sys/vfs/procfs/procfs_subr.c

   1 /*
   2  * Copyright (c) 1993 Jan-Simon Pendry
   3  * Copyright (c) 1993
   4  *      The Regents of the University of California.  All rights reserved.
   5  *
   6  * This code is derived from software contributed to Berkeley by
   7  * Jan-Simon Pendry.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  * 3. All advertising materials mentioning features or use of this software
  18  *    must display the following acknowledgement:
  19  *      This product includes software developed by the University of
  20  *      California, Berkeley and its contributors.
  21  * 4. Neither the name of the University nor the names of its contributors
  22  *    may be used to endorse or promote products derived from this software
  23  *    without specific prior written permission.
  24  *
  25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  35  * SUCH DAMAGE.
  36  *
  37  *      @(#)procfs_subr.c       8.6 (Berkeley) 5/14/95
  38  *
  39  * $FreeBSD: src/sys/miscfs/procfs/procfs_subr.c,v 1.26.2.3 2002/02/18 21:28:04 des Exp $
  40  * $DragonFly: src/sys/vfs/procfs/procfs_subr.c,v 1.18 2007/08/25 23:27:02 corecode Exp $
  41  */
  42
  43 #include <sys/param.h>
  44 #include <sys/systm.h>
  45 #include <sys/sysctl.h>
  46 #include <sys/proc.h>
  47 #include <sys/mount.h>
  48 #include <sys/vnode.h>
  49 #include <sys/malloc.h>
  50
  51 #include <vfs/procfs/procfs.h>
  52
  53 #define PFS_HSIZE       256
  54 #define PFS_HMASK       (PFS_HSIZE - 1)
  55
  56 static struct pfsnode *pfshead[PFS_HSIZE];
  57 static int pfsvplock;
  58
  59 #define PFSHASH(pid)    &pfshead[(pid) & PFS_HMASK]
  60
  61 /*
  62  * Allocate a pfsnode/vnode pair.  If no error occurs the returned vnode
  63  * will be referenced and exclusively locked.
  64  *
  65  * The pid, pfs_type, and mount point uniquely identify a pfsnode.
  66  * The mount point is needed because someone might mount this filesystem
  67  * twice.
  68  *
  69  * All pfsnodes are maintained on a singly-linked list.  new nodes are
  70  * only allocated when they cannot be found on this list.  entries on
  71  * the list are removed when the vfs reclaim entry is called.
  72  *
  73  * A single lock is kept for the entire list.  this is needed because the
  74  * getnewvnode() function can block waiting for a vnode to become free,
  75  * in which case there may be more than one process trying to get the same
  76  * vnode.  this lock is only taken if we are going to call getnewvnode,
  77  * since the kernel itself is single-threaded.
  78  *
  79  * If an entry is found on the list, then call vget() to take a reference
  80  * and obtain the lock.  This will properly re-reference the vnode if it
  81  * had gotten onto the free list.
  82  */
  83 int
  84 procfs_allocvp(struct mount *mp, struct vnode **vpp, long pid, pfstype pfs_type)
  85 {
  86         struct pfsnode *pfs;
  87         struct vnode *vp;
  88         struct pfsnode **pp;
  89         int error;
  90
  91         pp = PFSHASH(pid);
  92 loop:
  93         for (pfs = *pp; pfs; pfs = pfs->pfs_next) {
  94                 if (pfs->pfs_pid == pid && pfs->pfs_type == pfs_type &&
  95                     PFSTOV(pfs)->v_mount == mp) {
  96                         vp = PFSTOV(pfs);
  97                         if (vget(vp, LK_EXCLUSIVE))
  98                                 goto loop;
  99
 100                         /*
 101                          * Make sure the vnode is still in the cache after
 102                          * getting the interlock to avoid racing a free.
