sys/emulation/linux/i386/linprocfs/linprocfs_vnops.c

   1 /*
   2  * Copyright (c) 2000 Dag-Erling Coïdan Smørgrav
   3  * Copyright (c) 1999 Pierre Beyssac
   4  * Copyright (c) 1993, 1995 Jan-Simon Pendry
   5  * Copyright (c) 1993, 1995
   6  *      The Regents of the University of California.  All rights reserved.
   7  *
   8  * This code is derived from software contributed to Berkeley by
   9  * Jan-Simon Pendry.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  * 1. Redistributions of source code must retain the above copyright
  15  *    notice, this list of conditions and the following disclaimer.
  16  * 2. Redistributions in binary form must reproduce the above copyright
  17  *    notice, this list of conditions and the following disclaimer in the
  18  *    documentation and/or other materials provided with the distribution.
  19  * 3. All advertising materials mentioning features or use of this software
  20  *    must display the following acknowledgement:
  21  *      This product includes software developed by the University of
  22  *      California, Berkeley and its contributors.
  23  * 4. Neither the name of the University nor the names of its contributors
  24  *    may be used to endorse or promote products derived from this software
  25  *    without specific prior written permission.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  37  * SUCH DAMAGE.
  38  *
  39  *      @(#)procfs_vnops.c      8.18 (Berkeley) 5/21/95
  40  *
  41  * $FreeBSD: src/sys/i386/linux/linprocfs/linprocfs_vnops.c,v 1.3.2.5 2001/08/12 14:29:19 rwatson Exp $
  42  * $DragonFly: src/sys/emulation/linux/i386/linprocfs/linprocfs_vnops.c,v 1.26 2006/01/31 02:15:10 corecode Exp $
  43  */
  44
  45 /*
  46  * procfs vnode interface
  47  */
  48
  49 #include <sys/param.h>
  50 #include <sys/systm.h>
  51 #include <sys/time.h>
  52 #include <sys/kernel.h>
  53 #include <sys/lock.h>
  54 #include <sys/fcntl.h>
  55 #include <sys/proc.h>
  56 #include <sys/signalvar.h>
  57 #include <sys/vnode.h>
  58 #include <sys/mount.h>
  59 #include <sys/namei.h>
  60 #include <sys/dirent.h>
  61 #include <sys/malloc.h>
  62 #include <machine/reg.h>
  63 #include <vm/vm_zone.h>
  64 #include "linprocfs.h"
  65 #include <sys/pioctl.h>
  66
  67 #include <machine/limits.h>
  68
  69 extern struct vnode *procfs_findtextvp (struct proc *);
  70
  71 static int      linprocfs_access (struct vop_access_args *);
  72 static int      linprocfs_badop (void);
  73 static int      linprocfs_bmap (struct vop_bmap_args *);
  74 static int      linprocfs_close (struct vop_close_args *);
  75 static int      linprocfs_getattr (struct vop_getattr_args *);
  76 static int      linprocfs_inactive (struct vop_inactive_args *);
  77 static int      linprocfs_ioctl (struct vop_ioctl_args *);
  78 static int      linprocfs_lookup (struct vop_old_lookup_args *);
  79 static int      linprocfs_open (struct vop_open_args *);
  80 static int      linprocfs_print (struct vop_print_args *);
  81 static int      linprocfs_readdir (struct vop_readdir_args *);
  82 static int      linprocfs_readlink (struct vop_readlink_args *);
  83 static int      linprocfs_reclaim (struct vop_reclaim_args *);
  84 static int      linprocfs_setattr (struct vop_setattr_args *);
  85
  86 static int      linprocfs_readdir_proc(struct vop_readdir_args *);
  87 static int      linprocfs_readdir_root(struct vop_readdir_args *);
  88
  89 /*
  90  * This is a list of the valid names in the
  91  * process-specific sub-directories.  It is
  92  * used in linprocfs_lookup and linprocfs_readdir
  93  */
  94 static struct proc_target {
  95         u_char  pt_type;
  96         u_char  pt_namlen;
  97         char    *pt_name;
  98         pfstype pt_pfstype;
  99         int     (*pt_valid) (struct proc *p);
 100 } proc_targets[] = {
 101 #define N(s) sizeof(s)-1, s
 102         /*        name          type            validp */
 103         { DT_DIR, N("."),       Pproc,          NULL },
 104         { DT_DIR, N(".."),      Proot,          NULL },
 105         { DT_REG, N("mem"),     Pmem,           NULL },
 106         { DT_LNK, N("exe"),     Pexe,           NULL },
 107         { DT_REG, N("stat"),    Pprocstat,      NULL },
 108         { DT_REG, N("status"),  Pprocstatus,    NULL },
 109 #undef N
 110 };
 111 static const int nproc_targets = sizeof(proc_targets) / sizeof(proc_targets[0]);
 112
 113 static pid_t atopid (const char *, u_int);
 114
 115 /*
 116  * set things up for doing i/o on
 117  * the pfsnode (vp).  (vp) is locked
 118  * on entry, and should be left locked
 119  * on exit.
