1 /* 2 * Copyright (c) 1993, 1995 Jan-Simon Pendry 3 * Copyright (c) 1993, 1995 4 * The Regents of the University of California. All rights reserved. 5 * 6 * This code is derived from software contributed to Berkeley by 7 * Jan-Simon Pendry. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. All advertising materials mentioning features or use of this software 18 * must display the following acknowledgement: 19 * This product includes software developed by the University of 20 * California, Berkeley and its contributors. 21 * 4. Neither the name of the University nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 * 37 * @(#)procfs_vnops.c 8.18 (Berkeley) 5/21/95 38 * 39 * $FreeBSD: src/sys/miscfs/procfs/procfs_vnops.c,v 1.76.2.7 2002/01/22 17:22:59 nectar Exp $ 40 */ 41 42 /* 43 * procfs vnode interface 44 */ 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/time.h> 49 #include <sys/kernel.h> 50 #include <sys/lock.h> 51 #include <sys/fcntl.h> 52 #include <sys/proc.h> 53 #include <sys/priv.h> 54 #include <sys/signalvar.h> 55 #include <sys/vnode.h> 56 #include <sys/uio.h> 57 #include <sys/mount.h> 58 #include <sys/namei.h> 59 #include <sys/dirent.h> 60 #include <sys/malloc.h> 61 #include <sys/reg.h> 62 #include <vm/vm_zone.h> 63 #include <vfs/procfs/procfs.h> 64 #include <sys/pioctl.h> 65 66 #include <sys/spinlock2.h> 67 68 #include <machine/limits.h> 69 70 static int procfs_access (struct vop_access_args *); 71 static int procfs_badop (struct vop_generic_args *); 72 static int procfs_bmap (struct vop_bmap_args *); 73 static int procfs_close (struct vop_close_args *); 74 static int procfs_getattr (struct vop_getattr_args *); 75 static int procfs_inactive (struct vop_inactive_args *); 76 static int procfs_ioctl (struct vop_ioctl_args *); 77 static int procfs_lookup (struct vop_old_lookup_args *); 78 static int procfs_open (struct vop_open_args *); 79 static int procfs_print (struct vop_print_args *); 80 static int procfs_readdir (struct vop_readdir_args *); 81 static int procfs_readlink (struct vop_readlink_args *); 82 static int procfs_reclaim (struct vop_reclaim_args *); 83 static int procfs_setattr (struct vop_setattr_args *); 84 85 static int procfs_readdir_proc(struct vop_readdir_args *); 86 static int procfs_readdir_root(struct vop_readdir_args *); 87 88 /* 89 * procfs vnode operations. 90 */ 91 struct vop_ops procfs_vnode_vops = { 92 .vop_default = vop_defaultop, 93 .vop_access = procfs_access, 94 .vop_advlock = (void *)procfs_badop, 95 .vop_bmap = procfs_bmap, 96 .vop_close = procfs_close, 97 .vop_old_create = (void *)procfs_badop, 98 .vop_getattr = procfs_getattr, 99 .vop_inactive = procfs_inactive, 100 .vop_old_link = (void *)procfs_badop, 101 .vop_old_lookup = procfs_lookup, 102 .vop_old_mkdir = (void *)procfs_badop, 103 .vop_old_mknod = (void *)procfs_badop, 104 .vop_open = procfs_open, 105 .vop_pathconf = vop_stdpathconf, 106 .vop_print = procfs_print, 107 .vop_read = procfs_rw, 108 .vop_readdir = procfs_readdir, 109 .vop_readlink = procfs_readlink, 110 .vop_reclaim = procfs_reclaim, 111 .vop_old_remove = (void *)procfs_badop, 112 .vop_old_rename = (void *)procfs_badop, 113 .vop_old_rmdir = (void *)procfs_badop, 114 .vop_setattr = procfs_setattr, 115 .vop_old_symlink = (void *)procfs_badop, 116 .vop_write = (void *)procfs_rw, 117 .vop_ioctl = procfs_ioctl 118 }; 119 120 121 /* 122 * This is a list of the valid names in the 123 * process-specific sub-directories. It is 124 * used in procfs_lookup and procfs_readdir 125 */ 126 static struct proc_target { 127 u_char pt_type; 128 u_char pt_namlen; 129 char *pt_name; 130 