sysv_shm.c revision 132776
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: head/sys/kern/sysv_shm.c 132776 2004-07-28 06:42:41Z kan $"); 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/module.h> 48#include <sys/mutex.h> 49#include <sys/resourcevar.h> 50#include <sys/stat.h> 51#include <sys/syscall.h> 52#include <sys/syscallsubr.h> 53#include <sys/sysent.h> 54#include <sys/sysproto.h> 55#include <sys/jail.h> 56 57#include <vm/vm.h> 58#include <vm/vm_param.h> 59#include <vm/pmap.h> 60#include <vm/vm_object.h> 61#include <vm/vm_map.h> 62#include <vm/vm_page.h> 63#include <vm/vm_pager.h> 64 65static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments"); 66 67struct oshmctl_args; 68static int oshmctl(struct thread *td, struct oshmctl_args *uap); 69 70static int shmget_allocate_segment(struct thread *td, 71 struct shmget_args *uap, int mode); 72static int shmget_existing(struct thread *td, struct shmget_args *uap, 73 int mode, int segnum); 74 75/* XXX casting to (sy_call_t *) is bogus, as usual. */ 76static sy_call_t *shmcalls[] = { 77 (sy_call_t *)shmat, (sy_call_t *)oshmctl, 78 (sy_call_t *)shmdt, (sy_call_t *)shmget, 79 (sy_call_t *)shmctl 80}; 81 82#define SHMSEG_FREE 0x0200 83#define SHMSEG_REMOVED 0x0400 84#define SHMSEG_ALLOCATED 0x0800 85#define SHMSEG_WANTED 0x1000 86 87static int shm_last_free, shm_nused, shm_committed, shmalloced; 88static struct shmid_ds *shmsegs; 89 90struct shmmap_state { 91 vm_offset_t va; 92 int shmid; 93}; 94 95static void shm_deallocate_segment(struct shmid_ds *); 96static int shm_find_segment_by_key(key_t); 97static struct shmid_ds *shm_find_segment_by_shmid(int); 98static struct shmid_ds *shm_find_segment_by_shmidx(int); 99static int shm_delete_mapping(struct vmspace *vm, struct shmmap_state *); 100static void shmrealloc(void); 101static void shminit(void); 102static int sysvshm_modload(struct module *, int, void *); 103static int shmunload(void); 104static void shmexit_myhook(struct vmspace *vm); 105static void shmfork_myhook(struct proc *p1, struct proc *p2); 106static int sysctl_shmsegs(SYSCTL_HANDLER_ARGS); 107 108/* 109 * Tuneable values. 110 */ 111#ifndef SHMMAXPGS 112#define SHMMAXPGS 8192 /* Note: sysv shared memory is swap backed. */ 113#endif 114#ifndef SHMMAX 115#define SHMMAX (SHMMAXPGS*PAGE_SIZE) 116#endif 117#ifndef SHMMIN 118#define SHMMIN 1 119#endif 120#ifndef SHMMNI 121#define SHMMNI 192 122#endif 123#ifndef SHMSEG 124#define SHMSEG 128 125#endif 126#ifndef SHMALL 127#define SHMALL (SHMMAXPGS) 128#endif 129 130struct shminfo shminfo = { 131 SHMMAX, 132 SHMMIN, 133 SHMMNI, 134 SHMSEG, 135 SHMALL 136}; 137 138static int shm_use_phys; 139static int shm_allow_removed; 140 141SYSCTL_DECL(_kern_ipc); 142SYSCTL_INT(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RW, &shminfo.shmmax, 0, ""); 143SYSCTL_INT(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RW, &shminfo.shmmin, 