Home | History | Annotate | Line # | Download | only in kern
kern_ktrace.c revision 1.76
      1 /*	$NetBSD: kern_ktrace.c,v 1.76 2003/08/04 18:53:06 drochner 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 1.76 2003/08/04 18:53:06 drochner 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 proc *, int, int, int, struct file *);
     63 void	ktrinitheader(struct ktr_header *, struct proc *, int);
     64 int	ktrops(struct proc *, struct proc *, int, int, struct file *);
     65 int	ktrsetchildren(struct proc *, struct proc *, int, int,
     66 	    struct file *);
     67 int	ktrwrite(struct proc *, 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, p, type)
    119 	struct ktr_header *kth;
    120 	struct proc *p;
    121 	int type;
    122 {
    123 
    124 	memset(kth, 0, sizeof(*kth));
    125 	kth->ktr_type = type;
    126 	microtime(&kth->ktr_time);
    127 	kth->ktr_pid = p->p_pid;
    128 	memcpy(kth->ktr_comm, p->p_comm, MAXCOMLEN);
    129 }
    130 
    131 void
    132 ktrsyscall(p, code, realcode, callp, args)
    133 	struct proc *p;
    134 	register_t code;
    135 	register_t realcode;
    136 	const struct sysent *callp;
    137 	register_t args[];
    138 {
    139 	struct ktr_header kth;
    140 	struct ktr_syscall *ktp;
    141 	register_t *argp;
    142 	int argsize;
    143 	size_t len;
    144 	u_int i;
    145 
    146 	if (callp == NULL)
    147 		callp = p->p_emul->e_sysent;
    148 
    149 	argsize = callp[code].sy_argsize;
    150 	len = sizeof(struct ktr_syscall) + argsize;
    151 
    152 	p->p_traceflag |= KTRFAC_ACTIVE;
    153 	ktrinitheader(&kth, p, KTR_SYSCALL);
    154 	ktp = malloc(len, M_TEMP, M_WAITOK);
    155 	ktp->ktr_code = realcode;
    156 	ktp->ktr_argsize = argsize;
    157 	argp = (register_t *)((char *)ktp + sizeof(struct ktr_syscall));
    158 	for (i = 0; i < (argsize / sizeof(*argp)); i++)
    159 		*argp++ = args[i];
    160 	kth.ktr_buf = (caddr_t)ktp;
    161 	kth.ktr_len = len;
    162 	(void) ktrwrite(p, &kth);
    163 	free(ktp, M_TEMP);
    164 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    165 }
    166 
    167 void
    168 ktrsysret(p, code, error, retval)
    169 	struct proc *p;
    170 	register_t code;
    171 	int error;
    172 	register_t *retval;
    173 {
    174 	struct ktr_header kth;
    175 	struct ktr_sysret ktp;
    176 
    177 	p->p_traceflag |= KTRFAC_ACTIVE;
    178 	ktrinitheader(&kth, p, KTR_SYSRET);
    179 	ktp.ktr_code = code;
    180 	ktp.ktr_eosys = 0;			/* XXX unused */
    181 	ktp.ktr_error = error;
    182 	ktp.ktr_retval = retval ? retval[0] : 0;
    183 	ktp.ktr_retval_1 = retval ? retval[1] : 0;
    184 
    185 	kth.ktr_buf = (caddr_t)&ktp;
    186 	kth.ktr_len = sizeof(struct ktr_sysret);
    187 
    188 	(void) ktrwrite(p, &kth);
    189 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    190 }
    191 
    192 void
    193 ktrnamei(p, path)
    194 	struct proc *p;
    195 	char *path;
    196 {
    197 	struct ktr_header kth;
    198 
    199 	p->p_traceflag |= KTRFAC_ACTIVE;
    200 	ktrinitheader(&kth, p, KTR_NAMEI);
    201 	kth.ktr_len = strlen(path);
    202 	kth.ktr_buf = path;
    203 
    204 	(void) ktrwrite(p, &kth);
    205 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    206 }
    207 
    208 void
    209 ktremul(p)
    210 	struct proc *p;
