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 1.46 2007/11/20 21:03:50 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/priv.h> 55 #include <sys/signalvar.h> 56 #include <sys/vnode.h> 57 #include <sys/uio.h> 58 #include <sys/mount.h> 59 #include <sys/namei.h> 60 #include <sys/dirent.h> 61 #include <sys/malloc.h> 62 #include <sys/reg.h> 63 #include <vm/vm_zone.h> 64 #include <vfs/procfs/procfs.h> 65 #include <sys/pioctl.h> 66 67 #include <sys/spinlock2.h> 68 69 #include <machine/limits.h> 70 71 static int procfs_access (struct vop_access_args *); 72 static int procfs_badop (struct vop_generic_args *); 73 static int procfs_bmap (struct vop_bmap_args *); 74 static int procfs_close (struct vop_close_args *); 75 static int procfs_getattr (struct vop_getattr_args *); 76 static int procfs_inactive (struct vop_inactive_args *); 77 static int procfs_ioctl (struct vop_ioctl_args *); 78 static int procfs_lookup (struct vop_old_lookup_args *); 79 static int procfs_open (struct vop_open_args *); 80 static int procfs_print (struct vop_print_args *); 81 static int procfs_readdir (struct vop_readdir_args *); 82 static int procfs_readlink (struct vop_readlink_args *); 83 static int procfs_reclaim (struct vop_reclaim_args *); 84 static int procfs_setattr (struct vop_setattr_args *); 85 86 static int procfs_readdir_proc(struct vop_readdir_args *); 87 static int procfs_readdir_root(struct vop_readdir_args *); 88 89 /* 90 * procfs vnode operations. 91 */ 92 struct vop_ops procfs_vnode_vops = { 93 .vop_default = vop_defaultop, 94 .vop_access = procfs_access, 95 .vop_advlock = (void *)procfs_badop, 96 .vop_bmap = procfs_bmap, 97 .vop_close = procfs_close, 98 .vop_old_create = (void *)procfs_badop, 99 .vop_getattr = procfs_getattr, 100 .vop_inactive = procfs_inactive, 101 .vop_old_link = (void *)procfs_badop, 102 .vop_old_lookup = procfs_lookup, 103 .vop_old_mkdir = (void *)procfs_badop, 104 .vop_old_mknod = (void *)procfs_badop, 105 .vop_open = procfs_open, 106 .vop_pathconf = vop_stdpathconf, 107 .vop_print = procfs_print, 108 .vop_read = procfs_rw, 109 .vop_readdir = procfs_readdir, 110 .vop_readlink = procfs_readlink, 111 .vop_reclaim = procfs_reclaim, 112 .vop_old_remove = (void *)procfs_badop, 113 .vop_old_rename = (void *)procfs_badop, 114 .vop_old_rmdir = (void *)procfs_badop, 115 .vop_setattr = procfs_setattr, 116 .vop_old_symlink = (void *)procfs_badop, 117 .vop_write = (void *)procfs_rw, 118 .vop_ioctl = procfs_ioctl 119 }; 120 121 122 /* 123 * This is a list of the valid names in the 124 * process-specific sub-directories. It is 125 * used in procfs_lookup and procfs_readdir 126 */ 127 static struct proc_target { 128 u_char pt_type; 129 u_char pt_namlen; 130 char *pt_name; 131 pfstype pt_pfstype; 132 int (*pt_valid) (struct lwp *p); 133 } proc_targets[] = { 134 #define N(s) sizeof(s)-1, s 135 /* name type validp */ 136 { DT_DIR, N("."), Pproc, NULL }, 137 { DT_DIR, N(".."), Proot, NULL }, 138 { DT_REG, N("mem"), Pmem, NULL }, 139 { DT_REG, N("regs"), Pregs, procfs_validregs }, 140 { DT_REG, N("fpregs"), Pfpregs, procfs_validfpregs }, 141 { DT_REG, N("dbregs"), Pdbregs, procfs_validdbregs }, 142 { DT_REG, N("ctl"), Pctl, NULL }, 143 { DT_REG, N("status"), Pstatus, NULL }, 144 { DT_REG, N("note"), Pnote, NULL }, 145 { DT_REG, N("notepg"), Pnotepg, NULL }, 146 { DT_REG, N("map"), Pmap, procfs_validmap }, 147 { DT_REG, N("etype"), Ptype, procfs_validtype }, 148 { DT_REG, N("cmdline"), Pcmdline, NULL }, 149 { DT_REG, N("rlimit"), Prlimit, NULL }, 150 { DT_LNK, N("file"), Pfile, NULL }, 151 #undef N 152 }; 153 static const int nproc_targets = sizeof(proc_targets) / sizeof(proc_targets[0]); 154 155 static pid_t atopid (const char *, u_int); 156 157 /* 158 * set things up for doing i/o on 159 * the pfsnode (vp). (vp) is locked 160 * on entry, and should be left locked 161 * on exit. 162 * 163 * for procfs we don't need to do anything 164 * in particular for i/o. all that is done 165 * is to support exclusive open on process 166 * memory images. 167 * 168 * procfs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred, 169 * struct file *a_fp) 170 */ 171 static int 172 procfs_open(struct vop_open_args *ap) 173 { 174 struct pfsnode *pfs = VTOPFS(ap->a_vp); 175 struct proc *p1, *p2; 176 177 p2 = 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 return (ENOENT); 182 183 switch (pfs->pfs_type) { 184 case Pmem: 185 if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) || 186 ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE))) 187 return (EBUSY); 188 189 p1 = curproc; 190 KKASSERT(p1); 191 /* Can't trace a process that's currently exec'ing. */ 192 if ((p2->p_flag & P_INEXEC) != 0) 193 return EAGAIN; 194 if (!CHECKIO(p1, p2) || p_trespass(ap->a_cred, p2->p_ucred)) 195 return (EPERM); 196 197 if (ap->a_mode & FWRITE) 198 pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL); 199 200 break; 201 202 default: 203 break; 204 } 205 206 return (vop_stdopen(ap)); 207 } 208 209 /* 210 * close the pfsnode (vp) after doing i/o. 211 * (vp) is not locked on entry or exit. 212 * 213 * nothing to do for procfs other than undo 214 * any exclusive open flag (see _open above). 215 * 216 * procfs_close(struct vnode *a_vp, int a_fflag, struct ucred *a_cred) 217 */ 218 static int 219 procfs_close(struct vop_close_args *ap) 220 { 221 struct pfsnode *pfs = VTOPFS(ap->a_vp); 222 struct proc *p; 223 224 switch (pfs->pfs_type) { 225 case Pmem: 226 if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL)) 227 pfs->pfs_flags &= ~(FWRITE|O_EXCL); 228 /* 229 * v_opencount determines the last real close on the vnode. 230 * 231 * If this is the last close, then it checks to see if 232 * the target process has PF_LINGER set in p_pfsflags, 233 * if this is *not* the case, then the process' stop flags 234 * are cleared, and the process is woken up. This is 235 * to help prevent the case where a process has been 236 * told to stop on an event, but then the requesting process 237 * has gone away or forgotten about it. 238 */ 239 if ((ap->a_vp->v_opencount < 2) 240 && (p = pfind(pfs->pfs_pid)) 241 && !