pfstype pt_pfstype; 131 int (*pt_valid) (struct lwp *p); 132 } proc_targets[] = { 133 #define N(s) sizeof(s)-1, s 134 /* name type validp */ 135 { DT_DIR, N("."), Pproc, NULL }, 136 { DT_DIR, N(".."), Proot, NULL }, 137 { DT_REG, N("mem"), Pmem, NULL }, 138 { DT_REG, N("regs"), Pregs, procfs_validregs }, 139 { DT_REG, N("fpregs"), Pfpregs, procfs_validfpregs }, 140 { DT_REG, N("dbregs"), Pdbregs, procfs_validdbregs }, 141 { DT_REG, N("ctl"), Pctl, NULL }, 142 { DT_REG, N("status"), Pstatus, NULL }, 143 { DT_REG, N("note"), Pnote, NULL }, 144 { DT_REG, N("notepg"), Pnotepg, NULL }, 145 { DT_REG, N("map"), Pmap, procfs_validmap }, 146 { DT_REG, N("etype"), Ptype, procfs_validtype }, 147 { DT_REG, N("cmdline"), Pcmdline, NULL }, 148 { DT_REG, N("rlimit"), Prlimit, NULL }, 149 { DT_LNK, N("file"), Pfile, NULL }, 150 #undef N 151 }; 152 static const int nproc_targets = NELEM(proc_targets); 153 154 static pid_t atopid (const char *, u_int); 155 156 /* 157 * set things up for doing i/o on 158 * the pfsnode (vp). (vp) is locked 159 * on entry, and should be left locked 160 * on exit. 161 * 162 * for procfs we don't need to do anything 163 * in particular for i/o. all that is done 164 * is to support exclusive open on process 165 * memory images. 166 * 167 * procfs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred, 168 * struct file *a_fp) 169 */ 170 static int 171 procfs_open(struct vop_open_args *ap) 172 { 173 struct pfsnode *pfs = VTOPFS(ap->a_vp); 174 struct proc *p1, *p2; 175 int error; 176 177 p2 = pfs_pfind(pfs->pfs_pid); 178 if (p2 == NULL) 179 return (ENOENT); 180 if (pfs->pfs_pid && !PRISON_CHECK(ap->a_cred, p2->p_ucred)) { 181 error = ENOENT; 182 goto done; 183 } 184 185 switch (pfs->pfs_type) { 186 case Pmem: 187 if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) || 188 ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE))) { 189 error = EBUSY; 190 goto done; 191 } 192 193 p1 = curproc; 194 KKASSERT(p1); 195 /* Can't trace a process that's currently exec'ing. */ 196 if ((p2->p_flags & P_INEXEC) != 0) { 197 error = EAGAIN; 198 goto done; 199 } 200 if (!CHECKIO(p1, p2) || p_trespass(ap->a_cred, p2->p_ucred)) { 201 error = EPERM; 202 goto done; 203 } 204 205 if (ap->a_mode & FWRITE) 206 pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL); 207 208 break; 209 210 default: 211 break; 212 } 213 error = vop_stdopen(ap); 214 done: 215 pfs_pdone(p2); 216 return error; 217 } 218 219 /* 220 * close the pfsnode (vp) after doing i/o. 221 * (vp) is not locked on entry or exit. 222 * 223 * nothing to do for procfs other than undo 224 * any exclusive open flag (see _open above). 225 * 226 * procfs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred) 227 */ 228 static int 229 procfs_close(struct vop_close_args *ap) 230 { 231 struct pfsnode *pfs = VTOPFS(ap->a_vp); 232 struct proc *p; 233 234 switch (pfs->pfs_type) { 235 case Pmem: 236 if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL)) 237 pfs->pfs_flags &= ~(FWRITE|O_EXCL); 238 /* 239 * v_opencount determines the last real close on the vnode. 240 * 241 * If this is the last close, then it checks to see if 242 * the target process has PF_LINGER set in p_pfsflags, 243 * if this is *not* the case, then the process' stop flags 244 * are cleared, and the process is woken up. This is 245 * to help prevent the case where a process has been 246 * told to stop on an event, but then the requesting process 247 * has gone away or forgotten about it. 248 */ 249 p = NULL; 250 if ((ap->a_vp->v_opencount < 2) 251 && (p = pfs_pfind(pfs->pfs_pid)) 252 && !