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 1.42 2007/11/20 21:03:48 dillon 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/priv.h>
  57 #include <sys/signalvar.h>
  58 #include <sys/vnode.h>
  59 #include <sys/mount.h>
  60 #include <sys/namei.h>
  61 #include <sys/dirent.h>
  62 #include <sys/malloc.h>
  63 #include <sys/reg.h>
  64 #include <sys/jail.h>
  65 #include <vm/vm_zone.h>
  66 #include "linprocfs.h"
  67 #include <sys/pioctl.h>
  68
  69 #include <machine/limits.h>
  70
  71 extern struct vnode *procfs_findtextvp (struct proc *);
  72
  73 static int      linprocfs_access (struct vop_access_args *);
  74 static int      linprocfs_badop (struct vop_generic_args *);
  75 static int      linprocfs_bmap (struct vop_bmap_args *);
  76 static int      linprocfs_close (struct vop_close_args *);
  77 static int      linprocfs_getattr (struct vop_getattr_args *);
  78 static int      linprocfs_inactive (struct vop_inactive_args *);
  79 static int      linprocfs_ioctl (struct vop_ioctl_args *);
  80 static int      linprocfs_lookup (struct vop_old_lookup_args *);
  81 static int      linprocfs_open (struct vop_open_args *);
  82 static int      linprocfs_print (struct vop_print_args *);
  83 static int      linprocfs_readdir (struct vop_readdir_args *);
  84 static int      linprocfs_readlink (struct vop_readlink_args *);
  85 static int      linprocfs_reclaim (struct vop_reclaim_args *);
  86 static int      linprocfs_setattr (struct vop_setattr_args *);
  87
  88 static int      linprocfs_readdir_proc(struct vop_readdir_args *);
  89 static int      linprocfs_readdir_root(struct vop_readdir_args *);
  90 static int      linprocfs_readdir_net(struct vop_readdir_args *ap);
  91 static int      linprocfs_readdir_sys(struct vop_readdir_args *ap);
  92 static int      linprocfs_readdir_syskernel(struct vop_readdir_args *ap);
  93
  94 /*
  95  * procfs vnode operations.
  96  */
  97 struct vop_ops linprocfs_vnode_vops = {
  98         .vop_default =          vop_defaultop,
  99         .vop_access =           linprocfs_access,
 100         .vop_advlock =          (void *)linprocfs_badop,
 101         .vop_bmap =             linprocfs_bmap,
 102         .vop_close =            linprocfs_close,
 103         .vop_old_create =       (void *)linprocfs_badop,
 104         .vop_getattr =          linprocfs_getattr,
 105         .vop_inactive =         linprocfs_inactive,
 106         .vop_old_link =         (void *)linprocfs_badop,
 107         .vop_old_lookup =       linprocfs_lookup,
 108         .vop_old_mkdir =        (void *)linprocfs_badop,
 109         .vop_old_mknod =        (void *)linprocfs_badop,
 110         .vop_open =             linprocfs_open,
 111         .vop_pathconf =         vop_stdpathconf,
 112         .vop_print =            linprocfs_print,
 113         .vop_read =             (void *)linprocfs_rw,
 114         .vop_readdir =          linprocfs_readdir,
 115         .vop_readlink =         linprocfs_readlink,
 116         .vop_reclaim =          linprocfs_reclaim,
 117         .vop_old_remove =       (void *)linprocfs_badop,
 118         .vop_old_rename =       (void *)linprocfs_badop,
 119         .vop_old_rmdir =        (void *)linprocfs_badop,
 120         .vop_setattr =          linprocfs_setattr,
 121         .vop_old_symlink =      (void *)linprocfs_badop,
 122         .vop_write =            (void *)linprocfs_rw,
 123         .vop_ioctl =            linprocfs_ioctl
 124 };
 125
 126 /*
 127  * This is a list of the valid names in the
 128  * process-specific sub-directories.  It is
 129  * used in linprocfs_lookup and linprocfs_readdir
 130  */
 131 static struct proc_target {
 132         u_char  pt_type;
 133         u_char  pt_namlen;
 134         char    *pt_name;
 135         pfstype pt_pfstype;
 136         int     (*pt_valid) (struct proc *p);
 137 } proc_targets[] = {
 138 #define N(s) sizeof(s)-1, s
 139         /*        name          type            validp */
 140         { DT_DIR, N("."),       Pproc,          NULL },
 141         { DT_DIR, N(".."),      Proot,          NULL },
 142         { DT_REG, N("mem"),     Pmem,           NULL },
 143
 144         { DT_LNK, N("exe"),     Pexe,           NULL },
 145         { DT_LNK, N("cwd"),     Pcwd,           NULL },
 146         { DT_LNK, N("root"),    Pprocroot,      NULL },
 147         { DT_LNK, N("fd"),      Pfd,            NULL },
 148
 149         { DT_REG, N("stat"),    Pprocstat,      NULL },
 150         { DT_REG, N("status"),  Pprocstatus,    NULL },
 151         { DT_REG, N("maps"),    Pmaps,          NULL },
 152         { DT_REG, N("statm"),   Pstatm,         NULL },
 153 #if 0
 154         { DT_REG, N("cmdline"), Pcmdline,       NULL },
 155         { DT_REG, N("environ"), Penviron,       NULL },
 156 #endif
 157 #undef N
 158 };
 159 static const int nproc_targets = sizeof(proc_targets) / sizeof(proc_targets[0]);
 160
 161 static pid_t atopid (const char *, u_int);
 162
 163 /*
 164  * set things up for doing i/o on
 165  * the pfsnode (vp).  (vp) is locked
 166  * on entry, and should be left locked
 167  * on exit.
 168  *
 169  * for procfs we don't need to do anything
 170  * in particular for i/o.  all that is done
 171  * is to support exclusive open on process
 172  * memory images.
 173  */
 174 static int
 175 linprocfs_open(struct vop_open_args *ap)
 176 {
 177         struct pfsnode *pfs = VTOPFS(ap->a_vp);
 178         struct proc *p2;
 179
 180         p2 = PFIND(pfs->pfs_pid);
 181         if (p2 == NULL)
 182                 return (ENOENT);
 183         if (pfs->pfs_pid && !PRISON_CHECK(ap->a_cred, p2->p_ucred))
 184                 return (ENOENT);
 185
 186         switch (pfs->pfs_type) {
 187         case Pmem:
 188                 if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) ||
 189                     ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE)))
 190                         return (EBUSY);
 191
 192                 if (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         default:
 200                 break;
 201         }
 202
 203         return (vop_stdopen(ap));
 204 }
 205
 206 /*
 207  * close the pfsnode (vp) after doing i/o.