0, ""); 144SYSCTL_INT(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RDTUN, &shminfo.shmmni, 0, ""); 145SYSCTL_INT(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RDTUN, &shminfo.shmseg, 0, ""); 146SYSCTL_INT(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RW, &shminfo.shmall, 0, ""); 147SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RW, 148 &shm_use_phys, 0, ""); 149SYSCTL_INT(_kern_ipc, OID_AUTO, shm_allow_removed, CTLFLAG_RW, 150 &shm_allow_removed, 0, ""); 151SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLFLAG_RD, 152 NULL, 0, sysctl_shmsegs, "", ""); 153 154static int 155shm_find_segment_by_key(key) 156 key_t key; 157{ 158 int i; 159 160 for (i = 0; i < shmalloced; i++) 161 if ((shmsegs[i].shm_perm.mode & SHMSEG_ALLOCATED) && 162 shmsegs[i].shm_perm.key == key) 163 return (i); 164 return (-1); 165} 166 167static struct shmid_ds * 168shm_find_segment_by_shmid(int shmid) 169{ 170 int segnum; 171 struct shmid_ds *shmseg; 172 173 segnum = IPCID_TO_IX(shmid); 174 if (segnum < 0 || segnum >= shmalloced) 175 return (NULL); 176 shmseg = &shmsegs[segnum]; 177 if ((shmseg->shm_perm.mode & SHMSEG_ALLOCATED) == 0 || 178 (!shm_allow_removed && 179 (shmseg->shm_perm.mode & SHMSEG_REMOVED) != 0) || 180 shmseg->shm_perm.seq != IPCID_TO_SEQ(shmid)) 181 return (NULL); 182 return (shmseg); 183} 184 185static struct shmid_ds * 186shm_find_segment_by_shmidx(int segnum) 187{ 188 struct shmid_ds *shmseg; 189 190 if (segnum < 0 || segnum >= shmalloced) 191 return (NULL); 192 shmseg = &shmsegs[segnum]; 193 if ((shmseg->shm_perm.mode & SHMSEG_ALLOCATED) == 0 || 194 (!shm_allow_removed && 195 (shmseg->shm_perm.mode & SHMSEG_REMOVED) != 0)) 196 return (NULL); 197 return (shmseg); 198} 199 200static void 201shm_deallocate_segment(shmseg) 202 struct shmid_ds *shmseg; 203{ 204 size_t size; 205 206 GIANT_REQUIRED; 207 208 vm_object_deallocate(shmseg->shm_internal); 209 shmseg->shm_internal = NULL; 210 size = round_page(shmseg->shm_segsz); 211 shm_committed -= btoc(size); 212 shm_nused--; 213 shmseg->shm_perm.mode = SHMSEG_FREE; 214} 215 216static int 217shm_delete_mapping(struct vmspace *vm, struct shmmap_state *shmmap_s) 218{ 219 struct shmid_ds *shmseg; 220 int segnum, result; 221 size_t size; 222 223 GIANT_REQUIRED; 224 225 segnum = IPCID_TO_IX(shmmap_s->shmid); 226 shmseg = &shmsegs[segnum]; 227 size = round_page(shmseg->shm_segsz); 228 result = vm_map_remove(&vm->vm_map, shmmap_s->va, shmmap_s->va + size); 229 if (result != KERN_SUCCESS) 230 return (EINVAL); 231 shmmap_s->shmid = -1; 232 shmseg->shm_dtime = time_second; 233 if ((--shmseg->shm_nattch <= 0) && 234 (shmseg->shm_perm.mode & SHMSEG_REMOVED)) { 235 shm_deallocate_segment(shmseg); 236 shm_last_free = segnum; 237 } 238 return (0); 239} 240 241#ifndef _SYS_SYSPROTO_H_ 242struct shmdt_args { 243 const void *shmaddr; 