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 * $DragonFly: src/sys/vfs/procfs/procfs_vnops.c,v 1.42 2007/02/19 01:14:24 corecode Exp $ 41 */ 42 43 /* 44 * procfs vnode interface 45 */ 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/time.h> 50 #include <sys/kernel.h> 51 #include <sys/lock.h> 52 #include <sys/fcntl.h> 53 #include <sys/proc.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 <machine/limits.h> 67 68 static int procfs_access (struct vop_access_args *); 69 static int procfs_badop (struct vop_generic_args *); 70 static int procfs_bmap (struct vop_bmap_args *); 71 static int procfs_close (struct vop_close_args *); 72 static int procfs_getattr (struct vop_getattr_args *); 73 static int procfs_inactive (struct vop_inactive_args *); 74 static int procfs_ioctl (struct vop_ioctl_args *); 75 static int procfs_lookup (struct vop_old_lookup_args *); 76 static int procfs_open (struct vop_open_args *); 77 static int procfs_print (struct vop_print_args *); 78 static int procfs_readdir (struct vop_readdir_args *); 79 static int procfs_readlink (struct vop_readlink_args *); 80 static int procfs_reclaim (struct vop_reclaim_args *); 81 static int procfs_setattr (struct vop_setattr_args *); 82 83 static int procfs_readdir_proc(struct vop_readdir_args *); 84 static int procfs_readdir_root(struct vop_readdir_args *); 85 86 /* 87 * procfs vnode operations. 88 */ 89 struct vop_ops procfs_vnode_vops = { 90 .vop_default = vop_defaultop, 91 .vop_access = procfs_access, 92 .vop_advlock = (void *)procfs_badop, 93 .vop_bmap = procfs_bmap, 94 .vop_close = procfs_close, 95 .vop_old_create = (void *)procfs_badop, 96 .vop_getattr = procfs_getattr, 97 .vop_inactive = procfs_inactive, 98 .vop_old_link = (void *)procfs_badop, 99 .vop_old_lookup = procfs_lookup, 100 .vop_old_mkdir = (void *)procfs_badop, 101 .vop_old_mknod = (void *)procfs_badop, 102 .vop_open = procfs_open, 103 .vop_pathconf = vop_stdpathconf, 104 .vop_print = procfs_print, 105 .vop_read = procfs_rw, 106 .vop_readdir = procfs_readdir, 107 .vop_readlink = procfs_readlink, 108 .vop_reclaim = procfs_reclaim, 109 .vop_old_remove = (void *)procfs_badop, 110 .vop_old_rename = (void *)procfs_badop, 111 .vop_old_rmdir = (void *)procfs_badop, 112 .vop_setattr = procfs_setattr, 113 .vop_old_symlink = (void *)procfs_badop, 114 .vop_write = (void *)procfs_rw, 115 .vop_ioctl = procfs_ioctl 116 }; 117 118 119 /* 120 * This is a list of the valid names in the 121 * process-specific sub-directories. It is 122 * used in procfs_lookup and procfs_readdir 123 */ 124 static struct proc_target { 125 u_char pt_type; 126 u_char pt_namlen; 127 char *pt_name; 128 pfstype pt_pfstype; 129 int (*pt_valid) (struct lwp *p); 130 } proc_targets[] = { 131 #define N(s) sizeof(s)-1, s 132 /* name type validp */ 133 { DT_DIR, N("."), Pproc, NULL }, 134 { DT_DIR, N(".."), Proot, NULL }, 135 { DT_REG, N("mem"), Pmem, NULL }, 136 { DT_REG, N("regs"), Pregs, procfs_validregs }, 137 { DT_REG, N("fpregs"), Pfpregs, procfs_validfpregs }, 138 { DT_REG, N("dbregs"), Pdbregs, procfs_validdbregs }, 139 { DT_REG, N("ctl"), Pctl, NULL }, 140 { DT_REG, N("status"), Pstatus, NULL }, 141 { DT_REG, N("note"), Pnote, NULL }, 142 { DT_REG, N("notepg"), Pnotepg, NULL }, 143 { DT_REG, N("map"), Pmap, procfs_validmap }, 144 { DT_REG, N("etype"), Ptype, procfs_validtype }, 145 { DT_REG, N("cmdline"), Pcmdline, NULL }, 146 { DT_REG, N("rlimit"), Prlimit, NULL }, 147 { DT_LNK, N("file"), Pfile, NULL }, 148 #undef N 149 }; 150 static const int nproc_targets = sizeof(proc_targets) / sizeof(proc_targets[0]); 151 152 static pid_t atopid (const char *, u_int); 153 154 /* 155 * set things up for doing i/o on 156 * the pfsnode (vp). (vp) is locked 157 * on entry, and should be left locked 158 * on exit. 159 * 160 * for procfs we don't need to do anything 161 * in particular for i/o. all that is done 162 * is to support exclusive open on process 163 * memory images. 