arch/sparc/kernel/sys_sparc_64.c

   1 /* linux/arch/sparc64/kernel/sys_sparc.c
   2  *
   3  * This file contains various random system calls that
   4  * have a non-standard calling sequence on the Linux/sparc
   5  * platform.
   6  */
   7
   8 #include <linux/errno.h>
   9 #include <linux/types.h>
  10 #include <linux/sched.h>
  11 #include <linux/fs.h>
  12 #include <linux/file.h>
  13 #include <linux/mm.h>
  14 #include <linux/sem.h>
  15 #include <linux/msg.h>
  16 #include <linux/shm.h>
  17 #include <linux/stat.h>
  18 #include <linux/mman.h>
  19 #include <linux/utsname.h>
  20 #include <linux/smp.h>
  21 #include <linux/slab.h>
  22 #include <linux/syscalls.h>
  23 #include <linux/ipc.h>
  24 #include <linux/personality.h>
  25 #include <linux/random.h>
  26 #include <linux/export.h>
  27 #include <linux/context_tracking.h>
  28
  29 #include <asm/uaccess.h>
  30 #include <asm/utrap.h>
  31 #include <asm/unistd.h>
  32
  33 #include "entry.h"
  34 #include "systbls.h"
  35
  36 /* #define DEBUG_UNIMP_SYSCALL */
  37
  38 asmlinkage unsigned long sys_getpagesize(void)
  39 {
  40         return PAGE_SIZE;
  41 }
  42
  43 /* Does addr --> addr+len fall within 4GB of the VA-space hole or
  44  * overflow past the end of the 64-bit address space?
  45  */
  46 static inline int invalid_64bit_range(unsigned long addr, unsigned long len)
  47 {
  48         unsigned long va_exclude_start, va_exclude_end;
  49
  50         va_exclude_start = VA_EXCLUDE_START;
  51         va_exclude_end   = VA_EXCLUDE_END;
  52
  53         if (unlikely(len >= va_exclude_start))
  54                 return 1;
  55
  56         if (unlikely((addr + len) < addr))
  57                 return 1;
  58
  59         if (unlikely((addr >= va_exclude_start && addr < va_exclude_end) ||
  60                      ((addr + len) >= va_exclude_start &&
  61                       (addr + len) < va_exclude_end)))
  62                 return 1;
  63
  64         return 0;
  65 }
  66
  67 /* These functions differ from the default implementations in
  68  * mm/mmap.c in two ways:
  69  *
  70  * 1) For file backed MAP_SHARED mmap()'s we D-cache color align,
  71  *    for fixed such mappings we just validate what the user gave us.
  72  * 2) For 64-bit tasks we avoid mapping anything within 4GB of
  73  *    the spitfire/niagara VA-hole.
  74  */
  75
  76 static inline unsigned long COLOR_ALIGN(unsigned long addr,
  77                                          unsigned long pgoff)
  78 {
  79         unsigned long base = (addr+SHMLBA-1)&~(SHMLBA-1);
  80         unsigned long off = (pgoff<<PAGE_SHIFT) & (SHMLBA-1);
  81
  82         return base + off;
  83 }
  84
  85 unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr, unsigned long len, unsigned long pgoff, unsigned long flags)
  86 {
  87         struct mm_struct *mm = current->mm;
  88         struct vm_area_struct * vma;
  89         unsigned long task_size = TASK_SIZE;
  90         int do_color_align;
  91         struct vm_unmapped_area_info info;
  92
  93         if (flags & MAP_FIXED) {
  94                 /* We do not accept a shared mapping if it would violate
  95                  * cache aliasing constraints.