(p->p_pfsflags & PF_LINGER)) { 242 spin_lock(&p->p_spin); 243 p->p_stops = 0; 244 p->p_step = 0; 245 spin_unlock(&p->p_spin); 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 = priv_check_cred(ap->a_cred, PRIV_ROOT, 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 /* 307 * NOTE: syscall entry deals with stopevents and may run without 308 * the MP lock. 309 */ 310 psp = (struct procfs_status *)ap->a_data; 311 psp->flags = procp->p_pfsflags; 312 psp->events = procp->p_stops; 313 spin_lock(&procp->p_spin); 314 if (procp->p_step) { 315 psp->state = 0; 316 psp->why = procp->p_stype; 317 psp->val = procp->p_xstat; 318 spin_unlock(&procp->p_spin); 319 } else { 320 psp->state = 1; 321 spin_unlock(&procp->p_spin); 322 psp->why = 0; /* Not defined values */ 323 psp->val = 0; /* Not defined values */ 324 } 325 break; 326 case PIOCWAIT: 327 /* 328 * NOTE: syscall entry deals with stopevents and may run without 329 * the MP lock. 330 */ 331 psp = (struct procfs_status *)ap->a_data; 332 spin_lock(&procp->p_spin); 333 while (procp->p_step == 0) { 334 tsleep_interlock(&procp->p_stype, PCATCH); 335 spin_unlock(&procp->p_spin); 336 error = tsleep(&procp->p_stype, PCATCH | PINTERLOCKED, "piocwait", 0); 337 if (error) 338 return error; 339 spin_lock(&procp->p_spin); 340 } 341 spin_unlock(&procp->p_spin); 342 psp->state = 1; /* It stopped */ 343 psp->flags = procp->p_pfsflags; 344 psp->events = procp->p_stops; 345 psp->why = procp->p_stype; /* why it stopped */ 346 psp->val = procp->p_xstat; /* any extra info */ 347 break; 348 case PIOCCONT: /* Restart a proc */ 349 /* 350 * NOTE: syscall entry deals with stopevents and may run without 351 * the MP lock. However, the caller is presumably interlocked 352 * by having waited. 353 */ 354 if (procp->p_step == 0) 355 return EINVAL; /* Can only start a stopped process */ 356 if ((signo = *(int*)ap->a_data) != 0) { 357 if (signo >= NSIG || signo <= 0) 358 return EINVAL; 359 ksignal(procp, signo); 360 } 361 procp->p_step = 0; 362 wakeup(&procp->p_step); 363 break; 364 default: 365 return (ENOTTY); 366 } 367 return 0; 368 } 369 370 /* 371 * do block mapping for pfsnode (vp). 372 * since we don't use the buffer cache 373 * for procfs this function should never 374 * be called. in any case, it's not clear 375 * what part of the kernel ever makes use 376 * of this function. for sanity, this is the 377 * usual no-op bmap, although returning 378 * (EIO) would be a reasonable alternative. 379 * 380 * XXX mmap assumes buffer cache operation 381 * 382 * procfs_bmap(struct vnode *a_vp, off_t a_loffset, 383 * off_t *a_doffsetp, int *a_runp, int *a_runb) 384 */ 385 static int 386 procfs_bmap(struct vop_bmap_args *ap) 387 { 388 if (ap->a_doffsetp != NULL) 389 *ap->a_doffsetp = ap->a_loffset; 390 if (ap->a_runp != NULL) 391 *ap->a_runp = 0; 392 if (ap->a_runb != NULL) 393 *ap->a_runb = 0; 394 return (0); 395 } 396 397 /* 398 * procfs_inactive is called when the pfsnode 399 * is vrele'd and the reference count goes 400 * to zero. (vp) will be on the vnode free 401 * list, so to get it back vget() must be 402 * used. 403 * 404 * (vp) is locked on entry, but must be unlocked on exit. 405 * 406 * procfs_inactive(struct vnode *a_vp) 407 */ 408 static int 409 procfs_inactive(struct vop_inactive_args *ap) 410 { 411 struct pfsnode *pfs = VTOPFS(ap->a_vp); 412 413 if (pfs->pfs_pid & PFS_DEAD) 414 vrecycle(ap->a_vp); 415 return (0); 416 } 417 418 /* 419 * _reclaim is called when getnewvnode() 420 * wants to make use of an entry on the vnode 421 * free list. at this time the filesystem needs 422 * to free any private data and remove the node 423 * from any private lists. 