164 * 165 * procfs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred) 166 */ 167 static int 168 procfs_open(struct vop_open_args *ap) 169 { 170 struct pfsnode *pfs = VTOPFS(ap->a_vp); 171 struct proc *p1, *p2; 172 173 p2 = PFIND(pfs->pfs_pid); 174 if (p2 == NULL) 175 return (ENOENT); 176 if (pfs->pfs_pid && !PRISON_CHECK(ap->a_cred, p2->p_ucred)) 177 return (ENOENT); 178 179 switch (pfs->pfs_type) { 180 case Pmem: 181 if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) || 182 ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE))) 183 return (EBUSY); 184 185 p1 = curproc; 186 KKASSERT(p1); 187 /* Can't trace a process that's currently exec'ing. */ 188 if ((p2->p_flag & P_INEXEC) != 0) 189 return EAGAIN; 190 if (!CHECKIO(p1, p2) || p_trespass(ap->a_cred, p2->p_ucred)) 191 return (EPERM); 192 193 if (ap->a_mode & FWRITE) 194 pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL); 195 196 break; 197 198 default: 199 break; 200 } 201 202 return (vop_stdopen(ap)); 203 } 204 205 /* 206 * close the pfsnode (vp) after doing i/o. 207 * (vp) is not locked on entry or exit. 208 * 209 * nothing to do for procfs other than undo 210 * any exclusive open flag (see _open above). 211 * 212 * procfs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred) 213 */ 214 static int 215 procfs_close(struct vop_close_args *ap) 216 { 217 struct pfsnode *pfs = VTOPFS(ap->a_vp); 218 struct proc *p; 219 220 switch (pfs->pfs_type) { 221 case Pmem: 222 if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL)) 223 pfs->pfs_flags &= ~(FWRITE|O_EXCL); 224 /* 225 * This rather complicated-looking code is trying to 226 * determine if this was the last close on this particular 227 * vnode. While one would expect v_usecount to be 1 at 228 * that point, it seems that (according to John Dyson) 229 * the VM system will bump up the usecount. So: if the 230 * usecount is 2, and VOBJBUF is set, then this is really 231 * the last close. Otherwise, if the usecount is < 2 232 * then it is definitely the last close. 233 * If this is the last close, then it checks to see if 234 * the target process has PF_LINGER set in p_pfsflags, 235 * if this is *not* the case, then the process' stop flags 236 * are cleared, and the process is woken up. This is 237 * to help prevent the case where a process has been 238 * told to stop on an event, but then the requesting process 239 * has gone away or forgotten about it. 240 */ 241 if ((ap->a_vp->v_usecount < 2) 242 && (p = pfind(pfs->pfs_pid)) 243 && !(p->p_pfsflags & PF_LINGER)) { 244 p->p_stops = 0; 245 p->p_step = 0; 246 wakeup(&p->p_step); 247 } 248 break; 249 default: 250 break; 251 } 252 253 return (vop_stdclose(ap)); 254 } 255 256 /* 257 * do an ioctl operation on a pfsnode (vp). 