    211 {
    212 	struct ktr_header kth;
    213 	const char *emul = p->p_emul->e_name;
    214 
    215 	p->p_traceflag |= KTRFAC_ACTIVE;
    216 	ktrinitheader(&kth, p, KTR_EMUL);
    217 	kth.ktr_len = strlen(emul);
    218 	kth.ktr_buf = (caddr_t)emul;
    219 
    220 	(void) ktrwrite(p, &kth);
    221 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    222 }
    223 
    224 void
    225 ktrkmem(struct proc *p, int ktr, const void *buf, size_t len)
    226 {
    227 	struct ktr_header kth;
    228 
    229 	p->p_traceflag |= KTRFAC_ACTIVE;
    230 	ktrinitheader(&kth, p, ktr);
    231 	kth.ktr_len = len;
    232 	kth.ktr_buf = buf;
    233 
    234 	(void)ktrwrite(p, &kth);
    235 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    236 }
    237 
    238 void
    239 ktrgenio(p, fd, rw, iov, len, error)
    240 	struct proc *p;
    241 	int fd;
    242 	enum uio_rw rw;
    243 	struct iovec *iov;
    244 	int len;
    245 	int error;
    246 {
    247 	struct ktr_header kth;
    248 	struct ktr_genio *ktp;
    249 	caddr_t cp;
    250 	int resid = len, cnt;
    251 	int buflen;
    252 
    253 	if (error)
    254 		return;
    255 
    256 	p->p_traceflag |= KTRFAC_ACTIVE;
    257 
    258 	buflen = min(PAGE_SIZE, len + sizeof(struct ktr_genio));
    259 
    260 	ktrinitheader(&kth, p, KTR_GENIO);
    261 	ktp = malloc(buflen, M_TEMP, M_WAITOK);
    262 	ktp->ktr_fd = fd;
    263 	ktp->ktr_rw = rw;
    264 
    265 	kth.ktr_buf = (caddr_t)ktp;
    266 
    267 	cp = (caddr_t)((char *)ktp + sizeof(struct ktr_genio));
    268 	buflen -= sizeof(struct ktr_genio);
    269 
    270 	while (resid > 0) {
    271 #if 0 /* XXX NJWLWP */
    272 		KDASSERT(p->p_cpu != NULL);
    273 		KDASSERT(p->p_cpu == curcpu());
    274 #endif
    275 		/* XXX NJWLWP */
    276 		if (curcpu()->ci_schedstate.spc_flags & SPCF_SHOULDYIELD)
    277 			preempt(1);
    278 
    279 		cnt = min(iov->iov_len, buflen);
    280 		if (cnt > resid)
    281 			cnt = resid;
    282 		if (copyin(iov->iov_base, cp, cnt))
    283 			break;
    284 
    285 		kth.ktr_len = cnt + sizeof(struct ktr_genio);
    286 
    287 		if (__predict_false(ktrwrite(p, &kth) != 0))
    288 			break;
    289 
    290 		iov->iov_base = (caddr_t)iov->iov_base + cnt;
    291 		iov->iov_len -= cnt;
    292 
    293 		if (iov->iov_len == 0)
    294 			iov++;
    295 
    296 		resid -= cnt;
    297 	}
    298 
    299 	free(ktp, M_TEMP);
    300 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    301 }
    302 
    303 void
    304 ktrpsig(p, sig, action, mask, code)
    305 	struct proc *p;
    306 	int sig;
    307 	sig_t action;
    308 	sigset_t *mask;
    309 	int code;
    310 {
    311 	struct ktr_header kth;
    312 	struct ktr_psig	kp;
    313 
    314 	p->p_traceflag |= KTRFAC_ACTIVE;
    315 	ktrinitheader(&kth, p, KTR_PSIG);
    316 	kp.signo = (char)sig;
    317 	kp.action = action;
    318 	kp.mask = *mask;
    319 	kp.code = code;
    320 	kth.ktr_buf = (caddr_t)&kp;
    321 	kth.ktr_len = sizeof(struct ktr_psig);
    322 
    323 	(void) ktrwrite(p, &kth);
    324 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    325 }
    326 
    327 void
    328 ktrcsw(p, out, user)
    329 	struct proc *p;
    330 	int out;
    331 	int user;
    332 {
    333 	struct ktr_header kth;