424 * 425 * procfs_reclaim(struct vnode *a_vp) 426 */ 427 static int 428 procfs_reclaim(struct vop_reclaim_args *ap) 429 { 430 return (procfs_freevp(ap->a_vp)); 431 } 432 433 /* 434 * _print is used for debugging. 435 * just print a readable description 436 * of (vp). 437 * 438 * procfs_print(struct vnode *a_vp) 439 */ 440 static int 441 procfs_print(struct vop_print_args *ap) 442 { 443 struct pfsnode *pfs = VTOPFS(ap->a_vp); 444 445 kprintf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n", 446 pfs->pfs_type, (long)pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags); 447 return (0); 448 } 449 450 /* 451 * generic entry point for unsupported operations 452 */ 453 static int 454 procfs_badop(struct vop_generic_args *ap) 455 { 456 return (EIO); 457 } 458 459 /* 460 * Invent attributes for pfsnode (vp) and store 461 * them in (vap). 462 * Directories lengths are returned as zero since 463 * any real length would require the genuine size 464 * to be computed, and nothing cares anyway. 465 * 466 * this is relatively minimal for procfs. 467 * 468 * procfs_getattr(struct vnode *a_vp, struct vattr *a_vap) 469 */ 470 static int 471 procfs_getattr(struct vop_getattr_args *ap) 472 { 473 struct pfsnode *pfs = VTOPFS(ap->a_vp); 474 struct vattr *vap = ap->a_vap; 475 struct proc *procp; 476 int error; 477 478 /* 479 * First make sure that the process and its credentials 480 * still exist. 481 */ 482 switch (pfs->pfs_type) { 483 case Proot: 484 case Pcurproc: 485 procp = 0; 486 break; 487 488 default: 489 procp = PFIND(pfs->pfs_pid); 490 if (procp == NULL || procp->p_ucred == NULL) 491 return (ENOENT); 492 } 493 494 error = 0; 495 496 /* start by zeroing out the attributes */ 497 VATTR_NULL(vap); 498 499 /* next do all the common fields */ 500 vap->va_type = ap->a_vp->v_type; 501 vap->va_mode = pfs->pfs_mode; 502 vap->va_fileid = pfs->pfs_fileno; 503 vap->va_flags = 0; 504 vap->va_blocksize = PAGE_SIZE; 505 vap->va_bytes = vap->va_size = 0; 506 vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0]; 507 508 /* 509 * Make all times be current TOD. 510 * It would be possible to get the process start 511 * time from the p_stat structure, but there's 512 * no "file creation" time stamp anyway, and the 513 * p_stat structure is not addressible if u. gets 514 * swapped out for that process. 515 */ 516 nanotime(&vap->va_ctime); 517 vap->va_atime = vap->va_mtime = vap->va_ctime; 518 519 /* 520 * If the process has exercised some setuid or setgid 521 * privilege, then rip away read/write permission so 522 * that only root can gain access. 