258 * (vp) is not locked on entry or exit. 259 */ 260 static int 261 procfs_ioctl(struct vop_ioctl_args *ap) 262 { 263 struct pfsnode *pfs = VTOPFS(ap->a_vp); 264 struct proc *procp; 265 struct proc *p; 266 int error; 267 int signo; 268 struct procfs_status *psp; 269 unsigned char flags; 270 271 procp = pfind(pfs->pfs_pid); 272 if (procp == NULL) 273 return ENOTTY; 274 p = curproc; 275 if (p == NULL) 276 return EINVAL; 277 278 /* Can't trace a process that's currently exec'ing. */ 279 if ((procp->p_flag & P_INEXEC) != 0) 280 return EAGAIN; 281 if (!CHECKIO(p, procp) || p_trespass(ap->a_cred, procp->p_ucred)) 282 return EPERM; 283 284 switch (ap->a_command) { 285 case PIOCBIS: 286 procp->p_stops |= *(unsigned int*)ap->a_data; 287 break; 288 case PIOCBIC: 289 procp->p_stops &= ~*(unsigned int*)ap->a_data; 290 break; 291 case PIOCSFL: 292 /* 293 * NFLAGS is "non-suser_xxx flags" -- currently, only 294 * PFS_ISUGID ("ignore set u/g id"); 295 */ 296 #define NFLAGS (PF_ISUGID) 297 flags = (unsigned char)*(unsigned int*)ap->a_data; 298 if (flags & NFLAGS && (error = suser_cred(ap->a_cred, 0))) 299 return error; 300 procp->p_pfsflags = flags; 301 break; 302 case PIOCGFL: 303 *(unsigned int*)ap->a_data = (unsigned int)procp->p_pfsflags; 304 break; 305 case PIOCSTATUS: 306 psp = (struct procfs_status *)ap->a_data; 307 psp->state = (procp->p_step == 0); 308 psp->flags = procp->p_pfsflags; 309 psp->events = procp->p_stops; 310 if (procp->p_step) { 311 psp->why = procp->p_stype; 312 psp->val = procp->p_xstat; 313 } else { 314 psp->why = psp->val = 0; /* Not defined values */ 315 } 316 break; 317 case PIOCWAIT: 318 psp = (struct procfs_status *)ap->a_data; 319 if (procp->p_step == 0) { 320 error = tsleep(&procp->p_stype, PCATCH, "piocwait", 0); 321 if (error) 322 return error; 323 } 324 psp->state = 1; /* It stopped */ 325 psp->flags = procp->p_pfsflags; 326 psp->events = procp->p_stops; 327 psp->why = procp->p_stype; /* why it stopped */ 328 psp->val = procp->p_xstat; /* any extra info */ 329 break; 330 case PIOCCONT: /* Restart a proc */ 331 if (procp->p_step == 0) 332 return EINVAL; /* Can only start a stopped process */ 333 if ((signo = *(int*)ap->a_data) != 0) { 334 if (signo >= NSIG || signo <= 0) 335 return EINVAL; 336 ksignal(procp, signo); 337 } 338 procp->p_step = 0; 339 wakeup(&procp->p_step); 340 break; 341 default: 342 return (ENOTTY); 343 } 344 return 0; 345 } 346 347 /* 348 * do block mapping for pfsnode (vp). 349 * since we don't use the buffer cache 350 * for procfs this function should never 351 * be called. in any case, it's not clear 352 * what part of the kernel ever makes use 353 * of this function. for sanity, this is the 354 * usual no-op bmap, although returning 355 * (EIO) would be a reasonable alternative. 356 * 357 * procfs_bmap(struct vnode *a_vp, off_t a_loffset, struct vnode **a_vpp, 358 * off_t *a_doffsetp, int *a_runp) 359 */ 360 static int 361 procfs_bmap(struct vop_bmap_args *ap) 362 { 363 if (ap->a_vpp != NULL) 364 *ap->a_vpp = ap->a_vp; 365 if (ap->a_doffsetp != NULL) 366 *ap->a_doffsetp = ap->a_loffset; 367 if (ap->a_runp != NULL) 368 *ap->a_runp = 0; 369 if (ap->a_runb != NULL) 370 *ap->a_runb = 0; 371 return (0); 372 } 373 374 /* 375 * procfs_inactive is called when the pfsnode 376 * is vrele'd and the reference count goes 377 * to zero. (vp) will be on the vnode free 378 * list, so to get it back vget() must be 379 * used. 380 * 381 * (vp) is locked on entry, but must be unlocked on exit. 382 * 383 * procfs_inactive(struct vnode *a_vp, struct thread *a_td) 384 */ 385 static int 386 procfs_inactive(struct vop_inactive_args *ap) 387 { 388 /*struct vnode *vp = ap->a_vp;*/ 389 390 return (0); 391 } 392 393 /* 394 * _reclaim is called when getnewvnode() 395 * wants to make use of an entry on the vnode 396 * free list. at this time the filesystem needs 397 * to free any private data and remove the node 398 * from any private lists. 