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                 pfs->pfs_mode = (VREAD|VEXEC) |
 151                                 (VREAD|VEXEC) >> 3 |
 152                                 (VREAD|VEXEC) >> 6;
 153                 vp->v_type = VDIR;
 154                 vsetflags(vp, VROOT);
 155                 break;
 156
 157         case Pself:     /* /proc/self = lr--r--r-- */
 158                 pfs->pfs_mode = (VREAD) |
 159                                 (VREAD >> 3) |
 160                                 (VREAD >> 6);
 161                 vp->v_type = VLNK;
 162                 break;
 163
 164         case Pproc:
 165                 pfs->pfs_mode = (VREAD|VEXEC) |
 166                                 (VREAD|VEXEC) >> 3 |
 167                                 (VREAD|VEXEC) >> 6;
 168                 vp->v_type = VDIR;
 169                 break;
 170
 171         case Pexe:
 172                 pfs->pfs_mode = (VREAD|VEXEC) |
 173                                 (VREAD|VEXEC) >> 3 |
 174                                 (VREAD|VEXEC) >> 6;
 175                 vp->v_type = VLNK;
 176                 break;
 177
 178         case Pmem:
 179                 pfs->pfs_mode = (VREAD|VWRITE) |
 180                                 (VREAD) >> 3;
 181                 vp->v_type = VREG;
 182                 break;
 183
 184         case Pprocstat:
 185         case Pprocstatus:
 186                 /* fallthrough */
 187
 188         case Pmeminfo:
 189         case Pcpuinfo:
 190         case Pstat:
 191         case Puptime:
 192         case Pversion:
 193         case Ploadavg:
 194                 pfs->pfs_mode = (VREAD) |
 195                                 (VREAD >> 3) |
 196                                 (VREAD >> 6);
 197                 vp->v_type = VREG;
 198                 break;
 199
 200         default:
 201                 panic("linprocfs_allocvp");
 202         }
 203
 204         /* add to procfs vnode list */
 205         for (pp = &pfshead[pid & PFSHMASK]; *pp; pp = &(*pp)->pfs_next)
 206                 continue;
 207         *pp = pfs;
 208
 209         vx_unlock(vp);  /* vnode ready to roll! */
 210
 211 out:
 212         pfsvplock &= ~PROCFS_LOCKED;
 213
 214         if (pfsvplock & PROCFS_WANT) {
 215                 pfsvplock &= ~PROCFS_WANT;
 216                 wakeup((caddr_t) &pfsvplock);
 217         }
 218         lwkt_reltoken(&ilock);
 219
 220         return (error);
 221 }
 222
 223 int
 224 linprocfs_freevp(struct vnode *vp)
 225 {
 226         struct pfsnode **pfspp;
 227         struct pfsnode *pfs = VTOPFS(vp);
 228         lwkt_tokref ilock;
 229
 230         lwkt_gettoken(&ilock, &pfs_token);
 231         pfspp = &pfshead[pfs->pfs_pid & PFSHMASK];
 232         while (*pfspp != pfs) {
 233                 KKASSERT(*pfspp != NULL);
 234                 pfspp = &(*pfspp)->pfs_next;
 235         }
 236         *pfspp = pfs->pfs_next;
 237         lwkt_reltoken(&ilock);
 238         FREE(vp->v_data, M_TEMP);
 239         vp->v_data = NULL;
 240         return (0);
 241 }
 242
 243 int
 244 linprocfs_rw(struct vop_read_args *ap)
 245 {
 246         struct vnode *vp = ap->a_vp;
 247         struct uio *uio = ap->a_uio;
 248         struct thread *td = uio->uio_td;
 249         struct pfsnode *pfs = VTOPFS(vp);
 250         struct proc *p;
 251         struct proc *curp;
 252         struct lwp *lp;
 253         int rtval;
 254
 255         curp = td->td_proc;
 256         KKASSERT(curp);
 257
 258         p = PFIND(pfs->pfs_pid);
 259         if (p == 0)
 260                 return (EINVAL);
 261         if (p->p_pid == 1 && securelevel > 0 && uio->uio_rw == UIO_WRITE)
 262                 return (EACCES);
 263         lp = FIRST_LWP_IN_PROC(p);
 264         LWPHOLD(lp);
 265
 266         while (pfs->pfs_lockowner) {
 267                 tsleep(&pfs->pfs_lockowner, 0, "pfslck", 0);
 268         }
 269         pfs->pfs_lockowner = curthread;
 270
 271         switch (pfs->pfs_type) {
 272         case Pmem:
 273                 rtval = procfs_domem(curp, lp, pfs, uio);
 274                 break;
 275         case Pprocstat:
 276                 rtval = linprocfs_doprocstat(curp, p, pfs, uio);
 277                 break;
 278         case Pprocstatus:
 279                 rtval = linprocfs_doprocstatus(curp, p, pfs, uio);
 280                 break;
 281         case Pmeminfo:
 282                 rtval = linprocfs_domeminfo(curp, p, pfs, uio);
 283                 break;
 284         case Pcpuinfo:
 285                 rtval = linprocfs_docpuinfo(curp, p, pfs, uio);
 286                 break;
 287         case Pstat:
 288                 rtval = linprocfs_dostat(curp, p, pfs, uio);
 289                 break;
 290         case Puptime:
 291                 rtval = linprocfs_douptime(curp, p, pfs, uio);
 292                 break;
 293         case Pversion:
 294                 rtval = linprocfs_doversion(curp, p, pfs, uio);
 295                 break;
 296         case Ploadavg:
 297                 rtval = linprocfs_doloadavg(curp, p, pfs, uio);
 298                 break;
 299         default:
 300                 rtval = EOPNOTSUPP;
 301                 break;
 302         }
 303         LWPRELE(lp);
 304         pfs->pfs_lockowner = NULL;
 305         wakeup(&pfs->pfs_lockowner);
 306         return rtval;
 307 }
 308
 309 #if 0
 310 /*
 311  * Get a string from userland into (buf).  Strip a trailing
 312  * nl character (to allow easy access from the shell).
 313  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
 314  * will automatically add a nul char at the end.
 315  *
 316  * Returns 0 on success or the following errors
 317  *
 318  * EINVAL:    file offset is non-zero.
 319  * EMSGSIZE:  message is longer than kernel buffer
 320  * EFAULT:    user i/o buffer is not addressable
 321  */
 322 int
 323 vfs_getuserstr(struct uio *uio, char *buf, int *buflenp)
 324 {
 325         int xlen;
 326         int error;
 327
 328         if (uio->uio_offset != 0)
 329                 return (EINVAL);
 330
 331         xlen = *buflenp;
 332
 333         /* must be able to read the whole string in one go */
 334         if (xlen < uio->uio_resid)
 335                 return (EMSGSIZE);
 336         xlen = uio->uio_resid;
 337
 338         if ((error = uiomove(buf, xlen, uio)) != 0)
 339                 return (error);
 340
 341         /* allow multiple writes without seeks */
 342         uio->uio_offset = 0;
 343
 344         /* cleanup string and remove trailing newline */
 345         buf[xlen] = '\0';
 346         xlen = strlen(buf);
 347         if (xlen > 0 && buf[xlen-1] == '\n')
 348                 buf[--xlen] = '\0';
 349         *buflenp = xlen;
 350
 351         return (0);
 352 }
 353
 354 vfs_namemap_t *
 355 vfs_findname(vfs_namemap_t *nm, char *buf, int buflen)
 356 {
 357
 358         for (; nm->nm_name; nm++)
 359                 if (bcmp(buf, nm->nm_name, buflen+1) == 0)
 360                         return (nm);
 361
 362         return (0);
 363 }
 364 #endif
 365
 366 void
 367 linprocfs_init(void)
 368 {
 369         lwkt_token_init(&pfs_token);
 370 }
 371
 372 void
 373 linprocfs_exit(struct thread *td)
 374 {
 375         lwkt_tokref ilock;
 376         struct pfsnode *pfs;
 377         struct vnode *vp;
 378         pid_t pid;
 379
 380         KKASSERT(td->td_proc);
 381         pid = td->td_proc->p_pid;
 382
 383         /*
 384          * Remove all the procfs vnodes associated with an exiting process.
 385          */
 386         lwkt_gettoken(&ilock, &pfs_token);
 387 restart:
 388         for (pfs = pfshead[pid & PFSHMASK]; pfs; pfs = pfs->pfs_next) {
 389                 if (pfs->pfs_pid == pid) {
 390                         vp = PFSTOV(pfs);
 391                         vx_get(vp);
 392                         vgone_vxlocked(vp);
 393                         vx_put(vp);
 394                         goto restart;
 395                 }
 396         }
 397         lwkt_reltoken(&ilock);
 398         lwkt_token_uninit(&pfs_token);
 399 }
 400