(p->p_pfsflags & PF_LINGER)) { 253 spin_lock(&p->p_spin); 254 p->p_stops = 0; 255 p->p_step = 0; 256 spin_unlock(&p->p_spin); 257 wakeup(&p->p_step); 258 } 259 pfs_pdone(p); 260 break; 261 default: 262 break; 263 } 264 265 return (vop_stdclose(ap)); 266 } 267 268 /* 269 * do an ioctl operation on a pfsnode (vp). 270 * (vp) is not locked on entry or exit. 271 */ 272 static int 273 procfs_ioctl(struct vop_ioctl_args *ap) 274 { 275 struct pfsnode *pfs = VTOPFS(ap->a_vp); 276 struct proc *procp; 277 struct proc *p; 278 int error; 279 int signo; 280 struct procfs_status *psp; 281 unsigned char flags; 282 283 procp = pfs_pfind(pfs->pfs_pid); 284 if (procp == NULL) 285 return ENOTTY; 286 p = curproc; 287 if (p == NULL) { 288 error = EINVAL; 289 goto done; 290 } 291 292 /* Can't trace a process that's currently exec'ing. */ 293 if ((procp->p_flags & P_INEXEC) != 0) { 294 error = EAGAIN; 295 goto done; 296 } 297 if (!CHECKIO(p, procp) || p_trespass(ap->a_cred, procp->p_ucred)) { 298 error = EPERM; 299 goto done; 300 } 301 302 switch (ap->a_command) { 303 case PIOCBIS: 304 procp->p_stops |= *(unsigned int*)ap->a_data; 305 break; 306 case PIOCBIC: 307 procp->p_stops &= ~*(unsigned int*)ap->a_data; 308 break; 309 case PIOCSFL: 310 /* 311 * NFLAGS is "non-suser_xxx flags" -- currently, only 312 * PFS_ISUGID ("ignore set u/g id"); 313 */ 314 #define NFLAGS (PF_ISUGID) 315 flags = (unsigned char)*(unsigned int*)ap->a_data; 316 if (flags & NFLAGS && (error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0))) 317 goto done; 318 procp->p_pfsflags = flags; 319 break; 320 case PIOCGFL: 321 *(unsigned int*)ap->a_data = (unsigned int)procp->p_pfsflags; 322 break; 323 case PIOCSTATUS: 324 /* 325 * NOTE: syscall entry deals with stopevents and may run without 326 * the MP lock. 327 */ 328 psp = (struct procfs_status *)ap->a_data; 329 psp->flags = procp->p_pfsflags; 330 psp->events = procp->p_stops; 331 spin_lock(&procp->p_spin); 332 if (procp->p_step) { 333 psp->state = 0; 334 psp->why = procp->p_stype; 335 psp->val = procp->p_xstat; 336 spin_unlock(&procp->p_spin); 337 } else { 338 psp->state = 1; 339 spin_unlock(&procp->p_spin); 340 psp->why = 0; /* Not defined values */ 341 psp->val = 0; /* Not defined values */ 342 } 343 break; 344 case PIOCWAIT: 345 /* 346 * NOTE: syscall entry deals with stopevents and may run without 347 * the MP lock. 348 */ 349 psp = (struct procfs_status *)ap->a_data; 350 spin_lock(&procp->p_spin); 351 while (procp->p_step == 0) { 352 tsleep_interlock(&procp->p_stype, PCATCH); 353 spin_unlock(&procp->p_spin); 354 error = tsleep(&procp->p_stype, PCATCH | PINTERLOCKED, "piocwait", 0); 355 if (error) 356 goto done; 357 spin_lock(&procp->p_spin); 358 } 359 spin_unlock(&procp->p_spin); 360 psp->state = 1; /* It stopped */ 361 psp->flags = procp->p_pfsflags; 362 psp->events = procp->p_stops; 363 psp->why = procp->p_stype; /* why it stopped */ 364 psp->val = procp->p_xstat; /* any extra info */ 365 break; 366 case PIOCCONT: /* Restart a proc */ 367 /* 368 * NOTE: syscall entry deals with stopevents and may run without 369 * the MP lock. However, the caller is presumably interlocked 370 * by having waited. 371 */ 372 if (procp->p_step == 0) { 373 error = EINVAL; /* Can only start a stopped process */ 374 goto done; 375 } 376 if ((signo = *(int*)ap->a_data) != 0) { 377 if (signo >= NSIG || signo <= 0) { 378 error = EINVAL; 379 goto done; 380 } 381 ksignal(procp, signo); 382 } 383 procp->p_step = 0; 384 wakeup(&procp->p_step); 385 break; 386 default: 387 error = ENOTTY; 388 goto done; 389 } 390 error = 0; 391 done: 392 pfs_pdone(procp); 393 return 0; 394 } 395 396 /* 397 * do block mapping for pfsnode (vp). 