1 /* $NetBSD: sysv_shm.c,v 1.23 1994/07/04 23:25:12 glass Exp $ */ 2 /* 3 * Copyright (c) 1994 Adam Glass and Charles Hannum. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by Adam Glass and Charles 16 * Hannum. 17 * 4. The names of the authors may not be used to endorse or promote products 18 * derived from this software without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR 21 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 22 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 23 * IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT, 24 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 25 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 26 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 27 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 28 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 29 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_compat.h" 36 #include "opt_sysvipc.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/kernel.h> 41 #include <sys/lock.h> 42 #include <sys/sysctl.h> 43 #include <sys/shm.h> 44 #include <sys/proc.h> 45 #include <sys/malloc.h> 46 #include <sys/mman.h> 47 #include <sys/mutex.h> 48 #include <sys/stat.h> 49 #include <sys/syscall.h> 50 #include <sys/syscallsubr.h> 51 #include <sys/sysent.h> 52 #include <sys/sysproto.h> 53 #include <sys/jail.h> 54 55 #include <vm/vm.h> 56 #include <vm/vm_param.h> 57 #include <vm/pmap.h> 58 #include <vm/vm_object.h> 59 #include <vm/vm_map.h> 60 #include <vm/vm_page.h> 61 #include <vm/vm_pager.h> 62 63 static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments"); 64 65 struct oshmctl_args; 66 static int oshmctl(struct thread *td, struct oshmctl_args *uap); 67 68 static int shmget_allocate_segment(struct thread *td, 69 struct shmget_args *uap, int mode); 70 static int shmget_existing(struct thread *td, struct shmget_args *uap, 71 int mode, int segnum); 72 73 /* XXX casting to (sy_call_t *) is bogus, as usual. */ 74 static sy_call_t *shmcalls[] = { 75 (sy_call_t *)shmat, (sy_call_t *)oshmctl, 76 (sy_call_t *)shmdt, (sy_call_t *)shmget, 77 (sy_call_t *)shmctl 78 }; 79 80 #define SHMSEG_FREE 0x0200 81 #define SHMSEG_REMOVED 0x0400 82 #define SHMSEG_ALLOCATED 0x0800 83 #define SHMSEG_WANTED 0x1000 84 85 static int shm_last_free, shm_nused, shm_committed, shmalloced; 86 static struct shmid_ds *shmsegs; 87 88 struct shm_handle { 89 /* vm_offset_t kva; */ 90 vm_object_t shm_object; 91 }; 92 93 struct shmmap_state { 94 vm_offset_t va; 95 int shmid; 96 }; 97 98 static void shm_deallocate_segment(struct shmid_ds *); 99 static int shm_find_segment_by_key(key_t); 100 static struct shmid_ds *shm_find_segment_by_shmid(int); 101 static struct shmid_ds *shm_find_segment_by_shmidx(int); 102 static int shm_delete_mapping(struct vmspace *vm, struct shmmap_state *); 103 static void shmrealloc(void); 104 static void shminit(void); 105 static int sysvshm_modload(struct module *, int, void *); 106 static int shmunload(void); 107 static void shmexit_myhook(struct vmspace *vm); 108 static void shmfork_myhook(struct proc *p1, struct proc *p2); 109 static int sysctl_shmsegs(SYSCTL_HANDLER_ARGS); 110 111 /* 112 * Tuneable values. 