Chris Zankel | 5a0015d | 2005-06-23 22:01:16 -0700 | [diff] [blame^] | 1 | /* |
| 2 | * arch/xtensa/kernel/syscall.c |
| 3 | * |
| 4 | * This file is subject to the terms and conditions of the GNU General Public |
| 5 | * License. See the file "COPYING" in the main directory of this archive |
| 6 | * for more details. |
| 7 | * |
| 8 | * Copyright (C) 2001 - 2005 Tensilica Inc. |
| 9 | * Copyright (C) 2000 Silicon Graphics, Inc. |
| 10 | * Copyright (C) 1995 - 2000 by Ralf Baechle |
| 11 | * |
| 12 | * Joe Taylor <joe@tensilica.com, joetylr@yahoo.com> |
| 13 | * Marc Gauthier <marc@tensilica.com, marc@alumni.uwaterloo.ca> |
| 14 | * Chris Zankel <chris@zankel.net> |
| 15 | * Kevin Chea |
| 16 | * |
| 17 | */ |
| 18 | |
| 19 | #define DEBUG 0 |
| 20 | |
| 21 | #include <linux/config.h> |
| 22 | #include <linux/linkage.h> |
| 23 | #include <linux/mm.h> |
| 24 | #include <linux/smp.h> |
| 25 | #include <linux/smp_lock.h> |
| 26 | #include <linux/mman.h> |
| 27 | #include <linux/sched.h> |
| 28 | #include <linux/file.h> |
| 29 | #include <linux/slab.h> |
| 30 | #include <linux/utsname.h> |
| 31 | #include <linux/unistd.h> |
| 32 | #include <linux/stringify.h> |
| 33 | #include <linux/syscalls.h> |
| 34 | #include <linux/sem.h> |
| 35 | #include <linux/msg.h> |
| 36 | #include <linux/shm.h> |
| 37 | #include <linux/errno.h> |
| 38 | #include <asm/ptrace.h> |
| 39 | #include <asm/signal.h> |
| 40 | #include <asm/uaccess.h> |
| 41 | #include <asm/hardirq.h> |
| 42 | #include <asm/mman.h> |
| 43 | #include <asm/shmparam.h> |
| 44 | #include <asm/page.h> |
| 45 | #include <asm/ipc.h> |
| 46 | |
| 47 | extern void do_syscall_trace(void); |
| 48 | typedef int (*syscall_t)(void *a0,...); |
| 49 | extern int (*do_syscalls)(struct pt_regs *regs, syscall_t fun, |
| 50 | int narg); |
| 51 | extern syscall_t sys_call_table[]; |
| 52 | extern unsigned char sys_narg_table[]; |
| 53 | |
| 54 | /* |
| 55 | * sys_pipe() is the normal C calling standard for creating a pipe. It's not |
| 56 | * the way unix traditional does this, though. |
| 57 | */ |
| 58 | |
| 59 | int sys_pipe(int __user *userfds) |
| 60 | { |
| 61 | int fd[2]; |
| 62 | int error; |
| 63 | |
| 64 | error = do_pipe(fd); |
| 65 | if (!error) { |
| 66 | if (copy_to_user(userfds, fd, 2 * sizeof(int))) |
| 67 | error = -EFAULT; |
| 68 | } |
| 69 | return error; |
| 70 | } |
| 71 | |
| 72 | /* |
| 73 | * Common code for old and new mmaps. |
| 74 | */ |
| 75 | |
| 76 | static inline long do_mmap2(unsigned long addr, unsigned long len, |
| 77 | unsigned long prot, unsigned long flags, |
| 78 | unsigned long fd, unsigned long pgoff) |
| 79 | { |
| 80 | int error = -EBADF; |
| 81 | struct file * file = NULL; |
| 82 | |
| 83 | flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE); |
| 84 | if (!(flags & MAP_ANONYMOUS)) { |
| 85 | file = fget(fd); |
| 86 | if (!file) |
| 87 | goto out; |
| 88 | } |
| 89 | |
| 90 | down_write(¤t->mm->mmap_sem); |
| 91 | error = do_mmap_pgoff(file, addr, len, prot, flags, pgoff); |
| 92 | up_write(¤t->mm->mmap_sem); |
| 93 | |
| 94 | if (file) |
| 95 | fput(file); |
| 96 | out: |
| 97 | return error; |
| 98 | } |
| 99 | |
| 100 | unsigned long old_mmap(unsigned long addr, size_t len, int prot, |
| 101 | int flags, int fd, off_t offset) |
| 102 | { |
