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 1.43 2007/05/06 19:23:35 dillon 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 * v_opencount determines the last real close on the vnode. 226 * 227 * If this is the last close, then it checks to see if 228 * the target process has PF_LINGER set in p_pfsflags, 229 * if this is *not* the case, then the process' stop flags 230 * are cleared, and the process is woken up. This is 231 * to help prevent the case where a process has been 232 * told to stop on an event, but then the requesting process 233 * has gone away or forgotten about it. 234 */ 235 if ((ap->a_vp->v_opencount < 2) 236 && (p = pfind(pfs->pfs_pid)) 237 && !(p->p_pfsflags & PF_LINGER)) { 238 p->p_stops = 0; 239 p->p_step = 0; 240 wakeup(&p->p_step); 241 } 242 break; 243 default: 244 break; 245 } 246 247 return (vop_stdclose(ap)); 248 } 249 250 /* 251 * do an ioctl operation on a pfsnode (vp). 252 * (vp) is not locked on entry or exit. 253 */ 254 static int 255 procfs_ioctl(struct vop_ioctl_args *ap) 256 { 257 struct pfsnode *pfs = VTOPFS(ap->a_vp); 258 struct proc *procp; 259 struct proc *p; 260 int error; 261 int signo; 262 struct procfs_status *psp; 263 unsigned char flags; 264 265 procp = pfind(pfs->pfs_pid); 266 if (procp == NULL) 267 return ENOTTY; 268 p = curproc; 269 if (p == NULL) 270 return EINVAL; 271 272 /* Can't trace a process that's currently exec'ing. */ 273 if ((procp->p_flag & P_INEXEC) != 0) 274 return EAGAIN; 275 if (!CHECKIO(p, procp) || p_trespass(ap->a_cred, procp->p_ucred)) 276 return EPERM; 277 278 switch (ap->a_command) { 279 case PIOCBIS: 280 procp->p_stops |= *(unsigned int*)ap->a_data; 281 break; 282 case PIOCBIC: 283 procp->p_stops &= ~*(unsigned int*)ap->a_data; 284 break; 285 case PIOCSFL: 286 /* 287 * NFLAGS is "non-suser_xxx flags" -- currently, only 288 * PFS_ISUGID ("ignore set u/g id"); 289 */ 290 #define NFLAGS (PF_ISUGID) 291 flags = (unsigned char)*(unsigned int*)ap->a_data; 292 if (flags & NFLAGS && (error = suser_cred(ap->a_cred, 0))) 293 return error; 294 procp->p_pfsflags = flags; 295 break; 296 case PIOCGFL: 297 *(unsigned int*)ap->a_data = (unsigned int)procp->p_pfsflags; 298 break; 299 case PIOCSTATUS: 300 psp = (struct procfs_status *)ap->a_data; 301 psp->state = (procp->p_step == 0); 302 psp->flags = procp->p_pfsflags; 303 psp->events = procp->p_stops; 304 if (procp->p_step) { 305 psp->why = procp->p_stype; 306 psp->val = procp->p_xstat; 307 } else { 308 psp->why = psp->val = 0; /* Not defined values */ 309 } 310 break; 311 case PIOCWAIT: 312 psp = (struct procfs_status *)ap->a_data; 313 if (procp->p_step == 0) { 314 error = tsleep(&procp->p_stype, PCATCH, "piocwait", 0); 315 if (error) 316 return error; 317 } 318 psp->state = 1; /* It stopped */ 319 psp->flags = procp->p_pfsflags; 320 psp->events = procp->p_stops; 321 psp->why = procp->p_stype; /* why it stopped */ 322 psp->val = procp->p_xstat; /* any extra info */ 323 break; 324 case PIOCCONT: /* Restart a proc */ 325 if (procp->p_step == 0) 326 return EINVAL; /* Can only start a stopped process */ 327 if ((signo = *(int*)ap->a_data) != 0) { 328 if (signo >= NSIG || signo <= 0) 329 return EINVAL; 330 ksignal(procp, signo); 331 } 332 procp->p_step = 0; 333 wakeup(&procp->p_step); 334 break; 335 default: 336 return (ENOTTY); 337 } 338 return 0; 339 } 340 341 /* 342 * do block mapping for pfsnode (vp). 343 * since we don't use the buffer cache 344 * for procfs this function should never 345 * be called. in any case, it's not clear 346 * what part of the kernel ever makes use 347 * of this function. for sanity, this is the 348 * usual no-op bmap, although returning 349 * (EIO) would be a reasonable alternative. 