399 * 400 * procfs_reclaim(struct vnode *a_vp) 401 */ 402 static int 403 procfs_reclaim(struct vop_reclaim_args *ap) 404 { 405 return (procfs_freevp(ap->a_vp)); 406 } 407 408 /* 409 * _print is used for debugging. 410 * just print a readable description 411 * of (vp). 412 * 413 * procfs_print(struct vnode *a_vp) 414 */ 415 static int 416 procfs_print(struct vop_print_args *ap) 417 { 418 struct pfsnode *pfs = VTOPFS(ap->a_vp); 419 420 kprintf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n", 421 pfs->pfs_type, (long)pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags); 422 return (0); 423 } 424 425 /* 426 * generic entry point for unsupported operations 427 */ 428 static int 429 procfs_badop(struct vop_generic_args *ap) 430 { 431 return (EIO); 432 } 433 434 /* 435 * Invent attributes for pfsnode (vp) and store 436 * them in (vap). 437 * Directories lengths are returned as zero since 438 * any real length would require the genuine size 439 * to be computed, and nothing cares anyway. 440 * 441 * this is relatively minimal for procfs. 442 * 443 * procfs_getattr(struct vnode *a_vp, struct vattr *a_vap, 444 * struct ucred *a_cred, struct thread *a_td) 445 */ 446 static int 447 procfs_getattr(struct vop_getattr_args *ap) 448 { 449 struct pfsnode *pfs = VTOPFS(ap->a_vp); 450 struct vattr *vap = ap->a_vap; 451 struct proc *procp; 452 int error; 453 454 /* 455 * First make sure that the process and its credentials 456 * still exist. 457 */ 458 switch (pfs->pfs_type) { 459 case Proot: 460 case Pcurproc: 461 procp = 0; 462 break; 463 464 default: 465 procp = PFIND(pfs->pfs_pid); 466 if (procp == NULL || procp->p_ucred == NULL) 467 return (ENOENT); 468 } 469 470 error = 0; 471 472 /* start by zeroing out the attributes */ 473 VATTR_NULL(vap); 474 475 /* next do all the common fields */ 476 vap->va_type = ap->a_vp->v_type; 477 vap->va_mode = pfs->pfs_mode; 478 vap->va_fileid = pfs->pfs_fileno; 479 vap->va_flags = 0; 480 vap->va_blocksize = PAGE_SIZE; 481 vap->va_bytes = vap->va_size = 0; 482 vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0]; 483 484 /* 485 * Make all times be current TOD. 486 * It would be possible to get the process start 487 * time from the p_stat structure, but there's 488 * no "file creation" time stamp anyway, and the 489 * p_stat structure is not addressible if u. gets 490 * swapped out for that process. 491 */ 492 nanotime(&vap->va_ctime); 493 vap->va_atime = vap->va_mtime = vap->va_ctime; 494 495 /* 496 * If the process has exercised some setuid or setgid 497 * privilege, then rip away read/write permission so 498 * that only root can gain access. 