 120  *
 121  * for procfs we don't need to do anything
 122  * in particular for i/o.  all that is done
 123  * is to support exclusive open on process
 124  * memory images.
 125  */
 126 static int
 127 linprocfs_open(struct vop_open_args *ap)
 128 {
 129         struct pfsnode *pfs = VTOPFS(ap->a_vp);
 130         struct proc *p2;
 131
 132         p2 = PFIND(pfs->pfs_pid);
 133         if (p2 == NULL)
 134                 return (ENOENT);
 135         if (pfs->pfs_pid && !PRISON_CHECK(ap->a_cred, p2->p_ucred))
 136                 return (ENOENT);
 137
 138         switch (pfs->pfs_type) {
 139         case Pmem:
 140                 if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) ||
 141                     ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE)))
 142                         return (EBUSY);
 143
 144                 if (p_trespass(ap->a_cred, p2->p_ucred))
 145                         return (EPERM);
 146
 147                 if (ap->a_mode & FWRITE)
 148                         pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL);
 149
 150                 return (0);
 151
 152         default:
 153                 break;
 154         }
 155
 156         return (0);
 157 }
 158
 159 /*
 160  * close the pfsnode (vp) after doing i/o.
 161  * (vp) is not locked on entry or exit.
 162  *
 163  * nothing to do for procfs other than undo
 164  * any exclusive open flag (see _open above).
 165  */
 166 static int
 167 linprocfs_close(struct vop_close_args *ap)
 168 {
 169         struct pfsnode *pfs = VTOPFS(ap->a_vp);
 170         struct proc *p;
 171
 172         switch (pfs->pfs_type) {
 173         case Pmem:
 174                 if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
 175                         pfs->pfs_flags &= ~(FWRITE|O_EXCL);
 176                 /*
 177                  * If this is the last close, then it checks to see if
 178                  * the target process has PF_LINGER set in p_pfsflags,
 179                  * if this is *not* the case, then the process' stop flags
 180                  * are cleared, and the process is woken up.  This is
 181                  * to help prevent the case where a process has been
 182                  * told to stop on an event, but then the requesting process
 183                  * has gone away or forgotten about it.
 184                  */
 185                 if ((ap->a_vp->v_usecount < 2)
 186                     && (p = pfind(pfs->pfs_pid))
 187                     && !(p->p_pfsflags & PF_LINGER)) {
 188                         p->p_stops = 0;
 189                         p->p_step = 0;
 190                         wakeup(&p->p_step);
 191                 }
 192                 break;
 193         default:
 194                 break;
 195         }
 196
 197         return (0);
 198 }
 199
 200 /*
 201  * do an ioctl operation on a pfsnode (vp).
 202  * (vp) is not locked on entry or exit.
 203  */
 204 static int
 205 linprocfs_ioctl(struct vop_ioctl_args *ap)
 206 {
 207         struct pfsnode *pfs = VTOPFS(ap->a_vp);
 208         struct proc *procp;
 209         struct thread *td;
 210         int error;
 211         int signo;
 212         struct procfs_status *psp;
 213         unsigned char flags;
 214
 215         td = ap->a_td;
 216         procp = pfind(pfs->pfs_pid);
 217         if (procp == NULL) {
 218                 return ENOTTY;
 219         }
 220
 221         if (p_trespass(ap->a_cred, procp->p_ucred))
 222                 return EPERM;
 223
 224         switch (ap->a_command) {
 225         case PIOCBIS:
 226           procp->p_stops |= *(unsigned int*)ap->a_data;
 227           break;
 228         case PIOCBIC:
 229           procp->p_stops &= ~*(unsigned int*)ap->a_data;
 230           break;
 231         case PIOCSFL:
 232           /*
 233            * NFLAGS is "non-suser_xxx flags" -- currently, only
 234            * PFS_ISUGID ("ignore set u/g id");
 235            */
 236 #define NFLAGS  (PF_ISUGID)
 237           flags = (unsigned char)*(unsigned int*)ap->a_data;
 238           if (flags & NFLAGS && (error = suser_cred(ap->a_cred, 0)))
 239             return error;
 240           procp->p_pfsflags = flags;
 241           break;
 242         case PIOCGFL:
 243           *(unsigned int*)ap->a_data = (unsigned int)procp->p_pfsflags;
 244         case PIOCSTATUS:
 245           psp = (struct procfs_status *)ap->a_data;
