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 1.12 2004/12/17 00:18:34 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 #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 thread *td = curthread; /* XXX */ 87 struct pfsnode *pfs; 88 struct vnode *vp; 89 struct pfsnode **pp; 90 int error; 91 92 pp = PFSHASH(pid); 93 loop: 94 for (pfs = *pp; pfs; pfs = pfs->pfs_next) { 95 if (pfs->pfs_pid == pid && pfs->pfs_type == pfs_type && 96 PFSTOV(pfs)->v_mount == mp) { 97 vp = PFSTOV(pfs); 98 if (vget(vp, LK_EXCLUSIVE, td)) 99 goto loop; 100 101 /* 102 * Make sure the vnode is still in the cache after 103 * getting the interlock to avoid racing a free. 104 */ 105 for (pfs = *pp; pfs; pfs = pfs->pfs_next) { 106 if (PFSTOV(pfs) == vp && 107 pfs->pfs_pid == pid && 108 pfs->pfs_type == pfs_type && 109 PFSTOV(pfs)->v_mount == mp) { 110 break; 111 } 112 } 113 if (pfs == NULL || PFSTOV(pfs) != vp) { 114 vput(vp); 115 goto loop; 116 117 } 118 *vpp = vp; 119 return (0); 120 } 121 } 122 123 /* 124 * otherwise lock the vp list while we call getnewvnode 125 * since that can block. 126 */ 127 if (pfsvplock & PROCFS_LOCKED) { 128 pfsvplock |= PROCFS_WANT; 129 (void) tsleep((caddr_t) &pfsvplock, 0, "pfsavp", 0); 130 goto loop; 131 } 132 pfsvplock |= PROCFS_LOCKED; 133 134 /* 135 * Do the MALLOC before the getnewvnode since doing so afterward 136 * might cause a bogus v_data pointer to get dereferenced 137 * elsewhere if MALLOC should block. 138 * 139 * XXX this may not matter anymore since getnewvnode now returns 140 * a VX locked vnode. 141 */ 142 MALLOC(pfs, struct pfsnode *, sizeof(struct pfsnode), M_TEMP, M_WAITOK); 143 144 error = getnewvnode(VT_PROCFS, mp, vpp, 0, 0); 145 if (error) { 146 free(pfs, M_TEMP); 147 goto out; 148 } 149 vp = *vpp; 150 151 vp->v_data = pfs; 152 153 pfs->pfs_next = 0; 154 pfs->pfs_pid = (pid_t) pid; 155 pfs->pfs_type = pfs_type; 156 pfs->pfs_vnode = vp; 157 pfs->pfs_flags = 0; 158 pfs->pfs_lockowner = 0; 159 pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type); 160 161 switch (pfs_type) { 162 case Proot: /* /proc = dr-xr-xr-x */ 163 pfs->pfs_mode = (VREAD|VEXEC) | 164 (VREAD|VEXEC) >> 3 | 165 (VREAD|VEXEC) >> 6; 166 vp->v_type = VDIR; 167 vp->v_flag = VROOT; 168 break; 169 170 case Pcurproc: /* /proc/curproc = lr--r--r-- */ 171 pfs->pfs_mode = (VREAD) | 172 (VREAD >> 3) | 173 (VREAD >> 6); 174 vp->v_type = VLNK; 175 break; 176 177 case Pproc: 178 pfs->pfs_mode = (VREAD|VEXEC) | 179 (VREAD|VEXEC) >> 3 | 180 (VREAD|VEXEC) >> 6; 181 vp->v_type = VDIR; 182 break; 183 184 case Pfile: 185 pfs->pfs_mode = (VREAD|VEXEC) | 186 (VREAD|VEXEC) >> 3 | 187 (VREAD|VEXEC) >> 6; 188 vp->v_type = VLNK; 189 break; 190 191 case Pmem: 192 pfs->pfs_mode = (VREAD|VWRITE); 193 vp->v_type = VREG; 194 break; 195 196 case Pregs: 197 case Pfpregs: 198 case Pdbregs: 199 pfs->pfs_mode = (VREAD|VWRITE); 200 vp->v_type = VREG; 201 break; 202 203 case Pctl: 204 case Pnote: 205 case Pnotepg: 206 pfs->pfs_mode = (VWRITE); 207 vp->v_type = VREG; 208 break; 209 210 case Ptype: 211 case Pmap: 212 case Pstatus: 213 case Pcmdline: 214 case Prlimit: 215 pfs->pfs_mode = (VREAD) | 216 (VREAD >> 3) | 217 (VREAD >> 6); 218 vp->v_type = VREG; 219 break; 220 221 default: 222 panic("procfs_allocvp"); 223 } 224 225 /* add to procfs vnode list */ 226 pfs->pfs_next = *pp; 227 *pp = pfs; 228 229 out: 230 pfsvplock &= ~PROCFS_LOCKED; 231 232 if (pfsvplock & PROCFS_WANT) { 233 pfsvplock &= ~PROCFS_WANT; 234 wakeup((caddr_t) &pfsvplock); 235 } 236 237 return (error); 238 } 239 240 int 241 procfs_freevp(struct vnode *vp) 242 { 243 struct pfsnode **pfspp; 244 struct pfsnode *pfs; 245 246 pfs = VTOPFS(vp); 247 vp->v_data = NULL; 248 249 pfspp = PFSHASH(pfs->pfs_pid); 250 while (*pfspp != pfs && *pfspp) 251 pfspp = &(*pfspp)->pfs_next; 252 KKASSERT(*pfspp); 253 *pfspp = pfs->pfs_next; 254 pfs->pfs_next = NULL; 255 free(pfs, M_TEMP); 256 return (0); 257 } 258 259 int 260 procfs_rw(struct vop_read_args *ap) 261 { 262 struct vnode *vp = ap->a_vp; 263 struct uio *uio = ap->a_uio; 264 struct thread *curtd = uio->uio_td; 265 struct proc *curp; 266 struct pfsnode *pfs = VTOPFS(vp); 267 struct proc *p; 268 int rtval; 269 270 if (curtd == NULL) 271 return (EINVAL); 272 if ((curp = curtd->td_proc) == NULL) /* XXX */ 273 return (EINVAL); 274 275 p = PFIND(pfs->pfs_pid); 276 if (p == NULL) 277 return (EINVAL); 278 if (p->p_pid == 1 && securelevel > 0 && uio->uio_rw == UIO_WRITE) 279 return (EACCES); 280 281 while (pfs->pfs_lockowner) { 282 tsleep(&pfs->pfs_lockowner, 0, "pfslck", 0); 283 } 284 pfs->pfs_lockowner = curproc->p_pid; 285 286 switch (pfs->pfs_type) { 287 case Pnote: 288 case Pnotepg: 289 rtval = procfs_donote(curp, p, pfs, uio); 290 break; 291 292 case Pregs: 293 rtval = procfs_doregs(curp, p, pfs, uio); 294 break; 295 296 case Pfpregs: 297 rtval = procfs_dofpregs(curp, p, pfs, uio); 298 break; 299 300 case Pdbregs: 301 rtval = procfs_dodbregs(curp, p, pfs, uio); 302 break; 303 304 case Pctl: 305 rtval = procfs_doctl(curp, p, pfs, uio); 306 break; 307 308 case Pstatus: 309 rtval = procfs_dostatus(curp, p, pfs, uio); 310 break; 311 312 case Pmap: 313 rtval = procfs_domap(curp, p, pfs, uio); 314 break; 315 316 case Pmem: 317 rtval = procfs_domem(curp, p, pfs, uio); 318 break; 319 320 case Ptype: 321 rtval = procfs_dotype(curp, p, pfs, uio); 322 break; 323 324 case Pcmdline: 325 rtval = procfs_docmdline(curp, p, pfs, uio); 326 break; 327 328 case Prlimit: 329 rtval = procfs_dorlimit(curp, p, pfs, uio); 330 break; 331 332 default: 333 rtval = EOPNOTSUPP; 334 break; 335 } 336 pfs->pfs_lockowner = 0; 337 wakeup(&pfs->pfs_lockowner); 338 return rtval; 339 } 340 341 /* 342 * Get a string from userland into (buf). Strip a trailing 343 * nl character (to allow easy access from the shell). 344 * The buffer should be *buflenp + 1 chars long. vfs_getuserstr 345 * will automatically add a nul char at the end. 346 * 347 * Returns 0 on success or the following errors 348 * 349 * EINVAL: file offset is non-zero. 350 * EMSGSIZE: message is longer than kernel buffer 351 * EFAULT: user i/o buffer is not addressable 352 */ 353 int 354 vfs_getuserstr(struct uio *uio, char *buf, int *buflenp) 355 { 356 int xlen; 357 int error; 358 359 if (uio->uio_offset != 0) 360 return (EINVAL); 361 362 xlen = *buflenp; 363 364 /* must be able to read the whole string in one go */ 365 if (xlen < uio->uio_resid) 366 return (EMSGSIZE); 367 xlen = uio->uio_resid; 368 369 if ((error = uiomove(buf, xlen, uio)) != 0) 370 return (error); 371 372 /* allow multiple writes without seeks */ 373 uio->uio_offset = 0; 374 375 /* cleanup string and remove trailing newline */ 376 buf[xlen] = '\0'; 377 xlen = strlen(buf); 378 if (xlen > 0 && buf[xlen-1] == '\n') 379 buf[--xlen] = '\0'; 380 *buflenp = xlen; 381 382 return (0); 383 } 384 385 vfs_namemap_t * 386 vfs_findname(vfs_namemap_t *nm, char *buf, int buflen) 387 { 388 389 for (; nm->nm_name; nm++) 390 if (bcmp(buf, nm->nm_name, buflen+1) == 0) 391 return (nm); 392 393 return (0); 394 } 395 396 void 397 procfs_exit(struct thread *td) 398 { 399 struct pfsnode *pfs; 400 struct vnode *vp; 401 pid_t pid; 402 403 KKASSERT(td->td_proc); 404 pid = td->td_proc->p_pid; 405 406 /* 407 * The reason for this loop is not obvious -- basicly, 408 * procfs_freevp(), which is called via vgone() (eventually), 409 * removes the specified procfs node from the pfshead list. 410 * It does this by *pfsp = pfs->pfs_next, meaning that it 411 * overwrites the node. So when we do pfs = pfs->next, we 412 * end up skipping the node that replaces the one that was 413 * vgone'd. Since it may have been the last one on the list, 414 * it may also have been set to null -- but *our* pfs pointer, 415 * here, doesn't see this. So the loop starts from the beginning 416 * again. 417 * 418 * This is not a for() loop because the final event 419 * would be "pfs = pfs->pfs_next"; in the case where 420 * pfs is set to pfshead again, that would mean that 421 * pfshead is skipped over. 422 * 423 */ 424 again: 425 pfs = *PFSHASH(pid); 426 while (pfs) { 427 if (pfs->pfs_pid == pid) { 428 vp = PFSTOV(pfs); 429 if (vx_lock(vp) == 0) { 430 vgone(vp); 431 vx_unlock(vp); 432 } 433 goto again; 434 } 435 pfs = pfs->pfs_next; 436 } 437 } 438 439