 208  * (vp) is not locked on entry or exit.
 209  *
 210  * nothing to do for procfs other than undo
 211  * any exclusive open flag (see _open above).
 212  */
 213 static int
 214 linprocfs_close(struct vop_close_args *ap)
 215 {
 216         struct pfsnode *pfs = VTOPFS(ap->a_vp);
 217         struct proc *p;
 218
 219         switch (pfs->pfs_type) {
 220         case Pmem:
 221                 if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
 222                         pfs->pfs_flags &= ~(FWRITE|O_EXCL);
 223                 /*
 224                  * If this is the last close, then it checks to see if
 225                  * the target process has PF_LINGER set in p_pfsflags,
 226                  * if this is *not* the case, then the process' stop flags
 227                  * are cleared, and the process is woken up.  This is
 228                  * to help prevent the case where a process has been
 229                  * told to stop on an event, but then the requesting process
 230                  * has gone away or forgotten about it.
 231                  */
 232                 if ((ap->a_vp->v_opencount < 2)
 233                     && (p = pfind(pfs->pfs_pid))
 234                     && !(p->p_pfsflags & PF_LINGER)) {
 235                         p->p_stops = 0;
 236                         p->p_step = 0;
 237                         wakeup(&p->p_step);
 238                 }
 239                 break;
 240         default:
 241                 break;
 242         }
 243         return (vop_stdclose(ap));
 244 }
 245
 246 /*
 247  * do an ioctl operation on a pfsnode (vp).
 248  * (vp) is not locked on entry or exit.
 249  */
 250 static int
 251 linprocfs_ioctl(struct vop_ioctl_args *ap)
 252 {
 253         struct pfsnode *pfs = VTOPFS(ap->a_vp);
 254         struct proc *procp;
 255         int error;
 256         int signo;
 257         struct procfs_status *psp;
 258         unsigned char flags;
 259
 260         procp = pfind(pfs->pfs_pid);
 261         if (procp == NULL) {
 262                 return ENOTTY;
 263         }
 264
 265         if (p_trespass(ap->a_cred, procp->p_ucred))
 266                 return EPERM;
 267
 268         switch (ap->a_command) {
 269         case PIOCBIS:
 270           procp->p_stops |= *(unsigned int*)ap->a_data;
 271           break;
 272         case PIOCBIC:
 273           procp->p_stops &= ~*(unsigned int*)ap->a_data;
 274           break;
 275         case PIOCSFL:
 276           /*
 277            * NFLAGS is "non-suser_xxx flags" -- currently, only
 278            * PFS_ISUGID ("ignore set u/g id");
 279            */
 280 #define NFLAGS  (PF_ISUGID)
 281           flags = (unsigned char)*(unsigned int*)ap->a_data;
 282           if (flags & NFLAGS && (error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0)))
 283             return error;
 284           procp->p_pfsflags = flags;
 285           break;
 286         case PIOCGFL:
 287           *(unsigned int*)ap->a_data = (unsigned int)procp->p_pfsflags;
 288         case PIOCSTATUS:
 289           psp = (struct procfs_status *)ap->a_data;
 290           psp->state = (procp->p_step == 0);
 291           psp->flags = procp->p_pfsflags;
 292           psp->events = procp->p_stops;
 293           if (procp->p_step) {
 294             psp->why = procp->p_stype;
 295             psp->val = procp->p_xstat;
 296           } else {
 297             psp->why = psp->val = 0;    /* Not defined values */
 298           }
 299           break;
 300         case PIOCWAIT:
 301           psp = (struct procfs_status *)ap->a_data;
 302           if (procp->p_step == 0) {
 303             error = tsleep(&procp->p_stype, PCATCH, "piocwait", 0);
 304             if (error)
 305               return error;
 306           }
 307           psp->state = 1;       /* It stopped */
 308           psp->flags = procp->p_pfsflags;
 309           psp->events = procp->p_stops;
 310           psp->why = procp->p_stype;    /* why it stopped */
 311           psp->val = procp->p_xstat;    /* any extra info */
 312           break;
 313         case PIOCCONT:  /* Restart a proc */
 314           if (procp->p_step == 0)
 315             return EINVAL;      /* Can only start a stopped process */
 316           if ((signo = *(int*)ap->a_data) != 0) {
 317             if (signo >= NSIG || signo <= 0)
 318               return EINVAL;
 319             ksignal(procp, signo);
 320           }
 321           procp->p_step = 0;
 322           wakeup(&procp->p_step);
 323           break;
 324         default:
 325           return (ENOTTY);
 326         }
 327         return 0;
 328 }
 329
 330 /*
 331  * do block mapping for pfsnode (vp).
 332  * since we don't use the buffer cache
 333  * for procfs this function should never
 334  * be called.  in any case, it's not clear
 335  * what part of the kernel ever makes use
 336  * of this function.  for sanity, this is the
 337  * usual no-op bmap, although returning
 338  * (EIO) would be a reasonable alternative.
 339  */
 340 static int
 341 linprocfs_bmap(struct vop_bmap_args *ap)
 342 {
 343         if (ap->a_doffsetp != NULL)
 344                 *ap->a_doffsetp = ap->a_loffset;
 345         if (ap->a_runp != NULL)
 346                 *ap->a_runp = 0;
 347         if (ap->a_runb != NULL)
 348                 *ap->a_runb = 0;
 349         return (0);
 350 }
 351
 352 /*
 353  * linprocfs_inactive is called when the pfsnode
 354  * is vrele'd and the reference count is about
 355  * to go to zero.  (vp) will be on the vnode free
 356  * list, so to get it back vget() must be
 357  * used.
 358  *
 359  * (vp) is locked on entry and must remain locked
 360  *      on exit.
 361  */
 362 static int
 363 linprocfs_inactive(struct vop_inactive_args *ap)
 364 {
 365         /*struct vnode *vp = ap->a_vp;*/
 366
 367         return (0);
 368 }
 369
 370 /*
 371  * _reclaim is called when getnewvnode()
 372  * wants to make use of an entry on the vnode
 373  * free list.  at this time the filesystem needs
 374  * to free any private data and remove the node
 375  * from any private lists.
 376  */
 377 static int
 378 linprocfs_reclaim(struct vop_reclaim_args *ap)
 379 {
 380         return (linprocfs_freevp(ap->a_vp));
 381 }
 382
 383 /*
 384  * _print is used for debugging.
 385  * just print a readable description
 386  * of (vp).