 246           psp->state = (procp->p_step == 0);
 247           psp->flags = procp->p_pfsflags;
 248           psp->events = procp->p_stops;
 249           if (procp->p_step) {
 250             psp->why = procp->p_stype;
 251             psp->val = procp->p_xstat;
 252           } else {
 253             psp->why = psp->val = 0;    /* Not defined values */
 254           }
 255           break;
 256         case PIOCWAIT:
 257           psp = (struct procfs_status *)ap->a_data;
 258           if (procp->p_step == 0) {
 259             error = tsleep(&procp->p_stype, PCATCH, "piocwait", 0);
 260             if (error)
 261               return error;
 262           }
 263           psp->state = 1;       /* It stopped */
 264           psp->flags = procp->p_pfsflags;
 265           psp->events = procp->p_stops;
 266           psp->why = procp->p_stype;    /* why it stopped */
 267           psp->val = procp->p_xstat;    /* any extra info */
 268           break;
 269         case PIOCCONT:  /* Restart a proc */
 270           if (procp->p_step == 0)
 271             return EINVAL;      /* Can only start a stopped process */
 272           if ((signo = *(int*)ap->a_data) != 0) {
 273             if (signo >= NSIG || signo <= 0)
 274               return EINVAL;
 275             psignal(procp, signo);
 276           }
 277           procp->p_step = 0;
 278           wakeup(&procp->p_step);
 279           break;
 280         default:
 281           return (ENOTTY);
 282         }
 283         return 0;
 284 }
 285
 286 /*
 287  * do block mapping for pfsnode (vp).
 288  * since we don't use the buffer cache
 289  * for procfs this function should never
 290  * be called.  in any case, it's not clear
 291  * what part of the kernel ever makes use
 292  * of this function.  for sanity, this is the
 293  * usual no-op bmap, although returning
 294  * (EIO) would be a reasonable alternative.
 295  */
 296 static int
 297 linprocfs_bmap(struct vop_bmap_args *ap)
 298 {
 299
 300         if (ap->a_vpp != NULL)
 301                 *ap->a_vpp = ap->a_vp;
 302         if (ap->a_bnp != NULL)
 303                 *ap->a_bnp = ap->a_bn;
 304         if (ap->a_runp != NULL)
 305                 *ap->a_runp = 0;
 306         return (0);
 307 }
 308
 309 /*
 310  * linprocfs_inactive is called when the pfsnode
 311  * is vrele'd and the reference count is about
 312  * to go to zero.  (vp) will be on the vnode free
 313  * list, so to get it back vget() must be
 314  * used.
 315  *
 316  * (vp) is locked on entry and must remain locked
 317  *      on exit.
 318  */
 319 static int
 320 linprocfs_inactive(struct vop_inactive_args *ap)
 321 {
 322         /*struct vnode *vp = ap->a_vp;*/
 323
 324         return (0);
 325 }
 326
 327 /*
 328  * _reclaim is called when getnewvnode()
 329  * wants to make use of an entry on the vnode
 330  * free list.  at this time the filesystem needs
 331  * to free any private data and remove the node
 332  * from any private lists.
 333  */
 334 static int
 335 linprocfs_reclaim(struct vop_reclaim_args *ap)
 336 {
 337         return (linprocfs_freevp(ap->a_vp));
 338 }
 339
 340 /*
 341  * _print is used for debugging.
 342  * just print a readable description
 343  * of (vp).
 344  */
 345 static int
 346 linprocfs_print(struct vop_print_args *ap)
 347 {
 348         struct pfsnode *pfs = VTOPFS(ap->a_vp);
 349
 350         printf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n",
 351             pfs->pfs_type, (long)pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
 352         return (0);
 353 }
 354
 355 /*
 356  * generic entry point for unsupported operations
 357  */
 358 static int
 359 linprocfs_badop(void)
 360 {
 361
 362         return (EIO);
 363 }
 364
 365 /*
 366  * Invent attributes for pfsnode (vp) and store
 367  * them in (vap).
 368  * Directories lengths are returned as zero since
 369  * any real length would require the genuine size
 370  * to be computed, and nothing cares anyway.
 371  *
 372  * this is relatively minimal for procfs.
 373  */
 374 static int
 375 linprocfs_getattr(struct vop_getattr_args *ap)
 376 {
 377         struct pfsnode *pfs = VTOPFS(ap->a_vp);
 378         struct vattr *vap = ap->a_vap;
 379         struct proc *procp;
 380         int error;
 381
 382         /*
 383          * First make sure that the process and its credentials
 384          * still exist.
