Home | History | Annotate | Line # | Download | only in kern
kern_ktrace.c revision 1.74.2.2 
      1 /*	$NetBSD: kern_ktrace.c,v 1.74.2.2 2003/08/19 19:40:51 skrll Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1989, 1993
      5  *	The Regents of the University of California.  All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. All advertising materials mentioning features or use of this software
     16  *    must display the following acknowledgement:
     17  *	This product includes software developed by the University of
     18  *	California, Berkeley and its contributors.
     19  * 4. Neither the name of the University nor the names of its contributors
     20  *    may be used to endorse or promote products derived from this software
     21  *    without specific prior written permission.
     22  *
     23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     33  * SUCH DAMAGE.
     34  *
     35  *	@(#)kern_ktrace.c	8.5 (Berkeley) 5/14/95
     36  */
     37 
     38 #include <sys/cdefs.h>
     39 __KERNEL_RCSID(0, "$NetBSD: kern_ktrace.c,v 1.74.2.2 2003/08/19 19:40:51 skrll Exp $");
     40 
     41 #include "opt_ktrace.h"
     42 #include "opt_compat_mach.h"
     43 
     44 #include <sys/param.h>
     45 #include <sys/systm.h>
     46 #include <sys/proc.h>
     47 #include <sys/file.h>
     48 #include <sys/namei.h>
     49 #include <sys/vnode.h>
     50 #include <sys/ktrace.h>
     51 #include <sys/malloc.h>
     52 #include <sys/syslog.h>
     53 #include <sys/filedesc.h>
     54 #include <sys/ioctl.h>
     55 
     56 #include <sys/mount.h>
     57 #include <sys/sa.h>
     58 #include <sys/syscallargs.h>
     59 
     60 #ifdef KTRACE
     61 
     62 int	ktrace_common(struct lwp *, int, int, int, struct file *);
     63 void	ktrinitheader(struct ktr_header *, struct lwp *, int);
     64 int	ktrops(struct lwp *, struct proc *, int, int, struct file *);
     65 int	ktrsetchildren(struct lwp *, struct proc *, int, int,
     66 	    struct file *);
     67 int	ktrwrite(struct lwp *, struct ktr_header *);
     68 int	ktrcanset(struct proc *, struct proc *);
     69 int	ktrsamefile(struct file *, struct file *);
     70 
     71 /*
     72  * "deep" compare of two files for the purposes of clearing a trace.
     73  * Returns true if they're the same open file, or if they point at the
     74  * same underlying vnode/socket.
     75  */
     76 
     77 int
     78 ktrsamefile(f1, f2)
     79 	struct file *f1;
     80 	struct file *f2;
     81 {
     82 	return ((f1 == f2) ||
     83 	    ((f1 != NULL) && (f2 != NULL) &&
     84 		(f1->f_type == f2->f_type) &&
     85 		(f1->f_data == f2->f_data)));
     86 }
     87 
     88 void
     89 ktrderef(p)
     90 	struct proc *p;
     91 {
     92 	struct file *fp = p->p_tracep;
     93 	p->p_traceflag = 0;
     94 	if (fp == NULL)
     95 		return;
     96 	simple_lock(&fp->f_slock);
     97 	FILE_USE(fp);
     98 
     99 	/*