113 */ 114 #ifndef SHMMAXPGS 115 #define SHMMAXPGS 8192 /* Note: sysv shared memory is swap backed. */ 116 #endif 117 #ifndef SHMMAX 118 #define SHMMAX (SHMMAXPGS*PAGE_SIZE) 119 #endif 120 #ifndef SHMMIN 121 #define SHMMIN 1 122 #endif 123 #ifndef SHMMNI 124 #define SHMMNI 192 125 #endif 126 #ifndef SHMSEG 127 #define SHMSEG 128 128 #endif 129 #ifndef SHMALL 130 #define SHMALL (SHMMAXPGS) 131 #endif 132 133 struct shminfo shminfo = { 134 SHMMAX, 135 SHMMIN, 136 SHMMNI, 137 SHMSEG, 138 SHMALL 139 }; 140 141 static int shm_use_phys; 142 static int shm_allow_removed; 143 144 SYSCTL_DECL(_kern_ipc); 145 SYSCTL_INT(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RW, &shminfo.shmmax, 0, ""); 146 SYSCTL_INT(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RW, &shminfo.shmmin, 0, ""); 147 SYSCTL_INT(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RDTUN, &shminfo.shmmni, 0, ""); 148 SYSCTL_INT(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RDTUN, &shminfo.shmseg, 0, ""); 149 SYSCTL_INT(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RW, &shminfo.shmall, 0, ""); 150 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RW, 151 &shm_use_phys, 0, ""); 152 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_allow_removed, CTLFLAG_RW, 153 &shm_allow_removed, 0, ""); 154 SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLFLAG_RD, 155 NULL, 0, sysctl_shmsegs, "", ""); 156 157 static int 158 shm_find_segment_by_key(key) 159 key_t key; 160 { 161 int i; 162 163 for (i = 0; i < shmalloced; i++) 164 if ((shmsegs[i].shm_perm.mode & SHMSEG_ALLOCATED) && 165 shmsegs[i].shm_perm.key == key) 166 return (i); 167 return (-1); 168 } 169 170 static struct shmid_ds * 171 shm_find_segment_by_shmid(int shmid) 172 { 173 int segnum; 174 struct shmid_ds *shmseg; 175 176 segnum = IPCID_TO_IX(shmid); 177 if (segnum < 0 || segnum >= shmalloced) 178 return (NULL); 179 shmseg = &shmsegs[segnum]; 180 if ((shmseg->shm_perm.mode & SHMSEG_ALLOCATED) == 0 || 181 (!shm_allow_removed && 182 (shmseg->shm_perm.mode & SHMSEG_REMOVED) != 0) || 183 shmseg->shm_perm.seq != IPCID_TO_SEQ(shmid)) 184 return (NULL); 185 return (shmseg); 186 } 187 188 static struct shmid_ds * 189 shm_find_segment_by_shmidx(int segnum) 190 { 191 struct shmid_ds *shmseg; 192 193 if (segnum < 0 || segnum >= shmalloced) 194 return (NULL); 195 shmseg = &shmsegs[segnum]; 196 if ((shmseg->shm_perm.mode & SHMSEG_ALLOCATED) == 0 || 197 (!shm_allow_removed && 198 (shmseg->shm_perm.mode & SHMSEG_REMOVED) != 0)) 199 return (NULL); 200 return (shmseg); 201 } 202 203 static void 204 shm_deallocate_segment(shmseg) 205 struct shmid_ds *shmseg; 206 { 207 struct shm_handle *shm_handle; 208 size_t size; 209 210 GIANT_REQUIRED; 211 212 shm_handle = shmseg->shm_internal; 213 vm_object_deallocate(shm_handle->shm_object); 214 free(shm_handle, M_SHM); 215 shmseg->shm_internal = NULL; 216 size = round_page(shmseg->shm_segsz); 217 shm_committed -= btoc(size); 