350 * 351 * procfs_bmap(struct vnode *a_vp, off_t a_loffset, struct vnode **a_vpp, 352 * off_t *a_doffsetp, int *a_runp) 353 */ 354 static int 355 procfs_bmap(struct vop_bmap_args *ap) 356 { 357 if (ap->a_vpp != NULL) 358 *ap->a_vpp = ap->a_vp; 359 if (ap->a_doffsetp != NULL) 360 *ap->a_doffsetp = ap->a_loffset; 361 if (ap->a_runp != NULL) 362 *ap->a_runp = 0; 363 if (ap->a_runb != NULL) 364 *ap->a_runb = 0; 365 return (0); 366 } 367 368 /* 369 * procfs_inactive is called when the pfsnode 370 * is vrele'd and the reference count goes 371 * to zero. (vp) will be on the vnode free 372 * list, so to get it back vget() must be 373 * used. 374 * 375 * (vp) is locked on entry, but must be unlocked on exit. 376 * 377 * procfs_inactive(struct vnode *a_vp, struct thread *a_td) 378 */ 379 static int 380 procfs_inactive(struct vop_inactive_args *ap) 381 { 382 /*struct vnode *vp = ap->a_vp;*/ 383 384 return (0); 385 } 386 387 /* 388 * _reclaim is called when getnewvnode() 389 * wants to make use of an entry on the vnode 390 * free list. at this time the filesystem needs 391 * to free any private data and remove the node 392 * from any private lists. 393 * 394 * procfs_reclaim(struct vnode *a_vp) 395 */ 396 static int 397 procfs_reclaim(struct vop_reclaim_args *ap) 398 { 399 return (procfs_freevp(ap->a_vp)); 400 } 401 402 /* 403 * _print is used for debugging. 404 * just print a readable description 405 * of (vp). 406 * 407 * procfs_print(struct vnode *a_vp) 408 */ 409 static int 410 procfs_print(struct vop_print_args *ap) 411 { 412 struct pfsnode *pfs = VTOPFS(ap->a_vp); 413 414 kprintf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n", 415 pfs->pfs_type, (long)pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags); 416 return (0); 417 } 418 419 /* 420 * generic entry point for unsupported operations 421 */ 422 static int 423 procfs_badop(struct vop_generic_args *ap) 424 { 425 return (EIO); 426 } 427 428 /* 429 * Invent attributes for pfsnode (vp) and store 430 * them in (vap). 431 * Directories lengths are returned as zero since 432 * any real length would require the genuine size 433 * to be computed, and nothing cares anyway. 434 * 435 * this is relatively minimal for procfs. 436 * 437 * procfs_getattr(struct vnode *a_vp, struct vattr *a_vap, 438 * struct ucred *a_cred, struct thread *a_td) 439 */ 440 static int 441 procfs_getattr(struct vop_getattr_args *ap) 442 { 443 struct pfsnode *pfs = VTOPFS(ap->a_vp); 444 struct vattr *vap = ap->a_vap; 445 struct proc *procp; 446 int error; 447 448 /* 449 * First make sure that the process and its credentials 450 * still exist. 451 */ 452 switch (pfs->pfs_type) { 453 case Proot: 454 case Pcurproc: 455 procp = 0; 456 break; 457 458 default: 459 procp = PFIND(pfs->pfs_pid); 460 if (procp == NULL || procp->p_ucred == NULL) 461 return (ENOENT); 462 } 463 464 error = 0; 465 466 /* start by zeroing out the attributes */ 467 VATTR_NULL(vap); 468 469 /* next do all the common fields */ 470 vap->va_type = ap->a_vp->v_type; 471 vap->va_mode = pfs->pfs_mode; 472 vap->va_fileid = pfs->pfs_fileno; 473 vap->va_flags = 0; 474 vap->va_blocksize = PAGE_SIZE; 475 vap->va_bytes = vap->va_size = 0; 476 vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0]; 477 478 /* 479 * Make all times be current TOD. 480 * It would be possible to get the process start 481 * time from the p_stat structure, but there's 482 * no "file creation" time stamp anyway, and the 483 * p_stat structure is not addressible if u. gets 484 * swapped out for that process. 