    100 	 * ktrace file descriptor can't be watched (are not visible to
    101 	 * userspace), so no kqueue stuff here
    102 	 */
    103 	closef(fp, NULL);
    104 
    105 	p->p_tracep = NULL;
    106 }
    107 
    108 void
    109 ktradref(p)
    110 	struct proc *p;
    111 {
    112 	struct file *fp = p->p_tracep;
    113 
    114 	fp->f_count++;
    115 }
    116 
    117 void
    118 ktrinitheader(kth, l, type)
    119 	struct ktr_header *kth;
    120 	struct lwp *l;
    121 	int type;
    122 {
    123 	struct proc *p = l->l_proc;
    124 
    125 	memset(kth, 0, sizeof(*kth));
    126 	kth->ktr_type = type;
    127 	microtime(&kth->ktr_time);
    128 	kth->ktr_pid = p->p_pid;
    129 	memcpy(kth->ktr_comm, p->p_comm, MAXCOMLEN);
    130 }
    131 
    132 void
    133 ktrsyscall(l, code, realcode, callp, args)
    134 	struct lwp *l;
    135 	register_t code;
    136 	register_t realcode;
    137 	const struct sysent *callp;
    138 	register_t args[];
    139 {
    140 	struct proc *p = l->l_proc;
    141 	struct ktr_syscall *ktp;
    142 	struct ktr_header kth;
    143 	register_t *argp;
    144 	int argsize;
    145 	size_t len;
    146 	u_int i;
    147 
    148 	if (callp == NULL)
    149 		callp = p->p_emul->e_sysent;
    150 
    151 	argsize = callp[code].sy_narg * sizeof (register_t);
    152 	len = sizeof(struct ktr_syscall) + argsize;
    153 
    154 	p->p_traceflag |= KTRFAC_ACTIVE;
    155 	ktrinitheader(&kth, l, KTR_SYSCALL);
    156 	ktp = malloc(len, M_TEMP, M_WAITOK);
    157 	ktp->ktr_code = realcode;
    158 	ktp->ktr_argsize = argsize;
    159 	argp = (register_t *)((char *)ktp + sizeof(struct ktr_syscall));
    160 	for (i = 0; i < (argsize / sizeof(*argp)); i++)
    161 		*argp++ = args[i];
    162 	kth.ktr_buf = (caddr_t)ktp;
    163 	kth.ktr_len = len;
    164 	(void) ktrwrite(l, &kth);
    165 	free(ktp, M_TEMP);
    166 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    167 }
    168 
    169 void
    170 ktrsysret(l, code, error, retval)
    171 	struct lwp *l;
    172 	register_t code;
    173 	int error;
    174 	register_t *retval;
    175 {
    176 	struct proc *p = l->l_proc;
    177 	struct ktr_header kth;
    178 	struct ktr_sysret ktp;
    179 
    180 	p->p_traceflag |= KTRFAC_ACTIVE;
    181 	ktrinitheader(&kth, l, KTR_SYSRET);
    182 	ktp.ktr_code = code;
    183 	ktp.ktr_eosys = 0;			/* XXX unused */
    184 	ktp.ktr_error = error;
    185 	ktp.ktr_retval = retval ? retval[0] : 0;
    186 	ktp.ktr_retval_1 = retval ? retval[1] : 0;
    187 
    188 	kth.ktr_buf = (caddr_t)&ktp;
    189 	kth.ktr_len = sizeof(struct ktr_sysret);
    190 
    191 	(void) ktrwrite(l, &kth);
    192 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    193 }
    194 
    195 void
    196 ktrnamei(l, path)
    197 	struct lwp *l;
    198 	char *path;
    199 {
    200 	struct proc *p = l->l_proc;
    201 	struct ktr_header kth;
    202 
    203 	p->p_traceflag |= KTRFAC_ACTIVE;
    204 	ktrinitheader(&kth, l, KTR_NAMEI);