| 103 | return do_mmap2(addr, len, prot, flags, fd, offset >> PAGE_SHIFT); |
| 104 | } |
| 105 | |
| 106 | long sys_mmap2(unsigned long addr, unsigned long len, unsigned long prot, |
| 107 | unsigned long flags, unsigned long fd, unsigned long pgoff) |
| 108 | { |
| 109 | return do_mmap2(addr, len, prot, flags, fd, pgoff); |
| 110 | } |
| 111 | |
| 112 | int sys_fork(struct pt_regs *regs) |
| 113 | { |
| 114 | return do_fork(SIGCHLD, regs->areg[1], regs, 0, NULL, NULL); |
| 115 | } |
| 116 | |
| 117 | int sys_vfork(struct pt_regs *regs) |
| 118 | { |
| 119 | return do_fork(CLONE_VFORK|CLONE_VM|SIGCHLD, regs->areg[1], |
| 120 | regs, 0, NULL, NULL); |
| 121 | } |
| 122 | |
| 123 | int sys_clone(struct pt_regs *regs) |
| 124 | { |
| 125 | unsigned long clone_flags; |
| 126 | unsigned long newsp; |
| 127 | int __user *parent_tidptr, *child_tidptr; |
| 128 | clone_flags = regs->areg[4]; |
| 129 | newsp = regs->areg[3]; |
| 130 | parent_tidptr = (int __user *)regs->areg[5]; |
| 131 | child_tidptr = (int __user *)regs->areg[6]; |
| 132 | if (!newsp) |
| 133 | newsp = regs->areg[1]; |
| 134 | return do_fork(clone_flags,newsp,regs,0,parent_tidptr,child_tidptr); |
| 135 | } |
| 136 | |
| 137 | /* |
| 138 | * sys_execve() executes a new program. |
| 139 | */ |
| 140 | |
| 141 | int sys_execve(struct pt_regs *regs) |
| 142 | { |
| 143 | int error; |
| 144 | char * filename; |
| 145 | |
| 146 | filename = getname((char *) (long)regs->areg[5]); |
| 147 | error = PTR_ERR(filename); |
| 148 | if (IS_ERR(filename)) |
| 149 | goto out; |
| 150 | error = do_execve(filename, (char **) (long)regs->areg[3], |
| 151 | (char **) (long)regs->areg[4], regs); |
| 152 | putname(filename); |
| 153 | |
| 154 | out: |
| 155 | return error; |
| 156 | } |
| 157 | |
| 158 | int sys_uname(struct old_utsname * name) |
| 159 | { |
| 160 | if (name && !copy_to_user(name, &system_utsname, sizeof (*name))) |
| 161 | return 0; |
| 162 | return -EFAULT; |
| 163 | } |
| 164 | |
| 165 | int sys_olduname(struct oldold_utsname * name) |
| 166 | { |
| 167 | int error; |
| 168 | |
| 169 | if (!name) |
| 170 | return -EFAULT; |
| 171 | if (!access_ok(VERIFY_WRITE,name,sizeof(struct oldold_utsname))) |
| 172 | return -EFAULT; |
| 173 | |
| 174 | error = __copy_to_user(&name->sysname,&system_utsname.sysname,__OLD_UTS_LEN); |
| 175 | error -= __put_user(0,name->sysname+__OLD_UTS_LEN); |
| 176 | error -= __copy_to_user(&name->nodename,&system_utsname.nodename,__OLD_UTS_LEN); |
| 177 | error -= __put_user(0,name->nodename+__OLD_UTS_LEN); |
| 178 | error -= __copy_to_user(&name->release,&system_utsname.release,__OLD_UTS_LEN); |
| 179 | error -= __put_user(0,name->release+__OLD_UTS_LEN); |
| 180 | error -= __copy_to_user(&name->version,&system_utsname.version,__OLD_UTS_LEN); |
| 181 | error -= __put_user(0,name->version+__OLD_UTS_LEN); |
| 182 | error -= __copy_to_user(&name->machine,&system_utsname.machine,__OLD_UTS_LEN); |
| 183 | error -= __put_user(0,name->machine+__OLD_UTS_LEN); |
| 184 | |
| 185 | return error ? -EFAULT : 0; |
| 186 | } |
| 187 | |
| 188 | |
| 189 | /* |
| 190 | * Build the string table for the builtin "poor man's strace". |
| 191 | */ |
| 192 | |
| 193 | #if DEBUG |
| 194 | #define SYSCALL(fun, narg) #fun, |
| 195 | static char *sfnames[] = { |
| 196 | #include "syscalls.h" |
| 197 | }; |
| 198 | #undef SYS |