218 shm_nused--; 219 shmseg->shm_perm.mode = SHMSEG_FREE; 220 } 221 222 static int 223 shm_delete_mapping(struct vmspace *vm, struct shmmap_state *shmmap_s) 224 { 225 struct shmid_ds *shmseg; 226 int segnum, result; 227 size_t size; 228 229 GIANT_REQUIRED; 230 231 segnum = IPCID_TO_IX(shmmap_s->shmid); 232 shmseg = &shmsegs[segnum]; 233 size = round_page(shmseg->shm_segsz); 234 result = vm_map_remove(&vm->vm_map, shmmap_s->va, shmmap_s->va + size); 235 if (result != KERN_SUCCESS) 236 return (EINVAL); 237 shmmap_s->shmid = -1; 238 shmseg->shm_dtime = time_second; 239 if ((--shmseg->shm_nattch <= 0) && 240 (shmseg->shm_perm.mode & SHMSEG_REMOVED)) { 241 shm_deallocate_segment(shmseg); 242 shm_last_free = segnum; 243 } 244 return (0); 245 } 246 247 #ifndef _SYS_SYSPROTO_H_ 248 struct shmdt_args { 249 const void *shmaddr; 250 }; 251 #endif 252 253 /* 254 * MPSAFE 255 */ 256 int 257 shmdt(td, uap) 258 struct thread *td; 259 struct shmdt_args *uap; 260 { 261 struct proc *p = td->td_proc; 262 struct shmmap_state *shmmap_s; 263 int i; 264 int error = 0; 265 266 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 267 return (ENOSYS); 268 mtx_lock(&Giant); 269 shmmap_s = p->p_vmspace->vm_shm; 270 if (shmmap_s == NULL) { 271 error = EINVAL; 272 goto done2; 273 } 274 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) { 275 if (shmmap_s->shmid != -1 && 276 shmmap_s->va == (vm_offset_t)uap->shmaddr) { 277 break; 278 } 279 } 280 if (i == shminfo.shmseg) { 281 error = EINVAL; 282 goto done2; 283 } 284 error = shm_delete_mapping(p->p_vmspace, shmmap_s); 285 done2: 286 mtx_unlock(&Giant); 287 return (error); 288 } 289 290 #ifndef _SYS_SYSPROTO_H_ 291 struct shmat_args { 292 int shmid; 293 const void *shmaddr; 294 int shmflg; 295 }; 296 #endif 297 298 /* 299 * MPSAFE 300 */ 301 int 302 kern_shmat(td, shmid, shmaddr, shmflg) 303 struct thread *td; 304 int shmid; 305 const void *shmaddr; 306 int shmflg; 307 { 308 struct proc *p = td->td_proc; 309 int i, flags; 310 struct shmid_ds *shmseg; 311 struct shmmap_state *shmmap_s = NULL; 312 struct shm_handle *shm_handle; 313 vm_offset_t attach_va; 314 vm_prot_t prot; 315 vm_size_t size; 316 int rv; 317 int error = 0; 318 319 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 320 return (ENOSYS); 321 mtx_lock(&Giant); 322 shmmap_s = p->p_vmspace->vm_shm; 323 if (shmmap_s == NULL) { 324 size = shminfo.shmseg * sizeof(struct shmmap_state); 325 shmmap_s = malloc(size, M_SHM, M_WAITOK); 326 for (i = 0; i < shminfo.shmseg; i++) 327 shmmap_s[i].shmid = -1; 328 p->p_vmspace->vm_shm = shmmap_s; 329 } 330 shmseg = shm_find_segment_by_shmid(shmid); 331 if (shmseg == NULL) { 332 error = EINVAL; 333 goto done2; 334 } 335 error = ipcperm(td, &shmseg->shm_perm, 336 (shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W); 337 if (error) 338 goto done2; 339 for (i = 0; i < shminfo.shmseg; i++) { 340 if (shmmap_s->shmid == -1) 341 break; 342 shmmap_s++; 343 } 344 if (i >= shminfo.shmseg) { 345 error = EMFILE; 346 goto done2; 347 } 348 size = round_page(shmseg->shm_segsz); 349 #ifdef