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