| 199 | #endif |
| 200 | |
| 201 | void system_call (struct pt_regs *regs) |
| 202 | { |
| 203 | syscall_t syscall; |
| 204 | unsigned long parm0, parm1, parm2, parm3, parm4, parm5; |
| 205 | int nargs, res; |
| 206 | unsigned int syscallnr; |
| 207 | int ps; |
| 208 | |
| 209 | #if DEBUG |
| 210 | int i; |
| 211 | unsigned long parms[6]; |
| 212 | char *sysname; |
| 213 | #endif |
| 214 | |
| 215 | regs->syscall = regs->areg[2]; |
| 216 | |
| 217 | do_syscall_trace(); |
| 218 | |
| 219 | /* Have to load after syscall_trace because strace |
| 220 | * sometimes changes regs->syscall. |
| 221 | */ |
| 222 | syscallnr = regs->syscall; |
| 223 | |
| 224 | parm0 = parm1 = parm2 = parm3 = parm4 = parm5 = 0; |
| 225 | |
| 226 | /* Restore interrupt level to syscall invoker's. |
| 227 | * If this were in assembly, we wouldn't disable |
| 228 | * interrupts in the first place: |
| 229 | */ |
| 230 | local_save_flags (ps); |
| 231 | local_irq_restore((ps & ~XCHAL_PS_INTLEVEL_MASK) | |
| 232 | (regs->ps & XCHAL_PS_INTLEVEL_MASK) ); |
| 233 | |
| 234 | if (syscallnr > __NR_Linux_syscalls) { |
| 235 | regs->areg[2] = -ENOSYS; |
| 236 | return; |
| 237 | } |
| 238 | |
| 239 | syscall = sys_call_table[syscallnr]; |
| 240 | nargs = sys_narg_table[syscallnr]; |
| 241 | |
| 242 | if (syscall == NULL) { |
| 243 | regs->areg[2] = -ENOSYS; |
| 244 | return; |
| 245 | } |
| 246 | |
| 247 | /* There shouldn't be more than six arguments in the table! */ |
| 248 | |
| 249 | if (nargs > 6) |
| 250 | panic("Internal error - too many syscall arguments (%d)!\n", |
| 251 | nargs); |
| 252 | |
| 253 | /* Linux takes system-call arguments in registers. The ABI |
| 254 | * and Xtensa software conventions require the system-call |
| 255 | * number in a2. If an argument exists in a2, we move it to |
| 256 | * the next available register. Note that for improved |
| 257 | * efficiency, we do NOT shift all parameters down one |
| 258 | * register to maintain the original order. |
| 259 | * |
| 260 | * At best case (zero arguments), we just write the syscall |
| 261 | * number to a2. At worst case (1 to 6 arguments), we move |
| 262 | * the argument in a2 to the next available register, then |
| 263 | * write the syscall number to a2. |
| 264 | * |
| 265 | * For clarity, the following truth table enumerates all |
| 266 | * possibilities. |
| 267 | * |
| 268 | * arguments syscall number arg0, arg1, arg2, arg3, arg4, arg5 |
| 269 | * --------- -------------- ---------------------------------- |
| 270 | * 0 a2 |
| 271 | * 1 a2 a3 |
| 272 | * 2 a2 a4, a3 |
| 273 | * 3 a2 a5, a3, a4 |
| 274 | * 4 a2 a6, a3, a4, a5 |
| 275 | * 5 a2 a7, a3, a4, a5, a6 |
| 276 | * 6 a2 a8, a3, a4, a5, a6, a7 |
| 277 | */ |
| 278 | if (nargs) { |
| 279 | parm0 = regs->areg[nargs+2]; |
| 280 | parm1 = regs->areg[3]; |
| 281 | parm2 = regs->areg[4]; |
| 282 | parm3 = regs->areg[5]; |
| 283 | parm4 = regs->areg[6]; |
| 284 | parm5 = regs->areg[7]; |
| 285 | } else /* nargs == 0 */ |
| 286 | parm0 = (unsigned long) regs; |
| 287 | |
| 288 | #if DEBUG |
| 289 | parms[0] = parm0; |
| 290 | parms[1] = parm1; |
| 291 | parms[2] = parm2; |
| 292 | parms[3] = parm3; |
| 293 | parms[4] = parm4; |
| 294 | parms[5] = parm5; |
| 295 | |
| 296 | sysname = sfnames[syscallnr]; |
| 297 | if (strncmp(sysname, "sys_", 4) == 0) |