499 */ 500 switch (pfs->pfs_type) { 501 case Pctl: 502 case Pregs: 503 case Pfpregs: 504 case Pdbregs: 505 case Pmem: 506 if (procp->p_flag & P_SUGID) 507 vap->va_mode &= ~((VREAD|VWRITE)| 508 ((VREAD|VWRITE)>>3)| 509 ((VREAD|VWRITE)>>6)); 510 break; 511 default: 512 break; 513 } 514 515 /* 516 * now do the object specific fields 517 * 518 * The size could be set from struct reg, but it's hardly 519 * worth the trouble, and it puts some (potentially) machine 520 * dependent data into this machine-independent code. If it 521 * becomes important then this function should break out into 522 * a per-file stat function in the corresponding .c file. 523 */ 524 525 vap->va_nlink = 1; 526 if (procp) { 527 vap->va_uid = procp->p_ucred->cr_uid; 528 vap->va_gid = procp->p_ucred->cr_gid; 529 } 530 531 switch (pfs->pfs_type) { 532 case Proot: 533 /* 534 * Set nlink to 1 to tell fts(3) we don't actually know. 535 */ 536 vap->va_nlink = 1; 537 vap->va_uid = 0; 538 vap->va_gid = 0; 539 vap->va_size = vap->va_bytes = DEV_BSIZE; 540 break; 541 542 case Pcurproc: { 543 char buf[16]; /* should be enough */ 544 vap->va_uid = 0; 545 vap->va_gid = 0; 546 vap->va_size = vap->va_bytes = 547 ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid); 548 break; 549 } 550 551 case Pproc: 552 vap->va_nlink = nproc_targets; 553 vap->va_size = vap->va_bytes = DEV_BSIZE; 554 break; 555 556 case Pfile: { 557 char *fullpath, *freepath; 558 error = vn_fullpath(procp, NULL, &fullpath, &freepath); 559 if (error == 0) { 560 vap->va_size = strlen(fullpath); 561 kfree(freepath, M_TEMP); 562 } else { 563 vap->va_size = sizeof("unknown") - 1; 564 error = 0; 565 } 566 vap->va_bytes = vap->va_size; 567 break; 568 } 569 570 case Pmem: 571 /* 572 * If we denied owner access earlier, then we have to 573 * change the owner to root - otherwise 'ps' and friends 574 * will break even though they are setgid kmem. *SIGH* 575 */ 576 if (procp->p_flag & P_SUGID) 577 vap->va_uid = 0; 578 else 579 vap->va_uid = procp->p_ucred->cr_uid; 580 break; 581 582 case Pregs: 583 vap->va_bytes = vap->va_size = sizeof(struct reg); 584 break; 585 586 case Pfpregs: 587 vap->va_bytes = vap->va_size = sizeof(struct fpreg); 588 break; 589 590 case Pdbregs: 591 vap->va_bytes = vap->va_size = sizeof(struct dbreg); 592 break; 593 594 case Ptype: 595 case Pmap: 596 case Pctl: 597 case Pstatus: 598 case Pnote: 599 case Pnotepg: 600 case Pcmdline: 601 case Prlimit: 602 break; 603 604 default: 605 panic("procfs_getattr"); 606 } 607 608 return (error); 609 } 610 611 /* 612 * procfs_setattr(struct vnode *a_vp, struct vattr *a_vap, 613 * struct ucred *a_cred, struct thread *a_td) 614 */ 615 static int 616 procfs_setattr(struct vop_setattr_args *ap) 617 { 618 if (ap->a_vap->va_flags != VNOVAL) 619 return (EOPNOTSUPP); 620 621 /* 622 * just fake out attribute setting 623 * it's not good to generate an error 624 * return, otherwise things like creat() 625 * will fail when they try to set the 626 * file length to 0. worse, this means 627 * that echo $note > /proc/$pid/note will fail. 628 */ 629 630 return (0); 631 } 632 633 /* 634 * implement access checking. 635 * 636 * something very similar to this code is duplicated 637 * throughout the 4bsd kernel and should be moved 638 * into kern/vfs_subr.c sometime. 639 * 640 * actually, the check for super-user is slightly 641 * broken since it will allow read access to write-only 642 * objects. this doesn't cause any particular trouble 643 * but does mean that the i/o entry points need to check 644 * that the operation really does make sense. 645 * 646 * procfs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred, 647 * struct thread *a_td) 648 */ 649 static int 650 procfs_access(struct vop_access_args *ap) 651 { 652 struct vattr *vap; 653 struct vattr vattr; 654 int error; 655 656 /* 657 * If you're the super-user, 658 * you always get access. 