398 * since we don't use the buffer cache 399 * for procfs this function should never 400 * be called. in any case, it's not clear 401 * what part of the kernel ever makes use 402 * of this function. for sanity, this is the 403 * usual no-op bmap, although returning 404 * (EIO) would be a reasonable alternative. 405 * 406 * XXX mmap assumes buffer cache operation 407 * 408 * procfs_bmap(struct vnode *a_vp, off_t a_loffset, 409 * off_t *a_doffsetp, int *a_runp, int *a_runb) 410 */ 411 static int 412 procfs_bmap(struct vop_bmap_args *ap) 413 { 414 if (ap->a_doffsetp != NULL) 415 *ap->a_doffsetp = ap->a_loffset; 416 if (ap->a_runp != NULL) 417 *ap->a_runp = 0; 418 if (ap->a_runb != NULL) 419 *ap->a_runb = 0; 420 return (0); 421 } 422 423 /* 424 * procfs_inactive is called when the pfsnode 425 * is vrele'd and the reference count goes 426 * to zero. (vp) will be on the vnode free 427 * list, so to get it back vget() must be 428 * used. 429 * 430 * (vp) is locked on entry, but must be unlocked on exit. 431 * 432 * procfs_inactive(struct vnode *a_vp) 433 */ 434 static int 435 procfs_inactive(struct vop_inactive_args *ap) 436 { 437 struct pfsnode *pfs = VTOPFS(ap->a_vp); 438 439 if (pfs->pfs_pid & PFS_DEAD) 440 vrecycle(ap->a_vp); 441 return (0); 442 } 443 444 /* 445 * _reclaim is called when getnewvnode() 446 * wants to make use of an entry on the vnode 447 * free list. at this time the filesystem needs 448 * to free any private data and remove the node 449 * from any private lists. 450 * 451 * procfs_reclaim(struct vnode *a_vp) 452 */ 453 static int 454 procfs_reclaim(struct vop_reclaim_args *ap) 455 { 456 return (procfs_freevp(ap->a_vp)); 457 } 458 459 /* 460 * _print is used for debugging. 461 * just print a readable description 462 * of (vp). 463 * 464 * procfs_print(struct vnode *a_vp) 465 */ 466 static int 467 procfs_print(struct vop_print_args *ap) 468 { 469 struct pfsnode *pfs = VTOPFS(ap->a_vp); 470 471 kprintf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n", 472 pfs->pfs_type, (long)pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags); 473 return (0); 474 } 475 476 /* 477 * generic entry point for unsupported operations 478 */ 479 static int 480 procfs_badop(struct vop_generic_args *ap) 481 { 482 return (EIO); 483 } 484 485 /* 486 * Invent attributes for pfsnode (vp) and store 487 * them in (vap). 488 * Directories lengths are returned as zero since 489 * any real length would require the genuine size 490 * to be computed, and nothing cares anyway. 491 * 492 * this is relatively minimal for procfs. 493 * 494 * procfs_getattr(struct vnode *a_vp, struct vattr *a_vap) 495 */ 496 static int 497 procfs_getattr(struct vop_getattr_args *ap) 498 { 499 struct pfsnode *pfs = VTOPFS(ap->a_vp); 500 struct vattr *vap = ap->a_vap; 501 struct proc *procp; 502 int error; 503 504 /* 505 * First make sure that the process and its credentials 506 * still exist. 507 */ 508 switch (pfs->pfs_type) { 509 case Proot: 510 case Pcurproc: 511 procp = NULL; 512 break; 513 default: 514 procp = pfs_pfind(pfs->pfs_pid); 515 if (procp == NULL || procp->p_ucred == NULL) { 516 error = ENOENT; 517 goto done; 518 } 519 break; 520 } 521 522 error = 0; 523 524 /* start by zeroing out the attributes */ 525 VATTR_NULL(vap); 526 527 /* next do all the common fields */ 528 vap->va_type = ap->a_vp->v_type; 529 vap->va_mode = pfs->pfs_mode; 530 vap->va_fileid = pfs->pfs_fileno; 531 vap->va_flags = 0; 532 vap->va_blocksize = PAGE_SIZE; 533 vap->va_bytes = vap->va_size = 0; 534 vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0]; 535 536 /* 537 * Make all times be current TOD. 538 * It would be possible to get the process start 539 * time from the p_stat structure, but there's 540 * no "file creation" time stamp anyway, and the 541 * p_stat structure is not addressible if u. gets 542 * swapped out for that process. 