| 298 | sysname = sysname + 4; |
| 299 | |
| 300 | printk("\017SYSCALL:I:%x:%d:%s %s(", regs->pc, current->pid, |
| 301 | current->comm, sysname); |
| 302 | for (i = 0; i < nargs; i++) |
| 303 | printk((i>0) ? ", %#lx" : "%#lx", parms[i]); |
| 304 | printk(")\n"); |
| 305 | #endif |
| 306 | |
| 307 | res = syscall((void *)parm0, parm1, parm2, parm3, parm4, parm5); |
| 308 | |
| 309 | #if DEBUG |
| 310 | printk("\017SYSCALL:O:%d:%s %s(",current->pid, current->comm, sysname); |
| 311 | for (i = 0; i < nargs; i++) |
| 312 | printk((i>0) ? ", %#lx" : "%#lx", parms[i]); |
| 313 | if (res < 4096) |
| 314 | printk(") = %d\n", res); |
| 315 | else |
| 316 | printk(") = %#x\n", res); |
| 317 | #endif /* DEBUG */ |
| 318 | |
| 319 | regs->areg[2] = res; |
| 320 | do_syscall_trace(); |
| 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 | int sys_ipc (uint call, int first, int second, |
| 330 | int third, void __user *ptr, long fifth) |
| 331 | { |
| 332 | int version, ret; |
| 333 | |
| 334 | version = call >> 16; /* hack for backward compatibility */ |
| 335 | call &= 0xffff; |
| 336 | ret = -ENOSYS; |
| 337 | |
| 338 | switch (call) { |
| 339 | case SEMOP: |
| 340 | ret = sys_semtimedop (first, (struct sembuf __user *)ptr, |
| 341 | second, NULL); |
| 342 | break; |
| 343 | |
| 344 | case SEMTIMEDOP: |
| 345 | ret = sys_semtimedop (first, (struct sembuf __user *)ptr, |
| 346 | second, (const struct timespec *) fifth); |
| 347 | break; |
| 348 | |
| 349 | case SEMGET: |
| 350 | ret = sys_semget (first, second, third); |
| 351 | break; |
| 352 | |
| 353 | case SEMCTL: { |
| 354 | union semun fourth; |
| 355 | |
| 356 | if (ptr && !get_user(fourth.__pad, (void *__user *) ptr)) |
| 357 | ret = sys_semctl (first, second, third, fourth); |
| 358 | break; |
| 359 | } |
| 360 | |
| 361 | case MSGSND: |
| 362 | ret = sys_msgsnd (first, (struct msgbuf __user*) ptr, |
| 363 | second, third); |
| 364 | break; |
| 365 | |
| 366 | case MSGRCV: |
| 367 | switch (version) { |
| 368 | case 0: { |
| 369 | struct ipc_kludge tmp; |
| 370 | |
| 371 | if (ptr && !copy_from_user(&tmp, |
| 372 | (struct ipc_kludge *) ptr, |
| 373 | sizeof (tmp))) |
| 374 | ret = sys_msgrcv (first, tmp.msgp, second, |
| 375 | tmp.msgtyp, third); |
| 376 | break; |
| 377 | } |
| 378 | |
| 379 | default: |
| 380 | ret = sys_msgrcv (first, (struct msgbuf __user *) ptr, |
| 381 | second, 0, third); |
| 382 | break; |
| 383 | } |
| 384 | break; |
| 385 | |
| 386 | case MSGGET: |
| 387 | ret = sys_msgget ((key_t) first, second); |
| 388 | break; |
| 389 | |
| 390 | case MSGCTL: |
| 391 | ret = sys_msgctl (first, second, (struct msqid_ds __user*) ptr); |
| 392 | break; |
| 393 | |
| 394 | case SHMAT: { |
| 395 | ulong raddr; |
| 396 | ret = do_shmat (first, (char __user *) ptr, second, &raddr); |
| 397 | |
| 398 | if (!ret) |
| 399 | ret = put_user (raddr, (ulong __user *) third); |
| 400 | |
| 401 | break; |
| 402 | } |
| 403 | |
| 404 | case SHMDT: |
| 405 | ret = sys_shmdt ((char __user *)ptr); |
| 406 | break; |
| 407 | |
| 408 | case SHMGET: |
| 409 | ret = sys_shmget (first, second, third); |
| 410 | break; |
| 411 | |
| 412 | case SHMCTL: |
| 413 | ret = sys_shmctl (first, second, (struct shmid_ds __user*) ptr); |
| 414 | break; |
| 415 | } |
| 416 | return ret; |
| 417 | } |
| 418 | |