659 */ 660 if (ap->a_cred->cr_uid == 0) 661 return (0); 662 663 vap = &vattr; 664 error = VOP_GETATTR(ap->a_vp, vap); 665 if (error) 666 return (error); 667 668 /* 669 * Access check is based on only one of owner, group, public. 670 * If not owner, then check group. If not a member of the 671 * group, then check public access. 672 */ 673 if (ap->a_cred->cr_uid != vap->va_uid) { 674 gid_t *gp; 675 int i; 676 677 ap->a_mode >>= 3; 678 gp = ap->a_cred->cr_groups; 679 for (i = 0; i < ap->a_cred->cr_ngroups; i++, gp++) 680 if (vap->va_gid == *gp) 681 goto found; 682 ap->a_mode >>= 3; 683 found: 684 ; 685 } 686 687 if ((vap->va_mode & ap->a_mode) == ap->a_mode) 688 return (0); 689 690 return (EACCES); 691 } 692 693 /* 694 * lookup. this is incredibly complicated in the general case, however 695 * for most pseudo-filesystems very little needs to be done. 696 * 697 * procfs_lookup(struct vnode *a_dvp, struct vnode **a_vpp, 698 * struct componentname *a_cnp) 699 */ 700 static int 701 procfs_lookup(struct vop_old_lookup_args *ap) 702 { 703 struct componentname *cnp = ap->a_cnp; 704 struct vnode **vpp = ap->a_vpp; 705 struct vnode *dvp = ap->a_dvp; 706 char *pname = cnp->cn_nameptr; 707 /* struct proc *curp = cnp->cn_proc; */ 708 struct proc_target *pt; 709 pid_t pid; 710 struct pfsnode *pfs; 711 struct proc *p; 712 struct lwp *lp; 713 int i; 714 int error; 715 716 *vpp = NULL; 717 718 if (cnp->cn_nameiop == NAMEI_DELETE || cnp->cn_nameiop == NAMEI_RENAME) 719 return (EROFS); 720 721 error = 0; 722 if (cnp->cn_namelen == 1 && *pname == '.') { 723 *vpp = dvp; 724 vref(*vpp); 725 goto out; 726 } 727 728 pfs = VTOPFS(dvp); 729 switch (pfs->pfs_type) { 730 case Proot: 731 if (cnp->cn_flags & CNP_ISDOTDOT) 732 return (EIO); 733 734 if (CNEQ(cnp, "curproc", 7)) { 735 error = procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc); 736 goto out; 737 } 738 739 pid = atopid(pname, cnp->cn_namelen); 740 if (pid == NO_PID) 741 break; 742 743 p = PFIND(pid); 744 if (p == NULL) 745 break; 746 747 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred)) 748 break; 749 750 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 && 751 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid) 752 break; 753 754 error = procfs_allocvp(dvp->v_mount, vpp, pid, Pproc); 755 goto out; 756 757 case Pproc: 758 if (cnp->cn_flags & CNP_ISDOTDOT) { 759 error = procfs_root(dvp->v_mount, vpp); 760 goto out; 761 } 762 763 p = PFIND(pfs->pfs_pid); 764 if (p == NULL) 765 break; 766 /* XXX lwp */ 767 lp = FIRST_LWP_IN_PROC(p); 768 769 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred)) 770 break; 771 772 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 && 773 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid) 774 break; 775 776 for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) { 777 if (cnp->cn_namelen == pt->pt_namlen && 778 bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 && 779 (pt->pt_valid == NULL || (*pt->pt_valid)(lp))) 780 goto found; 781 } 782 break; 783 found: 784 error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid, 785 pt->pt_pfstype); 786 goto out; 787 788 default: 789 error = ENOTDIR; 790 goto out; 791 } 792 if (cnp->cn_nameiop == NAMEI_LOOKUP) 793 error = ENOENT; 794 else 795 error = EROFS; 796 /* 797 * If no error occured *vpp will hold a referenced locked vnode. 798 * dvp was passed to us locked and *vpp must be returned locked. 