2 * Copyright (c) 1993 Jan-Simon Pendry
4 * The Regents of the University of California. All rights reserved.
6 * This code is derived from software contributed to Berkeley by
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
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. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)procfs_subr.c 8.6 (Berkeley) 5/14/95
35 * $FreeBSD: src/sys/miscfs/procfs/procfs_subr.c,v 1.26.2.3 2002/02/18 21:28:04 des Exp $
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/sysctl.h>
42 #include <sys/mount.h>
43 #include <sys/vnode.h>
44 #include <sys/malloc.h>
45 #include <sys/thread2.h>
47 #include <vfs/procfs/procfs.h>
50 #define PFS_HMASK (PFS_HSIZE - 1)
52 static struct pfsnode *pfshead[PFS_HSIZE];
55 #define PFSHASH(pid) &pfshead[(pid) & PFS_HMASK]
58 * Allocate a pfsnode/vnode pair. If no error occurs the returned vnode
59 * will be referenced and exclusively locked.
61 * The pid, pfs_type, and mount point uniquely identify a pfsnode.
62 * The mount point is needed because someone might mount this filesystem
65 * All pfsnodes are maintained on a singly-linked list. new nodes are
66 * only allocated when they cannot be found on this list. entries on
67 * the list are removed when the vfs reclaim entry is called.
69 * A single lock is kept for the entire list. this is needed because the
70 * getnewvnode() function can block waiting for a vnode to become free,
71 * in which case there may be more than one process trying to get the same
72 * vnode. this lock is only taken if we are going to call getnewvnode,
73 * since the kernel itself is single-threaded.
75 * If an entry is found on the list, then call vget() to take a reference
76 * and obtain the lock. This will properly re-reference the vnode if it
77 * had gotten onto the free list.
80 procfs_allocvp(struct mount *mp, struct vnode **vpp, long pid, pfstype pfs_type)
89 for (pfs = *pp; pfs; pfs = pfs->pfs_next) {
90 if (pfs->pfs_pid == pid && pfs->pfs_type == pfs_type &&
91 PFSTOV(pfs)->v_mount == mp) {
94 if (vget(vp, LK_EXCLUSIVE)) {
100 * Make sure the vnode is still in the cache after
101 * getting the interlock to avoid racing a free.
103 for (pfs = *pp; pfs; pfs = pfs->pfs_next) {
104 if (PFSTOV(pfs) == vp &&
105 pfs->pfs_pid == pid &&
106 pfs->pfs_type == pfs_type &&
107 PFSTOV(pfs)->v_mount == mp) {
112 if (pfs == NULL || PFSTOV(pfs) != vp) {
117 KKASSERT(vp->v_data == pfs);
124 * otherwise lock the vp list while we call getnewvnode
125 * since that can block.
127 if (pfsvplock & PROCFS_LOCKED) {
128 pfsvplock |= PROCFS_WANT;
129 (void) tsleep((caddr_t) &pfsvplock, 0, "pfsavp", 0);
132 pfsvplock |= PROCFS_LOCKED;
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.
139 * XXX this may not matter anymore since getnewvnode now returns
142 pfs = kmalloc(sizeof(struct pfsnode), M_TEMP, M_WAITOK);
144 error = getnewvnode(VT_PROCFS, mp, vpp, 0, 0);
154 pfs->pfs_pid = (pid_t) pid;
155 pfs->pfs_type = pfs_type;
158 pfs->pfs_lockowner = 0;
159 pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type);
162 case Proot: /* /proc = dr-xr-xr-x */
163 pfs->pfs_mode = (VREAD|VEXEC) |
170 case Pcurproc: /* /proc/curproc = lr--r--r-- */
171 pfs->pfs_mode = (VREAD) |
178 pfs->pfs_mode = (VREAD|VEXEC) |
185 pfs->pfs_mode = (VREAD|VEXEC) |
192 pfs->pfs_mode = (VREAD|VWRITE);
199 pfs->pfs_mode = (VREAD|VWRITE);
206 pfs->pfs_mode = (VWRITE);
215 pfs->pfs_mode = (VREAD) |
222 panic("procfs_allocvp");
225 /* add to procfs vnode list */
230 pfsvplock &= ~PROCFS_LOCKED;
232 if (pfsvplock & PROCFS_WANT) {
233 pfsvplock &= ~PROCFS_WANT;
234 wakeup((caddr_t) &pfsvplock);
241 procfs_freevp(struct vnode *vp)
243 struct pfsnode **pfspp;
249 pfspp = PFSHASH(pfs->pfs_pid);
250 while (*pfspp != pfs && *pfspp)
251 pfspp = &(*pfspp)->pfs_next;
253 *pfspp = pfs->pfs_next;
254 pfs->pfs_next = NULL;
255 pfs->pfs_vnode = NULL;
261 * Try to find the calling pid. Note that pfind()
262 * now references the proc structure to be returned
263 * and needs to be released later with PRELE().
