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 */ 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/sysctl.h> 45 #include <sys/proc.h> 46 #include <sys/mount.h> 47 #include <sys/vnode.h> 48 #include <sys/malloc.h> 49 #include <sys/thread2.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 pfsnode *pfs; 87 struct vnode *vp; 88 struct pfsnode **pp; 89 int error; 90 91 pp = PFSHASH(pid); 92 loop: 93 for (pfs = *pp; pfs; pfs = pfs->pfs_next) { 94 if (pfs->pfs_pid == pid && pfs->pfs_type == pfs_type && 95 PFSTOV(pfs)->v_mount == mp) { 96 vp = PFSTOV(pfs); 97 vhold_interlocked(vp); 98 if (vget(vp, LK_EXCLUSIVE)) { 99 vdrop(vp); 100 goto loop; 101 } 102 103 /* 104 * Make sure the vnode is still in the cache after 105 * getting the interlock to avoid racing a free. 106 */ 107 for (pfs = *pp; pfs; pfs = pfs->pfs_next) { 108 if (PFSTOV(pfs) == vp && 109 pfs->pfs_pid == pid && 110 pfs->pfs_type == pfs_type && 111 PFSTOV(pfs)->v_mount == mp) { 112 break; 113 } 114 } 115 vdrop(vp); 116 if (pfs == NULL || PFSTOV(pfs) != vp) { 117 vput(vp); 118 goto loop; 119 120 } 121 KKASSERT(vp->v_data == pfs); 122 *vpp = vp; 123 return (0); 124 } 125 } 126 127 /* 128 * otherwise lock the vp list while we call getnewvnode 129 * since that can block. 130 */ 131 if (pfsvplock & PROCFS_LOCKED) { 132 pfsvplock |= PROCFS_WANT; 133 (void) tsleep((caddr_t) &pfsvplock, 0, "pfsavp", 0); 134 goto loop; 135 } 136 pfsvplock |= PROCFS_LOCKED; 137 138 /* 139 * Do the MALLOC before the getnewvnode since doing so afterward 140 * might cause a bogus v_data pointer to get dereferenced 141 * elsewhere if MALLOC should block. 142 * 143 * XXX this may not matter anymore since getnewvnode now returns 144 * a VX locked vnode. 145 */ 146 pfs = kmalloc(sizeof(struct pfsnode), M_TEMP, M_WAITOK); 147 148 error = getnewvnode(VT_PROCFS, mp, vpp, 0, 0); 149 if (error) { 150 kfree(pfs, M_TEMP); 151 goto out; 152 } 153 vp = *vpp; 154 155 vp->v_data = pfs; 156 157 pfs->pfs_next = 0; 158 pfs->pfs_pid = (pid_t) pid; 159 pfs->pfs_type = pfs_type; 160 pfs->pfs_vnode = vp; 161 pfs->pfs_flags = 0; 162 pfs->pfs_lockowner = 0; 163 pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type); 164 165 switch (pfs_type) { 166 case Proot: /* /proc = dr-xr-xr-x */ 167 pfs->pfs_mode = (VREAD|VEXEC) | 168 (VREAD|VEXEC) >> 3 | 169 (VREAD|VEXEC) >> 6; 170 vp->v_type = VDIR; 171 vp->v_flag = VROOT; 172 break; 173 174 case Pcurproc: /* /proc/curproc = lr--r--r-- */ 175 pfs->pfs_mode = (VREAD) | 176 (VREAD >> 3) | 177 (VREAD >> 6); 178 vp->v_type = VLNK; 179 break; 180 181 case Pproc: 182 pfs->pfs_mode = (VREAD|VEXEC) | 183 (VREAD|VEXEC) >> 3 | 184 (VREAD|VEXEC) >> 6; 185 vp->v_type = VDIR; 186 break; 187 188 case Pfile: 189 pfs->pfs_mode = (VREAD|VEXEC) | 190 (VREAD|VEXEC) >> 3 | 191 (VREAD|VEXEC) >> 6; 192 vp->v_type = VLNK; 193 break; 194 195 case Pmem: 196 pfs->pfs_mode = (VREAD|VWRITE); 197 vp->v_type = VREG; 198 break; 199 200 case Pregs: 201 case Pfpregs: 202 case Pdbregs: 203 pfs->pfs_mode = (VREAD|VWRITE); 204 vp->v_type = VREG; 205 break; 206 207 case Pctl: 208 case Pnote: 209 case Pnotepg: 210 pfs->pfs_mode = (VWRITE); 211 vp->v_type = VREG; 212 break; 213 214 case Ptype: 215 