485 */ 486 nanotime(&vap->va_ctime); 487 vap->va_atime = vap->va_mtime = vap->va_ctime; 488 489 /* 490 * If the process has exercised some setuid or setgid 491 * privilege, then rip away read/write permission so 492 * that only root can gain access. 493 */ 494 switch (pfs->pfs_type) { 495 case Pctl: 496 case Pregs: 497 case Pfpregs: 498 case Pdbregs: 499 case Pmem: 500 if (procp->p_flag & P_SUGID) 501 vap->va_mode &= ~((VREAD|VWRITE)| 502 ((VREAD|VWRITE)>>3)| 503 ((VREAD|VWRITE)>>6)); 504 break; 505 default: 506 break; 507 } 508 509 /* 510 * now do the object specific fields 511 * 512 * The size could be set from struct reg, but it's hardly 513 * worth the trouble, and it puts some (potentially) machine 514 * dependent data into this machine-independent code. If it 515 * becomes important then this function should break out into 516 * a per-file stat function in the corresponding .c file. 517 */ 518 519 vap->va_nlink = 1; 520 if (procp) { 521 vap->va_uid = procp->p_ucred->cr_uid; 522 vap->va_gid = procp->p_ucred->cr_gid; 523 } 524 525 switch (pfs->pfs_type) { 526 case Proot: 527 /* 528 * Set nlink to 1 to tell fts(3) we don't actually know. 529 */ 530 vap->va_nlink = 1; 531 vap->va_uid = 0; 532 vap->va_gid = 0; 533 vap->va_size = vap->va_bytes = DEV_BSIZE; 534 break; 535 536 case Pcurproc: { 537 char buf[16]; /* should be enough */ 538 vap->va_uid = 0; 539 vap->va_gid = 0; 540 vap->va_size = vap->va_bytes = 541 ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid); 542 break; 543 } 544 545 case Pproc: 546 vap->va_nlink = nproc_targets; 547 vap->va_size = vap->va_bytes = DEV_BSIZE; 548 break; 549 550 case Pfile: { 551 char *fullpath, *freepath; 552 error = vn_fullpath(procp, NULL, &fullpath, &freepath); 553 if (error == 0) { 554 vap->va_size = strlen(fullpath); 555 kfree(freepath, M_TEMP); 556 } else { 557 vap->va_size = sizeof("unknown") - 1; 558 error = 0; 559 } 560 vap->va_bytes = vap->va_size; 561 break; 562 } 563 564 case Pmem: 565 /* 566 * If we denied owner access earlier, then we have to 567 * change the owner to root - otherwise 'ps' and friends 568 * will break even though they are setgid kmem. *SIGH* 569 */ 570 if (procp->p_flag & P_SUGID) 571 vap->va_uid = 0; 572 else 573 vap->va_uid = procp->p_ucred->cr_uid; 574 break; 575 576 case Pregs: 577 vap->va_bytes = vap->va_size = sizeof(struct reg); 578 break; 579 580 case Pfpregs: 581 vap->va_bytes = vap->va_size = sizeof(struct fpreg); 582 break; 583 584 case Pdbregs: 585 vap->va_bytes = vap->va_size = sizeof(struct dbreg); 586 break; 587 588 case Ptype: 589 case Pmap: 590 case Pctl: 591 case Pstatus: 592 case Pnote: 593 case Pnotepg: 594 case Pcmdline: 595 case Prlimit: 596 break; 597 598 default: 599 panic("procfs_getattr"); 600 } 601 602 return (error); 603 } 604 605 /* 606 * procfs_setattr(struct vnode *a_vp, struct vattr *a_vap, 607 * struct ucred *a_cred, struct thread *a_td) 608 */ 609 static int 610 procfs_setattr(struct vop_setattr_args *ap) 611 { 612 if (ap->a_vap->va_flags != VNOVAL) 613 return (EOPNOTSUPP); 614 615 /* 616 * just fake out attribute setting 617 * it's not good to generate an error 618 * return, otherwise things like creat() 619 * will fail when they try to set the 620 * file length to 0. worse, this means 621 * that echo $note > /proc/$pid/note will fail. 622 */ 623 624 return (0); 625 } 626 627 /* 628 * implement access checking. 629 * 630 * something very similar to this code is duplicated 631 * throughout the 4bsd kernel and should be moved 632 * into kern/vfs_subr.c sometime. 