  96                  */
  97                 if ((flags & MAP_SHARED) &&
  98                     ((addr - (pgoff << PAGE_SHIFT)) & (SHMLBA - 1)))
  99                         return -EINVAL;
 100                 return addr;
 101         }
 102
 103         if (test_thread_flag(TIF_32BIT))
 104                 task_size = STACK_TOP32;
 105         if (unlikely(len > task_size || len >= VA_EXCLUDE_START))
 106                 return -ENOMEM;
 107
 108         do_color_align = 0;
 109         if (filp || (flags & MAP_SHARED))
 110                 do_color_align = 1;
 111
 112         if (addr) {
 113                 if (do_color_align)
 114                         addr = COLOR_ALIGN(addr, pgoff);
 115                 else
 116                         addr = PAGE_ALIGN(addr);
 117
 118                 vma = find_vma(mm, addr);
 119                 if (task_size - len >= addr &&
 120                     (!vma || addr + len <= vma->vm_start))
 121                         return addr;
 122         }
 123
 124         info.flags = 0;
 125         info.length = len;
 126         info.low_limit = TASK_UNMAPPED_BASE;
 127         info.high_limit = min(task_size, VA_EXCLUDE_START);
 128         info.align_mask = do_color_align ? (PAGE_MASK & (SHMLBA - 1)) : 0;
 129         info.align_offset = pgoff << PAGE_SHIFT;
 130         addr = vm_unmapped_area(&info);
 131
 132         if ((addr & ~PAGE_MASK) && task_size > VA_EXCLUDE_END) {
 133                 VM_BUG_ON(addr != -ENOMEM);
 134                 info.low_limit = VA_EXCLUDE_END;
 135                 info.high_limit = task_size;
 136                 addr = vm_unmapped_area(&info);
 137         }
 138
 139         return addr;
 140 }
 141
 142 unsigned long
 143 arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
 144                           const unsigned long len, const unsigned long pgoff,
 145                           const unsigned long flags)
 146 {
 147         struct vm_area_struct *vma;
 148         struct mm_struct *mm = current->mm;
 149         unsigned long task_size = STACK_TOP32;
 150         unsigned long addr = addr0;
 151         int do_color_align;
 152         struct vm_unmapped_area_info info;
 153
 154         /* This should only ever run for 32-bit processes.  */
 155         BUG_ON(!test_thread_flag(TIF_32BIT));
 156
 157         if (flags & MAP_FIXED) {
 158                 /* We do not accept a shared mapping if it would violate
 159                  * cache aliasing constraints.
 160                  */
 161                 if ((flags & MAP_SHARED) &&
 162                     ((addr - (pgoff << PAGE_SHIFT)) & (SHMLBA - 1)))
 163                         return -EINVAL;
 164                 return addr;
 165         }
 166
 167         if (unlikely(len > task_size))
 168                 return -ENOMEM;
 169
 170         do_color_align = 0;
 171         if (filp || (flags & MAP_SHARED))
 172                 do_color_align = 1;
 173
 174         /* requesting a specific address */
 175         if (addr) {
 176                 if (do_color_align)
 177                         addr = COLOR_ALIGN(addr, pgoff);
 178                 else
 179                         addr = PAGE_ALIGN(addr);
 180
 181                 vma = find_vma(mm, addr);
 182                 if (task_size - len >= addr &&
 183                     (!vma || addr + len <= vma->vm_start))
 184                         return addr;
 185         }
 186
 187         info.flags = VM_UNMAPPED_AREA_TOPDOWN;
 188         info.length = len;
 189         info.low_limit = PAGE_SIZE;
 190         info.high_limit = mm->mmap_base;
 191         info.align_mask = do_color_align ? (PAGE_MASK & (SHMLBA - 1)) : 0;
 192         info.align_offset = pgoff << PAGE_SHIFT;
 193         addr = vm_unmapped_area(&info);
 194
 195         /*
 196          * A failed mmap() very likely causes application failure,
 197          * so fall back to the bottom-up function here. This scenario
 198          * can happen with large stack limits and large mmap()
 199          * allocations.