VM_PROT_READ_IS_EXEC 350 prot = VM_PROT_READ | VM_PROT_EXECUTE; 351 #else 352 prot = VM_PROT_READ; 353 #endif 354 if ((shmflg & SHM_RDONLY) == 0) 355 prot |= VM_PROT_WRITE; 356 flags = MAP_ANON | MAP_SHARED; 357 if (shmaddr) { 358 flags |= MAP_FIXED; 359 if (shmflg & SHM_RND) { 360 attach_va = (vm_offset_t)shmaddr & ~(SHMLBA-1); 361 } else if (((vm_offset_t)shmaddr & (SHMLBA-1)) == 0) { 362 attach_va = (vm_offset_t)shmaddr; 363 } else { 364 error = EINVAL; 365 goto done2; 366 } 367 } else { 368 /* 369 * This is just a hint to vm_map_find() about where to 370 * put it. 371 */ 372 attach_va = round_page((vm_offset_t)p->p_vmspace->vm_taddr 373 + maxtsiz + maxdsiz); 374 } 375 376 shm_handle = shmseg->shm_internal; 377 vm_object_reference(shm_handle->shm_object); 378 rv = vm_map_find(&p->p_vmspace->vm_map, shm_handle->shm_object, 379 0, &attach_va, size, (flags & MAP_FIXED)?0:1, prot, prot, 0); 380 if (rv != KERN_SUCCESS) { 381 vm_object_deallocate(shm_handle->shm_object); 382 error = ENOMEM; 383 goto done2; 384 } 385 vm_map_inherit(&p->p_vmspace->vm_map, 386 attach_va, attach_va + size, VM_INHERIT_SHARE); 387 388 shmmap_s->va = attach_va; 389 shmmap_s->shmid = shmid; 390 shmseg->shm_lpid = p->p_pid; 391 shmseg->shm_atime = time_second; 392 shmseg->shm_nattch++; 393 td->td_retval[0] = attach_va; 394 done2: 395 mtx_unlock(&Giant); 396 return (error); 397 } 398 399 int 400 shmat(td, uap) 401 struct thread *td; 402 struct shmat_args *uap; 403 { 404 return kern_shmat(td, uap->shmid, uap->shmaddr, uap->shmflg); 405 } 406 407 struct oshmid_ds { 408 struct ipc_perm shm_perm; /* operation perms */ 409 int shm_segsz; /* size of segment (bytes) */ 410 u_short shm_cpid; /* pid, creator */ 411 u_short shm_lpid; /* pid, last operation */ 412 short shm_nattch; /* no. of current attaches */ 413 time_t shm_atime; /* last attach time */ 414 time_t shm_dtime; /* last detach time */ 415 time_t shm_ctime; /* last change time */ 416 void *shm_handle; /* internal handle for shm segment */ 417 }; 418 419 struct oshmctl_args { 420 int shmid; 421 int cmd; 422 struct oshmid_ds *ubuf; 423 }; 424 425 /* 426 * MPSAFE 427 */ 428 static int 429 oshmctl(td, uap) 430 struct thread *td; 431 struct oshmctl_args *uap; 432 { 433 #ifdef COMPAT_43 434 int error = 0; 435 struct shmid_ds *shmseg; 436 struct oshmid_ds outbuf; 437 438 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 439 return (ENOSYS); 440 mtx_lock(&Giant); 441 shmseg = shm_find_segment_by_shmid(uap->shmid); 442 if (shmseg == NULL) { 443 error = EINVAL; 444 goto done2; 445 } 446 switch (uap->cmd) { 447 case IPC_STAT: 448 error = ipcperm(td, &shmseg->shm_perm, IPC_R); 449 if (error) 450 goto done2; 451 outbuf.shm_perm = shmseg->shm_perm; 452 outbuf.shm_segsz = shmseg->shm_segsz; 453 outbuf.shm_cpid = shmseg->shm_cpid; 454 outbuf.shm_lpid = shmseg->shm_lpid; 455 outbuf.shm_nattch = shmseg->shm_nattch; 456 outbuf.shm_atime = shmseg->shm_atime; 457 outbuf.shm_dtime = shmseg->shm_dtime; 458 outbuf.shm_ctime = shmseg->shm_ctime; 459 outbuf.shm_handle = shmseg->shm_internal; 460 error = copyout(&outbuf, uap->ubuf, sizeof(outbuf)); 461 if (error) 462 goto done2; 463 break; 464 default: 465 /* XXX casting to (sy_call_t *) is bogus, as usual. */ 466 error = ((sy_call_t *)shmctl)(td, uap); 467 break; 468 } 469 done2: 470 mtx_unlock(&Giant); 471 return (error); 472 #else 473 return (EINVAL); 474 #endif 475 } 476 477 #ifndef _SYS_SYSPROTO_H_ 478 struct shmctl_args { 479 int shmid; 480 int cmd; 481 struct shmid_ds *buf; 482 }; 483 #endif 484 485 /* 486 * MPSAFE 487 */ 488 int 489 kern_shmctl(td, shmid, cmd, buf, bufsz) 490 struct thread *td; 491 int shmid; 492 int cmd; 493 void *buf; 494 size_t *bufsz; 495 { 496 int error = 0; 497 struct shmid_ds *shmseg; 498 499 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 500 return (ENOSYS); 501 502 mtx_lock(&Giant); 503 switch (cmd) { 504 case IPC_INFO: 505 memcpy(buf, &shminfo, sizeof(shminfo)); 506 if (bufsz) 507 *bufsz = sizeof(shminfo); 508 td->td_retval[0] = shmalloced; 509 goto done2; 510 case SHM_INFO: { 511 struct shm_info shm_info; 512 shm_info.used_ids = shm_nused; 513 shm_info.shm_rss = 0; /*XXX where to get from ? */ 514 shm_info.shm_tot = 0; /*XXX where to get from ? */ 515 shm_info.shm_swp = 0; /*XXX where to get from ? */ 516 shm_info.swap_attempts = 0; /*XXX where to get from ? */ 517 shm_info.swap_successes = 0; /*XXX where to get from ? */ 518 memcpy(buf, &shm_info, sizeof(shm_info)); 519 if (bufsz) 520 *bufsz = sizeof(shm_info); 521 td->td_retval[0] = shmalloced; 522 goto done2; 523 } 524 } 525 if (cmd == SHM_STAT) 526 shmseg = shm_find_segment_by_shmidx(shmid); 527 else 528 shmseg = shm_find_segment_by_shmid(shmid); 529 if (shmseg == NULL) { 530 error = EINVAL; 531 goto done2; 532 } 533 switch (cmd) { 534 case SHM_STAT: 535 case IPC_STAT: 536 error = ipcperm(td, &shmseg->shm_perm, IPC_R); 537 if (error) 538 goto done2; 539 memcpy(buf, shmseg, sizeof(struct shmid_ds)); 540 if (bufsz) 541 *bufsz = sizeof(struct shmid_ds); 542 if (cmd == SHM_STAT) 543 td->td_retval[0] = IXSEQ_TO_IPCID(shmid, shmseg->shm_perm); 544 break; 545 case IPC_SET: { 546 struct shmid_ds *shmid; 547 548 shmid = (struct shmid_ds *)buf; 549 error = ipcperm(td, &shmseg->shm_perm, IPC_M); 550 if (error) 551 goto done2; 552 shmseg->shm_perm.uid = shmid->shm_perm.uid; 553 shmseg->shm_perm.gid = shmid->shm_perm.gid; 554 shmseg->shm_perm.mode = 555 (shmseg->shm_perm.mode & ~ACCESSPERMS) | 556 (shmid->shm_perm.mode & ACCESSPERMS); 557 shmseg->shm_ctime = time_second; 558 break; 559 } 560 case IPC_RMID: 561 error = ipcperm(td, &shmseg->shm_perm, IPC_M); 562 if (error) 563 goto done2; 564 shmseg->shm_perm.key = IPC_PRIVATE; 565 shmseg->shm_perm.mode |= SHMSEG_REMOVED; 566 if (shmseg->shm_nattch <= 0) { 567 shm_deallocate_segment(shmseg); 568 shm_last_free = IPCID_TO_IX(shmid); 569 } 570 break; 571 #if 0 572 case SHM_LOCK: 573 case SHM_UNLOCK: 574 #endif 575 default: 576 error = EINVAL; 577 break; 578 } 579 done2: 580 mtx_unlock(&Giant); 581 return (error); 582 } 583 584 int 585 shmctl(td, uap) 586 struct thread *td; 587 struct shmctl_args *uap; 588 { 589 int error = 0; 590 