543 */ 544 nanotime(&vap->va_ctime); 545 vap->va_atime = vap->va_mtime = vap->va_ctime; 546 547 /* 548 * If the process has exercised some setuid or setgid 549 * privilege, then rip away read/write permission so 550 * that only root can gain access. 551 */ 552 switch (pfs->pfs_type) { 553 case Pctl: 554 case Pregs: 555 case Pfpregs: 556 case Pdbregs: 557 case Pmem: 558 if (procp->p_flags & P_SUGID) { 559 vap->va_mode &= ~((VREAD|VWRITE)| 560 ((VREAD|VWRITE)>>3)| 561 ((VREAD|VWRITE)>>6)); 562 } 563 break; 564 default: 565 break; 566 } 567 568 /* 569 * now do the object specific fields 570 * 571 * The size could be set from struct reg, but it's hardly 572 * worth the trouble, and it puts some (potentially) machine 573 * dependent data into this machine-independent code. If it 574 * becomes important then this function should break out into 575 * a per-file stat function in the corresponding .c file. 576 */ 577 578 vap->va_nlink = 1; 579 if (procp) { 580 if (procp->p_ucred) { 581 vap->va_uid = procp->p_ucred->cr_uid; 582 vap->va_gid = procp->p_ucred->cr_gid; 583 } else { 584 vap->va_uid = -1; 585 vap->va_gid = -1; 586 } 587 } 588 589 switch (pfs->pfs_type) { 590 case Proot: 591 /* 592 * Set nlink to 1 to tell fts(3) we don't actually know. 593 */ 594 vap->va_nlink = 1; 595 vap->va_uid = 0; 596 vap->va_gid = 0; 597 vap->va_size = vap->va_bytes = DEV_BSIZE; 598 break; 599 600 case Pcurproc: { 601 char buf[16]; /* should be enough */ 602 603 vap->va_uid = 0; 604 vap->va_gid = 0; 605 vap->va_size = ksnprintf(buf, sizeof(buf), 606 "%ld", (long)curproc->p_pid); 607 vap->va_bytes = vap->va_size; 608 break; 609 } 610 611 case Pproc: 612 vap->va_nlink = nproc_targets; 613 vap->va_size = vap->va_bytes = DEV_BSIZE; 614 break; 615 616 case Pfile: { 617 char *fullpath, *freepath; 618 619 if (procp->p_textnch.ncp) { 620 struct nchandle nch; 621 622 cache_copy(&procp->p_textnch, &nch); 623 error = cache_fullpath(procp, &nch, 624 &fullpath, &freepath, 0); 625 cache_drop(&nch); 626 } else { 627 error = EINVAL; 628 } 629 630 if (error == 0) { 631 vap->va_size = strlen(fullpath); 632 kfree(freepath, M_TEMP); 633 } else { 634 vap->va_size = sizeof("unknown") - 1; 635 error = 0; 636 } 637 vap->va_bytes = vap->va_size; 638 break; 639 } 640 641 case Pmem: 642 /* 643 * If we denied owner access earlier, then we have to 644 * change the owner to root - otherwise 'ps' and friends 645 * will break even though they are setgid kmem. *SIGH* 646 */ 647 if (procp->p_flags & P_SUGID) 648 vap->va_uid = 0; 649 else if (procp->p_ucred) 650 vap->va_uid = procp->p_ucred->cr_uid; 651 else 652 vap->va_uid = -1; 653 break; 654 655 case Pregs: 656 vap->va_bytes = vap->va_size = sizeof(struct reg); 657 break; 658 659 case Pfpregs: 660 vap->va_bytes = vap->va_size = sizeof(struct fpreg); 661 break; 662 663 case Pdbregs: 664 vap->va_bytes = vap->va_size = sizeof(struct dbreg); 665 break; 666 667 case Ptype: 668 case Pmap: 669 case Pctl: 670 case Pstatus: 671 case Pnote: 672 case Pnotepg: 673 case Pcmdline: 674 case Prlimit: 675 break; 676 677 default: 678 panic("procfs_getattr"); 679 } 680 done: 681 pfs_pdone(procp); 682 return (error); 683 } 684 685 /* 686 * procfs_setattr(struct vnode *a_vp, struct vattr *a_vap, 687 * struct ucred *a_cred) 688 */ 689 static int 690 procfs_setattr(struct vop_setattr_args *ap) 691 { 692 if (ap->a_vap->va_flags != VNOVAL) 693 return (EOPNOTSUPP); 694 695 /* 696 * just fake out attribute setting 697 * it's not good to generate an error 698 * return, otherwise things like creat() 699 * will fail when they try to set the 700 * file length to 0. worse, this means 701 * that echo $note > /proc/$pid/note will fail. 702 */ 703 704 return (0); 705 } 706 707 /* 708 * implement access checking. 