 200          */
 201         if (addr & ~PAGE_MASK) {
 202                 VM_BUG_ON(addr != -ENOMEM);
 203                 info.flags = 0;
 204                 info.low_limit = TASK_UNMAPPED_BASE;
 205                 info.high_limit = STACK_TOP32;
 206                 addr = vm_unmapped_area(&info);
 207         }
 208
 209         return addr;
 210 }
 211
 212 /* Try to align mapping such that we align it as much as possible. */
 213 unsigned long get_fb_unmapped_area(struct file *filp, unsigned long orig_addr, unsigned long len, unsigned long pgoff, unsigned long flags)
 214 {
 215         unsigned long align_goal, addr = -ENOMEM;
 216         unsigned long (*get_area)(struct file *, unsigned long,
 217                                   unsigned long, unsigned long, unsigned long);
 218
 219         get_area = current->mm->get_unmapped_area;
 220
 221         if (flags & MAP_FIXED) {
 222                 /* Ok, don't mess with it. */
 223                 return get_area(NULL, orig_addr, len, pgoff, flags);
 224         }
 225         flags &= ~MAP_SHARED;
 226
 227         align_goal = PAGE_SIZE;
 228         if (len >= (4UL * 1024 * 1024))
 229                 align_goal = (4UL * 1024 * 1024);
 230         else if (len >= (512UL * 1024))
 231                 align_goal = (512UL * 1024);
 232         else if (len >= (64UL * 1024))
 233                 align_goal = (64UL * 1024);
 234
 235         do {
 236                 addr = get_area(NULL, orig_addr, len + (align_goal - PAGE_SIZE), pgoff, flags);
 237                 if (!(addr & ~PAGE_MASK)) {
 238                         addr = (addr + (align_goal - 1UL)) & ~(align_goal - 1UL);
 239                         break;
 240                 }
 241
 242                 if (align_goal == (4UL * 1024 * 1024))
 243                         align_goal = (512UL * 1024);
 244                 else if (align_goal == (512UL * 1024))
 245                         align_goal = (64UL * 1024);
 246                 else
 247                         align_goal = PAGE_SIZE;
 248         } while ((addr & ~PAGE_MASK) && align_goal > PAGE_SIZE);
 249
 250         /* Mapping is smaller than 64K or larger areas could not
 251          * be obtained.
 252          */
 253         if (addr & ~PAGE_MASK)
 254                 addr = get_area(NULL, orig_addr, len, pgoff, flags);
 255
 256         return addr;
 257 }
 258 EXPORT_SYMBOL(get_fb_unmapped_area);
 259
 260 /* Essentially the same as PowerPC.  */
 261 static unsigned long mmap_rnd(void)
 262 {
 263         unsigned long rnd = 0UL;
 264
 265         if (current->flags & PF_RANDOMIZE) {
 266                 unsigned long val = get_random_int();
 267                 if (test_thread_flag(TIF_32BIT))
 268                         rnd = (val % (1UL << (23UL-PAGE_SHIFT)));
 269                 else
 270                         rnd = (val % (1UL << (30UL-PAGE_SHIFT)));
 271         }
 272         return rnd << PAGE_SHIFT;
 273 }
 274
 275 void arch_pick_mmap_layout(struct mm_struct *mm)
 276 {
 277         unsigned long random_factor = mmap_rnd();
 278         unsigned long gap;
 279
 280         /*
 281          * Fall back to the standard layout if the personality
 282          * bit is set, or if the expected stack growth is unlimited:
 283          */
 284         gap = rlimit(RLIMIT_STACK);
 285         if (!test_thread_flag(TIF_32BIT) ||
 286             (current->personality & ADDR_COMPAT_LAYOUT) ||
 287             gap == RLIM_INFINITY ||
 288             sysctl_legacy_va_layout) {
 289                 mm->mmap_base = TASK_UNMAPPED_BASE + random_factor;
 290                 mm->get_unmapped_area = arch_get_unmapped_area;
 291         } else {
 292                 /* We know it's 32-bit */
 293                 unsigned long task_size = STACK_TOP32;
 294
 295                 if (gap < 128 * 1024 * 1024)
 296                         gap = 128 * 1024 * 1024;
 297                 if (gap > (task_size / 6 * 5))
 298                         gap = (task_size / 6 * 5);
 299
 300                 mm->mmap_base = PAGE_ALIGN(task_size - gap - random_factor);
 301                 mm->get_unmapped_area = arch_get_unmapped_area_topdown;
 302         }
 303 }
 304
 305 /*
 306  * sys_pipe() is the normal C calling standard for creating
 307  * a pipe. It's not the way unix traditionally does this, though.
