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

   1 /*
   2  * Copyright (c) 2000 Dag-Erling Coïdan Smørgrav
   3  * Copyright (c) 1999 Pierre Beyssac
   4  * Copyright (c) 1993 Jan-Simon Pendry
   5  * Copyright (c) 1993
   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_subr.c       8.6 (Berkeley) 5/14/95
  40  *
  41  * $FreeBSD: src/sys/i386/linux/linprocfs/linprocfs_subr.c,v 1.3.2.4 2001/06/25 19:46:47 pirzyk Exp $
  42  * $DragonFly: src/sys/emulation/linux/i386/linprocfs/linprocfs_subr.c,v 1.23 2007/08/25 23:27:02 corecode Exp $
  43  */
  44
  45 #include <sys/param.h>
  46 #include <sys/systm.h>
  47 #include <sys/proc.h>
  48 #include <sys/vnode.h>
  49 #include <sys/malloc.h>
  50 #include <sys/mount.h>
  51 #include "linprocfs.h"
  52
  53 #define PFSHSIZE        256
  54 #define PFSHMASK        (PFSHSIZE - 1)
  55
  56 static struct pfsnode *pfshead[PFSHSIZE];
  57 static struct lwkt_token pfs_token;
  58 static int pfsvplock;
  59
  60 extern int procfs_domem (struct proc *, struct lwp *, struct pfsnode *pfsp, struct uio *uio);
  61
  62 /*
  63  * allocate a pfsnode/vnode pair.  the vnode is
  64  * referenced, but not locked.
  65  *
  66  * the pid, pfs_type, and mount point uniquely
  67  * identify a pfsnode.  the mount point is needed
  68  * because someone might mount this filesystem
  69  * twice.
  70  *
  71  * all pfsnodes are maintained on a singly-linked
  72  * list.  new nodes are only allocated when they cannot
  73  * be found on this list.  entries on the list are
  74  * removed when the vfs reclaim entry is called.
  75  *
  76  * a single lock is kept for the entire list.  this is
  77  * needed because the getnewvnode() function can block
  78  * waiting for a vnode to become free, in which case there
  79  * may be more than one process trying to get the same
  80  * vnode.  this lock is only taken if we are going to
  81  * call getnewvnode, since the kernel itself is single-threaded.
  82  *
  83  * if an entry is found on the list, then call vget() to
  84  * take a reference.  this is done because there may be
  85  * zero references to it and so it needs to removed from
  86  * the vnode free list.
  87  */
  88 int
  89 linprocfs_allocvp(struct mount *mp, struct vnode **vpp, long pid,
  90                   pfstype pfs_type)
  91 {
  92         struct pfsnode *pfs;
  93         struct vnode *vp;
  94         struct pfsnode **pp;
  95         lwkt_tokref ilock;
  96         int error;
  97
  98         lwkt_gettoken(&ilock, &pfs_token);
  99 loop:
 100         for (pfs = pfshead[pid & PFSHMASK]; pfs; pfs = pfs->pfs_next) {
 101                 vp = PFSTOV(pfs);
 102                 if (pfs->pfs_pid == pid &&
 103                     pfs->pfs_type == pfs_type &&
 104                     vp->v_mount == mp) {
 105                         if (vget(vp, LK_EXCLUSIVE|LK_SLEEPFAIL))
 106                                 goto loop;
 107                         *vpp = vp;
 108                         lwkt_reltoken(&ilock);
 109                         return (0);
 110                 }
 111         }
 112
 113         /*
 114          * otherwise lock the vp list while we call getnewvnode
 115          * since that can block.
 116          */
 117         if (pfsvplock & PROCFS_LOCKED) {
 118                 pfsvplock |= PROCFS_WANT;
 119                 (void) tsleep((caddr_t) &pfsvplock, 0, "pfsavp", 0);
 120                 goto loop;
 121         }
 122         pfsvplock |= PROCFS_LOCKED;
 123
 124         /*
 125          * Do the MALLOC before the getnewvnode since doing so afterward
 126          * might cause a bogus v_data pointer to get dereferenced
 127          * elsewhere if MALLOC should block.