523 */ 524 switch (pfs->pfs_type) { 525 case Pctl: 526 case Pregs: 527 case Pfpregs: 528 case Pdbregs: 529 case Pmem: 530 if (procp->p_flag & P_SUGID) 531 vap->va_mode &= ~((VREAD|VWRITE)| 532 ((VREAD|VWRITE)>>3)| 533 ((VREAD|VWRITE)>>6)); 534 break; 535 default: 536 break; 537 } 538 539 /* 540 * now do the object specific fields 541 * 542 * The size could be set from struct reg, but it's hardly 543 * worth the trouble, and it puts some (potentially) machine 544 * dependent data into this machine-independent code. If it 545 * becomes important then this function should break out into 546 * a per-file stat function in the corresponding .c file. 547 */ 548 549 vap->va_nlink = 1; 550 if (procp) { 551 vap->va_uid = procp->p_ucred->cr_uid; 552 vap->va_gid = procp->p_ucred->cr_gid; 553 } 554 555 switch (pfs->pfs_type) { 556 case Proot: 557 /* 558 * Set nlink to 1 to tell fts(3) we don't actually know. 559 */ 560 vap->va_nlink = 1; 561 vap->va_uid = 0; 562 vap->va_gid = 0; 563 vap->va_size = vap->va_bytes = DEV_BSIZE; 564 break; 565 566 case Pcurproc: { 567 char buf[16]; /* should be enough */ 568 vap->va_uid = 0; 569 vap->va_gid = 0; 570 vap->va_size = vap->va_bytes = 571 ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid); 572 break; 573 } 574 575 case Pproc: 576 vap->va_nlink = nproc_targets; 577 vap->va_size = vap->va_bytes = DEV_BSIZE; 578 break; 579 580 case Pfile: { 581 char *fullpath, *freepath; 582 error = cache_fullpath(procp, &procp->p_textnch, &fullpath, &freepath, 0); 583 if (error == 0) { 584 vap->va_size = strlen(fullpath); 585 kfree(freepath, M_TEMP); 586 } else { 587 vap->va_size = sizeof("unknown") - 1; 588 error = 0; 589 } 590 vap->va_bytes = vap->va_size; 591 break; 592 } 593 594 case Pmem: 595 /* 596 * If we denied owner access earlier, then we have to 597 * change the owner to root - otherwise 'ps' and friends 598 * will break even though they are setgid kmem. *SIGH* 599 */ 600 if (procp->p_flag & P_SUGID) 601 vap->va_uid = 0; 602 else 603 vap->va_uid = procp->p_ucred->cr_uid; 604 break; 605 606 case Pregs: 607 vap->va_bytes = vap->va_size = sizeof(struct reg); 608 break; 609 610 case Pfpregs: 611 vap->va_bytes = vap->va_size = sizeof(struct fpreg); 612 break; 613 614 case Pdbregs: 615 vap->va_bytes = vap->va_size = sizeof(struct dbreg); 616 break; 617 618 case Ptype: 619 case Pmap: 620 case Pctl: 621 case Pstatus: 622 case Pnote: 623 case Pnotepg: 624 case Pcmdline: 625 case Prlimit: 626 break; 627 628 default: 629 panic("procfs_getattr"); 630 } 631 632 return (error); 633 } 634 635 /* 636 * procfs_setattr(struct vnode *a_vp, struct vattr *a_vap, 637 * struct ucred *a_cred) 638 */ 639 static int 640 procfs_setattr(struct vop_setattr_args *ap) 641 { 642 if (ap->a_vap->va_flags != VNOVAL) 643 return (EOPNOTSUPP); 644 645 /* 646 * just fake out attribute setting 647 * it's not good to generate an error 648 * return, otherwise things like creat() 649 * will fail when they try to set the 650 * file length to 0. worse, this means 651 * that echo $note > /proc/$pid/note will fail. 652 */ 653 654 return (0); 655 } 656 657 /* 658 * implement access checking. 