 385          */
 386         switch (pfs->pfs_type) {
 387         case Proot:
 388         case Pself:
 389                 procp = 0;
 390                 break;
 391
 392         default:
 393                 procp = PFIND(pfs->pfs_pid);
 394                 if (procp == 0 || procp->p_ucred == NULL)
 395                         return (ENOENT);
 396         }
 397
 398         error = 0;
 399
 400         /* start by zeroing out the attributes */
 401         VATTR_NULL(vap);
 402
 403         /* next do all the common fields */
 404         vap->va_type = ap->a_vp->v_type;
 405         vap->va_mode = pfs->pfs_mode;
 406         vap->va_fileid = pfs->pfs_fileno;
 407         vap->va_flags = 0;
 408         vap->va_blocksize = PAGE_SIZE;
 409         vap->va_bytes = vap->va_size = 0;
 410         vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0];
 411
 412         /*
 413          * Make all times be current TOD.
 414          * It would be possible to get the process start
 415          * time from the p_stat structure, but there's
 416          * no "file creation" time stamp anyway, and the
 417          * p_stat structure is not addressible if u. gets
 418          * swapped out for that process.
 419          */
 420         nanotime(&vap->va_ctime);
 421         vap->va_atime = vap->va_mtime = vap->va_ctime;
 422
 423         /*
 424          * now do the object specific fields
 425          *
 426          * The size could be set from struct reg, but it's hardly
 427          * worth the trouble, and it puts some (potentially) machine
 428          * dependent data into this machine-independent code.  If it
 429          * becomes important then this function should break out into
 430          * a per-file stat function in the corresponding .c file.
 431          */
 432
 433         vap->va_nlink = 1;
 434         if (procp) {
 435                 vap->va_uid = procp->p_ucred->cr_uid;
 436                 vap->va_gid = procp->p_ucred->cr_gid;
 437         }
 438
 439         switch (pfs->pfs_type) {
 440         case Proot:
 441                 /*
 442                  * Set nlink to 1 to tell fts(3) we don't actually know.
 443                  */
 444                 vap->va_nlink = 1;
 445                 vap->va_uid = 0;
 446                 vap->va_gid = 0;
 447                 vap->va_size = vap->va_bytes = DEV_BSIZE;
 448                 break;
 449
 450         case Pself: {
 451                 char buf[16];           /* should be enough */
 452                 vap->va_uid = 0;
 453                 vap->va_gid = 0;
 454                 vap->va_size = vap->va_bytes =
 455                     snprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
 456                 break;
 457         }
 458
 459         case Pproc:
 460                 vap->va_nlink = nproc_targets;
 461                 vap->va_size = vap->va_bytes = DEV_BSIZE;
 462                 break;
 463
 464         case Pexe: {
 465                 char *fullpath, *freepath;
 466                 error = vn_fullpath(procp, NULL, &fullpath, &freepath);
 467                 if (error == 0) {
 468                         vap->va_size = strlen(fullpath);
 469                         free(freepath, M_TEMP);
 470                 } else {
 471                         vap->va_size = sizeof("unknown") - 1;
 472                         error = 0;
 473                 }
 474                 vap->va_bytes = vap->va_size;
 475                 break;
 476         }
 477
 478         case Pmeminfo:
 479         case Pcpuinfo:
 480         case Pstat:
 481         case Puptime:
 482         case Pversion:
 483         case Ploadavg:
 484                 vap->va_bytes = vap->va_size = 0;
 485                 vap->va_uid = 0;
 486                 vap->va_gid = 0;
 487                 break;
 488
 489         case Pmem:
 490                 /*
 491                  * If we denied owner access earlier, then we have to
 492                  * change the owner to root - otherwise 'ps' and friends
 493                  * will break even though they are setgid kmem. *SIGH*
 494                  */
 495                 if (procp->p_flag & P_SUGID)
 496                         vap->va_uid = 0;
 497                 else
 498                         vap->va_uid = procp->p_ucred->cr_uid;
 499                 break;
 500
 501         case Pprocstat:
 502         case Pprocstatus:
 503                 vap->va_bytes = vap->va_size = 0;
 504                 /* uid, gid are already set */
 505                 break;
 506
 507         default:
 508                 panic("linprocfs_getattr");
 509         }
 510
 511         return (error);
 512 }
 513
 514 static int
 515 linprocfs_setattr(struct vop_setattr_args *ap)
 516 {
 517
 518         if (ap->a_vap->va_flags != VNOVAL)
 519                 return (EOPNOTSUPP);
 520
 521         /*
 522          * just fake out attribute setting
 523          * it's not good to generate an error
 524          * return, otherwise things like creat()
 525          * will fail when they try to set the
 526          * file length to 0.  worse, this means
 527          * that echo $note > /proc/$pid/note will fail.