 103                          */
 104                         for (pfs = *pp; pfs; pfs = pfs->pfs_next) {
 105                                 if (PFSTOV(pfs) == vp &&
 106                                     pfs->pfs_pid == pid &&
 107                                     pfs->pfs_type == pfs_type &&
 108                                     PFSTOV(pfs)->v_mount == mp) {
 109                                         break;
 110                                 }
 111                         }
 112                         if (pfs == NULL || PFSTOV(pfs) != vp) {
 113                                 vput(vp);
 114                                 goto loop;
 115
 116                         }
 117                         *vpp = vp;
 118                         return (0);
 119                 }
 120         }
 121
 122         /*
 123          * otherwise lock the vp list while we call getnewvnode
 124          * since that can block.
 125          */
 126         if (pfsvplock & PROCFS_LOCKED) {
 127                 pfsvplock |= PROCFS_WANT;
 128                 (void) tsleep((caddr_t) &pfsvplock, 0, "pfsavp", 0);
 129                 goto loop;
 130         }
 131         pfsvplock |= PROCFS_LOCKED;
 132
 133         /*
 134          * Do the MALLOC before the getnewvnode since doing so afterward
 135          * might cause a bogus v_data pointer to get dereferenced
 136          * elsewhere if MALLOC should block.
 137          *
 138          * XXX this may not matter anymore since getnewvnode now returns
 139          * a VX locked vnode.
 140          */
 141         MALLOC(pfs, struct pfsnode *, sizeof(struct pfsnode), M_TEMP, M_WAITOK);
 142
 143         error = getnewvnode(VT_PROCFS, mp, vpp, 0, 0);
 144         if (error) {
 145                 kfree(pfs, M_TEMP);
 146                 goto out;
 147         }
 148         vp = *vpp;
 149
 150         vp->v_data = pfs;
 151
 152         pfs->pfs_next = 0;
 153         pfs->pfs_pid = (pid_t) pid;
 154         pfs->pfs_type = pfs_type;
 155         pfs->pfs_vnode = vp;
 156         pfs->pfs_flags = 0;
 157         pfs->pfs_lockowner = 0;
 158         pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type);
 159
 160         switch (pfs_type) {
 161         case Proot:     /* /proc = dr-xr-xr-x */
 162                 pfs->pfs_mode = (VREAD|VEXEC) |
 163                                 (VREAD|VEXEC) >> 3 |
 164                                 (VREAD|VEXEC) >> 6;
 165                 vp->v_type = VDIR;
 166                 vp->v_flag = VROOT;
 167                 break;
 168
 169         case Pcurproc:  /* /proc/curproc = lr--r--r-- */
 170                 pfs->pfs_mode = (VREAD) |
 171                                 (VREAD >> 3) |
 172                                 (VREAD >> 6);
 173                 vp->v_type = VLNK;
 174                 break;
 175
 176         case Pproc:
 177                 pfs->pfs_mode = (VREAD|VEXEC) |
 178                                 (VREAD|VEXEC) >> 3 |
 179                                 (VREAD|VEXEC) >> 6;
 180                 vp->v_type = VDIR;
 181                 break;
 182
 183         case Pfile:
 184                 pfs->pfs_mode = (VREAD|VEXEC) |
 185                                 (VREAD|VEXEC) >> 3 |
 186                                 (VREAD|VEXEC) >> 6;
 187                 vp->v_type = VLNK;
 188                 break;
 189
 190         case Pmem:
 191                 pfs->pfs_mode = (VREAD|VWRITE);
 192                 vp->v_type = VREG;
 193                 break;
 194
 195         case Pregs:
 196         case Pfpregs:
 197         case Pdbregs:
 198                 pfs->pfs_mode = (VREAD|VWRITE);
 199                 vp->v_type = VREG;
 200                 break;
 201
 202         case Pctl:
 203         case Pnote:
 204         case Pnotepg:
 205                 pfs->pfs_mode = (VWRITE);
 206                 vp->v_type = VREG;
 207                 break;
 208
 209         case Ptype:
 210         case Pmap:
 211         case Pstatus:
 212         case Pcmdline:
 213         case Prlimit:
 214                 pfs->pfs_mode = (VREAD) |
 215                                 (VREAD >> 3) |
 216                                 (VREAD >> 6);