    205 	kth.ktr_len = strlen(path);
    206 	kth.ktr_buf = path;
    207 
    208 	(void) ktrwrite(l, &kth);
    209 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    210 }
    211 
    212 void
    213 ktremul(l)
    214 	struct lwp *l;
    215 {
    216 	struct ktr_header kth;
    217 	const char *emul;
    218 	struct proc *p;
    219 
    220 	p = l->l_proc;
    221 	emul = p->p_emul->e_name;
    222 	p->p_traceflag |= KTRFAC_ACTIVE;
    223 	ktrinitheader(&kth, l, KTR_EMUL);
    224 	kth.ktr_len = strlen(emul);
    225 	kth.ktr_buf = (caddr_t)emul;
    226 
    227 	(void) ktrwrite(l, &kth);
    228 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    229 }
    230 
    231 void
    232 ktrgenio(l, fd, rw, iov, len, error)
    233 	struct lwp *l;
    234 	int fd;
    235 	enum uio_rw rw;
    236 	struct iovec *iov;
    237 	int len;
    238 	int error;
    239 {
    240 	struct proc *p = l->l_proc;
    241 	struct ktr_header kth;
    242 	struct ktr_genio *ktp;
    243 	int resid = len, cnt;
    244 	caddr_t cp;
    245 	int buflen;
    246 
    247 	if (error)
    248 		return;
    249 
    250 	p->p_traceflag |= KTRFAC_ACTIVE;
    251 
    252 	buflen = min(PAGE_SIZE, len + sizeof(struct ktr_genio));
    253 
    254 	ktrinitheader(&kth, l, KTR_GENIO);
    255 	ktp = malloc(buflen, M_TEMP, M_WAITOK);
    256 	ktp->ktr_fd = fd;
    257 	ktp->ktr_rw = rw;
    258 
    259 	kth.ktr_buf = (caddr_t)ktp;
    260 
    261 	cp = (caddr_t)((char *)ktp + sizeof(struct ktr_genio));
    262 	buflen -= sizeof(struct ktr_genio);
    263 
    264 	while (resid > 0) {
    265 #if 0 /* XXX NJWLWP */
    266 		KDASSERT(p->p_cpu != NULL);
    267 		KDASSERT(p->p_cpu == curcpu());
    268 #endif
    269 		/* XXX NJWLWP */
    270 		if (curcpu()->ci_schedstate.spc_flags & SPCF_SHOULDYIELD)
    271 			preempt(1);
    272 
    273 		cnt = min(iov->iov_len, buflen);
    274 		if (cnt > resid)
    275 			cnt = resid;
    276 		if (copyin(iov->iov_base, cp, cnt))
    277 			break;
    278 
    279 		kth.ktr_len = cnt + sizeof(struct ktr_genio);
    280 
    281 		if (__predict_false(ktrwrite(l, &kth) != 0))
    282 			break;
    283 
    284 		iov->iov_base = (caddr_t)iov->iov_base + cnt;
    285 		iov->iov_len -= cnt;
    286 
    287 		if (iov->iov_len == 0)
    288 			iov++;
    289 
    290 		resid -= cnt;
    291 	}
    292 
    293 	free(ktp, M_TEMP);
    294 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    295 }
    296 
    297 void
    298 ktrpsig(l, sig, action, mask, code)
    299 	struct lwp *l;
    300 	int sig;
    301 	sig_t action;
    302 	sigset_t *mask;
    303 	int code;
    304 {
    305 	struct proc *p = l->l_proc;
    306 	struct ktr_header kth;
    307 	struct ktr_psig	kp;
    308 
    309 	p->p_traceflag |= KTRFAC_ACTIVE;
    310 	ktrinitheader(&kth, l, KTR_PSIG);
    311 	kp.signo = (char)sig;
    312 	kp.action = action;
    313 	kp.mask = *mask;
    314 	kp.code = code;