 528          */
 529
 530         return (0);
 531 }
 532
 533 /*
 534  * implement access checking.
 535  *
 536  * something very similar to this code is duplicated
 537  * throughout the 4bsd kernel and should be moved
 538  * into kern/vfs_subr.c sometime.
 539  *
 540  * actually, the check for super-user is slightly
 541  * broken since it will allow read access to write-only
 542  * objects.  this doesn't cause any particular trouble
 543  * but does mean that the i/o entry points need to check
 544  * that the operation really does make sense.
 545  */
 546 static int
 547 linprocfs_access(struct vop_access_args *ap)
 548 {
 549         struct vattr *vap;
 550         struct vattr vattr;
 551         int error;
 552
 553         /*
 554          * If you're the super-user,
 555          * you always get access.
 556          */
 557         if (ap->a_cred->cr_uid == 0)
 558                 return (0);
 559
 560         vap = &vattr;
 561         error = VOP_GETATTR(ap->a_vp, vap, ap->a_td);
 562         if (error)
 563                 return (error);
 564
 565         /*
 566          * Access check is based on only one of owner, group, public.
 567          * If not owner, then check group. If not a member of the
 568          * group, then check public access.
 569          */
 570         if (ap->a_cred->cr_uid != vap->va_uid) {
 571                 gid_t *gp;
 572                 int i;
 573
 574                 ap->a_mode >>= 3;
 575                 gp = ap->a_cred->cr_groups;
 576                 for (i = 0; i < ap->a_cred->cr_ngroups; i++, gp++)
 577                         if (vap->va_gid == *gp)
 578                                 goto found;
 579                 ap->a_mode >>= 3;
 580 found:
 581                 ;
 582         }
 583
 584         if ((vap->va_mode & ap->a_mode) == ap->a_mode)
 585                 return (0);
 586
 587         return (EACCES);
 588 }
 589
 590 /*
 591  * lookup.  this is incredibly complicated in the general case, however
 592  * for most pseudo-filesystems very little needs to be done.
 593  */
 594 static int
 595 linprocfs_lookup(struct vop_old_lookup_args *ap)
 596 {
 597         struct componentname *cnp = ap->a_cnp;
 598         struct vnode **vpp = ap->a_vpp;
 599         struct vnode *dvp = ap->a_dvp;
 600         char *pname = cnp->cn_nameptr;
 601         struct proc_target *pt;
 602         pid_t pid;
 603         struct pfsnode *pfs;
 604         struct proc *p;
 605         int i;
 606         int error;
 607
 608         *vpp = NULL;
 609
 610         if (cnp->cn_nameiop == NAMEI_DELETE ||
 611             cnp->cn_nameiop == NAMEI_RENAME ||
 612             cnp->cn_nameiop == NAMEI_CREATE) {
 613                 return (EROFS);
 614         }
 615
 616         error = 0;
 617
 618         if (cnp->cn_namelen == 1 && *pname == '.') {
 619                 *vpp = dvp;
 620                 vref(*vpp);
 621                 goto out;
 622         }
 623
 624         pfs = VTOPFS(dvp);
 625         switch (pfs->pfs_type) {
 626         case Proot:
 627                 if (cnp->cn_flags & CNP_ISDOTDOT)
 628                         return (EIO);
 629
 630                 if (CNEQ(cnp, "self", 4)) {
 631                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pself);
 632                         goto out;
 633                 }
 634                 if (CNEQ(cnp, "meminfo", 7)) {
 635                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pmeminfo);
 636                         goto out;
 637                 }
 638                 if (CNEQ(cnp, "cpuinfo", 7)) {
 639                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pcpuinfo);
 640                         goto out;
 641                 }
 642                 if (CNEQ(cnp, "stat", 4)) {
 643                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pstat);
 644                         goto out;
 645                 }
 646                 if (CNEQ(cnp, "uptime", 6)) {
 647                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Puptime);
 648                         goto out;
 649                 }
 650                 if (CNEQ(cnp, "version", 7)) {
 651                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pversion);
 652                         goto out;
 653                 }
 654                 if (CNEQ(cnp, "loadavg", 7)) {
 655                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Ploadavg);
 656                         goto out;
 657                 }
 658
 659                 pid = atopid(pname, cnp->cn_namelen);
 660                 if (pid == NO_PID)
 661                         break;
 662
 663                 p = PFIND(pid);
 664                 if (p == 0)
 665                         break;
 666
 667                 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
 668                         break;
 669
 670                 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
 671                     ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
 672                         break;
 673
 674                 error = linprocfs_allocvp(dvp->v_mount, vpp, pid, Pproc);
 675                 goto out;
 676
 677         case Pproc:
 678                 if (cnp->cn_flags & CNP_ISDOTDOT) {
 679                         error = linprocfs_root(dvp->v_mount, vpp);
 680                         goto out;
 681                 }
 682
 683                 p = PFIND(pfs->pfs_pid);
 684                 if (p == 0)
 685                         break;
 686
 687                 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
 688                         break;
 689
 690                 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
 691                     ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
 692                         break;
 693
 694                 for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) {
 695                         if (cnp->cn_namelen == pt->pt_namlen &&
 696                             bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
 697                             (pt->pt_valid == NULL || (*pt->pt_valid)(p)))
 698                                 goto found;
 699                 }
 700                 break;
 701
 702         found:
 703                 error = linprocfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
 704                                         pt->pt_pfstype);
 705                 goto out;
 706
 707         default:
 708                 error = ENOTDIR;
 709                 goto out;
 710         }
 711
 712         if (cnp->cn_nameiop == NAMEI_LOOKUP)
 713                 error = ENOENT;
 714         else
 715                 error = EROFS;
 716
 717         /*
 718          * If no error occured *vpp will hold a referenced locked vnode.