659 * 660 * procfs_access(struct vnode *a_vp, int a_mode, struct ucred *a_cred) 661 */ 662 static int 663 procfs_access(struct vop_access_args *ap) 664 { 665 struct vattr vattr; 666 int error; 667 668 error = VOP_GETATTR(ap->a_vp, &vattr); 669 if (!error) 670 error = vop_helper_access(ap, vattr.va_uid, vattr.va_gid, 671 vattr.va_mode, 0); 672 return (error); 673 } 674 675 /* 676 * lookup. this is incredibly complicated in the general case, however 677 * for most pseudo-filesystems very little needs to be done. 678 * 679 * procfs_lookup(struct vnode *a_dvp, struct vnode **a_vpp, 680 * struct componentname *a_cnp) 681 */ 682 static int 683 procfs_lookup(struct vop_old_lookup_args *ap) 684 { 685 struct componentname *cnp = ap->a_cnp; 686 struct vnode **vpp = ap->a_vpp; 687 struct vnode *dvp = ap->a_dvp; 688 char *pname = cnp->cn_nameptr; 689 /* struct proc *curp = cnp->cn_proc; */ 690 struct proc_target *pt; 691 pid_t pid; 692 struct pfsnode *pfs; 693 struct proc *p; 694 struct lwp *lp; 695 int i; 696 int error; 697 698 *vpp = NULL; 699 700 if (cnp->cn_nameiop == NAMEI_DELETE || cnp->cn_nameiop == NAMEI_RENAME) 701 return (EROFS); 702 703 error = 0; 704 if (cnp->cn_namelen == 1 && *pname == '.') { 705 *vpp = dvp; 706 vref(*vpp); 707 goto out; 708 } 709 710 pfs = VTOPFS(dvp); 711 switch (pfs->pfs_type) { 712 case Proot: 713 if (cnp->cn_flags & CNP_ISDOTDOT) 714 return (EIO); 715 716 if (CNEQ(cnp, "curproc", 7)) { 717 error = procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc); 718 goto out; 719 } 720 721 pid = atopid(pname, cnp->cn_namelen); 722 if (pid == NO_PID) 723 break; 724 725 p = PFIND(pid); 726 if (p == NULL) 727 break; 728 729 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred)) 730 break; 731 732 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 && 733 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid) 734 break; 735 736 error = procfs_allocvp(dvp->v_mount, vpp, pid, Pproc); 737 goto out; 738 739 case Pproc: 740 if (cnp->cn_flags & CNP_ISDOTDOT) { 741 error = procfs_root(dvp->v_mount, vpp); 742 goto out; 743 } 744 745 p = PFIND(pfs->pfs_pid); 746 if (p == NULL) 747 break; 748 /* XXX lwp */ 749 lp = FIRST_LWP_IN_PROC(p); 750 751 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred)) 752 break; 753 754 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 && 755 ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid) 756 break; 757 758 for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) { 759 if (cnp->cn_namelen == pt->pt_namlen && 760 bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 && 761 (pt->pt_valid == NULL || (*pt->pt_valid)(lp))) 762 goto found; 763 } 764 break; 765 found: 766 error = procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid, 767 pt->pt_pfstype); 768 goto out; 769 770 default: 771 error = ENOTDIR; 772 goto out; 773 } 774 if (cnp->cn_nameiop == NAMEI_LOOKUP) 775 error = ENOENT; 776 else 777 error = EROFS; 778 /* 779 * If no error occured *vpp will hold a referenced locked vnode. 780 * dvp was passed to us locked and *vpp must be returned locked. 781 * If *vpp != dvp then we should unlock dvp if (1) this is not the 782 * last component or (2) CNP_LOCKPARENT is not set. 783 */ 784 out: 785 if (error == 0 && *vpp != dvp) { 786 if ((cnp->cn_flags & CNP_LOCKPARENT) == 0) { 787 cnp->cn_flags |= CNP_PDIRUNLOCK; 788 vn_unlock(dvp); 789 } 790 } 791 return (error); 792 } 793 794 /* 795 * Does this process have a text file? 796 */ 797 int 798 procfs_validfile(struct lwp *lp) 799 { 800 return (procfs_findtextvp(lp->lwp_proc) != NULLVP); 801 } 802 803 /* 804 * readdir() returns directory entries from pfsnode (vp). 