    315 	kth.ktr_buf = (caddr_t)&kp;
    316 	kth.ktr_len = sizeof(struct ktr_psig);
    317 
    318 	(void) ktrwrite(l, &kth);
    319 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    320 }
    321 
    322 void
    323 ktrcsw(l, out, user)
    324 	struct lwp *l;
    325 	int out;
    326 	int user;
    327 {
    328 	struct proc *p = l->l_proc;
    329 	struct ktr_header kth;
    330 	struct ktr_csw kc;
    331 
    332 	p->p_traceflag |= KTRFAC_ACTIVE;
    333 	ktrinitheader(&kth, l, KTR_CSW);
    334 	kc.out = out;
    335 	kc.user = user;
    336 	kth.ktr_buf = (caddr_t)&kc;
    337 	kth.ktr_len = sizeof(struct ktr_csw);
    338 
    339 	(void) ktrwrite(l, &kth);
    340 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    341 }
    342 
    343 void
    344 ktruser(l, id, addr, len, ustr)
    345 	struct lwp *l;
    346 	const char *id;
    347 	void *addr;
    348 	size_t len;
    349 	int ustr;
    350 {
    351 	struct proc *p = l->l_proc;
    352 	struct ktr_header kth;
    353 	struct ktr_user *ktp;
    354 	caddr_t user_dta;
    355 
    356 	p->p_traceflag |= KTRFAC_ACTIVE;
    357 	ktrinitheader(&kth, l, KTR_USER);
    358 	ktp = malloc(sizeof(struct ktr_user) + len, M_TEMP, M_WAITOK);
    359 	if (ustr) {
    360 		if (copyinstr(id, ktp->ktr_id, KTR_USER_MAXIDLEN, NULL) != 0)
    361 			ktp->ktr_id[0] = '\0';
    362 	} else
    363 		strncpy(ktp->ktr_id, id, KTR_USER_MAXIDLEN);
    364 	ktp->ktr_id[KTR_USER_MAXIDLEN-1] = '\0';
    365 
    366 	user_dta = (caddr_t) ((char *)ktp + sizeof(struct ktr_user));
    367 	if (copyin(addr, (void *) user_dta, len) != 0)
    368 		len = 0;
    369 
    370 	kth.ktr_buf = (void *)ktp;
    371 	kth.ktr_len = sizeof(struct ktr_user) + len;
    372 	(void) ktrwrite(l, &kth);
    373 
    374 	free(ktp, M_TEMP);
    375 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    376 
    377 }
    378 
    379 void
    380 ktrmmsg(l, msgh, size)
    381 	struct lwp *l;
    382 	const void *msgh;
    383 	size_t size;
    384 {
    385 	struct proc *p = l->l_proc;
    386 	struct ktr_header kth;
    387 	struct ktr_mmsg	*kp;
    388 
    389 	p->p_traceflag |= KTRFAC_ACTIVE;
    390 	ktrinitheader(&kth, l, KTR_MMSG);
    391 
    392 	kp = (struct ktr_mmsg *)msgh;
    393 	kth.ktr_buf = (caddr_t)kp;
    394 	kth.ktr_len = size;
    395 	(void) ktrwrite(l, &kth);
    396 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    397 }
    398 
    399 void
    400 ktrsaupcall(struct lwp *l, int type, int nevent, int nint, void *sas,
    401     void *ap)
    402 {
    403 	struct proc *p = l->l_proc;
    404 	struct ktr_header kth;
    405 	struct ktr_saupcall *ktp;
    406 	size_t len;
    407 	struct sa_t **sapp;
    408 	int i;
    409 
    410 	p->p_traceflag |= KTRFAC_ACTIVE;
    411 	ktrinitheader(&kth, l, KTR_SAUPCALL);
    412 
    413 	len = sizeof(struct ktr_saupcall);
    414 	ktp = malloc(len + sizeof(struct sa_t) * (nevent + nint + 1), M_TEMP,
    415 	    M_WAITOK);
    416 