    334 	struct ktr_csw kc;
    335 
    336 	p->p_traceflag |= KTRFAC_ACTIVE;
    337 	ktrinitheader(&kth, p, KTR_CSW);
    338 	kc.out = out;
    339 	kc.user = user;
    340 	kth.ktr_buf = (caddr_t)&kc;
    341 	kth.ktr_len = sizeof(struct ktr_csw);
    342 
    343 	(void) ktrwrite(p, &kth);
    344 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    345 }
    346 
    347 void
    348 ktruser(p, id, addr, len, ustr)
    349 	struct proc *p;
    350 	const char *id;
    351 	void *addr;
    352 	size_t len;
    353 	int ustr;
    354 {
    355 	struct ktr_header kth;
    356 	struct ktr_user *ktp;
    357 	caddr_t user_dta;
    358 
    359 	p->p_traceflag |= KTRFAC_ACTIVE;
    360 	ktrinitheader(&kth, p, KTR_USER);
    361 	ktp = malloc(sizeof(struct ktr_user) + len, M_TEMP, M_WAITOK);
    362 	if (ustr) {
    363 		if (copyinstr(id, ktp->ktr_id, KTR_USER_MAXIDLEN, NULL) != 0)
    364 			ktp->ktr_id[0] = '\0';
    365 	} else
    366 		strncpy(ktp->ktr_id, id, KTR_USER_MAXIDLEN);
    367 	ktp->ktr_id[KTR_USER_MAXIDLEN-1] = '\0';
    368 
    369 	user_dta = (caddr_t) ((char *)ktp + sizeof(struct ktr_user));
    370 	if (copyin(addr, (void *) user_dta, len) != 0)
    371 		len = 0;
    372 
    373 	kth.ktr_buf = (void *)ktp;
    374 	kth.ktr_len = sizeof(struct ktr_user) + len;
    375 	(void) ktrwrite(p, &kth);
    376 
    377 	free(ktp, M_TEMP);
    378 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    379 
    380 }
    381 
    382 void
    383 ktrmmsg(p, msgh, size)
    384 	struct proc *p;
    385 	const void *msgh;
    386 	size_t size;
    387 {
    388 	struct ktr_header kth;
    389 	struct ktr_mmsg	*kp;
    390 
    391 	p->p_traceflag |= KTRFAC_ACTIVE;
    392 	ktrinitheader(&kth, p, KTR_MMSG);
    393 
    394 	kp = (struct ktr_mmsg *)msgh;
    395 	kth.ktr_buf = (caddr_t)kp;
    396 	kth.ktr_len = size;
    397 	(void) ktrwrite(p, &kth);
    398 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    399 }
    400 
    401 /* Interface and common routines */
    402 
    403 int
    404 ktrace_common(curp, ops, facs, pid, fp)
    405 	struct proc *curp;
    406 	int ops;
    407 	int facs;
    408 	int pid;
    409 	struct file *fp;
    410 {
    411 	int ret = 0;
    412 	int error = 0;
    413 	int one = 1;
    414 	int descend;
    415 	struct proc *p;
    416 	struct pgrp *pg;
    417 
    418 	curp->p_traceflag |= KTRFAC_ACTIVE;
    419 	descend = ops & KTRFLAG_DESCEND;
    420 	facs = facs & ~((unsigned) KTRFAC_ROOT);
    421 
    422 	/*
    423 	 * Clear all uses of the tracefile
    424 	 */
    425 	if (KTROP(ops) == KTROP_CLEARFILE) {
    426 		proclist_lock_read();
    427 		for (p = LIST_FIRST(&allproc); p != NULL;
    428 		     p = LIST_NEXT(p, p_list)) {
    429 			if (ktrsamefile(p->p_tracep, fp)) {
    430 				if (ktrcanset(curp, p))
    431 					ktrderef(p);
    432 				else
    433 					error = EPERM;
    434 			}
    435 		}
    436 		proclist_unlock_read();
    437 		goto done;
    438 	}
    439 
    440 	/*
    441 	 * Mark fp non-blocking, to avoid problems from possible deadlocks.
    442 	 */
    443 
    444 	if (fp != NULL) {
    445 		fp->f_flag |= FNONBLOCK;
    446 		(*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&one, curp);
    447 	}
    448 
    449 	/*
    450 	 * need something to (un)trace (XXX - why is this here?)