244}; 245#endif 246 247/* 248 * MPSAFE 249 */ 250int 251shmdt(td, uap) 252 struct thread *td; 253 struct shmdt_args *uap; 254{ 255 struct proc *p = td->td_proc; 256 struct shmmap_state *shmmap_s; 257 int i; 258 int error = 0; 259 260 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 261 return (ENOSYS); 262 mtx_lock(&Giant); 263 shmmap_s = p->p_vmspace->vm_shm; 264 if (shmmap_s == NULL) { 265 error = EINVAL; 266 goto done2; 267 } 268 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) { 269 if (shmmap_s->shmid != -1 && 270 shmmap_s->va == (vm_offset_t)uap->shmaddr) { 271 break; 272 } 273 } 274 if (i == shminfo.shmseg) { 275 error = EINVAL; 276 goto done2; 277 } 278 error = shm_delete_mapping(p->p_vmspace, shmmap_s); 279done2: 280 mtx_unlock(&Giant); 281 return (error); 282} 283 284#ifndef _SYS_SYSPROTO_H_ 285struct shmat_args { 286 int shmid; 287 const void *shmaddr; 288 int shmflg; 289}; 290#endif 291 292/* 293 * MPSAFE 294 */ 295int 296kern_shmat(td, shmid, shmaddr, shmflg) 297 struct thread *td; 298 int shmid; 299 const void *shmaddr; 300 int shmflg; 301{ 302 struct proc *p = td->td_proc; 303 int i, flags; 304 struct shmid_ds *shmseg; 305 struct shmmap_state *shmmap_s = NULL; 306 vm_offset_t attach_va; 307 vm_prot_t prot; 308 vm_size_t size; 309 int rv; 310 int error = 0; 311 312 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 313 return (ENOSYS); 314 mtx_lock(&Giant); 315 shmmap_s = p->p_vmspace->vm_shm; 316 if (shmmap_s == NULL) { 317 size = shminfo.shmseg * sizeof(struct shmmap_state); 318 shmmap_s = malloc(size, M_SHM, M_WAITOK); 319 for (i = 0; i < shminfo.shmseg; i++) 320 shmmap_s[i].shmid = -1; 321 p->p_vmspace->vm_shm = shmmap_s; 322 } 323 shmseg = shm_find_segment_by_shmid(shmid); 324 if (shmseg == NULL) { 325 error = EINVAL; 326 goto done2; 327 } 328 error = ipcperm(td, &shmseg->shm_perm, 329 (shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W); 330 if (error) 331 goto done2; 332 for (i = 0; i < shminfo.shmseg; i++) { 333 if (shmmap_s->shmid == -1) 334 break; 335 shmmap_s++; 336 } 337 if (i >= shminfo.shmseg) { 338 error = EMFILE; 339 goto done2; 340 } 341 size = round_page(shmseg->shm_segsz); 342#ifdef VM_PROT_READ_IS_EXEC 343 prot = VM_PROT_READ | VM_PROT_EXECUTE; 344#else 345 prot = VM_PROT_READ; 346#endif 347 if ((shmflg & SHM_RDONLY) == 0) 348 prot |= VM_PROT_WRITE; 349 flags = MAP_ANON | MAP_SHARED; 350 if (shmaddr) { 351 flags |= MAP_FIXED; 352 if (shmflg & SHM_RND) { 353 attach_va = (vm_offset_t)shmaddr & ~(SHMLBA-1); 354 } else if (((vm_offset_t)shmaddr & (SHMLBA-1)) == 0) { 355 attach_va = (vm_offset_t)shmaddr; 356 } else { 357 error = EINVAL; 358 goto done2; 359 } 360 } else { 361 /* 362 * This is just a hint to vm_map_find() about where to 363 * put it. 364 */ 365 PROC_LOCK(p); 366 attach_va = round_page((vm_offset_t)p->p_vmspace->vm_daddr + 367 lim_max(p, RLIMIT_DATA)); 