    417 	ktp->ktr_type = type;
    418 	ktp->ktr_nevent = nevent;
    419 	ktp->ktr_nint = nint;
    420 	ktp->ktr_sas = sas;
    421 	ktp->ktr_ap = ap;
    422 	/*
    423 	 *  Copy the sa_t's
    424 	 */
    425 	sapp = (struct sa_t **) sas;
    426 
    427 	for (i = nevent + nint; i >= 0; i--) {
    428 		if (copyin(*sapp, (char *)ktp + len, sizeof(struct sa_t)) == 0)
    429 			len += sizeof(struct sa_t);
    430 		sapp++;
    431 	}
    432 
    433 	kth.ktr_buf = (void *)ktp;
    434 	kth.ktr_len = len;
    435 	(void) ktrwrite(l, &kth);
    436 
    437 	free(ktp, M_TEMP);
    438 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    439 }
    440 
    441 /* Interface and common routines */
    442 
    443 int
    444 ktrace_common(l, ops, facs, pid, fp)
    445 	struct lwp *l;
    446 	int ops;
    447 	int facs;
    448 	int pid;
    449 	struct file *fp;
    450 {
    451 	struct proc *curp = l->l_proc;
    452 	struct pgrp *pg;
    453 	struct proc *p;
    454 	int error = 0;
    455 	int ret = 0;
    456 	int one = 1;
    457 	int descend;
    458 
    459 	curp->p_traceflag |= KTRFAC_ACTIVE;
    460 	descend = ops & KTRFLAG_DESCEND;
    461 	facs = facs & ~((unsigned) KTRFAC_ROOT);
    462 
    463 	/*
    464 	 * Clear all uses of the tracefile
    465 	 */
    466 	if (KTROP(ops) == KTROP_CLEARFILE) {
    467 		proclist_lock_read();
    468 		for (p = LIST_FIRST(&allproc); p != NULL;
    469 		     p = LIST_NEXT(p, p_list)) {
    470 			if (ktrsamefile(p->p_tracep, fp)) {
    471 				if (ktrcanset(curp, p))
    472 					ktrderef(p);
    473 				else
    474 					error = EPERM;
    475 			}
    476 		}
    477 		proclist_unlock_read();
    478 		goto done;
    479 	}
    480 
    481 	/*
    482 	 * Mark fp non-blocking, to avoid problems from possible deadlocks.
    483 	 */
    484 
    485 	if (fp != NULL) {
    486 		fp->f_flag |= FNONBLOCK;
    487 		(*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&one, l);
    488 	}
    489 
    490 	/*
    491 	 * need something to (un)trace (XXX - why is this here?)
    492 	 */
    493 	if (!facs) {
    494 		error = EINVAL;
    495 		goto done;
    496 	}
    497 	/*
    498 	 * do it
    499 	 */
    500 	if (pid < 0) {
    501 		/*
    502 		 * by process group
    503 		 */
    504 		pg = pgfind(-pid);
    505 		if (pg == NULL) {
    506 			error = ESRCH;
    507 			goto done;
    508 		}
    509 		for (p = LIST_FIRST(&pg->pg_members); p != NULL;
    510 		     p = LIST_NEXT(p, p_pglist)) {
    511 			if (descend)
    512 				ret |= ktrsetchildren(l, p, ops, facs, fp);
    513 			else
    514 				ret |= ktrops(l, p, ops, facs, fp);
    515 		}
    516 
    517 	} else {
    518 		/*
    519 		 * by pid
    520 		 */
    521 		p = pfind(pid);
    522 		if (p == NULL) {
    523 			error = ESRCH;
    524 			goto done;
    525 		}
    526 		if (descend)
    527 			ret |= ktrsetchildren(l, p, ops, facs, fp);
    528 		else
    529 			ret |= ktrops(l, p, ops, facs, fp);
