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 1.2 2003/06/17 04:28:42 dillon 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 <miscfs/procfs/procfs.h> 52 53 static struct pfsnode *pfshead; 54 static int pfsvplock; 55 56 /* 57 * allocate a pfsnode/vnode pair. the vnode is 58 * referenced, but not locked. 59 * 60 * the pid, pfs_type, and mount point uniquely 61 * identify a pfsnode. the mount point is needed 62 * because someone might mount this filesystem 63 * twice. 64 * 65 * all pfsnodes are maintained on a singly-linked 66 * list. new nodes are only allocated when they cannot 67 * be found on this list. entries on the list are 68 * removed when the vfs reclaim entry is called. 69 * 70 * a single lock is kept for the entire list. this is 71 * needed because the getnewvnode() function can block 72 * waiting for a vnode to become free, in which case there 73 * may be more than one process trying to get the same 74 * vnode. this lock is only taken if we are going to 75 * call getnewvnode, since the kernel itself is single-threaded. 76 * 77 * if an entry is found on the list, then call vget() to 78 * take a reference. this is done because there may be 79 * zero references to it and so it needs to removed from 80 * the vnode free list. 81 */ 82 int 83 procfs_allocvp(mp, vpp, pid, pfs_type) 84 struct mount *mp; 85 struct vnode **vpp; 86 long pid; 87 pfstype pfs_type; 88 { 89 struct proc *p = curproc; /* XXX */ 90 struct pfsnode *pfs; 91 struct vnode *vp; 92 struct pfsnode **pp; 93 int error; 94 95 loop: 96 for (pfs = pfshead; pfs != 0; pfs = pfs->pfs_next) { 97 vp = PFSTOV(pfs); 98 if (pfs->pfs_pid == pid && 99 pfs->pfs_type == pfs_type && 100 vp->v_mount == mp) { 101 if (vget(vp, 0, p)) 102 goto loop; 103 *vpp = vp; 104 return (0); 105 } 106 } 107 108 /* 109 * otherwise lock the vp list while we call getnewvnode 110 * since that can block. 111 */ 112 if (pfsvplock & PROCFS_LOCKED) { 113 pfsvplock |= PROCFS_WANT; 114 (void) tsleep((caddr_t) &pfsvplock, PINOD, "pfsavp", 0); 115 goto loop; 116 } 117 pfsvplock |= PROCFS_LOCKED; 118 119 /* 120 * Do the MALLOC before the getnewvnode since doing so afterward 121 * might cause a bogus v_data pointer to get dereferenced 122 * elsewhere if MALLOC should block. 123 */ 124 MALLOC(pfs, struct pfsnode *, sizeof(struct pfsnode), M_TEMP, M_WAITOK); 125 126 if ((error = getnewvnode(VT_PROCFS, mp, procfs_vnodeop_p, vpp)) != 0) { 127 FREE(pfs, M_TEMP); 128 goto out; 129 } 130 vp = *vpp; 131 132 vp->v_data = pfs; 133 134 pfs->pfs_next = 0; 135 pfs->pfs_pid = (pid_t) pid; 136 pfs->pfs_type = pfs_type; 137 pfs->pfs_vnode = vp; 138 pfs->pfs_flags = 0; 139 pfs->pfs_lockowner = 0; 140 pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type); 141 142 switch (pfs_type) { 143 case Proot: /* /proc = dr-xr-xr-x */ 144 pfs->pfs_mode = (VREAD|VEXEC) | 145 (VREAD|VEXEC) >> 3 | 146 (VREAD|VEXEC) >> 6; 147 vp->v_type = VDIR; 