sys/vfs/procfs/procfs_vnops.c

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