 387  */
 388 static int
 389 linprocfs_print(struct vop_print_args *ap)
 390 {
 391         struct pfsnode *pfs = VTOPFS(ap->a_vp);
 392
 393         kprintf("tag VT_PROCFS, type %d, pid %ld, mode %x, flags %lx\n",
 394             pfs->pfs_type, (long)pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
 395         return (0);
 396 }
 397
 398 /*
 399  * generic entry point for unsupported operations
 400  */
 401 static int
 402 linprocfs_badop(struct vop_generic_args *ap __unused)
 403 {
 404
 405         return (EIO);
 406 }
 407
 408 /*
 409  * Invent attributes for pfsnode (vp) and store
 410  * them in (vap).
 411  * Directories lengths are returned as zero since
 412  * any real length would require the genuine size
 413  * to be computed, and nothing cares anyway.
 414  *
 415  * this is relatively minimal for procfs.
 416  */
 417 static int
 418 linprocfs_getattr(struct vop_getattr_args *ap)
 419 {
 420         struct pfsnode *pfs = VTOPFS(ap->a_vp);
 421         struct vattr *vap = ap->a_vap;
 422         struct proc *procp;
 423         int error;
 424
 425         /*
 426          * First make sure that the process and its credentials
 427          * still exist.
 428          */
 429         switch (pfs->pfs_type) {
 430         case Proot:
 431         case Pself:
 432                 procp = 0;
 433                 break;
 434
 435         default:
 436                 procp = PFIND(pfs->pfs_pid);
 437                 if (procp == 0 || procp->p_ucred == NULL)
 438                         return (ENOENT);
 439         }
 440
 441         error = 0;
 442
 443         /* start by zeroing out the attributes */
 444         VATTR_NULL(vap);
 445
 446         /* next do all the common fields */
 447         vap->va_type = ap->a_vp->v_type;
 448         vap->va_mode = pfs->pfs_mode;
 449         vap->va_fileid = pfs->pfs_fileno;
 450         vap->va_flags = 0;
 451         vap->va_blocksize = PAGE_SIZE;
 452         vap->va_bytes = vap->va_size = 0;
 453         vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0];
 454
 455         /*
 456          * Make all times be current TOD.
 457          * It would be possible to get the process start
 458          * time from the p_stat structure, but there's
 459          * no "file creation" time stamp anyway, and the
 460          * p_stat structure is not addressible if u. gets
 461          * swapped out for that process.
 462          */
 463         nanotime(&vap->va_ctime);
 464         vap->va_atime = vap->va_mtime = vap->va_ctime;
 465
 466         /*
 467          * now do the object specific fields
 468          *
 469          * The size could be set from struct reg, but it's hardly
 470          * worth the trouble, and it puts some (potentially) machine
 471          * dependent data into this machine-independent code.  If it
 472          * becomes important then this function should break out into
 473          * a per-file stat function in the corresponding .c file.
 474          */
 475
 476         vap->va_nlink = 1;
 477         if (procp) {
 478                 vap->va_uid = procp->p_ucred->cr_uid;
 479                 vap->va_gid = procp->p_ucred->cr_gid;
 480         }
 481
 482         switch (pfs->pfs_type) {
 483         case Proot:
 484         case Pnet:
 485         case Psys:
 486         case Psyskernel:
 487                 /*
 488                  * Set nlink to 1 to tell fts(3) we don't actually know.
 489                  */
 490                 vap->va_nlink = 1;
 491                 vap->va_uid = 0;
 492                 vap->va_gid = 0;
 493                 vap->va_size = vap->va_bytes = DEV_BSIZE;
 494                 break;
 495
 496         case Pself: {
 497                 char buf[16];           /* should be enough */
 498                 vap->va_uid = 0;
 499                 vap->va_gid = 0;
 500                 vap->va_size = vap->va_bytes =
 501                     ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
 502                 break;
 503         }
 504
 505         case Pproc:
 506                 vap->va_nlink = nproc_targets;
 507                 vap->va_size = vap->va_bytes = DEV_BSIZE;
 508                 break;
 509
 510         case Pexe: {
 511                 char *fullpath, *freepath;
 512                 error = cache_fullpath(procp, &procp->p_textnch, &fullpath, &freepath);
 513                 /* error = vn_fullpath(procp, NULL, &fullpath, &freepath); */
 514                 if (error == 0) {
 515                         vap->va_size = strlen(fullpath);
 516                         kfree(freepath, M_TEMP);
 517                 } else {
 518                         vap->va_size = sizeof("unknown") - 1;
 519                         error = 0;
 520                 }
 521                 vap->va_bytes = vap->va_size;
 522                 break;
 523         }
 524         case Pcwd: {
 525                 char *fullpath, *freepath;
 526                 error = cache_fullpath(procp, &procp->p_fd->fd_ncdir, &fullpath, &freepath);
 527                 if (error == 0) {
 528                         vap->va_size = strlen(fullpath);
 529                         kfree(freepath, M_TEMP);
 530                 } else {
 531                         vap->va_size = sizeof("unknown") - 1;
 532                         error = 0;
 533                 }
 534                 vap->va_bytes = vap->va_size;
 535                 break;
 536         }
 537         case Pprocroot: {
 538                 struct nchandle *nchp;
 539                 char *fullpath, *freepath;
 540                 nchp = jailed(procp->p_ucred) ? &procp->p_fd->fd_njdir : &procp->p_fd->fd_nrdir;
 541                 error = cache_fullpath(procp, nchp, &fullpath, &freepath);
 542                 if (error == 0) {
 543                         vap->va_size = strlen(fullpath);
 544                         kfree(freepath, M_TEMP);
 545                 } else {
 546                         vap->va_size = sizeof("unknown") - 1;
 547                         error = 0;
 548                 }
 549                 vap->va_bytes = vap->va_size;
 550                 break;
 551         }
 552         case Pfd: {
 553                 if (procp == curproc) {
 554                         vap->va_size = sizeof("/dev/fd") - 1;
 555                         error = 0;
 556                 } else {
 557                         vap->va_size = sizeof("unknown") - 1;
 558                         error = 0;
 559                 }
 560                 vap->va_bytes = vap->va_size;
 561                 break;
 562         }
 563
 564         case Pmeminfo:
 565         case Pcpuinfo:
 566         case Pmounts:
 567         case Pstat:
 568         case Puptime:
 569         case Pversion:
 570         case Ploadavg:
 571         case Pnetdev:
 572         case Pdevices:
 573         case Posrelease:
 574         case Postype:
 575         case Ppidmax:
 576                 vap->va_bytes = vap->va_size = 0;
 577                 vap->va_uid = 0;
 578                 vap->va_gid = 0;
 579                 break;
 580
 581         case Pmem:
 582                 /*
 583                  * If we denied owner access earlier, then we have to
 584                  * change the owner to root - otherwise 'ps' and friends
 585                  * will break even though they are setgid kmem. *SIGH*
 586                  */
 587                 if (procp->p_flag & P_SUGID)
 588                         vap->va_uid = 0;
 589                 else
 590                         vap->va_uid = procp->p_ucred->cr_uid;
 591                 break;
 592
 593         case Pprocstat:
 594         case Pprocstatus:
 595         case Pcmdline:
 596         case Penviron:
 597         case Pmaps:
 598         case Pstatm:
 599                 vap->va_bytes = vap->va_size = 0;
 600                 /* uid, gid are already set */
 601                 break;
 602
 603         default:
 604                 panic("linprocfs_getattr");
 605         }
 606
 607         return (error);
 608 }
 609
 610 static int
 611 linprocfs_setattr(struct vop_setattr_args *ap)
 612 {
 613
 614         if (ap->a_vap->va_flags != VNOVAL)
 615                 return (EOPNOTSUPP);
 616
 617         /*
 618          * just fake out attribute setting
 619          * it's not good to generate an error
 620          * return, otherwise things like creat()
 621          * will fail when they try to set the
 622          * file length to 0.  worse, this means
 623          * that echo $note > /proc/$pid/note will fail.