 719          * dvp was passed to us locked and *vpp must be returned locked
 720          * so if dvp != *vpp and CNP_LOCKPARENT is not set, unlock dvp.
 721          */
 722 out:
 723         if (error == 0) {
 724                 if (*vpp != dvp && (cnp->cn_flags & CNP_LOCKPARENT) == 0) {
 725                         cnp->cn_flags |= CNP_PDIRUNLOCK;
 726                         VOP_UNLOCK(dvp, 0, cnp->cn_td);
 727                 }
 728         }
 729         return (error);
 730 }
 731
 732 /*
 733  * Does this process have a text file?
 734  */
 735 int
 736 linprocfs_validfile(struct proc *p)
 737 {
 738
 739         return (procfs_findtextvp(p) != NULLVP);
 740 }
 741
 742 /*
 743  * readdir() returns directory entries from pfsnode (vp).
 744  *
 745  * We generate just one directory entry at a time, as it would probably
 746  * not pay off to buffer several entries locally to save uiomove calls.
 747  *
 748  * linprocfs_readdir(struct vnode *a_vp, struct uio *a_uio,
 749  *                   struct ucred *a_cred, int *a_eofflag,
 750  *                   int *a_ncookies, u_long **a_cookies)
 751  */
 752 static int
 753 linprocfs_readdir(struct vop_readdir_args *ap)
 754 {
 755         struct pfsnode *pfs;
 756         int error;
 757
 758         if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX)
 759                 return (EINVAL);
 760
 761         pfs = VTOPFS(ap->a_vp);
 762
 763         switch (pfs->pfs_type) {
 764         /*
 765          * this is for the process-specific sub-directories.
 766          * all that is needed to is copy out all the entries
 767          * from the procent[] table (top of this file).
 768          */
 769         case Pproc:
 770                 error = linprocfs_readdir_proc(ap);
 771                 break;
 772
 773         /*
 774          * this is for the root of the procfs filesystem
 775          * what is needed is a special entry for "self"
 776          * followed by an entry for each process on allproc
 777          */
 778
 779         case Proot:
 780                 error = linprocfs_readdir_root(ap);
 781                 break;
 782
 783         default:
 784                 error = ENOTDIR;
 785                 break;
 786         }
 787
 788         return (error);
 789 }
 790
 791 static int
 792 linprocfs_readdir_proc(struct vop_readdir_args *ap)
 793 {
 794         struct pfsnode *pfs;
 795         int error, i, retval;
 796         struct proc *p;
 797         struct proc_target *pt;
 798         struct uio *uio = ap->a_uio;
 799
 800         pfs = VTOPFS(ap->a_vp);
 801         p = PFIND(pfs->pfs_pid);
 802         if (p == NULL)
 803                 return(0);
 804         if (!PRISON_CHECK(ap->a_cred, p->p_ucred))
 805                 return(0);
 806
 807         error = 0;
 808         i = uio->uio_offset;
 809
 810         for (pt = &proc_targets[i];
 811              !error && uio->uio_resid > 0 && i < nproc_targets; pt++, i++) {
 812                 if (pt->pt_valid && (*pt->pt_valid)(p) == 0)
 813                         continue;
 814
 815                 retval = vop_write_dirent(&error, uio,
 816                     PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype), pt->pt_type,
 817                     pt->pt_namlen, pt->pt_name);
 818                 if (retval)
 819                         break;
 820         }
 821
 822         uio->uio_offset = i;
 823
 824         return(error);
 825 }
 826
 827 static int
 828 linprocfs_readdir_root(struct vop_readdir_args *ap)
 829 {
 830         int error, i, pcnt, retval;
 831         struct uio *uio = ap->a_uio;
 832         struct proc *p = LIST_FIRST(&allproc);
 833         ino_t d_ino;
 834         const char *d_name;
 835         char d_name_pid[20];
 836         size_t d_namlen;
 837         uint8_t d_type;
 838
 839         error = 0;
 840         i = uio->uio_offset;
 841         pcnt = 0;
 842
 843         for (; p && uio->uio_resid > 0 && !error; i++, pcnt++) {
 844                 switch (i) {
 845                 case 0:         /* `.' */
 846                         d_ino = PROCFS_FILENO(0, Proot);
 847                         d_name = ".";
 848                         d_namlen = 1;
 849                         d_type = DT_DIR;
 850                         break;
 851                 case 1:         /* `..' */
 852                         d_ino = PROCFS_FILENO(0, Proot);
 853                         d_name = "..";
 854                         d_namlen = 2;
 855                         d_type = DT_DIR;
 856                         break;
 857
 858                 case 2:
 859                         d_ino = PROCFS_FILENO(0, Proot);