 128          */
 129         MALLOC(pfs, struct pfsnode *, sizeof(struct pfsnode), M_TEMP, M_WAITOK);
 130
 131         error = getnewvnode(VT_PROCFS, mp, vpp, 0, 0);
 132         if (error) {
 133                 FREE(pfs, M_TEMP);
 134                 goto out;
 135         }
 136         vp = *vpp;
 137
 138         vp->v_data = pfs;
 139
 140         pfs->pfs_next = 0;
 141         pfs->pfs_pid = (pid_t) pid;
 142         pfs->pfs_type = pfs_type;
 143         pfs->pfs_vnode = vp;
 144         pfs->pfs_flags = 0;
 145         pfs->pfs_lockowner = NULL;
 146         pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type);
 147
 148         switch (pfs_type) {
 149         case Proot:     /* /proc = dr-xr-xr-x */
 150                 vsetflags(vp, VROOT);
 151                 /* fallthrough */
 152         case Pnet:
 153         case Psys:
 154         case Psyskernel:
 155                 pfs->pfs_mode = (VREAD|VEXEC) |
 156                                 (VREAD|VEXEC) >> 3 |
 157                                 (VREAD|VEXEC) >> 6;
 158                 vp->v_type = VDIR;
 159                 break;
 160
 161         case Pself:     /* /proc/self = lr--r--r-- */
 162                 pfs->pfs_mode = (VREAD) |
 163                                 (VREAD >> 3) |
 164                                 (VREAD >> 6);
 165                 vp->v_type = VLNK;
 166                 break;
 167
 168         case Pproc:
 169                 pfs->pfs_mode = (VREAD|VEXEC) |
 170                                 (VREAD|VEXEC) >> 3 |
 171                                 (VREAD|VEXEC) >> 6;
 172                 vp->v_type = VDIR;
 173                 break;
 174
 175         case Pexe:
 176         case Pcwd:
 177         case Pprocroot:
 178         case Pfd:
 179                 pfs->pfs_mode = (VREAD|VEXEC) |
 180                                 (VREAD|VEXEC) >> 3 |
 181                                 (VREAD|VEXEC) >> 6;
 182                 vp->v_type = VLNK;
 183                 break;
 184
 185         case Pmem:
 186                 pfs->pfs_mode = (VREAD|VWRITE) |
 187                                 (VREAD) >> 3;
 188                 vp->v_type = VREG;
 189                 break;
 190
 191         case Pprocstat:
 192         case Pprocstatus:
 193         case Pcmdline:
 194         case Penviron:
 195         case Pstatm:
 196                 /* fallthrough */
 197         case Pmaps:
 198         case Pmeminfo:
 199         case Pcpuinfo:
 200         case Pmounts:
 201         case Pstat:
 202         case Puptime:
 203         case Pversion:
 204         case Ploadavg:
 205         case Pdevices:
 206         case Pnetdev:
 207         case Posrelease:
 208         case Postype:
 209         case Ppidmax:
 210                 pfs->pfs_mode = (VREAD) |
 211                                 (VREAD >> 3) |
 212                                 (VREAD >> 6);
 213                 vp->v_type = VREG;
 214                 break;
 215
 216         default:
 217                 panic("linprocfs_allocvp");
 218         }
 219
 220         /* add to procfs vnode list */
 221         for (pp = &pfshead[pid & PFSHMASK]; *pp; pp = &(*pp)->pfs_next)
 222                 continue;
 223         *pp = pfs;
 224
 225         vx_unlock(vp);  /* vnode ready to roll! */
 226
 227 out:
 228         pfsvplock &= ~PROCFS_LOCKED;
 229
 230         if (pfsvplock & PROCFS_WANT) {
 231                 pfsvplock &= ~PROCFS_WANT;
 232                 wakeup((caddr_t) &pfsvplock);
 233         }
 234         lwkt_reltoken(&ilock);
 235
 236         return (error);
 237 }
 238
 239 int
 240 linprocfs_freevp(struct vnode *vp)
 241 {
 242         struct pfsnode **pfspp;
 243         struct pfsnode *pfs = VTOPFS(vp);
 244         lwkt_tokref ilock;
 245
 246         lwkt_gettoken(&ilock, &pfs_token);
 247         pfspp = &pfshead[pfs->pfs_pid & PFSHMASK];
 248         while (*pfspp != pfs) {
 249                 KKASSERT(*pfspp != NULL);
 250                 pfspp = &(*pfspp)->pfs_next;
 251         }
 252         *pfspp = pfs->pfs_next;
 253         lwkt_reltoken(&ilock);
 254         FREE(vp->v_data, M_TEMP);
 255         vp->v_data = NULL;
 256         return (0);
 257 }
 258
 259 int
 260 linprocfs_rw(struct vop_read_args *ap)
 261 {
 262         struct vnode *vp = ap->a_vp;
 263         struct uio *uio = ap->a_uio;
 264         struct thread *td = uio->uio_td;
 265         struct pfsnode *pfs = VTOPFS(vp);
 266         struct proc *p;
 267         struct proc *curp;
 268         struct lwp *lp;
 269         int rtval;
 270
 271         curp = td->td_proc;
 272         KKASSERT(curp);
 273
 274         p = PFIND(pfs->pfs_pid);
 275         if (p == 0)
 276                 return (EINVAL);
 277         if (p->p_pid == 1 && securelevel > 0 && uio->uio_rw == UIO_WRITE)
 278                 return (EACCES);
 279         lp = FIRST_LWP_IN_PROC(p);
 280         LWPHOLD(lp);
 281
 282         while (pfs->pfs_lockowner) {
 283                 tsleep(&pfs->pfs_lockowner, 0, "pfslck", 0);
 284         }
 285         pfs->pfs_lockowner = curthread;