148 vp->v_flag = VROOT; 149 break; 150 151 case Pcurproc: /* /proc/curproc = lr--r--r-- */ 152 pfs->pfs_mode = (VREAD) | 153 (VREAD >> 3) | 154 (VREAD >> 6); 155 vp->v_type = VLNK; 156 break; 157 158 case Pproc: 159 pfs->pfs_mode = (VREAD|VEXEC) | 160 (VREAD|VEXEC) >> 3 | 161 (VREAD|VEXEC) >> 6; 162 vp->v_type = VDIR; 163 break; 164 165 case Pfile: 166 pfs->pfs_mode = (VREAD|VEXEC) | 167 (VREAD|VEXEC) >> 3 | 168 (VREAD|VEXEC) >> 6; 169 vp->v_type = VLNK; 170 break; 171 172 case Pmem: 173 pfs->pfs_mode = (VREAD|VWRITE); 174 vp->v_type = VREG; 175 break; 176 177 case Pregs: 178 case Pfpregs: 179 case Pdbregs: 180 pfs->pfs_mode = (VREAD|VWRITE); 181 vp->v_type = VREG; 182 break; 183 184 case Pctl: 185 case Pnote: 186 case Pnotepg: 187 pfs->pfs_mode = (VWRITE); 188 vp->v_type = VREG; 189 break; 190 191 case Ptype: 192 case Pmap: 193 case Pstatus: 194 case Pcmdline: 195 case Prlimit: 196 pfs->pfs_mode = (VREAD) | 197 (VREAD >> 3) | 198 (VREAD >> 6); 199 vp->v_type = VREG; 200 break; 201 202 default: 203 panic("procfs_allocvp"); 204 } 205 206 /* add to procfs vnode list */ 207 for (pp = &pfshead; *pp; pp = &(*pp)->pfs_next) 208 continue; 209 *pp = pfs; 210 211 out: 212 pfsvplock &= ~PROCFS_LOCKED; 213 214 if (pfsvplock & PROCFS_WANT) { 215 pfsvplock &= ~PROCFS_WANT; 216 wakeup((caddr_t) &pfsvplock); 217 } 218 219 return (error); 220 } 221 222 int 223 procfs_freevp(vp) 224 struct vnode *vp; 225 { 226 struct pfsnode **pfspp; 227 struct pfsnode *pfs = VTOPFS(vp); 228 229 for (pfspp = &pfshead; *pfspp != 0; pfspp = &(*pfspp)->pfs_next) { 230 if (*pfspp == pfs) { 231 *pfspp = pfs->pfs_next; 232 break; 233 } 234 } 235 236 FREE(vp->v_data, M_TEMP); 237 vp->v_data = 0; 238 return (0); 239 } 240 241 int 242 procfs_rw(ap) 243 struct vop_read_args *ap; 244 { 245 struct vnode *vp = ap->a_vp; 246 struct uio *uio = ap->a_uio; 247 struct proc *curp = uio->uio_procp; 248 struct pfsnode *pfs = VTOPFS(vp); 249 struct proc *p; 250 int rtval; 251 252 p = PFIND(pfs->pfs_pid); 253 if (p == NULL) 254 return (EINVAL); 255 if (p->p_pid == 1 && securelevel > 0 && uio->uio_rw == UIO_WRITE) 256 return (EACCES); 257 258 while (pfs->pfs_lockowner) { 259 tsleep(&pfs->pfs_lockowner, PRIBIO, "pfslck", 0); 260 } 261 pfs->pfs_lockowner = curproc->p_pid; 262 263 switch (pfs->pfs_type) { 264 case Pnote: 265 case Pnotepg: 266 rtval = procfs_donote(curp, p, pfs, uio); 267 break; 268 269 case Pregs: 270 rtval = procfs_doregs(curp, p, pfs, uio); 271 break; 272 273 case Pfpregs: 274 rtval = procfs_dofpregs(curp, p, pfs, uio); 275 break; 276 277 case Pdbregs: 278 rtval = procfs_dodbregs(curp, p, pfs, uio); 279 break; 280 281 case Pctl: 282 rtval = procfs_doctl(curp, p, pfs, uio); 283 break; 284 285 case Pstatus: 286 rtval = procfs_dostatus(curp, p, pfs, uio); 287 break; 288 289 case Pmap: 290 rtval = procfs_domap(curp, p, pfs, uio); 291 break; 292 293 case Pmem: 294 rtval = procfs_domem(curp, p, pfs, uio); 