struct shmid_ds buf; 591 size_t bufsz; 592 593 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */ 594 if (uap->cmd == IPC_SET) { 595 if ((error = copyin(uap->buf, &buf, sizeof(struct shmid_ds)))) 596 goto done; 597 } 598 599 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz); 600 if (error) 601 goto done; 602 603 /* Cases in which we need to copyout */ 604 switch (uap->cmd) { 605 case IPC_INFO: 606 case SHM_INFO: 607 case SHM_STAT: 608 case IPC_STAT: 609 error = copyout(&buf, uap->buf, bufsz); 610 break; 611 } 612 613 done: 614 if (error) { 615 /* Invalidate the return value */ 616 td->td_retval[0] = -1; 617 } 618 return (error); 619 } 620 621 622 #ifndef _SYS_SYSPROTO_H_ 623 struct shmget_args { 624 key_t key; 625 size_t size; 626 int shmflg; 627 }; 628 #endif 629 630 static int 631 shmget_existing(td, uap, mode, segnum) 632 struct thread *td; 633 struct shmget_args *uap; 634 int mode; 635 int segnum; 636 { 637 struct shmid_ds *shmseg; 638 int error; 639 640 shmseg = &shmsegs[segnum]; 641 if (shmseg->shm_perm.mode & SHMSEG_REMOVED) { 642 /* 643 * This segment is in the process of being allocated. Wait 644 * until it's done, and look the key up again (in case the 645 * allocation failed or it was freed). 646 */ 647 shmseg->shm_perm.mode |= SHMSEG_WANTED; 648 error = tsleep(shmseg, PLOCK | PCATCH, "shmget", 0); 649 if (error) 650 return (error); 651 return (EAGAIN); 652 } 653 if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL)) 654 return (EEXIST); 655 error = ipcperm(td, &shmseg->shm_perm, mode); 656 if (error) 657 return (error); 658 if (uap->size && uap->size > shmseg->shm_segsz) 659 return (EINVAL); 660 td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->shm_perm); 661 return (0); 662 } 663 664 static int 665 shmget_allocate_segment(td, uap, mode) 666 struct thread *td; 667 struct shmget_args *uap; 668 int mode; 669 { 670 int i, segnum, shmid, size; 671 struct ucred *cred = td->td_ucred; 672 struct shmid_ds *shmseg; 673 struct shm_handle *shm_handle; 674 675 GIANT_REQUIRED; 676 677 if (uap->size < shminfo.shmmin || uap->size > shminfo.shmmax) 678 return (EINVAL); 679 if (shm_nused >= shminfo.shmmni) /* Any shmids left? */ 680 return (ENOSPC); 681 size = round_page(uap->size); 682 if (shm_committed + btoc(size) > shminfo.shmall) 683 return (ENOMEM); 684 if (shm_last_free < 0) { 685 shmrealloc(); /* Maybe expand the shmsegs[] array. */ 686 for (i = 0; i < shmalloced; i++) 687 if (shmsegs[i].shm_perm.mode & SHMSEG_FREE) 688 break; 689 if (i == shmalloced) 690 return (ENOSPC); 691 segnum = i; 692 } else { 693 segnum = shm_last_free; 694 shm_last_free = -1; 695 } 696 shmseg = &shmsegs[segnum]; 697 /* 698 * In case we sleep in malloc(), mark the segment present but deleted 699 * so that noone else tries to create the same key. 700 */ 701 shmseg->shm_perm.mode = SHMSEG_ALLOCATED | SHMSEG_REMOVED; 702 shmseg->shm_perm.key = uap->key; 703 shmseg->shm_perm.seq = (shmseg->shm_perm.seq + 1) & 0x7fff; 704 shm_handle = (struct shm_handle *) 705 malloc(sizeof(struct shm_handle), M_SHM, M_WAITOK); 706 