805 * 806 * We generate just one directory entry at a time, as it would probably 807 * not pay off to buffer several entries locally to save uiomove calls. 808 * 809 * procfs_readdir(struct vnode *a_vp, struct uio *a_uio, struct ucred *a_cred, 810 * int *a_eofflag, int *a_ncookies, off_t **a_cookies) 811 */ 812 static int 813 procfs_readdir(struct vop_readdir_args *ap) 814 { 815 struct pfsnode *pfs; 816 int error; 817 818 if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX) 819 return (EINVAL); 820 if ((error = vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY)) != 0) 821 return (error); 822 pfs = VTOPFS(ap->a_vp); 823 824 switch (pfs->pfs_type) { 825 case Pproc: 826 /* 827 * this is for the process-specific sub-directories. 828 * all that is needed to is copy out all the entries 829 * from the procent[] table (top of this file). 830 */ 831 error = procfs_readdir_proc(ap); 832 break; 833 case Proot: 834 /* 835 * this is for the root of the procfs filesystem 836 * what is needed is a special entry for "curproc" 837 * followed by an entry for each process on allproc 838 */ 839 error = procfs_readdir_root(ap); 840 break; 841 default: 842 error = ENOTDIR; 843 break; 844 } 845 846 vn_unlock(ap->a_vp); 847 return (error); 848 } 849 850 static int 851 procfs_readdir_proc(struct vop_readdir_args *ap) 852 { 853 struct pfsnode *pfs; 854 int error, i, retval; 855 struct proc *p; 856 struct lwp *lp; 857 struct proc_target *pt; 858 struct uio *uio = ap->a_uio; 859 860 pfs = VTOPFS(ap->a_vp); 861 p = PFIND(pfs->pfs_pid); 862 if (p == NULL) 863 return(0); 864 if (!PRISON_CHECK(ap->a_cred, p->p_ucred)) 865 return(0); 866 /* XXX lwp */ 867 lp = FIRST_LWP_IN_PROC(p); 868 869 error = 0; 870 i = (int)uio->uio_offset; 871 if (i < 0) 872 return (EINVAL); 873 874 for (pt = &proc_targets[i]; 875 !error && uio->uio_resid > 0 && i < nproc_targets; pt++, i++) { 876 if (pt->pt_valid && (*pt->pt_valid)(lp) == 0) 877 continue; 878 879 retval = vop_write_dirent(&error, uio, 880 PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype), pt->pt_type, 881 pt->pt_namlen, pt->pt_name); 882 if (retval) 883 break; 884 } 885 886 uio->uio_offset = (off_t)i; 887 888 return(0); 889 } 890 891 struct procfs_readdir_root_info { 892 int error; 893 int i; 894 int pcnt; 895 struct uio *uio; 896 struct ucred *cred; 897 }; 898 899 static int procfs_readdir_root_callback(struct proc *p, void *data); 900 901 static int 902 procfs_readdir_root(struct vop_readdir_args *ap) 903 { 904 struct procfs_readdir_root_info info; 905 struct uio *uio = ap->a_uio; 906 int res; 907 908 info.error = 0; 909 info.i = (int)uio->uio_offset; 910 911 if (info.i < 0) 912 return (EINVAL); 913 914 info.pcnt = 0; 915 info.uio = uio; 916 info.cred = ap->a_cred; 917 while (info.pcnt < 3) { 918 res = procfs_readdir_root_callback(NULL, &info); 919 if (res < 0) 920 break; 921 } 922 if (res >= 0) 923 allproc_scan(procfs_readdir_root_callback, &info); 924 uio->uio_offset = (off_t)info.i; 925 926 return (info.error); 927 } 928 929 static int 930 procfs_readdir_root_callback(struct proc *p, void *data) 931 { 932 struct procfs_readdir_root_info *info = data; 933 struct uio *uio; 