368 PROC_UNLOCK(p); 369 } 370 371 vm_object_reference(shmseg->shm_internal); 372 rv = vm_map_find(&p->p_vmspace->vm_map, shmseg->shm_internal, 373 0, &attach_va, size, (flags & MAP_FIXED)?0:1, prot, prot, 0); 374 if (rv != KERN_SUCCESS) { 375 vm_object_deallocate(shmseg->shm_internal); 376 error = ENOMEM; 377 goto done2; 378 } 379 vm_map_inherit(&p->p_vmspace->vm_map, 380 attach_va, attach_va + size, VM_INHERIT_SHARE); 381 382 shmmap_s->va = attach_va; 383 shmmap_s->shmid = shmid; 384 shmseg->shm_lpid = p->p_pid; 385 shmseg->shm_atime = time_second; 386 shmseg->shm_nattch++; 387 td->td_retval[0] = attach_va; 388done2: 389 mtx_unlock(&Giant); 390 return (error); 391} 392 393int 394shmat(td, uap) 395 struct thread *td; 396 struct shmat_args *uap; 397{ 398 return kern_shmat(td, uap->shmid, uap->shmaddr, uap->shmflg); 399} 400 401struct oshmid_ds { 402 struct ipc_perm shm_perm; /* operation perms */ 403 int shm_segsz; /* size of segment (bytes) */ 404 u_short shm_cpid; /* pid, creator */ 405 u_short shm_lpid; /* pid, last operation */ 406 short shm_nattch; /* no. of current attaches */ 407 time_t shm_atime; /* last attach time */ 408 time_t shm_dtime; /* last detach time */ 409 time_t shm_ctime; /* last change time */ 410 void *shm_handle; /* internal handle for shm segment */ 411}; 412 413struct oshmctl_args { 414 int shmid; 415 int cmd; 416 struct oshmid_ds *ubuf; 417}; 418 419/* 420 * MPSAFE 421 */ 422static int 423oshmctl(td, uap) 424 struct thread *td; 425 struct oshmctl_args *uap; 426{ 427#ifdef COMPAT_43 428 int error = 0; 429 struct shmid_ds *shmseg; 430 struct oshmid_ds outbuf; 431 432 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 433 return (ENOSYS); 434 mtx_lock(&Giant); 435 shmseg = shm_find_segment_by_shmid(uap->shmid); 436 if (shmseg == NULL) { 437 error = EINVAL; 438 goto done2; 439 } 440 switch (uap->cmd) { 441 case IPC_STAT: 442 error = ipcperm(td, &shmseg->shm_perm, IPC_R); 443 if (error) 444 goto done2; 445 outbuf.shm_perm = shmseg->shm_perm; 446 outbuf.shm_segsz = shmseg->shm_segsz; 447 outbuf.shm_cpid = shmseg->shm_cpid; 448 outbuf.shm_lpid = shmseg->shm_lpid; 449 outbuf.shm_nattch = shmseg->shm_nattch; 450 outbuf.shm_atime = shmseg->shm_atime; 451 outbuf.shm_dtime = shmseg->shm_dtime; 452 outbuf.shm_ctime = shmseg->shm_ctime; 453 outbuf.shm_handle = shmseg->shm_internal; 454 error = copyout(&outbuf, uap->ubuf, sizeof(outbuf)); 455 if (error) 456 goto done2; 457 break; 458 default: 459 error = shmctl(td, (struct shmctl_args *)uap); 460 break; 461 } 462done2: 463 mtx_unlock(&Giant); 464 return (error); 465#else 466 return (EINVAL); 467#endif 468} 469 470#ifndef _SYS_SYSPROTO_H_ 471struct shmctl_args { 472 int shmid; 473 int cmd; 474 struct shmid_ds *buf; 475}; 476#endif 477 478/* 479 * MPSAFE 480 */ 481int 482kern_shmctl(td, shmid, cmd, buf, bufsz) 483 struct thread *td; 484 int shmid; 