_read_trylock() might have succeeded in which
 360                  * case, we'll have missed the might_sleep() from down_read().
 361                  */
 362                 might_sleep();
 363 #ifdef CONFIG_DEBUG_VM
 364                 if (!user_mode(regs) && !search_exception_tables(regs->pc))
 365                         goto no_context;
 366 #endif
 367         }
 368
 369         fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
 370
 371         /*
 372          * If we need to retry but a fatal signal is pending, handle the
 373          * signal first. We do not need to release the mmap_sem because it
 374          * would already be released in __lock_page_or_retry in mm/filemap.c.
 375          */
 376         if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
 377                 return 0;
 378
 379         /*
 380          * Major/minor page fault accounting is only done on the initial
 381          * attempt. If we go through a retry, it is extremely likely that the
 382          * page will be found in page cache at that point.
 383          */
 384
 385         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
 386         if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
 387                 if (fault & VM_FAULT_MAJOR) {
 388                         tsk->maj_flt++;
 389                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
 390                                       addr);
 391                 } else {
 392                         tsk->min_flt++;
 393                         perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
 394                                       addr);
 395                 }
 396                 if (fault & VM_FAULT_RETRY) {
 397                         /*
 398                          * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
 399                          * starvation.
 400                          */
 401                         mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
 402                         mm_flags |= FAULT_FLAG_TRIED;
 403                         goto retry;
 404                 }
 405         }
 406
 407         up_read(&mm->mmap_sem);
 408
 409         /*
 410          * Handle the "normal" case first - VM_FAULT_MAJOR
 411          */
 412         if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
 413                               VM_FAULT_BADACCESS))))
 414                 return 0;
 415
 416         /*
 417          * If we are in kernel mode at this point, we have no context to
 418          * handle this fault with.
 419          */
 420         if (!user_mode(regs))
 421                 goto no_context;
 422
 423         if (fault & VM_FAULT_OOM) {
 424                 /*
 425                  * We ran out of memory, call the OOM killer, and return to
 426                  * userspace (which will retry the fault, or kill us if we got
 427                  * oom-killed).
 428                  */
 429                 pagefault_out_of_memory();
 430                 return 0;
 431         }
 432
 433         if (fault & VM_FAULT_SIGBUS) {
 434                 /*
 435                  * We had some memory, but were unable to successfully fix up
 436                  * this page fault.
 437                  */
 438                 sig = SIGBUS;
 439                 code = BUS_ADRERR;
 440         } else {
 441                 /*
 442                  * Something tried to access memory that isn't in our memory
 443                  * map.
 444                  */
 445                 sig = SIGSEGV;
 446                 code = fault == VM_FAULT_BADACCESS ?
 447                         SEGV_ACCERR : SEGV_MAPERR;
 448         }
 449
 450         __do_user_fault(tsk, addr, esr, sig, code, regs);
 451         return 0;
 452
 453 no_context:
 454         __do_kernel_fault(mm, addr, esr, regs);
 455         return 0;
 456 }
 457
 458 /*
 459  * First Level Translation Fault Handler
 460  *
 461  * We enter here because the first level page table doesn't contain a valid
 462  * entry for the address.
 463  *
 464  * If the address is in kernel space (>= TASK_SIZE), then we are probably
 465  * faulting in the vmalloc() area.
 466  *
 467  * If the init_task's first level page tables contains the relevant entry, we
 468  * copy the it to this task.  If not, we send the process a signal, fixup the
 469  * exception, or oops the kernel.
 470  *
 471  * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
 472  * or a critical region, and should only copy the information from the master
 473  * page table, nothing more.
 474  */
 475 static int __kprobes do_translation_fault(unsigned long addr,
 476                                           unsigned int esr,
 477                                           struct pt_regs *regs)
 478 {
 479         if (addr < TASK_SIZE)
 480                 return do_page_fault(addr, esr, regs);
 481
 482         do_bad_area(addr, esr, regs);
 483         return 0;
 484 }
 485
 486 static int do_alignment_fault(unsigned long addr, unsigned int esr,
 487                               struct pt_regs *regs)
 488 {
 489         do_bad_area(addr, esr, regs);
 490         return 0;
 491 }
 492
 493 /*
 494  * This abort handler always returns "fault".