case Pmap: 216 case Pstatus: 217 case Pcmdline: 218 case Prlimit: 219 pfs->pfs_mode = (VREAD) | 220 (VREAD >> 3) | 221 (VREAD >> 6); 222 vp->v_type = VREG; 223 break; 224 225 default: 226 panic("procfs_allocvp"); 227 } 228 229 /* add to procfs vnode list */ 230 pfs->pfs_next = *pp; 231 *pp = pfs; 232 233 out: 234 pfsvplock &= ~PROCFS_LOCKED; 235 236 if (pfsvplock & PROCFS_WANT) { 237 pfsvplock &= ~PROCFS_WANT; 238 wakeup((caddr_t) &pfsvplock); 239 } 240 241 return (error); 242 } 243 244 int 245 procfs_freevp(struct vnode *vp) 246 { 247 struct pfsnode **pfspp; 248 struct pfsnode *pfs; 249 250 pfs = VTOPFS(vp); 251 vp->v_data = NULL; 252 253 pfspp = PFSHASH(pfs->pfs_pid); 254 while (*pfspp != pfs && *pfspp) 255 pfspp = &(*pfspp)->pfs_next; 256 KKASSERT(*pfspp); 257 *pfspp = pfs->pfs_next; 258 pfs->pfs_next = NULL; 259 pfs->pfs_vnode = NULL; 260 kfree(pfs, M_TEMP); 261 return (0); 262 } 263 264 /* 265 * Try to find the calling pid. Note that pfind() 266 * now references the proc structure to be returned 267 * and needs to be released later with PRELE(). 268 */ 269 struct proc * 270 pfs_pfind(pid_t pfs_pid) 271 { 272 struct proc *p = NULL; 273 274 if (pfs_pid == 0) { 275 p = &proc0; 276 PHOLD(p); 277 } else { 278 p = pfind(pfs_pid); 279 } 280 281 /* 282 * Make sure the process is not in the middle of exiting (where 283 * a lot of its structural members may wind up being NULL). If it 284 * is we give up on it. 285 */ 286 if (p) { 287 lwkt_gettoken(&p->p_token); 288 if (p->p_flags & P_WEXIT) { 289 lwkt_reltoken(&p->p_token); 290 PRELE(p); 291 p = NULL; 292 } 293 } 294 return p; 295 } 296 297 void 298 pfs_pdone(struct proc *p) 299 { 300 if (p) { 301 lwkt_reltoken(&p->p_token); 302 PRELE(p); 303 } 304 } 305 306 int 307 procfs_rw(struct vop_read_args *ap) 308 { 309 struct vnode *vp = ap->a_vp; 310 struct uio *uio = ap->a_uio; 311 struct thread *curtd = uio->uio_td; 312 struct proc *curp; 313 struct pfsnode *pfs = VTOPFS(vp); 314 struct proc *p; 315 struct lwp *lp; 316 int rtval; 317 318 if (curtd == NULL) 319 return (EINVAL); 320 if ((curp = curtd->td_proc) == NULL) /* XXX */ 321 return (EINVAL); 322 323 lwkt_gettoken(&proc_token); 324 p = pfs_pfind(pfs->pfs_pid); 325 if (p == NULL) { 326 rtval = (EINVAL); 327 goto out; 328 } 329 if (p->p_pid == 1 && securelevel > 0 && uio->uio_rw == UIO_WRITE) { 330 rtval = (EACCES); 331 goto out; 332 } 333 /* XXX lwp */ 334 lp = FIRST_LWP_IN_PROC(p); 335 LWPHOLD(lp); 336 337 while (pfs->pfs_lockowner) { 338 tsleep(&pfs->pfs_lockowner, 0, "pfslck", 0); 339 } 340 pfs->pfs_lockowner = curproc->p_pid; 341 342 switch (pfs->pfs_type) { 343 case Pnote: 344 case Pnotepg: 345 rtval = procfs_donote(curp, lp, pfs, uio); 346 break; 347 348 case Pregs: 349 rtval = procfs_doregs(curp, lp, pfs, uio); 350 break; 351 352 case Pfpregs: 353 rtval = procfs_dofpregs(curp, lp, pfs, uio); 354 break; 355 356 case Pdbregs: 357 rtval = procfs_dodbregs(curp, lp, pfs, uio); 358 break; 359 360 case Pctl: 361 rtval = procfs_doctl(curp, lp, pfs, uio); 362 break; 363 364 case Pstatus: 365 rtval = procfs_dostatus(curp, lp, pfs, uio); 366 break; 367 368 case Pmap: 369 rtval = procfs_domap(curp, lp, pfs, uio); 370 break; 371 