shmid = IXSEQ_TO_IPCID(segnum, shmseg->shm_perm); 707 708 /* 709 * We make sure that we have allocated a pager before we need 710 * to. 711 */ 712 if (shm_use_phys) { 713 shm_handle->shm_object = 714 vm_pager_allocate(OBJT_PHYS, 0, size, VM_PROT_DEFAULT, 0); 715 } else { 716 shm_handle->shm_object = 717 vm_pager_allocate(OBJT_SWAP, 0, size, VM_PROT_DEFAULT, 0); 718 } 719 VM_OBJECT_LOCK(shm_handle->shm_object); 720 vm_object_clear_flag(shm_handle->shm_object, OBJ_ONEMAPPING); 721 vm_object_set_flag(shm_handle->shm_object, OBJ_NOSPLIT); 722 VM_OBJECT_UNLOCK(shm_handle->shm_object); 723 724 shmseg->shm_internal = shm_handle; 725 shmseg->shm_perm.cuid = shmseg->shm_perm.uid = cred->cr_uid; 726 shmseg->shm_perm.cgid = shmseg->shm_perm.gid = cred->cr_gid; 727 shmseg->shm_perm.mode = (shmseg->shm_perm.mode & SHMSEG_WANTED) | 728 (mode & ACCESSPERMS) | SHMSEG_ALLOCATED; 729 shmseg->shm_segsz = uap->size; 730 shmseg->shm_cpid = td->td_proc->p_pid; 731 shmseg->shm_lpid = shmseg->shm_nattch = 0; 732 shmseg->shm_atime = shmseg->shm_dtime = 0; 733 shmseg->shm_ctime = time_second; 734 shm_committed += btoc(size); 735 shm_nused++; 736 if (shmseg->shm_perm.mode & SHMSEG_WANTED) { 737 /* 738 * Somebody else wanted this key while we were asleep. Wake 739 * them up now. 740 */ 741 shmseg->shm_perm.mode &= ~SHMSEG_WANTED; 742 wakeup(shmseg); 743 } 744 td->td_retval[0] = shmid; 745 return (0); 746 } 747 748 /* 749 * MPSAFE 750 */ 751 int 752 shmget(td, uap) 753 struct thread *td; 754 struct shmget_args *uap; 755 { 756 int segnum, mode; 757 int error; 758 759 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 760 return (ENOSYS); 761 mtx_lock(&Giant); 762 mode = uap->shmflg & ACCESSPERMS; 763 if (uap->key != IPC_PRIVATE) { 764 again: 765 segnum = shm_find_segment_by_key(uap->key); 766 if (segnum >= 0) { 767 error = shmget_existing(td, uap, mode, segnum); 768 if (error == EAGAIN) 769 goto again; 770 goto done2; 771 } 772 if ((uap->shmflg & IPC_CREAT) == 0) { 773 error = ENOENT; 774 goto done2; 775 } 776 } 777 error = shmget_allocate_segment(td, uap, mode); 778 done2: 779 mtx_unlock(&Giant); 780 return (error); 781 } 782 783 /* 784 * MPSAFE 785 */ 786 int 787 shmsys(td, uap) 788 struct thread *td; 789 /* XXX actually varargs. */ 790 struct shmsys_args /* { 791 int which; 792 int a2; 793 int a3; 794 int a4; 795 } */ *uap; 796 { 797 int error; 798 799 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 800 return (ENOSYS); 801 if (uap->which < 0 || 802 uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0])) 803 return (EINVAL); 804 mtx_lock(&Giant); 805 error = (*shmcalls[uap->which])(td, &uap->a2); 806 mtx_unlock(&Giant); 807 return (error); 808 } 809 810 static void 811 shmfork_myhook(p1, p2) 812 struct proc *p1, *p2; 813 { 814 struct shmmap_state *shmmap_s; 815 size_t size; 816 int i; 817 818 size = shminfo.shmseg * sizeof(struct shmmap_state); 819 shmmap_s = malloc(size, M_SHM, M_WAITOK); 820 bcopy(p1->p_vmspace->vm_shm, shmmap_s, size); 821 p2->p_vmspace->vm_shm = shmmap_s; 822 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) 