709 * 710 * procfs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred) 711 */ 712 static int 713 procfs_access(struct vop_access_args *ap) 714 { 715 struct vattr vattr; 716 int error; 717 718 error = VOP_GETATTR(ap->a_vp, &vattr); 719 if (!error) 720 error = vop_helper_access(ap, vattr.va_uid, vattr.va_gid, 721 vattr.va_mode, 0); 722 return (error); 723 } 724 725 /* 726 * lookup. this is incredibly complicated in the general case, however 727 * for most pseudo-filesystems very little needs to be done. 728 * 729 * procfs_lookup(struct vnode *a_dvp, struct vnode **a_vpp, 730 * struct componentname *a_cnp) 731 */ 732 static int 733 procfs_lookup(struct vop_old_lookup_args *ap) 734 { 735 struct componentname *cnp = ap->a_cnp; 736 struct vnode **vpp = ap->a_vpp; 737 struct vnode *dvp = ap->a_dvp; 738 char *pname = cnp->cn_nameptr; 739 /* struct proc *curp = cnp->cn_proc; */ 740 struct proc_target *pt; 741 pid_t pid; 742 struct pfsnode *pfs; 743 struct proc *p; 744 struct lwp *lp; 745 int i; 746 int error; 747 748 *vpp = NULL; 749 750 if (cnp->cn_nameiop == NAMEI_DELETE || cnp->cn_nameiop == NAMEI_RENAME) 751 return (EROFS); 752 753 p = NULL; 754 error = 0; 755 if (cnp->cn_namelen == 1 && *pname == '.') { 756 *vpp = dvp; 757 vref(*vpp); 758 goto out; 759 } 760 761 pfs = VTOPFS(dvp); 762 switch (pfs->pfs_type) { 763 case Proot: 764 if (cnp->cn_flags & CNP_ISDOTDOT) 765 return (EIO); 766 767 if (CNEQ(cnp, "curproc", 7)) { 768 error = procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc); 769 goto out; 770 } 771 772 pid = atopid(pname, cnp->cn_namelen); 773 if (pid == NO_PID) 774 break; 775 776 p = pfs_pfind(pid); 777 if (p == NULL) 778 break; 779 780 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred)) 781 break; 782 783 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 && 784 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid) 785 break; 786 787 error = procfs_allocvp(dvp->v_mount, vpp, pid, Pproc); 788 goto out; 789 790 case Pproc: 791 if (cnp->cn_flags & CNP_ISDOTDOT) { 792 error = procfs_root(dvp->v_mount, vpp); 793 goto out; 794 } 795 796 p = pfs_pfind(pfs->pfs_pid); 797 if (p == NULL) 798 break; 799 /* XXX lwp */ 800 lp = FIRST_LWP_IN_PROC(p); 801 802 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred)) 803 break; 804 805 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 && 806 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid) 807 break; 808 809 for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) { 810 if (cnp->cn_namelen == pt->pt_namlen && 811 bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 && 812 (pt->pt_valid == NULL || (*pt->pt_valid)(lp))) 813 goto found; 814 } 815 break; 816 found: 817 error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid, 818 pt->pt_pfstype); 819 goto out; 820 821 default: 822 error = ENOTDIR; 823 goto out; 824 } 825 if (cnp->cn_nameiop == NAMEI_LOOKUP) 826 error = ENOENT; 827 else 828 error = EROFS; 829 /* 830 * If no error occured *vpp will hold a referenced locked vnode. 831 * dvp was passed to us locked and *vpp must be returned locked. 