485 int cmd; 486 void *buf; 487 size_t *bufsz; 488{ 489 int error = 0; 490 struct shmid_ds *shmseg; 491 492 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 493 return (ENOSYS); 494 495 mtx_lock(&Giant); 496 switch (cmd) { 497 case IPC_INFO: 498 memcpy(buf, &shminfo, sizeof(shminfo)); 499 if (bufsz) 500 *bufsz = sizeof(shminfo); 501 td->td_retval[0] = shmalloced; 502 goto done2; 503 case SHM_INFO: { 504 struct shm_info shm_info; 505 shm_info.used_ids = shm_nused; 506 shm_info.shm_rss = 0; /*XXX where to get from ? */ 507 shm_info.shm_tot = 0; /*XXX where to get from ? */ 508 shm_info.shm_swp = 0; /*XXX where to get from ? */ 509 shm_info.swap_attempts = 0; /*XXX where to get from ? */ 510 shm_info.swap_successes = 0; /*XXX where to get from ? */ 511 memcpy(buf, &shm_info, sizeof(shm_info)); 512 if (bufsz) 513 *bufsz = sizeof(shm_info); 514 td->td_retval[0] = shmalloced; 515 goto done2; 516 } 517 } 518 if (cmd == SHM_STAT) 519 shmseg = shm_find_segment_by_shmidx(shmid); 520 else 521 shmseg = shm_find_segment_by_shmid(shmid); 522 if (shmseg == NULL) { 523 error = EINVAL; 524 goto done2; 525 } 526 switch (cmd) { 527 case SHM_STAT: 528 case IPC_STAT: 529 error = ipcperm(td, &shmseg->shm_perm, IPC_R); 530 if (error) 531 goto done2; 532 memcpy(buf, shmseg, sizeof(struct shmid_ds)); 533 if (bufsz) 534 *bufsz = sizeof(struct shmid_ds); 535 if (cmd == SHM_STAT) 536 td->td_retval[0] = IXSEQ_TO_IPCID(shmid, shmseg->shm_perm); 537 break; 538 case IPC_SET: { 539 struct shmid_ds *shmid; 540 541 shmid = (struct shmid_ds *)buf; 542 error = ipcperm(td, &shmseg->shm_perm, IPC_M); 543 if (error) 544 goto done2; 545 shmseg->shm_perm.uid = shmid->shm_perm.uid; 546 shmseg->shm_perm.gid = shmid->shm_perm.gid; 547 shmseg->shm_perm.mode = 548 (shmseg->shm_perm.mode & ~ACCESSPERMS) | 549 (shmid->shm_perm.mode & ACCESSPERMS); 550 shmseg->shm_ctime = time_second; 551 break; 552 } 553 case IPC_RMID: 554 error = ipcperm(td, &shmseg->shm_perm, IPC_M); 555 if (error) 556 goto done2; 557 shmseg->shm_perm.key = IPC_PRIVATE; 558 shmseg->shm_perm.mode |= SHMSEG_REMOVED; 559 if (shmseg->shm_nattch <= 0) { 560 shm_deallocate_segment(shmseg); 561 shm_last_free = IPCID_TO_IX(shmid); 562 } 563 break; 564#if 0 565 case SHM_LOCK: 566 case SHM_UNLOCK: 567#endif 568 default: 569 error = EINVAL; 570 break; 571 } 572done2: 573 mtx_unlock(&Giant); 574 return (error); 575} 576 577int 578shmctl(td, uap) 579 struct thread *td; 580 struct shmctl_args *uap; 581{ 582 int error = 0; 583 struct shmid_ds buf; 584 size_t bufsz; 585 586 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */ 587 if (uap->cmd == IPC_SET) { 588 if ((error = copyin(uap->buf, &buf, sizeof(struct shmid_ds)))) 589 goto done; 590 } 591 592 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz); 593 if (error) 594 goto done; 595 596 /* Cases in which we need to copyout */ 597 switch (uap->cmd) { 598 case IPC_INFO: 599 case SHM_INFO: 600 case SHM_STAT: 601 case IPC_STAT: 602 error = copyout(&buf, uap->buf, bufsz); 603 break; 604 } 605 606done: 607 if (error) { 608 /* Invalidate the return value */ 609 td->td_retval[0] = -1; 610 } 611 return (error); 612} 613 614 615#ifndef _SYS_SYSPROTO_H_ 616struct shmget_args { 617 key_t key; 618 size_t size; 619 int shmflg; 620}; 621#endif 622 623static int 624shmget_existing(td, uap, mode, segnum) 625 struct thread *td; 626 struct shmget_args *uap; 627 int mode; 628 int segnum; 629{ 630 struct shmid_ds *shmseg; 631 int error; 632 633 shmseg = &shmsegs[segnum]; 634 if (shmseg->shm_perm.mode & SHMSEG_REMOVED) { 635 /* 636 * This segment is in the process of being allocated. Wait 637 * until it's done, and look the key up again (in case the 638 * allocation failed or it was freed). 639 */ 640 shmseg->shm_perm.mode |= SHMSEG_WANTED; 641 error = tsleep(shmseg, PLOCK | PCATCH, "shmget", 0); 642 if (error) 643 return (error); 644 return (EAGAIN); 645 } 646 if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL)) 647 return (EEXIST); 648 error = ipcperm(td, &shmseg->shm_perm, mode); 649 if (error) 650 return (error); 651 if (uap->size && uap->size > shmseg->shm_segsz) 652 return (EINVAL); 653 td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->shm_perm); 654 return (0); 655} 656 657static int 658shmget_allocate_segment(td, uap, mode) 659 struct thread *td; 660 struct shmget_args *uap; 661 int mode; 662{ 663 int i, segnum, shmid, size; 664 struct ucred *cred = td->td_ucred; 665 struct shmid_ds *shmseg; 666 vm_object_t shm_object; 667 668 GIANT_REQUIRED; 669 670 if (uap->size < shminfo.shmmin || uap->size > shminfo.shmmax) 671 return (EINVAL); 672 if (shm_nused >= shminfo.shmmni) /* Any shmids left? */ 673 return (ENOSPC); 674 size = round_page(uap->size); 675 if (shm_committed + btoc(size) > shminfo.shmall) 676 return (ENOMEM); 677 if (shm_last_free < 0) { 678 shmrealloc(); /* Maybe expand the shmsegs[] array. */ 679 for (i = 0; i < shmalloced; i++) 680 if (shmsegs[i].shm_perm.mode & SHMSEG_FREE) 681 break; 682 if (i == shmalloced) 683 return (ENOSPC); 684 segnum = i; 685 } else { 686 segnum = shm_last_free; 687 shm_last_free = -1; 688 } 689 shmseg = &shmsegs[segnum]; 690 /* 691 * In case we sleep in malloc(), mark the segment present but deleted 692 * so that noone else tries to create the same key. 693 */ 694 shmseg->shm_perm.mode = SHMSEG_ALLOCATED | SHMSEG_REMOVED; 695 shmseg->shm_perm.key = uap->key; 696 shmseg->shm_perm.seq = (shmseg->shm_perm.seq + 1) & 0x7fff; 697 shmid = IXSEQ_TO_IPCID(segnum, shmseg->shm_perm); 698 699 /* 700 * We make sure that we have allocated a pager before we need 701 * to. 702 */ 703 if (shm_use_phys) { 704 shm_object = 705 vm_pager_allocate(OBJT_PHYS, 0, size, VM_PROT_DEFAULT, 0); 706 } else { 707 shm_object = 708 vm_pager_allocate(OBJT_SWAP, 