    530 	}
    531 	if (!ret)
    532 		error = EPERM;
    533 done:
    534 	curp->p_traceflag &= ~KTRFAC_ACTIVE;
    535 	return (error);
    536 }
    537 
    538 /*
    539  * ktrace system call
    540  */
    541 /* ARGSUSED */
    542 int
    543 sys_fktrace(l, v, retval)
    544 	struct lwp *l;
    545 	void *v;
    546 	register_t *retval;
    547 {
    548 	struct sys_fktrace_args /* {
    549 		syscallarg(int) fd;
    550 		syscallarg(int) ops;
    551 		syscallarg(int) facs;
    552 		syscallarg(int) pid;
    553 	} */ *uap = v;
    554 	struct proc *curp;
    555 	struct file *fp = NULL;
    556 	struct filedesc *fdp = l->l_proc->p_fd;
    557 	int error;
    558 
    559 	curp = l->l_proc;
    560 	fdp = curp->p_fd;
    561 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
    562 		return (EBADF);
    563 
    564 	FILE_USE(fp);
    565 
    566 	if ((fp->f_flag & FWRITE) == 0)
    567 		error = EBADF;
    568 	else
    569 		error = ktrace_common(l, SCARG(uap, ops),
    570 		    SCARG(uap, facs), SCARG(uap, pid), fp);
    571 
    572 	FILE_UNUSE(fp, l);
    573 
    574 	return error;
    575 }
    576 
    577 /*
    578  * ktrace system call
    579  */
    580 /* ARGSUSED */
    581 int
    582 sys_ktrace(l, v, retval)
    583 	struct lwp *l;
    584 	void *v;
    585 	register_t *retval;
    586 {
    587 	struct sys_ktrace_args /* {
    588 		syscallarg(const char *) fname;
    589 		syscallarg(int) ops;
    590 		syscallarg(int) facs;
    591 		syscallarg(int) pid;
    592 	} */ *uap = v;
    593 	struct proc *curp = l->l_proc;
    594 	struct vnode *vp = NULL;
    595 	struct file *fp = NULL;
    596 	int ops = SCARG(uap, ops);
    597 	struct nameidata nd;
    598 	int error = 0;
    599 	int fd;
    600 
    601 	ops = KTROP(ops) | (ops & KTRFLAG_DESCEND);
    602 
    603 	curp->p_traceflag |= KTRFAC_ACTIVE;
    604 	if ((ops & KTROP_CLEAR) == 0) {
    605 		/*
    606 		 * an operation which requires a file argument.
    607 		 */
    608 		NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, fname),
    609 		    l);
    610 		if ((error = vn_open(&nd, FREAD|FWRITE, 0)) != 0) {
    611 			curp->p_traceflag &= ~KTRFAC_ACTIVE;
    612 			return (error);
    613 		}
    614 		vp = nd.ni_vp;
    615 		VOP_UNLOCK(vp, 0);
    616 		if (vp->v_type != VREG) {
    617 			(void) vn_close(vp, FREAD|FWRITE, curp->p_ucred, l);
    618 			curp->p_traceflag &= ~KTRFAC_ACTIVE;
    619 			return (EACCES);
    620 		}
    621 		/*
    622 		 * XXX This uses up a file descriptor slot in the
    623 		 * tracing process for the duration of this syscall.
    624 		 * This is not expected to be a problem.  If
    625 		 * falloc(NULL, ...) DTRT we could skip that part, but
    626 		 * that would require changing its interface to allow
    627 		 * the caller to pass in a ucred..
    628 		 *
    629 		 * This will FILE_USE the fp it returns, if any.
    630 		 * Keep it in use until we return.