832 * If *vpp != dvp then we should unlock dvp if (1) this is not the 833 * last component or (2) CNP_LOCKPARENT is not set. 834 */ 835 out: 836 if (error == 0 && *vpp != dvp) { 837 if ((cnp->cn_flags & CNP_LOCKPARENT) == 0) { 838 cnp->cn_flags |= CNP_PDIRUNLOCK; 839 vn_unlock(dvp); 840 } 841 } 842 pfs_pdone(p); 843 return (error); 844 } 845 846 /* 847 * Does this process have a text file? 848 */ 849 int 850 procfs_validfile(struct lwp *lp) 851 { 852 return (procfs_findtextvp(lp->lwp_proc) != NULLVP); 853 } 854 855 /* 856 * readdir() returns directory entries from pfsnode (vp). 857 * 858 * We generate just one directory entry at a time, as it would probably 859 * not pay off to buffer several entries locally to save uiomove calls. 860 * 861 * procfs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred, 862 * int *a_eofflag, int *a_ncookies, off_t **a_cookies) 863 */ 864 static int 865 procfs_readdir(struct vop_readdir_args *ap) 866 { 867 struct pfsnode *pfs; 868 int error; 869 870 if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX) 871 return (EINVAL); 872 if ((error = vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY)) != 0) 873 return (error); 874 pfs = VTOPFS(ap->a_vp); 875 876 switch (pfs->pfs_type) { 877 case Pproc: 878 /* 879 * this is for the process-specific sub-directories. 880 * all that is needed to is copy out all the entries 881 * from the procent[] table (top of this file). 882 */ 883 error = procfs_readdir_proc(ap); 884 break; 885 case Proot: 886 /* 887 * this is for the root of the procfs filesystem 888 * what is needed is a special entry for "curproc" 889 * followed by an entry for each process on allproc 890 */ 891 error = procfs_readdir_root(ap); 892 break; 893 default: 894 error = ENOTDIR; 895 break; 896 } 897 898 vn_unlock(ap->a_vp); 899 return (error); 900 } 901 902 static int 903 procfs_readdir_proc(struct vop_readdir_args *ap) 904 { 905 struct pfsnode *pfs; 906 int error, i, retval; 907 struct proc *p; 908 struct lwp *lp; 909 struct proc_target *pt; 910 struct uio *uio = ap->a_uio; 911 912 pfs = VTOPFS(ap->a_vp); 913 p = pfs_pfind(pfs->pfs_pid); 914 if (p == NULL) 915 return(0); 916 if (!PRISON_CHECK(ap->a_cred, p->p_ucred)) { 917 error = 0; 918 goto done; 919 } 920 /* XXX lwp, not MPSAFE */ 921 lp = FIRST_LWP_IN_PROC(p); 922 923 error = 0; 924 i = (int)uio->uio_offset; 925 if (i < 0) { 926 error = EINVAL; 927 goto done; 928 } 929 930 for (pt = &proc_targets[i]; 931 !error && uio->uio_resid > 0 && i < nproc_targets; pt++, i++) { 932 if (pt->pt_valid && (*pt->pt_valid)(lp) == 0) 933 continue; 934 935 retval = vop_write_dirent(&error, uio, 936 PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype), pt->pt_type, 937 pt->pt_namlen, pt->pt_name); 938 if (retval) 939 break; 940 } 941 942 uio->uio_offset = (off_t)i; 943 error = 0; 944 done: 945 pfs_pdone(p); 946 return error; 947 } 948 949 struct procfs_readdir_root_info { 950 int error; 951 int i; 952 int pcnt; 953 struct uio *uio; 954 struct ucred *cred; 955 }; 956 957 static int procfs_readdir_root_callback(struct proc *p, void *data); 958 959 static int 960 procfs_readdir_root(struct vop_readdir_args *ap) 961 { 962 struct procfs_readdir_root_info info; 963 struct uio *uio = ap->a_uio; 964 int res; 965 966 info.error = 0; 967 info.i = (int)uio->uio_offset; 968 969 if (info.i < 0) 970 return (EINVAL); 971 972 info.pcnt = 0; 973 info.uio = uio; 974 info.cred = ap->a_cred; 975 while (info.pcnt < 3) { 976 res = procfs_readdir_root_callback(NULL, &info); 977 if (res < 0) 978 break; 979 } 980 if (res >= 0) 981 allproc_scan(procfs_readdir_root_callback, &info); 982 uio->uio_offset = (off_t)info.i; 983 984 return (info.error); 985 } 986 987 static int 988 