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 1.9 2004/08/17 18:57:35 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 <vfs/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(struct mount *mp, struct vnode **vpp, long pid, pfstype pfs_type) 84 { 85 struct thread *td = curthread; /* XXX */ 86 struct pfsnode *pfs; 87 struct vnode *vp; 88 struct pfsnode **pp; 89 int error; 90 91 loop: 92 for (pfs = pfshead; pfs != 0; pfs = pfs->pfs_next) { 93 vp = PFSTOV(pfs); 94 if (pfs->pfs_pid == pid && 95 pfs->pfs_type == pfs_type && 96 vp->v_mount == mp) { 97 if (vget(vp, NULL, 0, td)) 98 goto loop; 99 *vpp = vp; 100 return (0); 101 } 102 } 103 104 /* 105 * otherwise lock the vp list while we call getnewvnode 106 * since that can block. 107 */ 108 if (pfsvplock & PROCFS_LOCKED) { 109 pfsvplock |= PROCFS_WANT; 110 (void) tsleep((caddr_t) &pfsvplock, 0, "pfsavp", 0); 111 goto loop; 112 } 113 pfsvplock |= PROCFS_LOCKED; 114 115 /* 116 * Do the MALLOC before the getnewvnode since doing so afterward 117 * might cause a bogus v_data pointer to get dereferenced 118 * elsewhere if MALLOC should block. 119 */ 120 MALLOC(pfs, struct pfsnode *, sizeof(struct pfsnode), M_TEMP, M_WAITOK); 121 122 if ((error = getnewvnode(VT_PROCFS, mp, mp->mnt_vn_ops, vpp)) != 0) { 123 FREE(pfs, M_TEMP); 124 goto out; 125 } 126 vp = *vpp; 127 128 vp->v_data = pfs; 129 130 pfs->pfs_next = 0; 131 pfs->pfs_pid = (pid_t) pid; 132 pfs->pfs_type = pfs_type; 133 pfs->pfs_vnode = vp; 134 pfs->pfs_flags = 0; 135 pfs->pfs_lockowner = 0; 136 pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type); 137 138 switch (pfs_type) { 139 case Proot: /* /proc = dr-xr-xr-x */ 140 pfs->pfs_mode = (VREAD|VEXEC) | 141 (VREAD|VEXEC) >> 3 | 142 (VREAD|VEXEC) >> 6; 143 vp->v_type = VDIR; 144 vp->v_flag = VROOT; 145 break; 146 147 case Pcurproc: /* /proc/curproc = lr--r--r-- */ 148 pfs->pfs_mode = (VREAD) | 149 (VREAD >> 3) | 150 (VREAD >> 6); 151 vp->v_type = VLNK; 152 break; 153 154 case Pproc: 155 pfs->pfs_mode = (VREAD|VEXEC) | 156 (VREAD|VEXEC) >> 3 | 157 (VREAD|VEXEC) >> 6; 158 vp->v_type = VDIR; 159 break; 160 161 case Pfile: 162 pfs->pfs_mode = (VREAD|VEXEC) | 163 (VREAD|VEXEC) >> 3 | 164 (VREAD|VEXEC) >> 6; 165 vp->v_type = VLNK; 166 break; 167 168 case Pmem: 169 pfs->pfs_mode = (VREAD|VWRITE); 170 vp->v_type = VREG; 171 break; 172 173 case Pregs: 174 case Pfpregs: 175 case Pdbregs: 176 pfs->pfs_mode = (VREAD|VWRITE); 177 vp->v_type = VREG; 178 break; 179 180 case Pctl: 181 case Pnote: 182 case Pnotepg: 183 pfs->pfs_mode = (VWRITE); 184 vp->v_type = VREG; 185 break; 186 187 case Ptype: 188 case Pmap: 189 case Pstatus: 190 case Pcmdline: 191 case Prlimit: 192 pfs->pfs_mode = (VREAD) | 193 (VREAD >> 3) | 194 (VREAD >> 6); 195 vp->v_type = VREG; 196 break; 197 198 default: 199 panic("procfs_allocvp"); 200 } 201 202 /* add to procfs vnode list */ 203 for (pp = &pfshead; *pp; pp = &(*pp)->pfs_next) 204 continue; 205 *pp = pfs; 206 207 out: 208 pfsvplock &= ~PROCFS_LOCKED; 