934 int retval; 935 ino_t d_ino; 936 const char *d_name; 937 char d_name_pid[20]; 938 size_t d_namlen; 939 uint8_t d_type; 940 941 uio = info->uio; 942 943 if (uio->uio_resid <= 0 || info->error) 944 return(-1); 945 946 switch (info->pcnt) { 947 case 0: /* `.' */ 948 d_ino = PROCFS_FILENO(0, Proot); 949 d_name = "."; 950 d_namlen = 1; 951 d_type = DT_DIR; 952 break; 953 case 1: /* `..' */ 954 d_ino = PROCFS_FILENO(0, Proot); 955 d_name = ".."; 956 d_namlen = 2; 957 d_type = DT_DIR; 958 break; 959 960 case 2: 961 d_ino = PROCFS_FILENO(0, Pcurproc); 962 d_namlen = 7; 963 d_name = "curproc"; 964 d_type = DT_LNK; 965 break; 966 967 968 default: 969 if (!PRISON_CHECK(info->cred, p->p_ucred)) 970 return(0); 971 if (ps_showallprocs == 0 && 972 info->cred->cr_uid != 0 && 973 info->cred->cr_uid != p->p_ucred->cr_uid) { 974 return(0); 975 } 976 977 /* 978 * Skip entries we have already returned (optimization) 979 */ 980 if (info->pcnt < info->i) { 981 ++info->pcnt; 982 return(0); 983 } 984 985 d_ino = PROCFS_FILENO(p->p_pid, Pproc); 986 d_namlen = ksnprintf(d_name_pid, sizeof(d_name_pid), 987 "%ld", (long)p->p_pid); 988 d_name = d_name_pid; 989 d_type = DT_DIR; 990 break; 991 } 992 993 /* 994 * Skip entries we have already returned (optimization) 995 */ 996 if (info->pcnt < info->i) { 997 ++info->pcnt; 998 return(0); 999 } 1000 1001 retval = vop_write_dirent(&info->error, uio, 1002 d_ino, d_type, d_namlen, d_name); 1003 if (retval) 1004 return(-1); 1005 ++info->pcnt; 1006 ++info->i; 1007 return(0); 1008 } 1009 1010 /* 1011 * readlink reads the link of `curproc' or `file' 1012 */ 1013 static int 1014 procfs_readlink(struct vop_readlink_args *ap) 1015 { 1016 char buf[16]; /* should be enough */ 1017 struct proc *procp; 1018 struct vnode *vp = ap->a_vp; 1019 struct pfsnode *pfs = VTOPFS(vp); 1020 char *fullpath, *freepath; 1021 int error, len; 1022 1023 switch (pfs->pfs_type) { 1024 case Pcurproc: 1025 if (pfs->pfs_fileno != PROCFS_FILENO(0, Pcurproc)) 1026 return (EINVAL); 1027 1028 len = ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid); 1029 1030 return (uiomove(buf, len, ap->a_uio)); 1031 /* 1032 * There _should_ be no way for an entire process to disappear 1033 * from under us... 1034 */ 1035 case Pfile: 1036 procp = PFIND(pfs->pfs_pid); 1037 if (procp == NULL || procp->p_ucred == NULL) { 1038 kprintf("procfs_readlink: pid %d disappeared\n", 1039 pfs->pfs_pid); 1040 return (uiomove("unknown", sizeof("unknown") - 1, 1041 ap->a_uio)); 1042 } 1043 error = cache_fullpath(procp, &procp->p_textnch, &fullpath, &freepath, 0); 1044 if (error != 0) 1045 return (uiomove("unknown", sizeof("unknown") - 1, 1046 ap->a_uio)); 1047 error = uiomove(fullpath, strlen(fullpath), ap->a_uio); 1048 kfree(freepath, M_TEMP); 1049 return (error); 1050 default: 1051 return (EINVAL); 1052 } 1053 } 1054 1055 /* 1056 * convert decimal ascii to pid_t 1057 */ 1058 static pid_t 1059 atopid(const char *b, u_int len) 1060 { 1061 pid_t p = 0; 1062 1063 while (len--) { 1064 char c = *b++; 1065 if (c < '0' || c > '9') 1066 return (NO_PID); 1067 p = 10 * p + (c - '0'); 1068 if (p > PID_MAX) 1069 return (NO_PID); 1070 } 1071 1072 return (p); 1073 } 1074 1075