823 if (shmmap_s->shmid != -1) 824 shmsegs[IPCID_TO_IX(shmmap_s->shmid)].shm_nattch++; 825 } 826 827 static void 828 shmexit_myhook(struct vmspace *vm) 829 { 830 struct shmmap_state *base, *shm; 831 int i; 832 833 GIANT_REQUIRED; 834 835 if ((base = vm->vm_shm) != NULL) { 836 vm->vm_shm = NULL; 837 for (i = 0, shm = base; i < shminfo.shmseg; i++, shm++) { 838 if (shm->shmid != -1) 839 shm_delete_mapping(vm, shm); 840 } 841 free(base, M_SHM); 842 } 843 } 844 845 static void 846 shmrealloc(void) 847 { 848 int i; 849 struct shmid_ds *newsegs; 850 851 if (shmalloced >= shminfo.shmmni) 852 return; 853 854 newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, M_WAITOK); 855 if (newsegs == NULL) 856 return; 857 for (i = 0; i < shmalloced; i++) 858 bcopy(&shmsegs[i], &newsegs[i], sizeof(newsegs[0])); 859 for (; i < shminfo.shmmni; i++) { 860 shmsegs[i].shm_perm.mode = SHMSEG_FREE; 861 shmsegs[i].shm_perm.seq = 0; 862 } 863 free(shmsegs, M_SHM); 864 shmsegs = newsegs; 865 shmalloced = shminfo.shmmni; 866 } 867 868 static void 869 shminit() 870 { 871 int i; 872 873 TUNABLE_INT_FETCH("kern.ipc.shmmaxpgs", &shminfo.shmall); 874 for (i = PAGE_SIZE; i > 0; i--) { 875 shminfo.shmmax = shminfo.shmall * PAGE_SIZE; 876 if (shminfo.shmmax >= shminfo.shmall) 877 break; 878 } 879 TUNABLE_INT_FETCH("kern.ipc.shmmin", &shminfo.shmmin); 880 TUNABLE_INT_FETCH("kern.ipc.shmmni", &shminfo.shmmni); 881 TUNABLE_INT_FETCH("kern.ipc.shmseg", &shminfo.shmseg); 882 TUNABLE_INT_FETCH("kern.ipc.shm_use_phys", &shm_use_phys); 883 884 shmalloced = shminfo.shmmni; 885 shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, M_WAITOK); 886 if (shmsegs == NULL) 887 panic("cannot allocate initial memory for sysvshm"); 888 for (i = 0; i < shmalloced; i++) { 889 shmsegs[i].shm_perm.mode = SHMSEG_FREE; 890 shmsegs[i].shm_perm.seq = 0; 891 } 892 shm_last_free = 0; 893 shm_nused = 0; 894 shm_committed = 0; 895 shmexit_hook = &shmexit_myhook; 896 shmfork_hook = &shmfork_myhook; 897 } 898 899 static int 900 shmunload() 901 { 902 903 if (shm_nused > 0) 904 return (EBUSY); 905 906 free(shmsegs, M_SHM); 907 shmexit_hook = NULL; 908 shmfork_hook = NULL; 909 return (0); 910 } 911 912 static int 913 sysctl_shmsegs(SYSCTL_HANDLER_ARGS) 914 { 915 916 return (SYSCTL_OUT(req, shmsegs, shmalloced * sizeof(shmsegs[0]))); 917 } 918 919 static int 920 sysvshm_modload(struct module *module, int cmd, void *arg) 921 { 922 int error = 0; 923 924 switch (cmd) { 925 case MOD_LOAD: 926 shminit(); 927 break; 928 case MOD_UNLOAD: 929 error = shmunload(); 930 break; 931 case MOD_SHUTDOWN: 932 break; 933 default: 934 error = EINVAL; 935 break; 936 } 937 return (error); 938 } 939 940 static moduledata_t sysvshm_mod = { 941 "sysvshm", 942 &sysvshm_modload, 943 NULL 944 }; 945 946 SYSCALL_MODULE_HELPER(shmsys); 947 SYSCALL_MODULE_HELPER(shmat); 948 SYSCALL_MODULE_HELPER(shmctl); 949 SYSCALL_MODULE_HELPER(shmdt); 950 SYSCALL_MODULE_HELPER(shmget); 951 952 DECLARE_MODULE(sysvshm, sysvshm_mod, 953 SI_SUB_SYSV_SHM, SI_ORDER_FIRST); 954 MODULE_VERSION(sysvshm, 1); 955