    451 	 */
    452 	if (!facs) {
    453 		error = EINVAL;
    454 		goto done;
    455 	}
    456 	/*
    457 	 * do it
    458 	 */
    459 	if (pid < 0) {
    460 		/*
    461 		 * by process group
    462 		 */
    463 		pg = pgfind(-pid);
    464 		if (pg == NULL) {
    465 			error = ESRCH;
    466 			goto done;
    467 		}
    468 		for (p = LIST_FIRST(&pg->pg_members); p != NULL;
    469 		     p = LIST_NEXT(p, p_pglist)) {
    470 			if (descend)
    471 				ret |= ktrsetchildren(curp, p, ops, facs, fp);
    472 			else
    473 				ret |= ktrops(curp, p, ops, facs, fp);
    474 		}
    475 
    476 	} else {
    477 		/*
    478 		 * by pid
    479 		 */
    480 		p = pfind(pid);
    481 		if (p == NULL) {
    482 			error = ESRCH;
    483 			goto done;
    484 		}
    485 		if (descend)
    486 			ret |= ktrsetchildren(curp, p, ops, facs, fp);
    487 		else
    488 			ret |= ktrops(curp, p, ops, facs, fp);
    489 	}
    490 	if (!ret)
    491 		error = EPERM;
    492 done:
    493 	curp->p_traceflag &= ~KTRFAC_ACTIVE;
    494 	return (error);
    495 }
    496 
    497 /*
    498  * ktrace system call
    499  */
    500 /* ARGSUSED */
    501 int
    502 sys_fktrace(l, v, retval)
    503 	struct lwp *l;
    504 	void *v;
    505 	register_t *retval;
    506 {
    507 	struct sys_fktrace_args /* {
    508 		syscallarg(int) fd;
    509 		syscallarg(int) ops;
    510 		syscallarg(int) facs;
    511 		syscallarg(int) pid;
    512 	} */ *uap = v;
    513 	struct proc *curp = l->l_proc;
    514 	struct file *fp = NULL;
    515 	struct filedesc *fdp = curp->p_fd;
    516 	int error;
    517 
    518 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
    519 		return (EBADF);
    520 
    521 	FILE_USE(fp);
    522 
    523 	if ((fp->f_flag & FWRITE) == 0)
    524 		error = EBADF;
    525 	else
    526 		error = ktrace_common(curp, SCARG(uap, ops),
    527 		    SCARG(uap, facs), SCARG(uap, pid), fp);
    528 
    529 	FILE_UNUSE(fp, curp);
    530 
    531 	return error;
    532 }
    533 
    534 /*
    535  * ktrace system call
    536  */
    537 /* ARGSUSED */
    538 int
    539 sys_ktrace(l, v, retval)
    540 	struct lwp *l;
    541 	void *v;
    542 	register_t *retval;
    543 {
    544 	struct sys_ktrace_args /* {
    545 		syscallarg(const char *) fname;
    546 		syscallarg(int) ops;
    547 		syscallarg(int) facs;
    548 		syscallarg(int) pid;
    549 	} */ *uap = v;
    550 	struct proc *curp = l->l_proc;
    551 	struct vnode *vp = NULL;
    552 	struct file *fp = NULL;
    553 	int fd;
    554 	int ops = SCARG(uap, ops);
    555 	int error = 0;
    556 	struct nameidata nd;
    557 
    558 	ops = KTROP(ops) | (ops & KTRFLAG_DESCEND);
    559 
    560 	curp->p_traceflag |= KTRFAC_ACTIVE;
    561 	if ((ops & KTROP_CLEAR) == 0) {
    562 		/*
    563 		 * an operation which requires a file argument.
    564 		 */
    565 		NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, fname),
    566 		    curp);
    567 		if ((error = vn_open(&nd, FREAD|FWRITE, 0)) != 0) {
    568 			curp->p_traceflag &= ~KTRFAC_ACTIVE;
    569 			return (error);
    570 		}
    571 		vp = nd.ni_vp;
    572 		VOP_UNLOCK(vp, 0);
    573 		if (vp->v_type != VREG) {
    574 			(void) vn_close(vp, FREAD|FWRITE, curp->p_ucred, curp);
    575 			curp->p_traceflag &= ~KTRFAC_ACTIVE;
    576 			return (EACCES);
    577 		}
    578 		/*
    579 		 * XXX This uses up a file descriptor slot in the
    580 		 * tracing process for the duration of this syscall.