372 case Pmem: 373 rtval = procfs_domem(curp, lp, pfs, uio); 374 break; 375 376 case Ptype: 377 rtval = procfs_dotype(curp, lp, pfs, uio); 378 break; 379 380 case Pcmdline: 381 rtval = procfs_docmdline(curp, lp, pfs, uio); 382 break; 383 384 case Prlimit: 385 rtval = procfs_dorlimit(curp, lp, pfs, uio); 386 break; 387 388 default: 389 rtval = EOPNOTSUPP; 390 break; 391 } 392 LWPRELE(lp); 393 394 pfs->pfs_lockowner = 0; 395 wakeup(&pfs->pfs_lockowner); 396 397 out: 398 pfs_pdone(p); 399 lwkt_reltoken(&proc_token); 400 401 return rtval; 402 } 403 404 /* 405 * Get a string from userland into (buf). Strip a trailing 406 * nl character (to allow easy access from the shell). 407 * The buffer should be *buflenp + 1 chars long. vfs_getuserstr 408 * will automatically add a nul char at the end. 409 * 410 * Returns 0 on success or the following errors 411 * 412 * EINVAL: file offset is non-zero. 413 * EMSGSIZE: message is longer than kernel buffer 414 * EFAULT: user i/o buffer is not addressable 415 */ 416 int 417 vfs_getuserstr(struct uio *uio, char *buf, int *buflenp) 418 { 419 int xlen; 420 int error; 421 422 if (uio->uio_offset != 0) 423 return (EINVAL); 424 425 xlen = *buflenp; 426 427 /* must be able to read the whole string in one go */ 428 if (xlen < uio->uio_resid) 429 return (EMSGSIZE); 430 xlen = uio->uio_resid; 431 432 if ((error = uiomove(buf, xlen, uio)) != 0) 433 return (error); 434 435 /* allow multiple writes without seeks */ 436 uio->uio_offset = 0; 437 438 /* cleanup string and remove trailing newline */ 439 buf[xlen] = '\0'; 440 xlen = strlen(buf); 441 if (xlen > 0 && buf[xlen-1] == '\n') 442 buf[--xlen] = '\0'; 443 *buflenp = xlen; 444 445 return (0); 446 } 447 448 vfs_namemap_t * 449 vfs_findname(vfs_namemap_t *nm, char *buf, int buflen) 450 { 451 452 for (; nm->nm_name; nm++) 453 if (bcmp(buf, nm->nm_name, buflen+1) == 0) 454 return (nm); 455 456 return (0); 457 } 458 459 void 460 procfs_exit(struct thread *td) 461 { 462 struct pfsnode *pfs; 463 struct vnode *vp; 464 pid_t pid; 465 466 KKASSERT(td->td_proc); 467 pid = td->td_proc->p_pid; 468 469 /* 470 * NOTE: We can't just vgone() the vnode any more, not while 471 * it may potentially still be active. This will clean 472 * the vp and clear the mount and cause the new VOP subsystem 473 * to assert or panic when someone tries to do an operation 474 * on an open (exited) procfs descriptor. 475 * 476 * Prevent further operations on this pid by setting pfs_pid to -1. 477 * Note that a pfs_pid of 0 is used for nodes which do not track 478 * any particular pid. 479 * 480 * Use vx_get() to properly ref/lock a vp which may not have any 481 * refs and which may or may not already be reclaimed. vx_put() 482 * will then properly deactivate it and cause it to be recycled. 483 * 484 * The hash table can also get ripped out from under us when 485 * we block so take the easy way out and restart the scan. 486 */ 487 again: 488 pfs = *PFSHASH(pid); 489 while (pfs) { 490 if (pfs->pfs_pid == pid) { 491 vp = PFSTOV(pfs); 492 vx_get(vp); 493 pfs->pfs_pid |= PFS_DEAD; /* does not effect hash */ 494 vx_put(vp); 495 goto again; 496 } 497 pfs = pfs->pfs_next; 498 } 499 } 500 501