 495  */
 496 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
 497 {
 498         return 1;
 499 }
 500
 501 static const struct fault_info fault_info[] = {
 502         { do_bad,               SIGBUS,  0,             "ttbr address size fault"       },
 503         { do_bad,               SIGBUS,  0,             "level 1 address size fault"    },
 504         { do_bad,               SIGBUS,  0,             "level 2 address size fault"    },
 505         { do_bad,               SIGBUS,  0,             "level 3 address size fault"    },
 506         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 0 translation fault"     },
 507         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 1 translation fault"     },
 508         { do_translation_fault, SIGSEGV, SEGV_MAPERR,   "level 2 translation fault"     },
 509         { do_page_fault,        SIGSEGV, SEGV_MAPERR,   "level 3 translation fault"     },
 510         { do_bad,               SIGBUS,  0,             "unknown 8"                     },
 511         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 access flag fault"     },
 512         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 access flag fault"     },
 513         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 access flag fault"     },
 514         { do_bad,               SIGBUS,  0,             "unknown 12"                    },
 515         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 1 permission fault"      },
 516         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 2 permission fault"      },
 517         { do_page_fault,        SIGSEGV, SEGV_ACCERR,   "level 3 permission fault"      },
 518         { do_bad,               SIGBUS,  0,             "synchronous external abort"    },
 519         { do_bad,               SIGBUS,  0,             "unknown 17"                    },
 520         { do_bad,               SIGBUS,  0,             "unknown 18"                    },
 521         { do_bad,               SIGBUS,  0,             "unknown 19"                    },
 522         { do_bad,               SIGBUS,  0,             "synchronous abort (translation table walk)" },
 523         { do_bad,               SIGBUS,  0,             "synchronous abort (translation table walk)" },
 524         { do_bad,               SIGBUS,  0,             "synchronous abort (translation table walk)" },
 525         { do_bad,               SIGBUS,  0,             "synchronous abort (translation table walk)" },
 526         { do_bad,               SIGBUS,  0,             "synchronous parity error"      },
 527         { do_bad,               SIGBUS,  0,             "unknown 25"                    },
 528         { do_bad,               SIGBUS,  0,             "unknown 26"                    },
 529         { do_bad,               SIGBUS,  0,             "unknown 27"                    },
 530         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk)" },
 531         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk)" },
 532         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk)" },
 533         { do_bad,               SIGBUS,  0,             "synchronous parity error (translation table walk)" },
 534         { do_bad,               SIGBUS,  0,             "unknown 32"                    },
 535         { do_alignment_fault,   SIGBUS,  BUS_ADRALN,    "alignment fault"               },
 536         { do_bad,               SIGBUS,  0,             "unknown 34"                    },
 537         { do_bad,               SIGBUS,  0,             "unknown 35"                    },
 538         { do_bad,               SIGBUS,  0,             "unknown 36"                    },
 539         { do_bad,               SIGBUS,  0,             "unknown 37"                    },
 540         { do_bad,               SIGBUS,  0,             "unknown 38"                    },
 541         { do_bad,               SIGBUS,  0,             "unknown 39"                    },
 542         { do_bad,               SIGBUS,  0,             "unknown 40"                    },
 543         { do_bad,               SIGBUS,  0,             "unknown 41"                    },
 544         { do_bad,               SIGBUS,  0,             "unknown 42"                    },
 545         { do_bad,               SIGBUS,  0,             "unknown 43"                    },
 546         { do_bad,               SIGBUS,  0,             "unknown 44"                    },
 547         { do_bad,               SIGBUS,  0,             "unknown 45"                    },
 548         { do_bad,               SIGBUS,  0,             "unknown 46"                    },
 549         { do_bad,               SIGBUS,  0,             "unknown 47"                    },
 550         { do_bad,               SIGBUS,  0,             "TLB conflict abort"            },
 551         { do_bad,               SIGBUS,  0,             "unknown 49"                    },
 552         { do_bad,               SIGBUS,  0,             "unknown 50"                    },
 553         { do_bad,               SIGBUS,  0,             "unknown 51"                    },
 554         { do_bad,               SIGBUS,  0,             "implementation fault (lockdown abort)" },
 555         { do_bad,               SIGBUS,  0,             "implementation fault (unsupported exclusive)" },
 556         { do_bad,               SIGBUS,  0,             "unknown 54"                    },
 557         { do_bad,               SIGBUS,  0,             "unknown 55"                    },
 558         { do_bad,               SIGBUS,  0,             "unknown 56"                    },
 559         { do_bad,               SIGBUS,  0,             "unknown 57"                    },
 560         { do_bad,               SIGBUS,  0,             "unknown 58"                    },
 561         { do_bad,               SIGBUS,  0,             "unknown 59"                    },
 562         { do_bad,               SIGBUS,  0,             "unknown 60"                    },
 563         { do_bad,               SIGBUS,  0,             "section domain fault"          },
 564         { do_bad,               SIGBUS,  0,             "page domain fault"             },
 565         { do_bad,               SIGBUS,  0,             "unknown 63"                    },
 566 };
 567
 568 /*
 569  * Dispatch a data abort to the relevant handler.