 308  */
 309 SYSCALL_DEFINE1(sparc_pipe_real, struct pt_regs *, regs)
 310 {
 311         int fd[2];
 312         int error;
 313
 314         error = do_pipe_flags(fd, 0);
 315         if (error)
 316                 goto out;
 317         regs->u_regs[UREG_I1] = fd[1];
 318         error = fd[0];
 319 out:
 320         return error;
 321 }
 322
 323 /*
 324  * sys_ipc() is the de-multiplexer for the SysV IPC calls..
 325  *
 326  * This is really horribly ugly.
 327  */
 328
 329 SYSCALL_DEFINE6(sparc_ipc, unsigned int, call, int, first, unsigned long, second,
 330                 unsigned long, third, void __user *, ptr, long, fifth)
 331 {
 332         long err;
 333
 334         /* No need for backward compatibility. We can start fresh... */
 335         if (call <= SEMCTL) {
 336                 switch (call) {
 337                 case SEMOP:
 338                         err = sys_semtimedop(first, ptr,
 339                                              (unsigned)second, NULL);
 340                         goto out;
 341                 case SEMTIMEDOP:
 342                         err = sys_semtimedop(first, ptr, (unsigned)second,
 343                                 (const struct timespec __user *)
 344                                              (unsigned long) fifth);
 345                         goto out;
 346                 case SEMGET:
 347                         err = sys_semget(first, (int)second, (int)third);
 348                         goto out;
 349                 case SEMCTL: {
 350                         err = sys_semctl(first, second,
 351                                          (int)third | IPC_64,
 352                                          (unsigned long) ptr);
 353                         goto out;
 354                 }
 355                 default:
 356                         err = -ENOSYS;
 357                         goto out;
 358                 }
 359         }
 360         if (call <= MSGCTL) {
 361                 switch (call) {
 362                 case MSGSND:
 363                         err = sys_msgsnd(first, ptr, (size_t)second,
 364                                          (int)third);
 365                         goto out;
 366                 case MSGRCV:
 367                         err = sys_msgrcv(first, ptr, (size_t)second, fifth,
 368                                          (int)third);
 369                         goto out;
 370                 case MSGGET:
 371                         err = sys_msgget((key_t)first, (int)second);
 372                         goto out;
 373                 case MSGCTL:
 374                         err = sys_msgctl(first, (int)second | IPC_64, ptr);
 375                         goto out;
 376                 default:
 377                         err = -ENOSYS;
 378                         goto out;
 379                 }
 380         }
 381         if (call <= SHMCTL) {
 382                 switch (call) {
 383                 case SHMAT: {
 384                         ulong raddr;
 385                         err = do_shmat(first, ptr, (int)second, &raddr, SHMLBA);
 386                         if (!err) {
 387                                 if (put_user(raddr,
 388                                              (ulong __user *) third))
 389                                         err = -EFAULT;
 390                         }
 391                         goto out;
 392                 }
 393                 case SHMDT:
 394                         err = sys_shmdt(ptr);
 395                         goto out;
 396                 case SHMGET:
 397                         err = sys_shmget(first, (size_t)second, (int)third);
 398                         goto out;
 399                 case SHMCTL:
 400                         err = sys_shmctl(first, (int)second | IPC_64, ptr);
 401                         goto out;
 402                 default:
 403                         err = -ENOSYS;
 404                         goto out;
 405                 }
 406         } else {
 407                 err = -ENOSYS;
 408         }
 409 out:
 410         return err;
 411 }
 412
 413 SYSCALL_DEFINE1(sparc64_personality, unsigned long, personality)
 414 {
 415         int ret;