 624          */
 625
 626         return (0);
 627 }
 628
 629 /*
 630  * implement access checking.
 631  *
 632  * something very similar to this code is duplicated
 633  * throughout the 4bsd kernel and should be moved
 634  * into kern/vfs_subr.c sometime.
 635  *
 636  * actually, the check for super-user is slightly
 637  * broken since it will allow read access to write-only
 638  * objects.  this doesn't cause any particular trouble
 639  * but does mean that the i/o entry points need to check
 640  * that the operation really does make sense.
 641  */
 642 static int
 643 linprocfs_access(struct vop_access_args *ap)
 644 {
 645         struct vattr *vap;
 646         struct vattr vattr;
 647         int error;
 648
 649         /*
 650          * If you're the super-user,
 651          * you always get access.
 652          */
 653         if (ap->a_cred->cr_uid == 0)
 654                 return (0);
 655
 656         vap = &vattr;
 657         error = VOP_GETATTR(ap->a_vp, vap);
 658         if (error)
 659                 return (error);
 660
 661         /*
 662          * Access check is based on only one of owner, group, public.
 663          * If not owner, then check group. If not a member of the
 664          * group, then check public access.
 665          */
 666         if (ap->a_cred->cr_uid != vap->va_uid) {
 667                 gid_t *gp;
 668                 int i;
 669
 670                 ap->a_mode >>= 3;
 671                 gp = ap->a_cred->cr_groups;
 672                 for (i = 0; i < ap->a_cred->cr_ngroups; i++, gp++)
 673                         if (vap->va_gid == *gp)
 674                                 goto found;
 675                 ap->a_mode >>= 3;
 676 found:
 677                 ;
 678         }
 679
 680         if ((vap->va_mode & ap->a_mode) == ap->a_mode)
 681                 return (0);
 682
 683         return (EACCES);
 684 }
 685
 686 /*
 687  * lookup.  this is incredibly complicated in the general case, however
 688  * for most pseudo-filesystems very little needs to be done.
 689  */
 690 static int
 691 linprocfs_lookup(struct vop_old_lookup_args *ap)
 692 {
 693         struct componentname *cnp = ap->a_cnp;
 694         struct vnode **vpp = ap->a_vpp;
 695         struct vnode *dvp = ap->a_dvp;
 696         char *pname = cnp->cn_nameptr;
 697         struct proc_target *pt;
 698         pid_t pid;
 699         struct pfsnode *pfs;
 700         struct proc *p;
 701         int i;
 702         int error;
 703
 704         *vpp = NULL;
 705
 706         if (cnp->cn_nameiop == NAMEI_DELETE ||
 707             cnp->cn_nameiop == NAMEI_RENAME ||
 708             cnp->cn_nameiop == NAMEI_CREATE) {
 709                 return (EROFS);
 710         }
 711
 712         error = 0;
 713
 714         if (cnp->cn_namelen == 1 && *pname == '.') {
 715                 *vpp = dvp;
 716                 vref(*vpp);
 717                 goto out;
 718         }
 719
 720         pfs = VTOPFS(dvp);
 721         switch (pfs->pfs_type) {
 722         case Psys:
 723                 if (cnp->cn_flags & CNP_ISDOTDOT) {
 724                         error = linprocfs_root(dvp->v_mount, vpp);
 725                         goto out;
 726                 }
 727                 if (CNEQ(cnp, "kernel", 6)) {
 728                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Psyskernel);
 729                         goto out;
 730                 }
 731                 break;
 732         case Pnet:
 733                 if (cnp->cn_flags & CNP_ISDOTDOT) {
 734                         error = linprocfs_root(dvp->v_mount, vpp);
 735                         goto out;
 736                 }
 737                 if (CNEQ(cnp, "dev", 3)) {
 738                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pnetdev);
 739                         goto out;
 740                 }
 741                 break;
 742         case Psyskernel:
 743                 if (cnp->cn_flags & CNP_ISDOTDOT) {
 744                         /* XXX: this is wrong, wrong, wrong. */
 745                         error = linprocfs_root(dvp->v_mount, vpp);
 746                         goto out;
 747                 }
 748                 if (CNEQ(cnp, "osrelease", 9)) {
 749                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Posrelease);
 750                         goto out;
 751                 }
 752                 if (CNEQ(cnp, "ostype", 6)) {
 753                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Postype);
 754                         goto out;
 755                 }
 756                 if (CNEQ(cnp, "pid_max", 7)) {
 757                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Ppidmax);
 758                         goto out;
 759                 }
 760                 if (CNEQ(cnp, "version", 7)) {
 761                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pversion);
 762                         goto out;
 763                 }
 764                 break;
 765
 766         case Proot:
 767                 if (cnp->cn_flags & CNP_ISDOTDOT)
 768                         return (EIO);
 769
 770                 if (CNEQ(cnp, "self", 4)) {
 771                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pself);
 772                         goto out;
 773                 }
 774                 if (CNEQ(cnp, "meminfo", 7)) {
 775                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pmeminfo);
 776                         goto out;
 777                 }
 778                 if (CNEQ(cnp, "cpuinfo", 7)) {
 779                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pcpuinfo);
 780                         goto out;
 781                 }
 782                 if (CNEQ(cnp, "mounts", 6)) {
 783                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pmounts);
 784                         goto out;
 785                 }
 786                 if (CNEQ(cnp, "stat", 4)) {
 787                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pstat);