799 * If *vpp != dvp then we should unlock dvp if (1) this is not the 800 * last component or (2) CNP_LOCKPARENT is not set. 801 */ 802 out: 803 if (error == 0 && *vpp != dvp) { 804 if ((cnp->cn_flags & CNP_LOCKPARENT) == 0) { 805 cnp->cn_flags |= CNP_PDIRUNLOCK; 806 vn_unlock(dvp); 807 } 808 } 809 return (error); 810 } 811 812 /* 813 * Does this process have a text file? 814 */ 815 int 816 procfs_validfile(struct lwp *lp) 817 { 818 return (procfs_findtextvp(lp->lwp_proc) != NULLVP); 819 } 820 821 /* 822 * readdir() returns directory entries from pfsnode (vp). 823 * 824 * We generate just one directory entry at a time, as it would probably 825 * not pay off to buffer several entries locally to save uiomove calls. 826 * 827 * procfs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred, 828 * int *a_eofflag, int *a_ncookies, u_long **a_cookies) 829 */ 830 static int 831 procfs_readdir(struct vop_readdir_args *ap) 832 { 833 struct pfsnode *pfs; 834 int error; 835 836 if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX) 837 return (EINVAL); 838 if ((error = vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY)) != 0) 839 return (error); 840 pfs = VTOPFS(ap->a_vp); 841 842 switch (pfs->pfs_type) { 843 case Pproc: 844 /* 845 * this is for the process-specific sub-directories. 846 * all that is needed to is copy out all the entries 847 * from the procent[] table (top of this file). 848 */ 849 error = procfs_readdir_proc(ap); 850 break; 851 case Proot: 852 /* 853 * this is for the root of the procfs filesystem 854 * what is needed is a special entry for "curproc" 855 * followed by an entry for each process on allproc 856 */ 857 error = procfs_readdir_root(ap); 858 break; 859 default: 860 error = ENOTDIR; 861 break; 862 } 863 864 vn_unlock(ap->a_vp); 865 return (error); 866 } 867 868 static int 869 procfs_readdir_proc(struct vop_readdir_args *ap) 870 { 871 struct pfsnode *pfs; 872 int error, i, retval; 873 struct proc *p; 874 struct lwp *lp; 875 struct proc_target *pt; 876 struct uio *uio = ap->a_uio; 877 878 pfs = VTOPFS(ap->a_vp); 879 p = PFIND(pfs->pfs_pid); 880 if (p == NULL) 881 return(0); 882 if (!PRISON_CHECK(ap->a_cred, p->p_ucred)) 883 return(0); 884 /* XXX lwp */ 885 lp = FIRST_LWP_IN_PROC(p); 886 887 error = 0; 888 i = (int)uio->uio_offset; 889 if (i < 0) 890 return (EINVAL); 891 892 for (pt = &proc_targets[i]; 893 !error && uio->uio_resid > 0 && i < nproc_targets; pt++, i++) { 894 if (pt->pt_valid && (*pt->pt_valid)(lp) == 0) 895 continue; 896 897 retval = vop_write_dirent(&error, uio, 898 PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype), pt->pt_type, 899 pt->pt_namlen, pt->pt_name); 900 if (retval) 901 break; 902 } 903 904 uio->uio_offset = (off_t)i; 905 906 return(0); 907 } 908 909 struct procfs_readdir_root_info { 910 int error; 911 int i; 912 int pcnt; 913 struct uio *uio; 914 struct ucred *cred; 915 }; 916 917 static int procfs_readdir_root_callback(struct proc *p, void *data); 918 919 static int 920 procfs_readdir_root(struct vop_readdir_args *ap) 921 { 922 struct procfs_readdir_root_info info; 923 struct uio *uio = ap->a_uio; 924 int res; 925 926 info.error = 0; 927 info.i = (int)uio->uio_offset; 928 929 if (info.i < 0) 930 return (EINVAL); 931 932 info.pcnt = 0; 933 info.uio = uio; 934 info.cred = ap->a_cred; 935 while (info.pcnt < 3) { 936 res = procfs_readdir_root_callback(NULL, &info); 937 if (res < 0) 938 break; 939 } 940 if (res >= 0) 941 allproc_scan(procfs_readdir_root_callback, &info); 