 217                 vp->v_type = VREG;
 218                 break;
 219
 220         default:
 221                 panic("procfs_allocvp");
 222         }
 223
 224         /* add to procfs vnode list */
 225         pfs->pfs_next = *pp;
 226         *pp = pfs;
 227
 228 out:
 229         pfsvplock &= ~PROCFS_LOCKED;
 230
 231         if (pfsvplock & PROCFS_WANT) {
 232                 pfsvplock &= ~PROCFS_WANT;
 233                 wakeup((caddr_t) &pfsvplock);
 234         }
 235
 236         return (error);
 237 }
 238
 239 int
 240 procfs_freevp(struct vnode *vp)
 241 {
 242         struct pfsnode **pfspp;
 243         struct pfsnode *pfs;
 244
 245         pfs = VTOPFS(vp);
 246         vp->v_data = NULL;
 247
 248         pfspp = PFSHASH(pfs->pfs_pid);
 249         while (*pfspp != pfs && *pfspp)
 250                 pfspp = &(*pfspp)->pfs_next;
 251         KKASSERT(*pfspp);
 252         *pfspp = pfs->pfs_next;
 253         pfs->pfs_next = NULL;
 254         kfree(pfs, M_TEMP);
 255         return (0);
 256 }
 257
 258 int
 259 procfs_rw(struct vop_read_args *ap)
 260 {
 261         struct vnode *vp = ap->a_vp;
 262         struct uio *uio = ap->a_uio;
 263         struct thread *curtd = uio->uio_td;
 264         struct proc *curp;
 265         struct pfsnode *pfs = VTOPFS(vp);
 266         struct proc *p;
 267         struct lwp *lp;
 268         int rtval;
 269
 270         if (curtd == NULL)
 271                 return (EINVAL);
 272         if ((curp = curtd->td_proc) == NULL)    /* XXX */
 273                 return (EINVAL);
 274
 275         lwkt_gettoken(&proc_token);
 276
 277         p = pfind(pfs->pfs_pid);
 278         if (p == NULL) {
 279                 lwkt_reltoken(&proc_token);
 280                 return (EINVAL);
 281         }
 282         if (p->p_pid == 1 && securelevel > 0 && uio->uio_rw == UIO_WRITE) {
 283                 lwkt_reltoken(&proc_token);
 284                 PRELE(p);
 285                 return (EACCES);
 286         }
 287         /* XXX lwp */
 288         lp = FIRST_LWP_IN_PROC(p);
 289         LWPHOLD(lp);
 290
 291         while (pfs->pfs_lockowner) {
 292                 tsleep(&pfs->pfs_lockowner, 0, "pfslck", 0);
 293         }
 294         pfs->pfs_lockowner = curproc->p_pid;
 295
 296         lwkt_gettoken(&p->p_token);
 297
 298         switch (pfs->pfs_type) {
 299         case Pnote:
 300         case Pnotepg:
 301                 rtval = procfs_donote(curp, lp, pfs, uio);
 302                 break;
 303
 304         case Pregs:
 305                 rtval = procfs_doregs(curp, lp, pfs, uio);
 306                 break;
 307
 308         case Pfpregs:
 309                 rtval = procfs_dofpregs(curp, lp, pfs, uio);
 310                 break;
 311
 312         case Pdbregs:
 313                 rtval = procfs_dodbregs(curp, lp, pfs, uio);
 314                 break;
 315
 316         case Pctl:
 317                 rtval = procfs_doctl(curp, lp, pfs, uio);
 318                 break;
 319
 320         case Pstatus:
 321                 rtval = procfs_dostatus(curp, lp, pfs, uio);
 322                 break;
 323
 324         case Pmap:
 325                 rtval = procfs_domap(curp, lp, pfs, uio);
 326                 break;
 327
 328         case Pmem:
 329                 rtval = procfs_domem(curp, lp, pfs, uio);
 330                 break;
 331
 332         case Ptype:
 333                 rtval = procfs_dotype(curp, lp, pfs, uio);
 334                 break;
 335
 336         case Pcmdline:
 337                 rtval = procfs_docmdline(curp, lp, pfs, uio);
 338                 break;
 339
 340         case Prlimit:
 341                 rtval = procfs_dorlimit(curp, lp, pfs, uio);
 342                 break;
 343
 344         default:
 345                 rtval = EOPNOTSUPP;
 346                 break;
 347         }
 348         lwkt_reltoken(&p->p_token);
 349         LWPRELE(lp);
 350
 351         pfs->pfs_lockowner = 0;
 352         lwkt_reltoken(&proc_token);
 353         wakeup(&pfs->pfs_lockowner);
 354         PRELE(p);
 355
 356         return rtval;
 357 }
 358
 359 /*
 360  * Get a string from userland into (buf).  Strip a trailing
 361  * nl character (to allow easy access from the shell).