 788                         goto out;
 789                 }
 790                 if (CNEQ(cnp, "uptime", 6)) {
 791                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Puptime);
 792                         goto out;
 793                 }
 794                 if (CNEQ(cnp, "version", 7)) {
 795                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pversion);
 796                         goto out;
 797                 }
 798                 if (CNEQ(cnp, "loadavg", 7)) {
 799                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Ploadavg);
 800                         goto out;
 801                 }
 802                 if (CNEQ(cnp, "net", 3)) {
 803                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Pnet);
 804                         goto out;
 805                 }
 806                 if (CNEQ(cnp, "sys", 3)) {
 807                         error = linprocfs_allocvp(dvp->v_mount, vpp, 0, Psys);
 808                         goto out;
 809                 }
 810
 811                 pid = atopid(pname, cnp->cn_namelen);
 812                 if (pid == NO_PID)
 813                         break;
 814
 815                 p = PFIND(pid);
 816                 if (p == 0)
 817                         break;
 818
 819                 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
 820                         break;
 821
 822                 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
 823                     ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
 824                         break;
 825
 826                 error = linprocfs_allocvp(dvp->v_mount, vpp, pid, Pproc);
 827                 goto out;
 828
 829         case Pproc:
 830                 if (cnp->cn_flags & CNP_ISDOTDOT) {
 831                         error = linprocfs_root(dvp->v_mount, vpp);
 832                         goto out;
 833                 }
 834
 835                 p = PFIND(pfs->pfs_pid);
 836                 if (p == 0)
 837                         break;
 838
 839                 if (!PRISON_CHECK(ap->a_cnp->cn_cred, p->p_ucred))
 840                         break;
 841
 842                 if (ps_showallprocs == 0 && ap->a_cnp->cn_cred->cr_uid != 0 &&
 843                     ap->a_cnp->cn_cred->cr_uid != p->p_ucred->cr_uid)
 844                         break;
 845
 846                 for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) {
 847                         if (cnp->cn_namelen == pt->pt_namlen &&
 848                             bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 &&
 849                             (pt->pt_valid == NULL || (*pt->pt_valid)(p)))
 850                                 goto found;
 851                 }
 852                 break;
 853
 854         found:
 855                 error = linprocfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
 856                                         pt->pt_pfstype);
 857                 goto out;
 858
 859         default:
 860                 error = ENOTDIR;
 861                 goto out;
 862         }
 863
 864         if (cnp->cn_nameiop == NAMEI_LOOKUP)
 865                 error = ENOENT;
 866         else
 867                 error = EROFS;
 868
 869         /*
 870          * If no error occured *vpp will hold a referenced locked vnode.
 871          * dvp was passed to us locked and *vpp must be returned locked
 872          * so if dvp != *vpp and CNP_LOCKPARENT is not set, unlock dvp.
 873          */
 874 out:
 875         if (error == 0) {
 876                 if (*vpp != dvp && (cnp->cn_flags & CNP_LOCKPARENT) == 0) {
 877                         cnp->cn_flags |= CNP_PDIRUNLOCK;
 878                         vn_unlock(dvp);
 879                 }
 880         }
 881         return (error);
 882 }
 883
 884 /*
 885  * Does this process have a text file?
 886  */
 887 int
 888 linprocfs_validfile(struct proc *p)
 889 {
 890
 891         return (procfs_findtextvp(p) != NULLVP);
 892 }
 893
 894 /*
 895  * readdir() returns directory entries from pfsnode (vp).
 896  *
 897  * We generate just one directory entry at a time, as it would probably
 898  * not pay off to buffer several entries locally to save uiomove calls.
 899  *
 900  * linprocfs_readdir(struct vnode *a_vp, struct uio *a_uio,
 901  *                   struct ucred *a_cred, int *a_eofflag,
 902  *                   int *a_ncookies, off_t **a_cookies)
 903  */
 904 static int
 905 linprocfs_readdir(struct vop_readdir_args *ap)
 906 {
 907         struct pfsnode *pfs;
 908         int error;
 909
 910         if (ap->a_uio->uio_offset < 0 || ap->a_uio->uio_offset > INT_MAX)
 911                 return (EINVAL);
 912
 913         pfs = VTOPFS(ap->a_vp);
 914         if ((error = vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY)) != 0)
 915                 return (error);
 916
 917         switch (pfs->pfs_type) {
 918         case Pproc:
 919                 /*
 920                  * This is for the process-specific sub-directories.
 921                  * all that is needed to is copy out all the entries
 922                  * from the procent[] table (top of this file).
 923                  */
 924                 error = linprocfs_readdir_proc(ap);
 925                 break;
 926         case Proot:
 927                 /*
 928                  * This is for the root of the procfs filesystem
 929                  * what is needed is a special entry for "self"
 930                  * followed by an entry for each process on allproc
 931                  */
 932                 error = linprocfs_readdir_root(ap);
 933                 break;
 934         case Pnet:
 935                 error = linprocfs_readdir_net(ap);
 936                 break;
 937         case Psys:
 938                 error = linprocfs_readdir_sys(ap);
 939                 break;
 940         case Psyskernel:
 941                 error = linprocfs_readdir_syskernel(ap);
 942                 break;
 943         default:
 944                 error = ENOTDIR;