 860                         d_namlen = 4;
 861                         d_name = "self";
 862                         d_type = DT_LNK;
 863                         break;
 864
 865                 case 3:
 866                         d_ino = PROCFS_FILENO(0, Pmeminfo);
 867                         d_namlen = 7;
 868                         d_name = "meminfo";
 869                         d_type = DT_REG;
 870                         break;
 871
 872                 case 4:
 873                         d_ino = PROCFS_FILENO(0, Pcpuinfo);
 874                         d_namlen = 7;
 875                         d_name = "cpuinfo";
 876                         d_type = DT_REG;
 877                         break;
 878
 879                 case 5:
 880                         d_ino = PROCFS_FILENO(0, Pstat);
 881                         d_namlen = 4;
 882                         d_name = "stat";
 883                         d_type = DT_REG;
 884                         break;
 885
 886                 case 6:
 887                         d_ino = PROCFS_FILENO(0, Puptime);
 888                         d_namlen = 6;
 889                         d_name = "uptime";
 890                         d_type = DT_REG;
 891                         break;
 892
 893                 case 7:
 894                         d_ino = PROCFS_FILENO(0, Pversion);
 895                         d_namlen = 7;
 896                         d_name = "version";
 897                         d_type = DT_REG;
 898                         break;
 899
 900                 case 8:
 901                         d_ino = PROCFS_FILENO(0, Ploadavg);
 902                         d_namlen = 7;
 903                         d_name = "loadavg";
 904                         d_type = DT_REG;
 905                         break;
 906
 907                 default:
 908                         while (pcnt < i) {
 909                                 p = LIST_NEXT(p, p_list);
 910                                 if (!p)
 911                                         goto done;
 912                                 if (!PRISON_CHECK(ap->a_cred, p->p_ucred))
 913                                         continue;
 914                                 pcnt++;
 915                         }
 916                         while (!PRISON_CHECK(ap->a_cred, p->p_ucred)) {
 917                                 p = LIST_NEXT(p, p_list);
 918                                 if (!p)
 919                                         goto done;
 920                         }
 921                         if (ps_showallprocs == 0 &&
 922                             ap->a_cred->cr_uid != 0 &&
 923                             ap->a_cred->cr_uid != p->p_ucred->cr_uid) {
 924                                 p = LIST_NEXT(p, p_list);
 925                                 if (!p)
 926                                         goto done;
 927                                 continue;
 928                         }
 929                         d_ino = PROCFS_FILENO(p->p_pid, Pproc);
 930                         d_namlen = snprintf(d_name_pid, sizeof(d_name_pid),
 931                             "%ld", (long)p->p_pid);
 932                         d_name = d_name_pid;
 933                         d_type = DT_DIR;
 934                         p = LIST_NEXT(p, p_list);
 935                         break;
 936                 }
 937
 938                 if (p != NULL)
 939                         PHOLD(p);
 940                 retval = vop_write_dirent(&error, uio,
 941                     d_ino, d_type, d_namlen, d_name);
 942                 if (p != NULL)
 943                         PRELE(p);
 944                 if (retval)
 945                         break;
 946         }
 947
 948 done:
 949         uio->uio_offset = i;
 950         return(error);
 951 }
 952
 953 /*
 954  * readlink reads the link of `self' or `exe'
 955  */
 956 static int
 957 linprocfs_readlink(struct vop_readlink_args *ap)
 958 {
 959         char buf[16];           /* should be enough */
 960         struct proc *procp;
 961         struct vnode *vp = ap->a_vp;
 962         struct pfsnode *pfs = VTOPFS(vp);
 963         char *fullpath, *freepath;
 964         int error, len;
 965
 966         switch (pfs->pfs_type) {
 967         case Pself:
 968                 if (pfs->pfs_fileno != PROCFS_FILENO(0, Pself))
 969                         return (EINVAL);
 970
 971                 len = snprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
 972
 973                 return (uiomove(buf, len, ap->a_uio));
 974         /*
 975          * There _should_ be no way for an entire process to disappear
 976          * from under us...