0, size, VM_PROT_DEFAULT, 0); 709 } 710 VM_OBJECT_LOCK(shm_object); 711 vm_object_clear_flag(shm_object, OBJ_ONEMAPPING); 712 vm_object_set_flag(shm_object, OBJ_NOSPLIT); 713 VM_OBJECT_UNLOCK(shm_object); 714 715 shmseg->shm_internal = shm_object; 716 shmseg->shm_perm.cuid = shmseg->shm_perm.uid = cred->cr_uid; 717 shmseg->shm_perm.cgid = shmseg->shm_perm.gid = cred->cr_gid; 718 shmseg->shm_perm.mode = (shmseg->shm_perm.mode & SHMSEG_WANTED) | 719 (mode & ACCESSPERMS) | SHMSEG_ALLOCATED; 720 shmseg->shm_segsz = uap->size; 721 shmseg->shm_cpid = td->td_proc->p_pid; 722 shmseg->shm_lpid = shmseg->shm_nattch = 0; 723 shmseg->shm_atime = shmseg->shm_dtime = 0; 724 shmseg->shm_ctime = time_second; 725 shm_committed += btoc(size); 726 shm_nused++; 727 if (shmseg->shm_perm.mode & SHMSEG_WANTED) { 728 /* 729 * Somebody else wanted this key while we were asleep. Wake 730 * them up now. 731 */ 732 shmseg->shm_perm.mode &= ~SHMSEG_WANTED; 733 wakeup(shmseg); 734 } 735 td->td_retval[0] = shmid; 736 return (0); 737} 738 739/* 740 * MPSAFE 741 */ 742int 743shmget(td, uap) 744 struct thread *td; 745 struct shmget_args *uap; 746{ 747 int segnum, mode; 748 int error; 749 750 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 751 return (ENOSYS); 752 mtx_lock(&Giant); 753 mode = uap->shmflg & ACCESSPERMS; 754 if (uap->key != IPC_PRIVATE) { 755 again: 756 segnum = shm_find_segment_by_key(uap->key); 757 if (segnum >= 0) { 758 error = shmget_existing(td, uap, mode, segnum); 759 if (error == EAGAIN) 760 goto again; 761 goto done2; 762 } 763 if ((uap->shmflg & IPC_CREAT) == 0) { 764 error = ENOENT; 765 goto done2; 766 } 767 } 768 error = shmget_allocate_segment(td, uap, mode); 769done2: 770 mtx_unlock(&Giant); 771 return (error); 772} 773 774/* 775 * MPSAFE 776 */ 777int 778shmsys(td, uap) 779 struct thread *td; 780 /* XXX actually varargs. */ 781 struct shmsys_args /* { 782 int which; 783 int a2; 784 int a3; 785 int a4; 786 } */ *uap; 787{ 788 int error; 789 790 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 791 return (ENOSYS); 792 if (uap->which < 0 || 793 uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0])) 794 return (EINVAL); 795 mtx_lock(&Giant); 796 error = (*shmcalls[uap->which])(td, &uap->a2); 797 mtx_unlock(&Giant); 798 return (error); 799} 800 801static void 802shmfork_myhook(p1, p2) 803 struct proc *p1, *p2; 804{ 805 struct shmmap_state *shmmap_s; 806 size_t size; 807 int i; 808 809 size = shminfo.shmseg * sizeof(struct shmmap_state); 810 shmmap_s = malloc(size, M_SHM, M_WAITOK); 811 bcopy(p1->p_vmspace->vm_shm, shmmap_s, size); 812 p2->p_vmspace->vm_shm = shmmap_s; 813 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) 814 if (shmmap_s->shmid != -1) 815 shmsegs[IPCID_TO_IX(shmmap_s->shmid)].shm_nattch++; 816} 817 818static void 819shmexit_myhook(struct vmspace *vm) 820{ 821 struct