295 break; 296 297 case Ptype: 298 rtval = procfs_dotype(curp, p, pfs, uio); 299 break; 300 301 case Pcmdline: 302 rtval = procfs_docmdline(curp, p, pfs, uio); 303 break; 304 305 case Prlimit: 306 rtval = procfs_dorlimit(curp, p, pfs, uio); 307 break; 308 309 default: 310 rtval = EOPNOTSUPP; 311 break; 312 } 313 pfs->pfs_lockowner = 0; 314 wakeup(&pfs->pfs_lockowner); 315 return rtval; 316 } 317 318 /* 319 * Get a string from userland into (buf). Strip a trailing 320 * nl character (to allow easy access from the shell). 321 * The buffer should be *buflenp + 1 chars long. vfs_getuserstr 322 * will automatically add a nul char at the end. 323 * 324 * Returns 0 on success or the following errors 325 * 326 * EINVAL: file offset is non-zero. 327 * EMSGSIZE: message is longer than kernel buffer 328 * EFAULT: user i/o buffer is not addressable 329 */ 330 int 331 vfs_getuserstr(uio, buf, buflenp) 332 struct uio *uio; 333 char *buf; 334 int *buflenp; 335 { 336 int xlen; 337 int error; 338 339 if (uio->uio_offset != 0) 340 return (EINVAL); 341 342 xlen = *buflenp; 343 344 /* must be able to read the whole string in one go */ 345 if (xlen < uio->uio_resid) 346 return (EMSGSIZE); 347 xlen = uio->uio_resid; 348 349 if ((error = uiomove(buf, xlen, uio)) != 0) 350 return (error); 351 352 /* allow multiple writes without seeks */ 353 uio->uio_offset = 0; 354 355 /* cleanup string and remove trailing newline */ 356 buf[xlen] = '\0'; 357 xlen = strlen(buf); 358 if (xlen > 0 && buf[xlen-1] == '\n') 359 buf[--xlen] = '\0'; 360 *buflenp = xlen; 361 362 return (0); 363 } 364 365 vfs_namemap_t * 366 vfs_findname(nm, buf, buflen) 367 vfs_namemap_t *nm; 368 char *buf; 369 int buflen; 370 { 371 372 for (; nm->nm_name; nm++) 373 if (bcmp(buf, nm->nm_name, buflen+1) == 0) 374 return (nm); 375 376 return (0); 377 } 378 379 void 380 procfs_exit(struct proc *p) 381 { 382 struct pfsnode *pfs; 383 pid_t pid = p->p_pid; 384 385 /* 386 * The reason for this loop is not obvious -- basicly, 387 * procfs_freevp(), which is called via vgone() (eventually), 388 * removes the specified procfs node from the pfshead list. 389 * It does this by *pfsp = pfs->pfs_next, meaning that it 390 * overwrites the node. So when we do pfs = pfs->next, we 391 * end up skipping the node that replaces the one that was 392 * vgone'd. Since it may have been the last one on the list, 393 * it may also have been set to null -- but *our* pfs pointer, 394 * here, doesn't see this. So the loop starts from the beginning 395 * again. 396 * 397 * This is not a for() loop because the final event 398 * would be "pfs = pfs->pfs_next"; in the case where 399 * pfs is set to pfshead again, that would mean that 400 * pfshead is skipped over. 401 * 402 */ 403 pfs = pfshead; 404 while (pfs) { 405 if (pfs->pfs_pid == pid) { 406 vgone(PFSTOV(pfs)); 407 pfs = pfshead; 408 } else 409 pfs = pfs->pfs_next; 410 } 411 } 412