    631 		 */
    632 		if ((error = falloc(curp, &fp, &fd)) != 0)
    633 			goto done;
    634 
    635 		fp->f_flag = FWRITE|FAPPEND;
    636 		fp->f_type = DTYPE_VNODE;
    637 		fp->f_ops = &vnops;
    638 		fp->f_data = (caddr_t)vp;
    639 		FILE_SET_MATURE(fp);
    640 		vp = NULL;
    641 	}
    642 	error = ktrace_common(l, SCARG(uap, ops), SCARG(uap, facs),
    643 	    SCARG(uap, pid), fp);
    644 done:
    645 	if (vp != NULL)
    646 		(void) vn_close(vp, FWRITE, curp->p_ucred, l);
    647 	if (fp != NULL) {
    648 		FILE_UNUSE(fp, l);	/* release file */
    649 		fdrelease(l, fd); 	/* release fd table slot */
    650 	}
    651 	return (error);
    652 }
    653 
    654 int
    655 ktrops(l, p, ops, facs, fp)
    656 	struct lwp *l;
    657 	struct proc *p;
    658 	int ops;
    659 	int facs;
    660 	struct file *fp;
    661 {
    662 	struct proc *curp = l->l_proc;
    663 
    664 	if (!ktrcanset(curp, p))
    665 		return (0);
    666 	if (KTROP(ops) == KTROP_SET) {
    667 		if (p->p_tracep != fp) {
    668 			/*
    669 			 * if trace file already in use, relinquish
    670 			 */
    671 			ktrderef(p);
    672 			p->p_tracep = fp;
    673 			ktradref(p);
    674 		}
    675 		p->p_traceflag |= facs;
    676 		if (curp->p_ucred->cr_uid == 0)
    677 			p->p_traceflag |= KTRFAC_ROOT;
    678 	} else {
    679 		/* KTROP_CLEAR */
    680 		if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) {
    681 			/* no more tracing */
    682 			ktrderef(p);
    683 		}
    684 	}
    685 
    686 	/*
    687 	 * Emit an emulation record, every time there is a ktrace
    688 	 * change/attach request.
    689 	 */
    690 	if (KTRPOINT(p, KTR_EMUL))
    691 		ktremul(l);
    692 #ifdef __HAVE_SYSCALL_INTERN
    693 	(*p->p_emul->e_syscall_intern)(p);
    694 #endif
    695 
    696 	return (1);
    697 }
    698 
    699 int
    700 ktrsetchildren(l, top, ops, facs, fp)
    701 	struct lwp *l;
    702 	struct proc *top;
    703 	int ops;
    704 	int facs;
    705 	struct file *fp;
    706 {
    707 	struct proc *p;
    708 	int ret = 0;
    709 
    710 	p = top;
    711 	for (;;) {
    712 		ret |= ktrops(l, p, ops, facs, fp);
    713 		/*
    714 		 * If this process has children, descend to them next,
    715 		 * otherwise do any siblings, and if done with this level,
    716 		 * follow back up the tree (but not past top).
    717 		 */
    718 		if (LIST_FIRST(&p->p_children) != NULL)
    719 			p = LIST_FIRST(&p->p_children);
    720 		else for (;;) {
    721 			if (p == top)
    722 				return (ret);
    723 			if (LIST_NEXT(p, p_sibling) != NULL) {
    724 				p = LIST_NEXT(p, p_sibling);
    725 				break;
    726 			}
    727 			p = p->p_pptr;
    728 		}
    729 	}
    730 	/*NOTREACHED*/
    731 }
    732 
    733 int
    734 ktrwrite(l, kth)
    735 	struct lwp *l;
    736 	struct ktr_header *kth;
    737 {
    738 	struct iovec aiov[2];
    739 	int error, tries;
    740 	struct uio auio;
    741 	struct file *fp;
    742 	struct proc *p;
    743 
    744 	p = l->l_proc;
    745 	fp = p->p_tracep;
    746 
    747 	if (fp == NULL)
    748 		return 0;
    749 
    750 	auio.uio_iov = &aiov[0];
    751 	auio.uio_offset = 0;
    752 	auio.uio_segflg = UIO_SYSSPACE;
    753 	auio.uio_rw = UIO_WRITE;
    754 	aiov[0].iov_base = (caddr_t)kth;
    755 	aiov[0].iov_len = sizeof(struct ktr_header);
    756 	auio.uio_resid = sizeof(struct ktr_header);
    757 	auio.uio_iovcnt = 1;
    758 	auio.uio_lwp = (struct lwp *)0;
    759 	if (kth->ktr_len > 0) {
    760 		auio.uio_iovcnt++;
    761 		aiov[1].iov_base = kth->ktr_buf;
    762 		aiov[1].iov_len = kth->ktr_len;
    763 		auio.uio_resid += kth->ktr_len;
    764 	}
    765 	kth->ktr_buf = (caddr_t)l->l_lid;
    766 
    767 	simple_lock(&fp->f_slock);
    768 	FILE_USE(fp);
    769 
    770 	tries = 0;
    771 	do {
    772 		error = (*fp->f_ops->fo_write)(fp, &fp->f_offset, &auio,
    773 		    fp->f_cred, FOF_UPDATE_OFFSET);
    774 		if (error == EWOULDBLOCK)
    775 		  	preempt(1);
    776 		else
    777 			tries++;
    778 	} while ((error == EWOULDBLOCK) && (tries < 3));
    779 	FILE_UNUSE(fp, NULL);
    780 
    781 	if (__predict_true(error == 0))
    782 		return (0);
    783 	/*
    784 	 * If error encountered, give up tracing on this vnode.  Don't report
    785 	 * EPIPE as this can easily happen with fktrace()/ktruss.