 945                 break;
 946         }
 947         vn_unlock(ap->a_vp);
 948
 949         return (error);
 950 }
 951
 952 static int
 953 linprocfs_readdir_proc(struct vop_readdir_args *ap)
 954 {
 955         struct pfsnode *pfs;
 956         int error, i, retval;
 957         struct proc *p;
 958         struct proc_target *pt;
 959         struct uio *uio = ap->a_uio;
 960
 961         pfs = VTOPFS(ap->a_vp);
 962         p = PFIND(pfs->pfs_pid);
 963         if (p == NULL)
 964                 return(0);
 965         if (!PRISON_CHECK(ap->a_cred, p->p_ucred))
 966                 return(0);
 967
 968         error = 0;
 969         i = uio->uio_offset;
 970
 971         for (pt = &proc_targets[i];
 972              !error && uio->uio_resid > 0 && i < nproc_targets; pt++, i++) {
 973                 if (pt->pt_valid && (*pt->pt_valid)(p) == 0)
 974                         continue;
 975
 976                 retval = vop_write_dirent(&error, uio,
 977                     PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype), pt->pt_type,
 978                     pt->pt_namlen, pt->pt_name);
 979                 if (retval)
 980                         break;
 981         }
 982
 983         uio->uio_offset = i;
 984
 985         return(error);
 986 }
 987
 988 struct linprocfs_readdir_root_info {
 989         int error;
 990         int pcnt;
 991         int i;
 992         struct uio *uio;
 993         struct ucred *cred;
 994 };
 995
 996 /*
 997  * Scan the root directory by scanning all process
 998  */
 999 static int linprocfs_readdir_root_callback(struct proc *p, void *data);
1000
1001 static int
1002 linprocfs_readdir_root(struct vop_readdir_args *ap)
1003 {
1004         struct linprocfs_readdir_root_info info;
1005         struct uio *uio = ap->a_uio;
1006         int res = 0;
1007
1008         info.error = 0;
1009         info.i = uio->uio_offset;
1010         info.pcnt = 0;
1011         info.uio = uio;
1012         info.cred = ap->a_cred;
1013
1014         while (info.pcnt < 12) {
1015                 res = linprocfs_readdir_root_callback(NULL, &info);
1016                 if (res < 0)
1017                         break;
1018         }
1019         if (res >= 0)
1020                 allproc_scan(linprocfs_readdir_root_callback, &info);
1021
1022         uio->uio_offset = info.i;
1023         return(info.error);
1024 }
1025
1026 static int
1027 linprocfs_readdir_root_callback(struct proc *p, void *data)
1028 {
1029         struct linprocfs_readdir_root_info *info = data;
1030         int retval;
1031         struct uio *uio = info->uio;
1032         ino_t d_ino;
1033         const char *d_name;
1034         char d_name_pid[20];
1035         size_t d_namlen;
1036         uint8_t d_type;
1037
1038         switch (info->pcnt) {
1039         case 0:         /* `.' */
1040                 d_ino = PROCFS_FILENO(0, Proot);
1041                 d_name = ".";
1042                 d_namlen = 1;
1043                 d_type = DT_DIR;
1044                 break;
1045         case 1:         /* `..' */
1046                 d_ino = PROCFS_FILENO(0, Proot);
1047                 d_name = "..";
1048                 d_namlen = 2;
1049                 d_type = DT_DIR;
1050                 break;
1051
1052         case 2:
1053                 d_ino = PROCFS_FILENO(0, Proot);
1054                 d_namlen = 4;
1055                 d_name = "self";
1056                 d_type = DT_LNK;
1057                 break;
1058
1059         case 3:
1060                 d_ino = PROCFS_FILENO(0, Pmeminfo);
1061                 d_namlen = 7;
1062                 d_name = "meminfo";
1063                 d_type = DT_REG;
1064                 break;
1065
1066         case 4:
1067                 d_ino = PROCFS_FILENO(0, Pcpuinfo);
1068                 d_namlen = 7;
1069                 d_name = "cpuinfo";
1070                 d_type = DT_REG;
1071                 break;
1072
1073         case 5:
1074                 d_ino = PROCFS_FILENO(0, Pstat);
1075                 d_namlen = 4;
1076                 d_name = "stat";
1077                 d_type = DT_REG;
1078                 break;
1079
1080         case 6:
1081                 d_ino = PROCFS_FILENO(0, Puptime);
1082                 d_namlen = 6;
1083                 d_name = "uptime";
1084                 d_type = DT_REG;
1085                 break;
1086
1087         case 7:
1088                 d_ino = PROCFS_FILENO(0, Pversion);
1089                 d_namlen = 7;
1090                 d_name = "version";
1091                 d_type = DT_REG;
1092                 break;
1093
1094         case 8:
1095                 d_ino = PROCFS_FILENO(0, Ploadavg);
1096                 d_namlen = 7;
1097                 d_name = "loadavg";
1098                 d_type = DT_REG;
1099                 break;
1100         case 9:
1101                 d_ino = PROCFS_FILENO(0, Pnet);
1102                 d_namlen = 3;
1103                 d_name = "net";
1104                 d_type = DT_DIR;
1105                 break;
1106         case 10:
1107                 d_ino = PROCFS_FILENO(0, Psys);
1108                 d_namlen = 3;
1109                 d_name = "sys";
1110                 d_type = DT_DIR;
1111                 break;
1112         case 11:
1113                 d_ino = PROCFS_FILENO(0, Pmounts);
1114                 d_namlen = 6;
1115                 d_name = "mounts";
1116                 d_type = DT_DIR;
1117                 break;
1118 #if 0
1119         case 11:
1120                 d_ino = PROCFS_FILENO(0, Pdevices);
1121                 d_namlen = 7;
1122                 d_name = "devices";
1123                 d_type = DT_REG;
1124                 break;
1125 #endif
1126         default:
1127                 /*
1128                  * Ignore processes that aren't in our prison
1129                  */
1130                 if (PRISON_CHECK(info->cred, p->p_ucred) == 0)
1131                         return(0);
1132
1133                 /*
1134                  * Ignore processes that we do not want to be visible.