 416
 417         if (personality(current->personality) == PER_LINUX32 &&
 418             personality(personality) == PER_LINUX)
 419                 personality |= PER_LINUX32;
 420         ret = sys_personality(personality);
 421         if (personality(ret) == PER_LINUX32)
 422                 ret &= ~PER_LINUX32;
 423
 424         return ret;
 425 }
 426
 427 int sparc_mmap_check(unsigned long addr, unsigned long len)
 428 {
 429         if (test_thread_flag(TIF_32BIT)) {
 430                 if (len >= STACK_TOP32)
 431                         return -EINVAL;
 432
 433                 if (addr > STACK_TOP32 - len)
 434                         return -EINVAL;
 435         } else {
 436                 if (len >= VA_EXCLUDE_START)
 437                         return -EINVAL;
 438
 439                 if (invalid_64bit_range(addr, len))
 440                         return -EINVAL;
 441         }
 442
 443         return 0;
 444 }
 445
 446 /* Linux version of mmap */
 447 SYSCALL_DEFINE6(mmap, unsigned long, addr, unsigned long, len,
 448                 unsigned long, prot, unsigned long, flags, unsigned long, fd,
 449                 unsigned long, off)
 450 {
 451         unsigned long retval = -EINVAL;
 452
 453         if ((off + PAGE_ALIGN(len)) < off)
 454                 goto out;
 455         if (off & ~PAGE_MASK)
 456                 goto out;
 457         retval = sys_mmap_pgoff(addr, len, prot, flags, fd, off >> PAGE_SHIFT);
 458 out:
 459         return retval;
 460 }
 461
 462 SYSCALL_DEFINE2(64_munmap, unsigned long, addr, size_t, len)
 463 {
 464         if (invalid_64bit_range(addr, len))
 465                 return -EINVAL;
 466
 467         return vm_munmap(addr, len);
 468 }
 469
 470 SYSCALL_DEFINE5(64_mremap, unsigned long, addr, unsigned long, old_len,
 471                 unsigned long, new_len, unsigned long, flags,
 472                 unsigned long, new_addr)
 473 {
 474         if (test_thread_flag(TIF_32BIT))
 475                 return -EINVAL;
 476         return sys_mremap(addr, old_len, new_len, flags, new_addr);
 477 }
 478
 479 /* we come to here via sys_nis_syscall so it can setup the regs argument */
 480 asmlinkage unsigned long c_sys_nis_syscall(struct pt_regs *regs)
 481 {
 482         static int count;
 483
 484         /* Don't make the system unusable, if someone goes stuck */
 485         if (count++ > 5)
 486                 return -ENOSYS;
 487
 488         printk ("Unimplemented SPARC system call %ld\n",regs->u_regs[1]);
 489 #ifdef DEBUG_UNIMP_SYSCALL
 490         show_regs (regs);
 491 #endif
 492
 493         return -ENOSYS;
 494 }
 495
 496 /* #define DEBUG_SPARC_BREAKPOINT */
 497
 498 asmlinkage void sparc_breakpoint(struct pt_regs *regs)
 499 {
 500         enum ctx_state prev_state = exception_enter();
 501         siginfo_t info;
 502
 503         if (test_thread_flag(TIF_32BIT)) {
 504                 regs->tpc &= 0xffffffff;
 505                 regs->tnpc &= 0xffffffff;
 506         }
 507 #ifdef DEBUG_SPARC_BREAKPOINT
 508         printk ("TRAP: Entering kernel PC=%lx, nPC=%lx\n", regs->tpc, regs->tnpc);
 509 #endif
 510         info.si_signo = SIGTRAP;
 511         info.si_errno = 0;
 512         info.si_code = TRAP_BRKPT;
 513         info.si_addr = (void __user *)regs->tpc;
 514         info.si_trapno = 0;
 515         force_sig_info(SIGTRAP, &info, current);
 516 #ifdef DEBUG_SPARC_BREAKPOINT
 517         printk ("TRAP: Returning to space: PC=%lx nPC=%lx\n", regs->tpc, regs->tnpc);
 518 #endif
 519         exception_exit(prev_state);
 520 }
 521
 522 extern void check_pending(int signum);
 523
 524 SYSCALL_DEFINE2(getdomainname, char __user *, name, int, len)
 525 {
 526         int nlen, err;
 527
 528         if (len < 0)
 529                 return -EINVAL;
 530
 531         down_read(&uts_sem);
 532
 533         nlen = strlen(utsname()->domainname) + 1;
 534         err = -EINVAL;
 535         if (nlen > len)
 536                 goto out;
 537
 538         err = -EFAULT;
 539         if (!copy_to_user(name, utsname()->domainname, nlen))
 540                 err = 0;
 541
 542 out:
 543         up_read(&uts_sem);
 544         return err;
 545 }
 546
