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 #include <sys/thread2.h> 51 52 #include <vfs/procfs/procfs.h> 53 54 #define PFS_HSIZE 256 55 #define PFS_HMASK (PFS_HSIZE - 1) 56 57 static struct pfsnode *pfshead[PFS_HSIZE]; 58 static int pfsvplock; 59 60 #define PFSHASH(pid) &pfshead[(pid) & PFS_HMASK] 61 62 /* 63 * Allocate a pfsnode/vnode pair. If no error occurs the returned vnode 64 * will be referenced and exclusively locked. 65 * 66 * The pid, pfs_type, and mount point uniquely identify a pfsnode. 67 * The mount point is needed because someone might mount this filesystem 68 * twice. 69 * 70 * All pfsnodes are maintained on a singly-linked list. new nodes are 71 * only allocated when they cannot be found on this list. entries on 72 * the list are removed when the vfs reclaim entry is called. 73 * 74 * A single lock is kept for the entire list. this is needed because the 75 * getnewvnode() function can block waiting for a vnode to become free, 76 * in which case there may be more than one process trying to get the same 77 * vnode. this lock is only taken if we are going to call getnewvnode, 78 * since the kernel itself is single-threaded. 79 * 80 * If an entry is found on the list, then call vget() to take a reference 81 * and obtain the lock. This will properly re-reference the vnode if it 82 * had gotten onto the free list. 83 */ 84 int 85 procfs_allocvp(struct mount *mp, struct vnode **vpp, long pid, pfstype pfs_type) 86 { 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)) 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 kfree(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 kfree(pfs, M_TEMP); 256 return (0); 257 } 258 259 /* 260 * Try to find the calling pid. Note that pfind() 261 * now references the proc structure to be returned 262 * and needs to be released later with PRELE(). 263 */ 264 struct proc * 265 pfs_pfind(pid_t pfs_pid) 266 { 267 struct proc *p = NULL; 268 269 if (pfs_pid == 0) { 270 p = &proc0; 271 PHOLD(p); 272 } else { 273 p = pfind(pfs_pid); 274 } 275 276 return p; 277 } 278 279 int 280 procfs_rw(struct vop_read_args *ap) 281 { 282 struct vnode *vp = ap->a_vp; 283 struct uio *uio = ap->a_uio; 284 struct thread *curtd = uio->uio_td; 285 struct proc *curp; 286 struct pfsnode *pfs = VTOPFS(vp); 287 struct proc *p; 288 struct lwp *lp; 289 int rtval; 290 291 if (curtd == NULL) 292 return (EINVAL); 293 if ((curp = curtd->td_proc) == NULL) /* XXX */ 294 return (EINVAL); 295 296 lwkt_gettoken(&proc_token); 297 298 p = pfs_pfind(pfs->pfs_pid); 299 if (p == NULL) { 300 rtval = (EINVAL); 301 goto out; 302 } 303 if (p->p_pid == 1 && securelevel > 0 && uio->uio_rw == UIO_WRITE) { 304 rtval = (EACCES); 305 goto out; 306 } 307 /* XXX lwp */ 308 lp = FIRST_LWP_IN_PROC(p); 309 LWPHOLD(lp); 310 311 while (pfs->pfs_lockowner) { 312 tsleep(&pfs->pfs_lockowner, 0, "pfslck", 0); 313 } 314 pfs->pfs_lockowner = curproc->p_pid; 315 316 lwkt_gettoken(&p->p_token); 317 318 switch (pfs->pfs_type) { 319 case Pnote: 320 case Pnotepg: 321 rtval = procfs_donote(curp, lp, pfs, uio); 322 break; 323 324 case Pregs: 325 rtval = procfs_doregs(curp, lp, pfs, uio); 326 break; 327 328 case Pfpregs: 329 rtval = procfs_dofpregs(curp, lp, pfs, uio); 330 break; 331 332 case Pdbregs: 333 rtval = procfs_dodbregs(curp, lp, pfs, uio); 334 break; 335 336 case Pctl: 337 rtval = procfs_doctl(curp, lp, pfs, uio); 338 break; 339 340 case Pstatus: 341 rtval = procfs_dostatus(curp, lp, pfs, uio); 342 break; 343 344 case Pmap: 345 rtval = procfs_domap(curp, lp, pfs, uio); 346 break; 347 348 case Pmem: 349 rtval = procfs_domem(curp, lp, pfs, uio); 350 break; 351 352 case Ptype: 353 rtval = procfs_dotype(curp, lp, pfs, uio); 354 break; 355 356 case Pcmdline: 357 rtval = procfs_docmdline(curp, lp, pfs, uio); 358 break; 359 360 case Prlimit: 361 rtval = procfs_dorlimit(curp, lp, pfs, uio); 362 break; 363 364 default: 365 rtval = EOPNOTSUPP; 366 break; 367 } 368 lwkt_reltoken(&p->p_token); 369 LWPRELE(lp); 370 371 pfs->pfs_lockowner = 0; 372 lwkt_reltoken(&proc_token); 373 wakeup(&pfs->pfs_lockowner); 374 375 out: 376 if (LWKT_TOKEN_HELD(&proc_token)) 377 lwkt_reltoken(&proc_token); 378 if (p) 379 PRELE(p); 380 381 return rtval; 382 } 383 384 /* 385 * Get a string from userland into (buf). Strip a trailing 386 * nl character (to allow easy access from the shell). 387 * The buffer should be *buflenp + 1 chars long. vfs_getuserstr 388 * will automatically add a nul char at the end. 389 * 390 * Returns 0 on success or the following errors 391 * 392 * EINVAL: file offset is non-zero. 393 * EMSGSIZE: message is longer than kernel buffer 394 * EFAULT: user i/o buffer is not addressable 395 */ 396 int 397 vfs_getuserstr(struct uio *uio, char *buf, int *buflenp) 398 { 399 int xlen; 400 int error; 401 402 if (uio->uio_offset != 0) 403 return (EINVAL); 404 405 xlen = *buflenp; 406 407 /* must be able to read the whole string in one go */ 408 if (xlen < uio->uio_resid) 409 return (EMSGSIZE); 410 xlen = uio->uio_resid; 411 412 if ((error = uiomove(buf, xlen, uio)) != 0) 413 return (error); 414 415 /* allow multiple writes without seeks */ 416 uio->uio_offset = 0; 417 418 /* cleanup string and remove trailing newline */ 419 buf[xlen] = '\0'; 420 xlen = strlen(buf); 421 if (xlen > 0 && buf[xlen-1] == '\n') 422 buf[--xlen] = '\0'; 423 *buflenp = xlen; 424 425 return (0); 426 } 427 428 vfs_namemap_t * 429 vfs_findname(vfs_namemap_t *nm, char *buf, int buflen) 430 { 431 432 for (; nm->nm_name; nm++) 433 if (bcmp(buf, nm->nm_name, buflen+1) == 0) 434 return (nm); 435 436 return (0); 437 } 438 439 void 440 procfs_exit(struct thread *td) 441 { 442 struct pfsnode *pfs; 443 struct vnode *vp; 444 pid_t pid; 445 446 KKASSERT(td->td_proc); 447 pid = td->td_proc->p_pid; 448 449 /* 450 * NOTE: We can't just vgone() the vnode any more, not while 451 * it may potentially still be active. This will clean 452 * the vp and clear the mount and cause the new VOP subsystem 453 * to assert or panic when someone tries to do an operation 454 * on an open (exited) procfs descriptor. 455 * 456 * Prevent further operations on this pid by setting pfs_pid to -1. 457 * Note that a pfs_pid of 0 is used for nodes which do not track 458 * any particular pid. 459 * 460 * Use vx_get() to properly ref/lock a vp which may not have any 461 * refs and which may or may not already be reclaimed. vx_put() 462 * will then properly deactivate it and cause it to be recycled. 463 * 464 * The hash table can also get ripped out from under us when 465 * we block so take the easy way out and restart the scan. 466 */ 467 again: 468 pfs = *PFSHASH(pid); 469 while (pfs) { 470 if (pfs->pfs_pid == pid) { 471 vp = PFSTOV(pfs); 472 vx_get(vp); 473 pfs->pfs_pid |= PFS_DEAD; /* does not effect hash */ 474 vx_put(vp); 475 goto again; 476 } 477 pfs = pfs->pfs_next; 478 } 479 } 480 481