 362  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
 363  * will automatically add a nul char at the end.
 364  *
 365  * Returns 0 on success or the following errors
 366  *
 367  * EINVAL:    file offset is non-zero.
 368  * EMSGSIZE:  message is longer than kernel buffer
 369  * EFAULT:    user i/o buffer is not addressable
 370  */
 371 int
 372 vfs_getuserstr(struct uio *uio, char *buf, int *buflenp)
 373 {
 374         int xlen;
 375         int error;
 376
 377         if (uio->uio_offset != 0)
 378                 return (EINVAL);
 379
 380         xlen = *buflenp;
 381
 382         /* must be able to read the whole string in one go */
 383         if (xlen < uio->uio_resid)
 384                 return (EMSGSIZE);
 385         xlen = uio->uio_resid;
 386
 387         if ((error = uiomove(buf, xlen, uio)) != 0)
 388                 return (error);
 389
 390         /* allow multiple writes without seeks */
 391         uio->uio_offset = 0;
 392
 393         /* cleanup string and remove trailing newline */
 394         buf[xlen] = '\0';
 395         xlen = strlen(buf);
 396         if (xlen > 0 && buf[xlen-1] == '\n')
 397                 buf[--xlen] = '\0';
 398         *buflenp = xlen;
 399
 400         return (0);
 401 }
 402
 403 vfs_namemap_t *
 404 vfs_findname(vfs_namemap_t *nm, char *buf, int buflen)
 405 {
 406
 407         for (; nm->nm_name; nm++)
 408                 if (bcmp(buf, nm->nm_name, buflen+1) == 0)
 409                         return (nm);
 410
 411         return (0);
 412 }
 413
 414 void
 415 procfs_exit(struct thread *td)
 416 {
 417         struct pfsnode *pfs;
 418         struct vnode *vp;
 419         pid_t pid;
 420
 421         KKASSERT(td->td_proc);
 422         pid = td->td_proc->p_pid;
 423
 424         /*
 425          * NOTE: We can't just vgone() the vnode any more, not while
 426          *       it may potentially still be active.  This will clean
 427          *       the vp and clear the mount and cause the new VOP subsystem
 428          *       to assert or panic when someone tries to do an operation
 429          *       on an open (exited) procfs descriptor.
 430          *
 431          * Prevent further operations on this pid by setting pfs_pid to -1.
 432          * Note that a pfs_pid of 0 is used for nodes which do not track
 433          * any particular pid.
 434          *
 435          * Use vx_get() to properly ref/lock a vp which may not have any
 436          * refs and which may or may not already be reclaimed.  vx_put()
 437          * will then properly deactivate it and cause it to be recycled.
 438          *
 439          * The hash table can also get ripped out from under us when
 440          * we block so take the easy way out and restart the scan.
 441          */
 442 again:
 443         pfs = *PFSHASH(pid);
 444         while (pfs) {
 445                 if (pfs->pfs_pid == pid) {
 446                         vp = PFSTOV(pfs);
 447                         vx_get(vp);
 448                         pfs->pfs_pid |= PFS_DEAD; /* does not effect hash */
 449                         vx_put(vp);
 450                         goto again;
 451                 }
 452                 pfs = pfs->pfs_next;
 453         }
 454 }
 455