 286         switch (pfs->pfs_type) {
 287         case Pmem:
 288                 rtval = procfs_domem(curp, lp, pfs, uio);
 289                 break;
 290         case Pprocstat:
 291                 rtval = linprocfs_doprocstat(curp, p, pfs, uio);
 292                 break;
 293         case Pprocstatus:
 294                 rtval = linprocfs_doprocstatus(curp, p, pfs, uio);
 295                 break;
 296         case Pmeminfo:
 297                 rtval = linprocfs_domeminfo(curp, p, pfs, uio);
 298                 break;
 299         case Pcpuinfo:
 300                 rtval = linprocfs_docpuinfo(curp, p, pfs, uio);
 301                 break;
 302         case Pmounts:
 303                 rtval = linprocfs_domounts(curp, p, pfs, uio);
 304                 break;
 305         case Pstat:
 306                 rtval = linprocfs_dostat(curp, p, pfs, uio);
 307                 break;
 308         case Puptime:
 309                 rtval = linprocfs_douptime(curp, p, pfs, uio);
 310                 break;
 311         case Pversion:
 312                 rtval = linprocfs_doversion(curp, p, pfs, uio);
 313                 break;
 314         case Ploadavg:
 315                 rtval = linprocfs_doloadavg(curp, p, pfs, uio);
 316                 break;
 317         case Pnetdev:
 318                 rtval = linprocfs_donetdev(curp, p, pfs, uio);
 319                 break;
 320         case Pdevices:
 321                 rtval = linprocfs_dodevices(curp, p, pfs, uio);
 322                 break;
 323         case Posrelease:
 324                 rtval = linprocfs_doosrelease(curp, p, pfs, uio);
 325                 break;
 326         case Postype:
 327                 rtval = linprocfs_doostype(curp, p, pfs, uio);
 328                 break;
 329         case Ppidmax:
 330                 rtval = linprocfs_dopidmax(curp, p, pfs, uio);
 331                 break;
 332         case Pmaps:
 333                 rtval = linprocfs_domaps(curp, p, pfs, uio);
 334                 break;
 335         case Pstatm:
 336                 rtval = linprocfs_dostatm(curp, p, pfs, uio);
 337                 break;
 338         default:
 339                 rtval = EOPNOTSUPP;
 340                 break;
 341         }
 342         LWPRELE(lp);
 343         pfs->pfs_lockowner = NULL;
 344         wakeup(&pfs->pfs_lockowner);
 345         return rtval;
 346 }
 347
 348 #if 0
 349 /*
 350  * Get a string from userland into (buf).  Strip a trailing
 351  * nl character (to allow easy access from the shell).
 352  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
 353  * will automatically add a nul char at the end.
 354  *
 355  * Returns 0 on success or the following errors
 356  *
 357  * EINVAL:    file offset is non-zero.
 358  * EMSGSIZE:  message is longer than kernel buffer
 359  * EFAULT:    user i/o buffer is not addressable
 360  */
 361 int
 362 vfs_getuserstr(struct uio *uio, char *buf, int *buflenp)
 363 {
 364         int xlen;
 365         int error;
 366
 367         if (uio->uio_offset != 0)
 368                 return (EINVAL);
 369
 370         xlen = *buflenp;
 371
 372         /* must be able to read the whole string in one go */
 373         if (xlen < uio->uio_resid)
 374                 return (EMSGSIZE);
 375         xlen = uio->uio_resid;
 376
 377         if ((error = uiomove(buf, xlen, uio)) != 0)
 378                 return (error);
 379
 380         /* allow multiple writes without seeks */
 381         uio->uio_offset = 0;
 382
 383         /* cleanup string and remove trailing newline */
 384         buf[xlen] = '\0';
 385         xlen = strlen(buf);
 386         if (xlen > 0 && buf[xlen-1] == '\n')
 387                 buf[--xlen] = '\0';
 388         *buflenp = xlen;
 389
 390         return (0);
 391 }
 392
 393 vfs_namemap_t *
 394 vfs_findname(vfs_namemap_t *nm, char *buf, int buflen)
 395 {
 396
 397         for (; nm->nm_name; nm++)
 398                 if (bcmp(buf, nm->nm_name, buflen+1) == 0)
 399                         return (nm);
 400
 401         return (0);
 402 }
 403 #endif
 404
 405 void
 406 linprocfs_init(void)
 407 {
 408         lwkt_token_init(&pfs_token);
 409 }
 410
 411 void
 412 linprocfs_exit(struct thread *td)
 413 {
 414         lwkt_tokref ilock;
 415         struct pfsnode *pfs;
 416         struct vnode *vp;
 417         pid_t pid;
 418
 419         KKASSERT(td->td_proc);
 420         pid = td->td_proc->p_pid;
 421
 422         /*
 423          * Remove all the procfs vnodes associated with an exiting process.
 424          */
 425         lwkt_gettoken(&ilock, &pfs_token);
 426 restart:
 427         for (pfs = pfshead[pid & PFSHMASK]; pfs; pfs = pfs->pfs_next) {
 428                 if (pfs->pfs_pid == pid) {
 429                         vp = PFSTOV(pfs);
 430                         vx_get(vp);
 431                         vgone_vxlocked(vp);
 432                         vx_put(vp);
 433                         goto restart;
 434                 }
 435         }
 436         lwkt_reltoken(&ilock);
 437         lwkt_token_uninit(&pfs_token);
 438 }
 439