procfs_readdir_root_callback(struct proc *p, void *data) 989 { 990 struct procfs_readdir_root_info *info = data; 991 struct uio *uio; 992 int retval; 993 ino_t d_ino; 994 const char *d_name; 995 char d_name_pid[20]; 996 size_t d_namlen; 997 uint8_t d_type; 998 999 uio = info->uio; 1000 1001 if (uio->uio_resid <= 0 || info->error) 1002 return(-1); 1003 1004 switch (info->pcnt) { 1005 case 0: /* `.' */ 1006 d_ino = PROCFS_FILENO(0, Proot); 1007 d_name = "."; 1008 d_namlen = 1; 1009 d_type = DT_DIR; 1010 break; 1011 case 1: /* `..' */ 1012 d_ino = PROCFS_FILENO(0, Proot); 1013 d_name = ".."; 1014 d_namlen = 2; 1015 d_type = DT_DIR; 1016 break; 1017 1018 case 2: 1019 d_ino = PROCFS_FILENO(0, Pcurproc); 1020 d_namlen = 7; 1021 d_name = "curproc"; 1022 d_type = DT_LNK; 1023 break; 1024 1025 1026 default: 1027 if (!PRISON_CHECK(info->cred, p->p_ucred)) 1028 return(0); 1029 if (ps_showallprocs == 0 && 1030 info->cred->cr_uid != 0 && 1031 info->cred->cr_uid != p->p_ucred->cr_uid) { 1032 return(0); 1033 } 1034 1035 /* 1036 * Skip entries we have already returned (optimization) 1037 */ 1038 if (info->pcnt < info->i) { 1039 ++info->pcnt; 1040 return(0); 1041 } 1042 1043 d_ino = PROCFS_FILENO(p->p_pid, Pproc); 1044 d_namlen = ksnprintf(d_name_pid, sizeof(d_name_pid), 1045 "%ld", (long)p->p_pid); 1046 d_name = d_name_pid; 1047 d_type = DT_DIR; 1048 break; 1049 } 1050 1051 /* 1052 * Skip entries we have already returned (optimization) 1053 */ 1054 if (info->pcnt < info->i) { 1055 ++info->pcnt; 1056 return(0); 1057 } 1058 1059 retval = vop_write_dirent(&info->error, uio, 1060 d_ino, d_type, d_namlen, d_name); 1061 if (retval) 1062 return(-1); 1063 ++info->pcnt; 1064 ++info->i; 1065 return(0); 1066 } 1067 1068 /* 1069 * readlink reads the link of `curproc' or `file' 1070 */ 1071 static int 1072 procfs_readlink(struct vop_readlink_args *ap) 1073 { 1074 char buf[16]; /* should be enough */ 1075 struct proc *procp; 1076 struct vnode *vp = ap->a_vp; 1077 struct pfsnode *pfs = VTOPFS(vp); 1078 char *fullpath, *freepath; 1079 int error, len; 1080 1081 switch (pfs->pfs_type) { 1082 case Pcurproc: 1083 if (pfs->pfs_fileno != PROCFS_FILENO(0, Pcurproc)) 1084 return (EINVAL); 1085 1086 len = ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid); 1087 1088 return (uiomove(buf, len, ap->a_uio)); 1089 /* 1090 * There _should_ be no way for an entire process to disappear 1091 * from under us... 1092 */ 1093 case Pfile: 1094 procp = pfs_pfind(pfs->pfs_pid); 1095 if (procp == NULL || procp->p_ucred == NULL) { 1096 kprintf("procfs_readlink: pid %d disappeared\n", 1097 pfs->pfs_pid); 1098 pfs_pdone(procp); 1099 return (uiomove("unknown", sizeof("unknown") - 1, 1100 ap->a_uio)); 1101 } 1102 if (procp->p_textnch.ncp) { 1103 struct nchandle nch; 1104 1105 cache_copy(&procp->p_textnch, &nch); 1106 error = cache_fullpath(procp, &nch, 1107 &fullpath, &freepath, 0); 1108 cache_drop(&nch); 1109 } else { 1110 error = EINVAL; 1111 } 1112 1113 if (error != 0) { 1114 pfs_pdone(procp); 1115 return (uiomove("unknown", sizeof("unknown") - 1, 1116 ap->a_uio)); 1117 } 1118 error = uiomove(fullpath, strlen(fullpath), ap->a_uio); 1119 kfree(freepath, M_TEMP); 1120 pfs_pdone(procp); 1121 return (error); 1122 default: 1123 return (EINVAL); 1124 } 1125 } 1126 1127 /* 1128 * convert decimal ascii to pid_t 1129 */ 1130 static pid_t 1131 atopid(const char *b, u_int len) 1132 { 1133 pid_t p = 0; 1134 1135 while (len--) { 1136 char c = *b++; 1137 if (c < '0' || c > '9') 1138 return (NO_PID); 1139 p = 10 * p + (c - '0'); 1140 if (p > PID_MAX) 1141 return (NO_PID); 1142 } 1143 1144 return (p); 1145 } 1146 1147