    786 	 */
    787 	if (error != EPIPE)
    788 		log(LOG_NOTICE,
    789 		    "ktrace write failed, errno %d, tracing stopped\n",
    790 		    error);
    791 	proclist_lock_read();
    792 	for (p = LIST_FIRST(&allproc); p != NULL; p = LIST_NEXT(p, p_list)) {
    793 		if (ktrsamefile(p->p_tracep, fp))
    794 			ktrderef(p);
    795 	}
    796 	proclist_unlock_read();
    797 
    798 	return (error);
    799 }
    800 
    801 /*
    802  * Return true if caller has permission to set the ktracing state
    803  * of target.  Essentially, the target can't possess any
    804  * more permissions than the caller.  KTRFAC_ROOT signifies that
    805  * root previously set the tracing status on the target process, and
    806  * so, only root may further change it.
    807  *
    808  * TODO: check groups.  use caller effective gid.
    809  */
    810 int
    811 ktrcanset(callp, targetp)
    812 	struct proc *callp;
    813 	struct proc *targetp;
    814 {
    815 	struct pcred *caller = callp->p_cred;
    816 	struct pcred *target = targetp->p_cred;
    817 
    818 	if ((caller->pc_ucred->cr_uid == target->p_ruid &&
    819 	     target->p_ruid == target->p_svuid &&
    820 	     caller->p_rgid == target->p_rgid &&	/* XXX */
    821 	     target->p_rgid == target->p_svgid &&
    822 	     (targetp->p_traceflag & KTRFAC_ROOT) == 0 &&
    823 	     (targetp->p_flag & P_SUGID) == 0) ||
    824 	     caller->pc_ucred->cr_uid == 0)
    825 		return (1);
    826 
    827 	return (0);
    828 }
    829 #endif /* KTRACE */
    830 
    831 /*
    832  * Put user defined entry to ktrace records.
    833  */
    834 int
    835 sys_utrace(l, v, retval)
    836 	struct lwp *l;
    837 	void *v;
    838 	register_t *retval;
    839 {
    840 #ifdef KTRACE
    841 	struct sys_utrace_args /* {
    842 		syscallarg(const char *) label;
    843 		syscallarg(void *) addr;
    844 		syscallarg(size_t) len;
    845 	} */ *uap = v;
    846 	struct proc *p = l->l_proc;
    847 	if (!KTRPOINT(p, KTR_USER))
    848 		return (0);
    849 
    850 	if (SCARG(uap, len) > KTR_USER_MAXLEN)
    851 		return (EINVAL);
    852 
    853 	ktruser(l, SCARG(uap, label), SCARG(uap, addr), SCARG(uap, len), 1);
    854 
    855 	return (0);
    856 #else /* !KTRACE */
    857 	return ENOSYS;
    858 #endif /* KTRACE */
    859 }
    860