266 pfs_pfind(pid_t pfs_pid)
268 struct proc *p = NULL;
278 * Make sure the process is not in the middle of exiting (where
279 * a lot of its structural members may wind up being NULL). If it
280 * is we give up on it.
283 lwkt_gettoken(&p->p_token);
284 if (p->p_flags & P_POSTEXIT) {
285 lwkt_reltoken(&p->p_token);
294 pfs_zpfind(pid_t pfs_pid)
296 struct proc *p = NULL;
306 * Make sure the process is not in the middle of exiting (where
307 * a lot of its structural members may wind up being NULL). If it
308 * is we give up on it.
311 lwkt_gettoken(&p->p_token);
312 if (p->p_flags & P_POSTEXIT) {
313 lwkt_reltoken(&p->p_token);
322 pfs_pdone(struct proc *p)
325 lwkt_reltoken(&p->p_token);
331 procfs_rw(struct vop_read_args *ap)
333 struct vnode *vp = ap->a_vp;
334 struct uio *uio = ap->a_uio;
335 struct thread *curtd = uio->uio_td;
337 struct pfsnode *pfs = VTOPFS(vp);
344 if ((curp = curtd->td_proc) == NULL) /* XXX */
347 lwkt_gettoken(&proc_token);
348 p = pfs_pfind(pfs->pfs_pid);
353 if (p->p_pid == 1 && securelevel > 0 && uio->uio_rw == UIO_WRITE) {
358 lp = FIRST_LWP_IN_PROC(p);
361 while (pfs->pfs_lockowner) {
362 tsleep(&pfs->pfs_lockowner, 0, "pfslck", 0);
364 pfs->pfs_lockowner = curproc->p_pid;
366 switch (pfs->pfs_type) {
369 rtval = procfs_donote(curp, lp, pfs, uio);
373 rtval = procfs_doregs(curp, lp, pfs, uio);
377 rtval = procfs_dofpregs(curp, lp, pfs, uio);
381 rtval = procfs_dodbregs(curp, lp, pfs, uio);
385 rtval = procfs_doctl(curp, lp, pfs, uio);
389 rtval = procfs_dostatus(curp, lp, pfs, uio);
393 rtval = procfs_domap(curp, lp, pfs, uio);
397 rtval = procfs_domem(curp, lp, pfs, uio);
401 rtval = procfs_dotype(curp, lp, pfs, uio);
405 rtval = procfs_docmdline(curp, lp, pfs, uio);
409 rtval = procfs_dorlimit(curp, lp, pfs, uio);
418 pfs->pfs_lockowner = 0;
419 wakeup(&pfs->pfs_lockowner);
423 lwkt_reltoken(&proc_token);
429 * Get a string from userland into (buf). Strip a trailing
430 * nl character (to allow easy access from the shell).
431 * The buffer should be *buflenp + 1 chars long. vfs_getuserstr
432 * will automatically add a nul char at the end.
434 * Returns 0 on success or the following errors
436 * EINVAL: file offset is non-zero.
437 * EMSGSIZE: message is longer than kernel buffer
438 * EFAULT: user i/o buffer is not addressable
441 vfs_getuserstr(struct uio *uio, char *buf, int *buflenp)
446 if (uio->uio_offset != 0)
451 /* must be able to read the whole string in one go */
452 if (xlen < uio->uio_resid)
454 xlen = uio->uio_resid;
456 if ((error = uiomove(buf, xlen, uio)) != 0)
459 /* allow multiple writes without seeks */
462 /* cleanup string and remove trailing newline */
465 if (xlen > 0 && buf[xlen-1] == '\n')
473 vfs_findname(vfs_namemap_t *nm, char *buf, int buflen)
476 for (; nm->nm_name; nm++)
477 if (bcmp(buf, nm->nm_name, buflen+1) == 0)
484 procfs_exit(struct thread *td)
490 KKASSERT(td->td_proc);
491 pid = td->td_proc->p_pid;
494 * NOTE: We can't just vgone() the vnode any more, not while
495 * it may potentially still be active. This will clean
496 * the vp and clear the mount and cause the new VOP subsystem
497 * to assert or panic when someone tries to do an operation
498 * on an open (exited) procfs descriptor.
500 * Prevent further operations on this pid by setting pfs_pid to -1.
501 * Note that a pfs_pid of 0 is used for nodes which do not track
502 * any particular pid.
504 * Use vx_get() to properly ref/lock a vp which may not have any
505 * refs and which may or may not already be reclaimed. vx_put()
506 * will then properly deactivate it and cause it to be recycled.
508 * The hash table can also get ripped out from under us when
509 * we block so take the easy way out and restart the scan.
514 if (pfs->pfs_pid == pid) {
517 pfs->pfs_pid |= PFS_DEAD; /* does not effect hash */