 570  */
 571 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
 572                                          struct pt_regs *regs)
 573 {
 574         const struct fault_info *inf = esr_to_fault_info(esr);
 575         struct siginfo info;
 576
 577         if (!inf->fn(addr, esr, regs))
 578                 return;
 579
 580         pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
 581                  inf->name, esr, addr);
 582
 583         info.si_signo = inf->sig;
 584         info.si_errno = 0;
 585         info.si_code  = inf->code;
 586         info.si_addr  = (void __user *)addr;
 587         arm64_notify_die("", regs, &info, esr);
 588 }
 589
 590 /*
 591  * Handle stack alignment exceptions.
 592  */
 593 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
 594                                            unsigned int esr,
 595                                            struct pt_regs *regs)
 596 {
 597         struct siginfo info;
 598         struct task_struct *tsk = current;
 599
 600         if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
 601                 pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
 602                                     tsk->comm, task_pid_nr(tsk),
 603                                     esr_get_class_string(esr), (void *)regs->pc,
 604                                     (void *)regs->sp);
 605
 606         info.si_signo = SIGBUS;
 607         info.si_errno = 0;
 608         info.si_code  = BUS_ADRALN;
 609         info.si_addr  = (void __user *)addr;
 610         arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
 611 }
 612
 613 int __init early_brk64(unsigned long addr, unsigned int esr,
 614                        struct pt_regs *regs);
 615
 616 /*
 617  * __refdata because early_brk64 is __init, but the reference to it is
 618  * clobbered at arch_initcall time.
 619  * See traps.c and debug-monitors.c:debug_traps_init().
 620  */
 621 static struct fault_info __refdata debug_fault_info[] = {
 622         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware breakpoint"   },
 623         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware single-step"  },
 624         { do_bad,       SIGTRAP,        TRAP_HWBKPT,    "hardware watchpoint"   },
 625         { do_bad,       SIGBUS,         0,              "unknown 3"             },
 626         { do_bad,       SIGTRAP,        TRAP_BRKPT,     "aarch32 BKPT"          },
 627         { do_bad,       SIGTRAP,        0,              "aarch32 vector catch"  },
 628         { early_brk64,  SIGTRAP,        TRAP_BRKPT,     "aarch64 BRK"           },
 629         { do_bad,       SIGBUS,         0,              "unknown 7"             },
 630 };
 631
 632 void __init hook_debug_fault_code(int nr,
 633                                   int (*fn)(unsigned long, unsigned int, struct pt_regs *),
 634                                   int sig, int code, const char *name)
 635 {
 636         BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
 637
 638         debug_fault_info[nr].fn         = fn;
 639         debug_fault_info[nr].sig        = sig;
 640         debug_fault_info[nr].code       = code;
 641         debug_fault_info[nr].name       = name;
 642 }
 643
 644 asmlinkage int __exception do_debug_exception(unsigned long addr,
 645                                               unsigned int esr,
 646                                               struct pt_regs *regs)
 647 {
 648         const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
 649         struct siginfo info;
 650         int rv;
 651
 652         /*
 653          * Tell lockdep we disabled irqs in entry.S. Do nothing if they were
 654          * already disabled to preserve the last enabled/disabled addresses.
 655          */
 656         if (interrupts_enabled(regs))
 657                 trace_hardirqs_off();
 658
 659         if (!inf->fn(addr, esr, regs)) {
 660                 rv = 1;
 661         } else {
 662                 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
 663                          inf->name, esr, addr);
 664
 665                 info.si_signo = inf->sig;
 666                 info.si_errno = 0;
 667                 info.si_code  = inf->code;
 668                 info.si_addr  = (void __user *)addr;
 669                 arm64_notify_die("", regs, &info, 0);
 670                 rv = 0;
 671         }
 672
 673         if (interrupts_enabled(regs))
 674                 trace_hardirqs_on();
 675
 676         return rv;
 677 }
 678 NOKPROBE_SYMBOL(do_debug_exception);
 679
 680 #ifdef CONFIG_ARM64_PAN
 681 int cpu_enable_pan(void *__unused)
 682 {
 683         /*
 684          * We modify PSTATE. This won't work from irq context as the PSTATE
 685          * is discarded once we return from the exception.
 686          */
 687         WARN_ON_ONCE(in_interrupt());
 688
 689         config_sctlr_el1(SCTLR_EL1_SPAN, 0);
 690         asm(SET_PSTATE_PAN(1));
 691         return 0;
 692 }
 693 #endif /* CONFIG_ARM64_PAN */
 694
 695 #ifdef CONFIG_ARM64_UAO
 696 /*
 697  * Kernel threads have fs=KERNEL_DS by default, and don't need to call
 698  * set_fs(), devtmpfs in particular relies on this behaviour.
 699  * We need to enable the feature at runtime (instead of adding it to
 700  * PSR_MODE_EL1h) as the feature may not be implemented by the cpu.
 701  */
 702 int cpu_enable_uao(void *__unused)
 703 {
 704         asm(SET_PSTATE_UAO(1));
 705         return 0;
 706 }
 707 #endif /* CONFIG_ARM64_UAO */