    581 		 * This is not expected to be a problem.  If
    582 		 * falloc(NULL, ...) DTRT we could skip that part, but
    583 		 * that would require changing its interface to allow
    584 		 * the caller to pass in a ucred..
    585 		 *
    586 		 * This will FILE_USE the fp it returns, if any.
    587 		 * Keep it in use until we return.
    588 		 */
    589 		if ((error = falloc(curp, &fp, &fd)) != 0)
    590 			goto done;
    591 
    592 		fp->f_flag = FWRITE|FAPPEND;
    593 		fp->f_type = DTYPE_VNODE;
    594 		fp->f_ops = &vnops;
    595 		fp->f_data = (caddr_t)vp;
    596 		FILE_SET_MATURE(fp);
    597 		vp = NULL;
    598 	}
    599 	error = ktrace_common(curp, SCARG(uap, ops), SCARG(uap, facs),
    600 	    SCARG(uap, pid), fp);
    601 done:
    602 	if (vp != NULL)
    603 		(void) vn_close(vp, FWRITE, curp->p_ucred, curp);
    604 	if (fp != NULL) {
    605 		FILE_UNUSE(fp, curp);	/* release file */
    606 		fdrelease(curp, fd); 	/* release fd table slot */
    607 	}
    608 	return (error);
    609 }
    610 
    611 int
    612 ktrops(curp, p, ops, facs, fp)
    613 	struct proc *curp;
    614 	struct proc *p;
    615 	int ops;
    616 	int facs;
    617 	struct file *fp;
    618 {
    619 
    620 	if (!ktrcanset(curp, p))
    621 		return (0);
    622 	if (KTROP(ops) == KTROP_SET) {
    623 		if (p->p_tracep != fp) {
    624 			/*
    625 			 * if trace file already in use, relinquish
    626 			 */
    627 			ktrderef(p);
    628 			p->p_tracep = fp;
    629 			ktradref(p);
    630 		}
    631 		p->p_traceflag |= facs;
    632 		if (curp->p_ucred->cr_uid == 0)
    633 			p->p_traceflag |= KTRFAC_ROOT;
    634 	} else {
    635 		/* KTROP_CLEAR */
    636 		if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) {
    637 			/* no more tracing */
    638 			ktrderef(p);
    639 		}
    640 	}
    641 
    642 	/*
    643 	 * Emit an emulation record, every time there is a ktrace
    644 	 * change/attach request.
    645 	 */
    646 	if (KTRPOINT(p, KTR_EMUL))
    647 		ktremul(p);
    648 #ifdef __HAVE_SYSCALL_INTERN
    649 	(*p->p_emul->e_syscall_intern)(p);
    650 #endif
    651 
    652 	return (1);
    653 }
    654 
    655 int
    656 ktrsetchildren(curp, top, ops, facs, fp)
    657 	struct proc *curp;
    658 	struct proc *top;
    659 	int ops;
    660 	int facs;
    661 	struct file *fp;
    662 {
    663 	struct proc *p;
    664 	int ret = 0;
    665 
    666 	p = top;
    667 	for (;;) {
    668 		ret |= ktrops(curp, p, ops, facs, fp);
    669 		/*
    670 		 * If this process has children, descend to them next,
    671 		 * otherwise do any siblings, and if done with this level,
    672 		 * follow back up the tree (but not past top).