1135                  */
1136                 if (ps_showallprocs == 0 &&
1137                     info->cred->cr_uid != 0 &&
1138                     info->cred->cr_uid != p->p_ucred->cr_uid) {
1139                         return(0);
1140                 }
1141
1142                 /*
1143                  * Skip processes we have already read (optimization)
1144                  */
1145                 if (info->pcnt < info->i) {
1146                         ++info->pcnt;
1147                         return(0);
1148                 }
1149                 d_ino = PROCFS_FILENO(p->p_pid, Pproc);
1150                 d_namlen = ksnprintf(d_name_pid, sizeof(d_name_pid),
1151                     "%ld", (long)p->p_pid);
1152                 d_name = d_name_pid;
1153                 d_type = DT_DIR;
1154                 break;
1155         }
1156
1157         /*
1158          * Skip processes we have already read
1159          */
1160         if (info->pcnt < info->i) {
1161                 ++info->pcnt;
1162                 return(0);
1163         }
1164         retval = vop_write_dirent(&info->error, info->uio,
1165                                   d_ino, d_type, d_namlen, d_name);
1166         if (retval == 0) {
1167                 ++info->pcnt;   /* iterate proc candidates scanned */
1168                 ++info->i;      /* iterate entries written */
1169         }
1170         if (retval || info->error || uio->uio_resid <= 0)
1171                 return(-1);
1172         return(0);
1173 }
1174
1175 /*
1176  * Scan the root directory by scanning all process
1177  */
1178 static int linprocfs_readdir_net_callback(struct proc *p, void *data);
1179
1180 static int
1181 linprocfs_readdir_net(struct vop_readdir_args *ap)
1182 {
1183         struct linprocfs_readdir_root_info info;
1184         struct uio *uio = ap->a_uio;
1185         int res;
1186
1187         info.error = 0;
1188         info.i = uio->uio_offset;
1189         info.pcnt = 0;
1190         info.uio = uio;
1191         info.cred = ap->a_cred;
1192
1193         while (info.pcnt < 3) {
1194                 res = linprocfs_readdir_net_callback(NULL, &info);
1195                 if (res < 0)
1196                         break;
1197         }
1198
1199         uio->uio_offset = info.i;
1200         return(info.error);
1201 }
1202
1203 static int
1204 linprocfs_readdir_net_callback(struct proc *p, void *data)
1205 {
1206         struct linprocfs_readdir_root_info *info = data;
1207         int retval;
1208         struct uio *uio = info->uio;
1209         ino_t d_ino;
1210         const char *d_name;
1211         size_t d_namlen;
1212         uint8_t d_type;
1213
1214         switch (info->pcnt) {
1215         case 0:         /* `.' */
1216                 d_ino = PROCFS_FILENO(0, Pnet);
1217                 d_name = ".";
1218                 d_namlen = 1;
1219                 d_type = DT_DIR;
1220                 break;
1221         case 1:         /* `..' */
1222                 d_ino = PROCFS_FILENO(0, Proot);
1223                 d_name = "..";
1224                 d_namlen = 2;
1225                 d_type = DT_DIR;
1226                 break;
1227
1228         case 2:
1229                 d_ino = PROCFS_FILENO(0, Pnet);
1230                 d_namlen = 3;
1231                 d_name = "dev";
1232                 d_type = DT_REG;
1233                 break;
1234         default:
1235                 d_ino = 0;
1236                 d_namlen = 0;
1237                 d_name = NULL;
1238                 d_type = DT_REG;
1239                 break;
1240         }
1241
1242         /*
1243          * Skip processes we have already read
1244          */
1245         if (info->pcnt < info->i) {
1246                 ++info->pcnt;
1247                 return(0);
1248         }
1249         retval = vop_write_dirent(&info->error, info->uio,
1250                                   d_ino, d_type, d_namlen, d_name);
1251         if (retval == 0) {
1252                 ++info->pcnt;   /* iterate proc candidates scanned */
1253                 ++info->i;      /* iterate entries written */
1254         }
1255         if (retval || info->error || uio->uio_resid <= 0)
1256                 return(-1);
1257         return(0);
1258 }
1259
1260
1261
1262
1263
1264
1265
1266 /*
1267  * Scan the root directory by scanning all process
1268  */
1269 static int linprocfs_readdir_sys_callback(struct proc *p, void *data);
1270
1271 static int
1272 linprocfs_readdir_sys(struct vop_readdir_args *ap)
1273 {
1274         struct linprocfs_readdir_root_info info;
1275         struct uio *uio = ap->a_uio;
1276         int res;
1277
1278         info.error = 0;
1279         info.i = uio->uio_offset;
1280         info.pcnt = 0;
1281         info.uio = uio;
1282         info.cred = ap->a_cred;
1283
1284         while (info.pcnt < 3) {
1285                 res = linprocfs_readdir_sys_callback(NULL, &info);
1286                 if (res < 0)
1287                         break;
1288         }
1289
1290         uio->uio_offset = info.i;
1291         return(info.error);
1292 }
1293
1294 static int
1295 linprocfs_readdir_sys_callback(struct proc *p, void *data)
1296 {
1297         struct linprocfs_readdir_root_info *info = data;
1298         int retval;
1299         struct uio *uio = info->uio;
1300         ino_t d_ino;
1301         const char *d_name;
1302         size_t d_namlen;
1303         uint8_t d_type;
1304
1305         switch (info->pcnt) {
1306         case 0:         /* `.' */
1307                 d_ino = PROCFS_FILENO(0, Psys);
1308                 d_name = ".";
1309                 d_namlen = 1;
1310                 d_type = DT_DIR;
1311                 break;
1312         case 1:         /* `..' */
1313                 d_ino = PROCFS_FILENO(0, Proot);
1314                 d_name = "..";
1315                 d_namlen = 2;
1316                 d_type = DT_DIR;
1317                 break;
1318
1319         case 2:
1320                 d_ino = PROCFS_FILENO(0, Psyskernel);
1321                 d_namlen = 6;
1322                 d_name = "kernel";
1323                 d_type = DT_DIR;
1324                 break;
1325         default:
1326                 d_ino = 0;
1327                 d_namlen = 0;
1328                 d_name = NULL;
1329                 d_type = DT_REG;
1330                 break;
1331         }
1332
1333         /*
1334          * Skip processes we have already read
1335          */
1336         if (info->pcnt < info->i) {
1337                 ++info->pcnt;
1338                 return(0);
1339         }
1340         retval = vop_write_dirent(&info->error, info->uio,
1341                                   d_ino, d_type, d_namlen, d_name);
1342         if (retval == 0) {
1343                 ++info->pcnt;   /* iterate proc candidates scanned */
1344                 ++info->i;      /* iterate entries written */
1345         }
1346         if (retval || info->error || uio->uio_resid <= 0)
1347                 return(-1);
1348         return(0);
1349 }
1350
1351
1352
1353
1354
1355 /*
1356  * Scan the root directory by scanning all process
1357  */
1358 static int linprocfs_readdir_syskernel_callback(struct proc *p, void *data);
1359
1360 static int
1361 linprocfs_readdir_syskernel(struct vop_readdir_args *ap)
1362 {
1363         struct linprocfs_readdir_root_info info;
1364         struct uio *uio = ap->a_uio;
1365         int res;
1366
1367         info.error = 0;
1368         info.i = uio->uio_offset;
1369         info.pcnt = 0;
1370         info.uio = uio;
1371         info.cred = ap->a_cred;
1372
1373         while (info.pcnt < 6) {
1374                 res = linprocfs_readdir_syskernel_callback(NULL, &info);