942 uio->uio_offset = (off_t)info.i; 943 944 return (info.error); 945 } 946 947 static int 948 procfs_readdir_root_callback(struct proc *p, void *data) 949 { 950 struct procfs_readdir_root_info *info = data; 951 struct uio *uio; 952 int retval; 953 ino_t d_ino; 954 const char *d_name; 955 char d_name_pid[20]; 956 size_t d_namlen; 957 uint8_t d_type; 958 959 uio = info->uio; 960 961 if (uio->uio_resid <= 0 || info->error) 962 return(-1); 963 964 switch (info->pcnt) { 965 case 0: /* `.' */ 966 d_ino = PROCFS_FILENO(0, Proot); 967 d_name = "."; 968 d_namlen = 1; 969 d_type = DT_DIR; 970 break; 971 case 1: /* `..' */ 972 d_ino = PROCFS_FILENO(0, Proot); 973 d_name = ".."; 974 d_namlen = 2; 975 d_type = DT_DIR; 976 break; 977 978 case 2: 979 d_ino = PROCFS_FILENO(0, Pcurproc); 980 d_namlen = 7; 981 d_name = "curproc"; 982 d_type = DT_LNK; 983 break; 984 985 986 default: 987 if (!PRISON_CHECK(info->cred, p->p_ucred)) 988 return(0); 989 if (ps_showallprocs == 0 && 990 info->cred->cr_uid != 0 && 991 info->cred->cr_uid != p->p_ucred->cr_uid) { 992 return(0); 993 } 994 995 /* 996 * Skip entries we have already returned (optimization) 997 */ 998 if (info->pcnt < info->i) { 999 ++info->pcnt; 1000 return(0); 1001 } 1002 1003 d_ino = PROCFS_FILENO(p->p_pid, Pproc); 1004 d_namlen = ksnprintf(d_name_pid, sizeof(d_name_pid), 1005 "%ld", (long)p->p_pid); 1006 d_name = d_name_pid; 1007 d_type = DT_DIR; 1008 break; 1009 } 1010 1011 /* 1012 * Skip entries we have already returned (optimization) 1013 */ 1014 if (info->pcnt < info->i) { 1015 ++info->pcnt; 1016 return(0); 1017 } 1018 1019 retval = vop_write_dirent(&info->error, uio, 1020 d_ino, d_type, d_namlen, d_name); 1021 if (retval) 1022 return(-1); 1023 ++info->pcnt; 1024 ++info->i; 1025 return(0); 1026 } 1027 1028 /* 1029 * readlink reads the link of `curproc' or `file' 1030 */ 1031 static int 1032 procfs_readlink(struct vop_readlink_args *ap) 1033 { 1034 char buf[16]; /* should be enough */ 1035 struct proc *procp; 1036 struct vnode *vp = ap->a_vp; 1037 struct pfsnode *pfs = VTOPFS(vp); 1038 char *fullpath, *freepath; 1039 int error, len; 1040 1041 switch (pfs->pfs_type) { 1042 case Pcurproc: 1043 if (pfs->pfs_fileno != PROCFS_FILENO(0, Pcurproc)) 1044 return (EINVAL); 1045 1046 len = ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid); 1047 1048 return (uiomove(buf, len, ap->a_uio)); 1049 /* 1050 * There _should_ be no way for an entire process to disappear 1051 * from under us... 1052 */ 1053 case Pfile: 1054 procp = PFIND(pfs->pfs_pid); 1055 if (procp == NULL || procp->p_ucred == NULL) { 1056 kprintf("procfs_readlink: pid %d disappeared\n", 1057 pfs->pfs_pid); 1058 return (uiomove("unknown", sizeof("unknown") - 1, 1059 ap->a_uio)); 1060 } 1061 error = vn_fullpath(procp, NULL, &fullpath, &freepath); 1062 if (error != 0) 1063 return (uiomove("unknown", sizeof("unknown") - 1, 1064 ap->a_uio)); 1065 error = uiomove(fullpath, strlen(fullpath), ap->a_uio); 1066 kfree(freepath, M_TEMP); 1067 return (error); 1068 default: 1069 return (EINVAL); 1070 } 1071 } 1072 1073 /* 1074 * convert decimal ascii to pid_t 1075 */ 1076 static pid_t 1077 atopid(const char *b, u_int len) 1078 { 1079 pid_t p = 0; 1080 1081 while (len--) { 1082 char c = *b++; 1083 if (c < '0' || c > '9') 1084 return (NO_PID); 1085 p = 10 * p + (c - '0'); 1086 if (p > PID_MAX) 1087 return (NO_PID); 1088 } 1089 1090 return (p); 1091 } 1092 1093