 547 SYSCALL_DEFINE5(utrap_install, utrap_entry_t, type,
 548                 utrap_handler_t, new_p, utrap_handler_t, new_d,
 549                 utrap_handler_t __user *, old_p,
 550                 utrap_handler_t __user *, old_d)
 551 {
 552         if (type < UT_INSTRUCTION_EXCEPTION || type > UT_TRAP_INSTRUCTION_31)
 553                 return -EINVAL;
 554         if (new_p == (utrap_handler_t)(long)UTH_NOCHANGE) {
 555                 if (old_p) {
 556                         if (!current_thread_info()->utraps) {
 557                                 if (put_user(NULL, old_p))
 558                                         return -EFAULT;
 559                         } else {
 560                                 if (put_user((utrap_handler_t)(current_thread_info()->utraps[type]), old_p))
 561                                         return -EFAULT;
 562                         }
 563                 }
 564                 if (old_d) {
 565                         if (put_user(NULL, old_d))
 566                                 return -EFAULT;
 567                 }
 568                 return 0;
 569         }
 570         if (!current_thread_info()->utraps) {
 571                 current_thread_info()->utraps =
 572                         kzalloc((UT_TRAP_INSTRUCTION_31+1)*sizeof(long), GFP_KERNEL);
 573                 if (!current_thread_info()->utraps)
 574                         return -ENOMEM;
 575                 current_thread_info()->utraps[0] = 1;
 576         } else {
 577                 if ((utrap_handler_t)current_thread_info()->utraps[type] != new_p &&
 578                     current_thread_info()->utraps[0] > 1) {
 579                         unsigned long *p = current_thread_info()->utraps;
 580
 581                         current_thread_info()->utraps =
 582                                 kmalloc((UT_TRAP_INSTRUCTION_31+1)*sizeof(long),
 583                                         GFP_KERNEL);
 584                         if (!current_thread_info()->utraps) {
 585                                 current_thread_info()->utraps = p;
 586                                 return -ENOMEM;
 587                         }
 588                         p[0]--;
 589                         current_thread_info()->utraps[0] = 1;
 590                         memcpy(current_thread_info()->utraps+1, p+1,
 591                                UT_TRAP_INSTRUCTION_31*sizeof(long));
 592                 }
 593         }
 594         if (old_p) {
 595                 if (put_user((utrap_handler_t)(current_thread_info()->utraps[type]), old_p))
 596                         return -EFAULT;
 597         }
 598         if (old_d) {
 599                 if (put_user(NULL, old_d))
 600                         return -EFAULT;
 601         }
 602         current_thread_info()->utraps[type] = (long)new_p;
 603
 604         return 0;
 605 }
 606
 607 asmlinkage long sparc_memory_ordering(unsigned long model,
 608                                       struct pt_regs *regs)
 609 {
 610         if (model >= 3)
 611                 return -EINVAL;
 612         regs->tstate = (regs->tstate & ~TSTATE_MM) | (model << 14);
 613         return 0;
 614 }
 615
 616 SYSCALL_DEFINE5(rt_sigaction, int, sig, const struct sigaction __user *, act,
 617                 struct sigaction __user *, oact, void __user *, restorer,
 618                 size_t, sigsetsize)
 619 {
 620         struct k_sigaction new_ka, old_ka;
 621         int ret;
 622
 623         /* XXX: Don't preclude handling different sized sigset_t's.  */
 624         if (sigsetsize != sizeof(sigset_t))
 625                 return -EINVAL;
 626
 627         if (act) {
 628                 new_ka.ka_restorer = restorer;
 629                 if (copy_from_user(&new_ka.sa, act, sizeof(*act)))
 630                         return -EFAULT;
 631         }
 632
 633         ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
 634
 635         if (!ret && oact) {
 636                 if (copy_to_user(oact, &old_ka.sa, sizeof(*oact)))
 637                         return -EFAULT;
 638         }
 639
 640         return ret;
 641 }
 642
 643 asmlinkage long sys_kern_features(void)
 644 {
 645         return KERN_FEATURE_MIXED_MODE_STACK;
 646 }