633 * 634 * actually, the check for super-user is slightly 635 * broken since it will allow read access to write-only 636 * objects. this doesn't cause any particular trouble 637 * but does mean that the i/o entry points need to check 638 * that the operation really does make sense. 639 * 640 * procfs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred, 641 * struct thread *a_td) 642 */ 643 static int 644 procfs_access(struct vop_access_args *ap) 645 { 646 struct vattr *vap; 647 struct vattr vattr; 648 int error; 649 650 /* 651 * If you're the super-user, 652 * you always get access. 653 */ 654 if (ap->a_cred->cr_uid == 0) 655 return (0); 656 657 vap = &vattr; 658 error = VOP_GETATTR(ap->a_vp, vap); 659 if (error) 660 return (error); 661 662 /* 663 * Access check is based on only one of owner, group, public. 664 * If not owner, then check group. If not a member of the 665 * group, then check public access. 666 */ 667 if (ap->a_cred->cr_uid != vap->va_uid) { 668 gid_t *gp; 669 int i; 670 671 ap->a_mode >>= 3; 672 gp = ap->a_cred->cr_groups; 673 for (i = 0; i < ap->a_cred->cr_ngroups; i++, gp++) 674 if (vap->va_gid == *gp) 675 goto found; 676 ap->a_mode >>= 3; 677 found: 678 ; 679 } 680 681 if ((vap->va_mode & ap->a_mode) == ap->a_mode) 682 return (0); 683 684 return (EACCES); 685 } 686 687 /* 688 * lookup. this is incredibly complicated in the general case, however 689 * for most pseudo-filesystems very little needs to be done. 690 * 691 * procfs_lookup(struct vnode *a_dvp, struct vnode **a_vpp, 692 * struct componentname *a_cnp) 693 */ 694 static int 695 procfs_lookup(struct vop_old_lookup_args *ap) 696 { 697 struct componentname *cnp = ap->a_cnp; 698 struct vnode **vpp = ap->a_vpp; 699 struct vnode *dvp = ap->a_dvp; 700 char *pname = cnp->cn_nameptr; 701 /* struct proc *curp = cnp->cn_proc; */ 702 struct proc_target *pt; 703 pid_t pid; 704 struct pfsnode *pfs; 705 struct proc *p; 706 struct lwp *lp; 707 int i; 708 int error; 709 710 *vpp = NULL; 711 712 if (cnp->cn_nameiop == NAMEI_DELETE || cnp->cn_nameiop == NAMEI_RENAME) 713 return (EROFS); 714 715 error = 0; 716 if (cnp->cn_namelen == 1 && *pname == '.') { 717 *vpp = dvp; 718 vref(*vpp); 719 goto out; 720 } 721 722 pfs = VTOPFS(dvp); 723 switch (pfs->pfs_type) { 724 case Proot: 725 if (cnp->cn_flags & CNP_ISDOTDOT) 726 return (EIO); 727 728 if (CNEQ(cnp, "curproc", 7)) { 729 error = procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc); 730 goto out; 731 } 732 733 pid = atopid(pname, cnp->cn_namelen); 734 if (pid == NO_PID) 735 break; 736 737 p = PFIND(pid); 738 if (p == NULL) 739 break; 740 741 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred)) 742 break; 743 744 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 && 745 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid) 746 break; 747 748 error = procfs_allocvp(dvp->v_mount, vpp, pid, Pproc); 749 goto out; 750 751 case Pproc: 752 if (cnp->cn_flags & CNP_ISDOTDOT) { 753 error = procfs_root(dvp->v_mount, vpp); 754 goto out; 755 } 756 757 p = PFIND(pfs->pfs_pid); 758 if (p == NULL) 759 break; 760 /* XXX lwp */ 761 lp = FIRST_LWP_IN_PROC(p); 762 763 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred)) 764 break; 765 766 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 && 767 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid) 768 break; 769 770 for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) { 771 if (cnp->cn_namelen == pt->pt_namlen && 772 bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 && 773 (pt->pt_valid == NULL || (*pt->pt_valid)(lp))) 774 goto found; 775 } 776 break; 777 found: 778 error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid, 779 pt->pt_pfstype); 780 goto out; 781 782 default: 783 error = ENOTDIR; 784 goto out; 785 } 786 if (cnp->cn_nameiop == NAMEI_LOOKUP) 787 error = ENOENT; 788 else 789 error = EROFS; 790 /* 791 * If no error occured *vpp will hold a referenced locked vnode. 