shmmap_state *base, *shm; 822 int i; 823 824 if ((base = vm->vm_shm) != NULL) { 825 vm->vm_shm = NULL; 826 mtx_lock(&Giant); 827 for (i = 0, shm = base; i < shminfo.shmseg; i++, shm++) { 828 if (shm->shmid != -1) 829 shm_delete_mapping(vm, shm); 830 } 831 mtx_unlock(&Giant); 832 free(base, M_SHM); 833 } 834} 835 836static void 837shmrealloc(void) 838{ 839 int i; 840 struct shmid_ds *newsegs; 841 842 if (shmalloced >= shminfo.shmmni) 843 return; 844 845 newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, M_WAITOK); 846 if (newsegs == NULL) 847 return; 848 for (i = 0; i < shmalloced; i++) 849 bcopy(&shmsegs[i], &newsegs[i], sizeof(newsegs[0])); 850 for (; i < shminfo.shmmni; i++) { 851 shmsegs[i].shm_perm.mode = SHMSEG_FREE; 852 shmsegs[i].shm_perm.seq = 0; 853 } 854 free(shmsegs, M_SHM); 855 shmsegs = newsegs; 856 shmalloced = shminfo.shmmni; 857} 858 859static void 860shminit() 861{ 862 int i; 863 864 TUNABLE_INT_FETCH("kern.ipc.shmmaxpgs", &shminfo.shmall); 865 for (i = PAGE_SIZE; i > 0; i--) { 866 shminfo.shmmax = shminfo.shmall * PAGE_SIZE; 867 if (shminfo.shmmax >= shminfo.shmall) 868 break; 869 } 870 TUNABLE_INT_FETCH("kern.ipc.shmmin", &shminfo.shmmin); 871 TUNABLE_INT_FETCH("kern.ipc.shmmni", &shminfo.shmmni); 872 TUNABLE_INT_FETCH("kern.ipc.shmseg", &shminfo.shmseg); 873 TUNABLE_INT_FETCH("kern.ipc.shm_use_phys", &shm_use_phys); 874 875 shmalloced = shminfo.shmmni; 876 shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, M_WAITOK); 877 if (shmsegs == NULL) 878 panic("cannot allocate initial memory for sysvshm"); 879 for (i = 0; i < shmalloced; i++) { 880 shmsegs[i].shm_perm.mode = SHMSEG_FREE; 881 shmsegs[i].shm_perm.seq = 0; 882 } 883 shm_last_free = 0; 884 shm_nused = 0; 885 shm_committed = 0; 886 shmexit_hook = &shmexit_myhook; 887 shmfork_hook = &shmfork_myhook; 888} 889 890static int 891shmunload() 892{ 893 894 if (shm_nused > 0) 895 return (EBUSY); 896 897 free(shmsegs, M_SHM); 898 shmexit_hook = NULL; 899 shmfork_hook = NULL; 900 return (0); 901} 902 903static int 904sysctl_shmsegs(SYSCTL_HANDLER_ARGS) 905{ 906 907 return (SYSCTL_OUT(req, shmsegs, shmalloced * sizeof(shmsegs[0]))); 908} 909 910static int 911sysvshm_modload(struct module *module, int cmd, void *arg) 912{ 913 int error = 0; 914 915 switch (cmd) { 916 case MOD_LOAD: 917 shminit(); 918 break; 919 case MOD_UNLOAD: 920 error = shmunload(); 921 break; 922 case MOD_SHUTDOWN: 923 break; 924 default: 925 error = EINVAL; 926 break; 927 } 928 return (error); 929} 930 931static moduledata_t sysvshm_mod = { 932 "sysvshm", 933 &sysvshm_modload, 934 NULL 935}; 936 937SYSCALL_MODULE_HELPER(shmsys); 938SYSCALL_MODULE_HELPER(shmat); 939SYSCALL_MODULE_HELPER(shmctl); 940SYSCALL_MODULE_HELPER(shmdt); 941SYSCALL_MODULE_HELPER(shmget); 942 943DECLARE_MODULE(sysvshm, sysvshm_mod, 944 SI_SUB_SYSV_SHM, SI_ORDER_FIRST); 945MODULE_VERSION(sysvshm, 1); 946