    673 		 */
    674 		if (LIST_FIRST(&p->p_children) != NULL)
    675 			p = LIST_FIRST(&p->p_children);
    676 		else for (;;) {
    677 			if (p == top)
    678 				return (ret);
    679 			if (LIST_NEXT(p, p_sibling) != NULL) {
    680 				p = LIST_NEXT(p, p_sibling);
    681 				break;
    682 			}
    683 			p = p->p_pptr;
    684 		}
    685 	}
    686 	/*NOTREACHED*/
    687 }
    688 
    689 int
    690 ktrwrite(p, kth)
    691 	struct proc *p;
    692 	struct ktr_header *kth;
    693 {
    694 	struct uio auio;
    695 	struct iovec aiov[2];
    696 	int error, tries;
    697 	struct file *fp = p->p_tracep;
    698 
    699 	if (fp == NULL)
    700 		return 0;
    701 
    702 	auio.uio_iov = &aiov[0];
    703 	auio.uio_offset = 0;
    704 	auio.uio_segflg = UIO_SYSSPACE;
    705 	auio.uio_rw = UIO_WRITE;
    706 	aiov[0].iov_base = (caddr_t)kth;
    707 	aiov[0].iov_len = sizeof(struct ktr_header);
    708 	auio.uio_resid = sizeof(struct ktr_header);
    709 	auio.uio_iovcnt = 1;
    710 	auio.uio_procp = (struct proc *)0;
    711 	if (kth->ktr_len > 0) {
    712 		auio.uio_iovcnt++;
    713 		aiov[1].iov_base = (void *)kth->ktr_buf;
    714 		aiov[1].iov_len = kth->ktr_len;
    715 		auio.uio_resid += kth->ktr_len;
    716 	}
    717 
    718 	simple_lock(&fp->f_slock);
    719 	FILE_USE(fp);
    720 
    721 	tries = 0;
    722 	do {
    723 		error = (*fp->f_ops->fo_write)(fp, &fp->f_offset, &auio,
    724 		    fp->f_cred, FOF_UPDATE_OFFSET);
    725 		tries++;
    726 		if (error == EWOULDBLOCK)
    727 		  	preempt(1);
    728 	} while ((error == EWOULDBLOCK) && (tries < 3));
    729 	FILE_UNUSE(fp, NULL);
    730 
    731 	if (__predict_true(error == 0))
    732 		return (0);
    733 	/*
    734 	 * If error encountered, give up tracing on this vnode.  Don't report
    735 	 * EPIPE as this can easily happen with fktrace()/ktruss.
    736 	 */
    737 	if (error != EPIPE)
    738 		log(LOG_NOTICE,
    739 		    "ktrace write failed, errno %d, tracing stopped\n",
    740 		    error);
    741 	proclist_lock_read();
    742 	for (p = LIST_FIRST(&allproc); p != NULL; p = LIST_NEXT(p, p_list)) {
    743 		if (ktrsamefile(p->p_tracep, fp))
    744 			ktrderef(p);
    745 	}
    746 	proclist_unlock_read();
    747 
    748 	return (error);
    749 }
    750 
    751 /*
    752  * Return true if caller has permission to set the ktracing state
    753  * of target.  Essentially, the target can't possess any
    754  * more permissions than the caller.  KTRFAC_ROOT signifies that
    755  * root previously set the tracing status on the target process, and
    756  * so, only root may further change it.
    757  *
    758  * TODO: check groups.  use caller effective gid.
    759  */
    760 int
    761 ktrcanset(callp, targetp)
    762 	struct proc *callp;
    763 	struct proc *targetp;
    764 {
    765 	struct pcred *caller = callp->p_cred;
    766 	struct pcred *target = targetp->p_cred;
    767 
    768 	if ((caller->pc_ucred->cr_uid == target->p_ruid &&
    769 	     target->p_ruid == target->p_svuid &&
    770 	     caller->p_rgid == target->p_rgid &&	/* XXX */
    771 	     target->p_rgid == target->p_svgid &&
    772 	     (targetp->p_traceflag & KTRFAC_ROOT) == 0 &&
    773 	     (targetp->p_flag & P_SUGID) == 0) ||
    774 	     caller->pc_ucred->cr_uid == 0)
    775 		return (1);
    776 
    777 	return (0);
    778 }
    779 #endif /* KTRACE */
    780 
    781 /*
    782  * Put user defined entry to ktrace records.
    783  */
    784 int
    785 sys_utrace(l, v, retval)
    786 	struct lwp *l;
    787 	void *v;
    788 	register_t *retval;
    789 {
    790 #ifdef KTRACE
    791 	struct sys_utrace_args /* {
    792 		syscallarg(const char *) label;
    793 		syscallarg(void *) addr;
    794 		syscallarg(size_t) len;
    795 	} */ *uap = v;
    796 	struct proc *p = l->l_proc;
    797 	if (!KTRPOINT(p, KTR_USER))
    798 		return (0);
    799 
    800 	if (SCARG(uap, len) > KTR_USER_MAXLEN)
    801 		return (EINVAL);
    802 
    803 	ktruser(p, SCARG(uap, label), SCARG(uap, addr), SCARG(uap, len), 1);
    804 
    805 	return (0);
    806 #else /* !KTRACE */
    807 	return ENOSYS;
    808 #endif /* KTRACE */
    809 }
    810