792 * dvp was passed to us locked and *vpp must be returned locked. 793 * If *vpp != dvp then we should unlock dvp if (1) this is not the 794 * last component or (2) CNP_LOCKPARENT is not set. 795 */ 796 out: 797 if (error == 0 && *vpp != dvp) { 798 if ((cnp->cn_flags & CNP_LOCKPARENT) == 0) { 799 cnp->cn_flags |= CNP_PDIRUNLOCK; 800 vn_unlock(dvp); 801 } 802 } 803 return (error); 804 } 805 806 /* 807 * Does this process have a text file? 808 */ 809 int 810 procfs_validfile(struct lwp *lp) 811 { 812 return (procfs_findtextvp(lp->lwp_proc) != NULLVP); 813 } 814 815 /* 816 * readdir() returns directory entries from pfsnode (vp). 817 * 818 * We generate just one directory entry at a time, as it would probably 819 * not pay off to buffer several entries locally to save uiomove calls. 820 * 821 * procfs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred, 822 * int *a_eofflag, int *a_ncookies, u_long **a_cookies) 823 */ 824 static int 825 procfs_readdir(struct vop_readdir_args *ap) 826 { 827 struct pfsnode *pfs; 828 int error; 829 830 if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX) 831 return (EINVAL); 832 if ((error = vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY)) != 0) 833 return (error); 834 pfs = VTOPFS(ap->a_vp); 835 836 switch (pfs->pfs_type) { 837 case Pproc: 838 /* 839 * this is for the process-specific sub-directories. 840 * all that is needed to is copy out all the entries 841 * from the procent[] table (top of this file). 842 */ 843 error = procfs_readdir_proc(ap); 844 break; 845 case Proot: 846 /* 847 * this is for the root of the procfs filesystem 848 * what is needed is a special entry for "curproc" 849 * followed by an entry for each process on allproc 850 */ 851 error = procfs_readdir_root(ap); 852 break; 853 default: 854 error = ENOTDIR; 855 break; 856 } 857 858 vn_unlock(ap->a_vp); 859 return (error); 860 } 861 862 static int 863 procfs_readdir_proc(struct vop_readdir_args *ap) 864 { 865 struct pfsnode *pfs; 866 int error, i, retval; 867 struct proc *p; 868 struct lwp *lp; 869 struct proc_target *pt; 870 struct uio *uio = ap->a_uio; 871 872 pfs = VTOPFS(ap->a_vp); 873 p = PFIND(pfs->pfs_pid); 874 if (p == NULL) 875 return(0); 876 if (!PRISON_CHECK(ap->a_cred, p->p_ucred)) 877 return(0); 878 /* XXX lwp */ 879 lp = FIRST_LWP_IN_PROC(p); 880 881 error = 0; 882 i = (int)uio->uio_offset; 883 if (i < 0) 884 return (EINVAL); 885 886 for (pt = &proc_targets[i]; 887 !error && uio->uio_resid > 0 && i < nproc_targets; pt++, i++) { 888 if (pt->pt_valid && (*pt->pt_valid)(lp) == 0) 889 continue; 890 891 retval = vop_write_dirent(&error, uio, 892 PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype), pt->pt_type, 893 pt->pt_namlen, pt->pt_name); 894 if (retval) 895 break; 896 } 897 898 uio->uio_offset = (off_t)i; 899 900 return(0); 901 } 902 903 struct procfs_readdir_root_info { 904 int error; 905 int i; 906 int pcnt; 907 struct uio *uio; 908 struct ucred *cred; 909 }; 910 911 static int procfs_readdir_root_callback(struct proc *p, void *data); 912 913 static int 914 procfs_readdir_root(struct vop_readdir_args *ap) 915 { 916 struct procfs_readdir_root_info info; 917 struct uio *uio = ap->a_uio; 918 int res; 919 920 info.error = 0; 921 info.i = (int)uio->uio_offset; 922 923 if (info.i < 0) 924 return (EINVAL); 925 926 info.pcnt = 0; 927 info.uio = uio; 928 