 977          */
 978         case Pexe:
 979                 procp = PFIND(pfs->pfs_pid);
 980                 if (procp == NULL || procp->p_ucred == NULL) {
 981                         printf("linprocfs_readlink: pid %d disappeared\n",
 982                             pfs->pfs_pid);
 983                         return (uiomove("unknown", sizeof("unknown") - 1,
 984                             ap->a_uio));
 985                 }
 986                 error = vn_fullpath(procp, NULL, &fullpath, &freepath);
 987                 if (error != 0)
 988                         return (uiomove("unknown", sizeof("unknown") - 1,
 989                             ap->a_uio));
 990                 error = uiomove(fullpath, strlen(fullpath), ap->a_uio);
 991                 free(freepath, M_TEMP);
 992                 return (error);
 993         default:
 994                 return (EINVAL);
 995         }
 996 }
 997
 998 /*
 999  * convert decimal ascii to pid_t
1000  */
1001 static pid_t
1002 atopid(const char *b, u_int len)
1003 {
1004         pid_t p = 0;
1005
1006         while (len--) {
1007                 char c = *b++;
1008                 if (c < '0' || c > '9')
1009                         return (NO_PID);
1010                 p = 10 * p + (c - '0');
1011                 if (p > PID_MAX)
1012                         return (NO_PID);
1013         }
1014
1015         return (p);
1016 }
1017
1018 /*
1019  * procfs vnode operations.
1020  */
1021 struct vnodeopv_entry_desc linprocfs_vnodeop_entries[] = {
1022         { &vop_default_desc,            (vnodeopv_entry_t)vop_defaultop },
1023         { &vop_access_desc,             (vnodeopv_entry_t)linprocfs_access },
1024         { &vop_advlock_desc,            (vnodeopv_entry_t)linprocfs_badop },
1025         { &vop_bmap_desc,               (vnodeopv_entry_t)linprocfs_bmap },
1026         { &vop_close_desc,              (vnodeopv_entry_t)linprocfs_close },
1027         { &vop_old_create_desc,         (vnodeopv_entry_t)linprocfs_badop },
1028         { &vop_getattr_desc,            (vnodeopv_entry_t)linprocfs_getattr },
1029         { &vop_inactive_desc,           (vnodeopv_entry_t)linprocfs_inactive },
1030         { &vop_old_link_desc,           (vnodeopv_entry_t)linprocfs_badop },
1031         { &vop_old_lookup_desc,         (vnodeopv_entry_t)linprocfs_lookup },
1032         { &vop_old_mkdir_desc,          (vnodeopv_entry_t)linprocfs_badop },
1033         { &vop_old_mknod_desc,          (vnodeopv_entry_t)linprocfs_badop },
1034         { &vop_open_desc,               (vnodeopv_entry_t)linprocfs_open },
1035         { &vop_pathconf_desc,           (vnodeopv_entry_t)vop_stdpathconf },
1036         { &vop_print_desc,              (vnodeopv_entry_t)linprocfs_print },
1037         { &vop_read_desc,               (vnodeopv_entry_t)linprocfs_rw },
1038         { &vop_readdir_desc,            (vnodeopv_entry_t)linprocfs_readdir },
1039         { &vop_readlink_desc,           (vnodeopv_entry_t)linprocfs_readlink },
1040         { &vop_reclaim_desc,            (vnodeopv_entry_t)linprocfs_reclaim },
1041         { &vop_old_remove_desc,         (vnodeopv_entry_t)linprocfs_badop },
1042         { &vop_old_rename_desc,         (vnodeopv_entry_t)linprocfs_badop },
1043         { &vop_old_rmdir_desc,          (vnodeopv_entry_t)linprocfs_badop },
1044         { &vop_setattr_desc,            (vnodeopv_entry_t)linprocfs_setattr },
1045         { &vop_old_symlink_desc,        (vnodeopv_entry_t)linprocfs_badop },
1046         { &vop_write_desc,              (vnodeopv_entry_t)linprocfs_rw },
1047         { &vop_ioctl_desc,              (vnodeopv_entry_t)linprocfs_ioctl },
1048         { NULL, NULL }
1049 };