1375                 if (res < 0)
1376                         break;
1377         }
1378
1379         uio->uio_offset = info.i;
1380         return(info.error);
1381 }
1382
1383 static int
1384 linprocfs_readdir_syskernel_callback(struct proc *p, void *data)
1385 {
1386         struct linprocfs_readdir_root_info *info = data;
1387         int retval;
1388         struct uio *uio = info->uio;
1389         ino_t d_ino;
1390         const char *d_name;
1391         size_t d_namlen;
1392         uint8_t d_type;
1393
1394         switch (info->pcnt) {
1395         case 0:         /* `.' */
1396                 d_ino = PROCFS_FILENO(0, Psyskernel);
1397                 d_name = ".";
1398                 d_namlen = 1;
1399                 d_type = DT_DIR;
1400                 break;
1401         case 1:         /* `..' */
1402                 d_ino = PROCFS_FILENO(0, Psys);
1403                 d_name = "..";
1404                 d_namlen = 2;
1405                 d_type = DT_DIR;
1406                 break;
1407
1408         case 2:
1409                 d_ino = PROCFS_FILENO(0, Posrelease);
1410                 d_namlen = 9;
1411                 d_name = "osrelease";
1412                 d_type = DT_REG;
1413                 break;
1414
1415         case 3:
1416                 d_ino = PROCFS_FILENO(0, Postype);
1417                 d_namlen = 4;
1418                 d_name = "ostype";
1419                 d_type = DT_REG;
1420                 break;
1421
1422         case 4:
1423                 d_ino = PROCFS_FILENO(0, Pversion);
1424                 d_namlen = 7;
1425                 d_name = "version";
1426                 d_type = DT_REG;
1427                 break;
1428
1429         case 5:
1430                 d_ino = PROCFS_FILENO(0, Ppidmax);
1431                 d_namlen = 7;
1432                 d_name = "pid_max";
1433                 d_type = DT_REG;
1434                 break;
1435         default:
1436                 d_ino = 0;
1437                 d_namlen = 0;
1438                 d_name = NULL;
1439                 d_type = DT_REG;
1440                 break;
1441         }
1442
1443         /*
1444          * Skip processes we have already read
1445          */
1446         if (info->pcnt < info->i) {
1447                 ++info->pcnt;
1448                 return(0);
1449         }
1450         retval = vop_write_dirent(&info->error, info->uio,
1451                                   d_ino, d_type, d_namlen, d_name);
1452         if (retval == 0) {
1453                 ++info->pcnt;   /* iterate proc candidates scanned */
1454                 ++info->i;      /* iterate entries written */
1455         }
1456         if (retval || info->error || uio->uio_resid <= 0)
1457                 return(-1);
1458         return(0);
1459 }
1460
1461 /*
1462  * readlink reads the link of `self' or `exe'
1463  */
1464 static int
1465 linprocfs_readlink(struct vop_readlink_args *ap)
1466 {
1467         char buf[16];           /* should be enough */
1468         struct proc *procp;
1469         struct vnode *vp = ap->a_vp;
1470         struct nchandle *nchp;
1471         struct pfsnode *pfs = VTOPFS(vp);
1472         char *fullpath, *freepath;
1473         int error, len;
1474
1475         switch (pfs->pfs_type) {
1476         case Pself:
1477                 if (pfs->pfs_fileno != PROCFS_FILENO(0, Pself))
1478                         return (EINVAL);
1479
1480                 len = ksnprintf(buf, sizeof(buf), "%ld", (long)curproc->p_pid);
1481
1482                 return (uiomove(buf, len, ap->a_uio));
1483         /*
1484          * There _should_ be no way for an entire process to disappear
1485          * from under us...
1486          */
1487         case Pexe:
1488                 procp = PFIND(pfs->pfs_pid);
1489                 if (procp == NULL || procp->p_ucred == NULL) {
1490                         kprintf("linprocfs_readlink: pid %d disappeared\n",
1491                             pfs->pfs_pid);
1492                         return (uiomove("unknown", sizeof("unknown") - 1,
1493                             ap->a_uio));
1494                 }
1495                 error = cache_fullpath(procp, &procp->p_textnch, &fullpath, &freepath);
1496                 if (error != 0)
1497                         return (uiomove("unknown", sizeof("unknown") - 1,
1498                             ap->a_uio));
1499                 error = uiomove(fullpath, strlen(fullpath), ap->a_uio);
1500                 kfree(freepath, M_TEMP);
1501                 return (error);
1502         case Pcwd:
1503                 procp = PFIND(pfs->pfs_pid);
1504                 if (procp == NULL || procp->p_ucred == NULL) {
1505                         kprintf("linprocfs_readlink: pid %d disappeared\n",
1506                             pfs->pfs_pid);
1507                         return (uiomove("unknown", sizeof("unknown") - 1,
1508                             ap->a_uio));
1509                 }
1510                 error = cache_fullpath(procp, &procp->p_fd->fd_ncdir, &fullpath, &freepath);
1511                 if (error != 0)
1512                         return (uiomove("unknown", sizeof("unknown") - 1,
1513                             ap->a_uio));
1514                 error = uiomove(fullpath, strlen(fullpath), ap->a_uio);
1515                 kfree(freepath, M_TEMP);
1516                 return (error);
1517         case Pprocroot:
1518                 procp = PFIND(pfs->pfs_pid);
1519                 if (procp == NULL || procp->p_ucred == NULL) {
1520                         kprintf("linprocfs_readlink: pid %d disappeared\n",
1521                             pfs->pfs_pid);
1522                         return (uiomove("unknown", sizeof("unknown") - 1,
1523                             ap->a_uio));
1524                 }
1525                 nchp = jailed(procp->p_ucred) ? &procp->p_fd->fd_njdir : &procp->p_fd->fd_nrdir;
1526                 error = cache_fullpath(procp, nchp, &fullpath, &freepath);
1527                 if (error != 0)
1528                         return (uiomove("unknown", sizeof("unknown") - 1,
1529                             ap->a_uio));
1530                 error = uiomove(fullpath, strlen(fullpath), ap->a_uio);
1531                 kfree(freepath, M_TEMP);
1532                 return (error);
1533         case Pfd:
1534                 procp = PFIND(pfs->pfs_pid);
1535                 if (procp == NULL || procp->p_ucred == NULL) {
1536                         kprintf("linprocfs_readlink: pid %d disappeared\n",
1537                             pfs->pfs_pid);
1538                         return (uiomove("unknown", sizeof("unknown") - 1,
1539                             ap->a_uio));
1540                 }
1541                 if (procp == curproc) {
1542                         return (uiomove("/dev/fd", sizeof("/dev/fd") - 1,
1543                             ap->a_uio));
1544                 } else {
1545                         return (uiomove("unknown", sizeof("unknown") - 1,
1546                             ap->a_uio));
1547                 }
1548                 /* notreached */
1549                 break;
1550         default:
1551                 return (EINVAL);
1552         }
1553 }
1554
1555 /*
1556  * convert decimal ascii to pid_t
1557  */
1558 static pid_t
1559 atopid(const char *b, u_int len)
1560 {
1561         pid_t p = 0;
1562
1563         while (len--) {
1564                 char c = *b++;
1565                 if (c < '0' || c > '9')
1566                         return (NO_PID);
1567                 p = 10 * p + (c - '0');
1568                 if (p > PID_MAX)
1569                         return (NO_PID);
1570         }
1571
1572         return (p);
1573 }
1574