info.cred = ap->a_cred; 929 while (info.pcnt < 3) { 930 res = procfs_readdir_root_callback(NULL, &info); 931 if (res < 0) 932 break; 933 } 934 if (res >= 0) 935 allproc_scan(procfs_readdir_root_callback, &info); 936 uio->uio_offset = (off_t)info.i; 937 938 return (info.error); 939 } 940 941 static int 942 procfs_readdir_root_callback(struct proc *p, void *data) 943 { 944 struct procfs_readdir_root_info *info = data; 945 struct uio *uio; 946 int retval; 947 ino_t d_ino; 948 const char *d_name; 949 char d_name_pid[20]; 950 size_t d_namlen; 951 uint8_t d_type; 952 953 uio = info->uio; 954 955 if (uio->uio_resid <= 0 || info->error) 956 return(-1); 957 958 switch (info->pcnt) { 959 case 0: /* `.' */ 960 d_ino = PROCFS_FILENO(0, Proot); 961 d_name = "."; 962 d_namlen = 1; 963 d_type = DT_DIR; 964 break; 965 case 1: /* `..' */ 966 d_ino = PROCFS_FILENO(0, Proot); 967 d_name = ".."; 968 d_namlen = 2; 969 d_type = DT_DIR; 970 break; 971 972 case 2: 973 d_ino = PROCFS_FILENO(0, Pcurproc); 974 d_namlen = 7; 975 d_name = "curproc"; 976 d_type = DT_LNK; 977 break; 978 979 980 default: 981 if (!PRISON_CHECK(info->cred, p->p_ucred)) 982 return(0); 983 if (ps_showallprocs == 0 && 984 info->cred->cr_uid != 0 && 985 info->cred->cr_uid != p->p_ucred->cr_uid) { 986 return(0); 987 } 988 989 /* 990 * Skip entries we have already returned (optimization) 991 */ 992 if (info->pcnt < info->i) { 993 ++info->pcnt; 994 return(0); 995 } 996 997 d_ino = PROCFS_FILENO(p->p_pid, Pproc); 998 d_namlen = ksnprintf(d_name_pid, sizeof(d_name_pid), 999 "%ld", (long)p->p_pid); 1000 d_name = d_name_pid; 1001 d_type = DT_DIR; 1002 break; 1003 } 1004 1005 /* 1006 * Skip entries we have already returned (optimization) 1007 */ 1008 if (info->pcnt < info->i) { 1009 ++info->pcnt; 1010 return(0); 1011 } 1012 1013 retval = vop_write_dirent(&info->error, uio, 1014 d_ino, d_type, d_namlen, d_name); 1015 if (retval) 1016 return(-1); 1017 ++info->pcnt; 1018 ++info->i; 1019 return(0); 1020 } 1021 1022 /* 1023 * readlink reads the link of `curproc' or `file' 1024 */ 1025 static int 1026 procfs_readlink(struct vop_readlink_args *ap) 1027 { 1028 char buf[16]; /* should be enough */ 1029 struct proc *procp; 1030 struct vnode *vp = ap->a_vp; 1031 struct pfsnode *pfs = VTOPFS(vp); 1032 char *fullpath, *freepath; 1033 int error, len; 1034 1035 switch (pfs->pfs_type) { 1036 case Pcurproc: 1037 if (pfs->pfs_fileno != PROCFS_FILENO(0, Pcurproc)) 1038 return (EINVAL); 1039 1040 len = ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid); 1041 1042 return (uiomove(buf, len, ap->a_uio)); 1043 /* 1044 * There _should_ be no way for an entire process to disappear 1045 * from under us... 1046 */ 1047 case Pfile: 1048 procp = PFIND(pfs->pfs_pid); 1049 if (procp == NULL || procp->p_ucred == NULL) { 1050 kprintf("procfs_readlink: pid %d disappeared\n", 1051 pfs->pfs_pid); 1052 return (uiomove("unknown", sizeof("unknown") - 1, 1053 ap->a_uio)); 1054 } 1055 error = vn_fullpath(procp, NULL, &fullpath, &freepath); 1056 if (error != 0) 1057 return (uiomove("unknown", sizeof("unknown") - 1, 1058 ap->a_uio)); 1059 error = uiomove(fullpath, strlen(fullpath), ap->a_uio); 1060 kfree(freepath, M_TEMP); 1061 return (error); 1062 default: 1063 return (EINVAL); 1064 } 1065 } 1066 1067 /* 1068 * convert decimal ascii to pid_t 1069 */ 1070 static pid_t 1071 atopid(const char *b, u_int len) 1072 { 1073 pid_t p = 0; 1074 1075 while (len--) { 1076 char c = *b++; 1077 if (c < '0' || c > '9') 1078 return (NO_PID); 1079 p = 10 * p + (c - '0'); 1080 if (p > PID_MAX) 1081 return (NO_PID); 1082 } 1083 1084 return (p); 1085 } 1086 1087