209 210 if (pfsvplock & PROCFS_WANT) { 211 pfsvplock &= ~PROCFS_WANT; 212 wakeup((caddr_t) &pfsvplock); 213 } 214 215 return (error); 216 } 217 218 int 219 procfs_freevp(struct vnode *vp) 220 { 221 struct pfsnode **pfspp; 222 struct pfsnode *pfs = VTOPFS(vp); 223 224 for (pfspp = &pfshead; *pfspp != 0; pfspp = &(*pfspp)->pfs_next) { 225 if (*pfspp == pfs) { 226 *pfspp = pfs->pfs_next; 227 break; 228 } 229 } 230 231 FREE(vp->v_data, M_TEMP); 232 vp->v_data = 0; 233 return (0); 234 } 235 236 int 237 procfs_rw(struct vop_read_args *ap) 238 { 239 struct vnode *vp = ap->a_vp; 240 struct uio *uio = ap->a_uio; 241 struct thread *curtd = uio->uio_td; 242 struct proc *curp; 243 struct pfsnode *pfs = VTOPFS(vp); 244 struct proc *p; 245 int rtval; 246 247 if (curtd == NULL) 248 return (EINVAL); 249 if ((curp = curtd->td_proc) == NULL) /* XXX */ 250 return (EINVAL); 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, 0, "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(struct uio *uio, char *buf, int *buflenp) 332 { 333 int xlen; 334 int error; 335 336 if (uio->uio_offset != 0) 337 return (EINVAL); 338 339 xlen = *buflenp; 340 341 /* must be able to read the whole string in one go */ 342 if (xlen < uio->uio_resid) 343 return (EMSGSIZE); 344 xlen = uio->uio_resid; 345 346 if ((error = uiomove(buf, xlen, uio)) != 0) 347 return (error); 348 349 /* allow multiple writes without seeks */ 350 uio->uio_offset = 0; 351 352 /* cleanup string and remove trailing newline */ 353 buf[xlen] = '\0'; 354 xlen = strlen(buf); 355 if (xlen > 0 && buf[xlen-1] == '\n') 356 buf[--xlen] = '\0'; 357 *buflenp = xlen; 358 359 return (0); 360 } 361 362 vfs_namemap_t * 363 vfs_findname(vfs_namemap_t *nm, char *buf, int buflen) 364 { 365 366 for (; nm->nm_name; nm++) 367 if (bcmp(buf, nm->nm_name, buflen+1) == 0) 368 return (nm); 369 370 return (0); 371 } 372 373 void 374 procfs_exit(struct thread *td) 375 { 376 struct pfsnode *pfs; 377 pid_t pid; 378 379 KKASSERT(td->td_proc); 380 pid = td->td_proc->p_pid; 381 382 /* 383 * The reason for this loop is not obvious -- basicly, 384 * procfs_freevp(), which is called via vgone() (eventually), 385 * removes the specified procfs node from the pfshead list. 386 * It does this by *pfsp = pfs->pfs_next, meaning that it 387 * overwrites the node. So when we do pfs = pfs->next, we 388 * end up skipping the node that replaces the one that was 389 * vgone'd. Since it may have been the last one on the list, 390 * it may also have been set to null -- but *our* pfs pointer, 391 * here, doesn't see this. So the loop starts from the beginning 392 * again. 393 * 394 * This is not a for() loop because the final event 395 * would be "pfs = pfs->pfs_next"; in the case where 396 * pfs is set to pfshead again, that would mean that 397 * pfshead is skipped over. 398 * 399 */ 400 pfs = pfshead; 401 while (pfs) { 402 if (pfs->pfs_pid == pid) { 403 vgone(PFSTOV(pfs)); 404 pfs = pfshead; 405 } else 406 pfs = pfs->pfs_next; 407 } 408 } 409