xref: /illumos-gate/usr/src/cmd/mdb/common/modules/dtrace/dtrace.c (revision 45ede40b2394db7967e59f19288fae9b62efd4aa)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24  * Copyright (c) 2013 by Delphix. All rights reserved.
25  * Copyright 2019 Joyent, Inc.
26  */
27 
28 /*
29  * explicitly define DTRACE_ERRDEBUG to pull in definition of dtrace_errhash_t
30  * explicitly define _STDARG_H to avoid stdarg.h/varargs.h u/k defn conflict
31  */
32 #define	DTRACE_ERRDEBUG
33 #define	_STDARG_H
34 
35 #include <mdb/mdb_param.h>
36 #include <mdb/mdb_modapi.h>
37 #include <mdb/mdb_ks.h>
38 #include <sys/dtrace_impl.h>
39 #include <sys/vmem_impl.h>
40 #include <sys/ddi_impldefs.h>
41 #include <sys/sysmacros.h>
42 #include <sys/kobj.h>
43 #include <dtrace.h>
44 #include <alloca.h>
45 #include <ctype.h>
46 #include <errno.h>
47 #include <math.h>
48 #include <stdio.h>
49 #include <unistd.h>
50 
51 /*ARGSUSED*/
52 int
53 id2probe(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
54 {
55 	uintptr_t probe = 0;
56 	uintptr_t probes;
57 
58 	if (!(flags & DCMD_ADDRSPEC))
59 		return (DCMD_USAGE);
60 
61 	if (addr == DTRACE_IDNONE || addr > UINT32_MAX)
62 		goto out;
63 
64 	if (mdb_readvar(&probes, "dtrace_probes") == -1) {
65 		mdb_warn("failed to read 'dtrace_probes'");
66 		return (DCMD_ERR);
67 	}
68 
69 	probes += (addr - 1) * sizeof (dtrace_probe_t *);
70 
71 	if (mdb_vread(&probe, sizeof (uintptr_t), probes) == -1) {
72 		mdb_warn("failed to read dtrace_probes[%d]", addr - 1);
73 		return (DCMD_ERR);
74 	}
75 
76 out:
77 	mdb_printf("%p\n", probe);
78 	return (DCMD_OK);
79 }
80 
81 void
82 dtrace_help(void)
83 {
84 
85 	mdb_printf("Given a dtrace_state_t structure that represents a "
86 	    "DTrace consumer, prints\n"
87 	    "dtrace(1M)-like output for in-kernel DTrace data.  (The "
88 	    "dtrace_state_t\n"
89 	    "structures for all DTrace consumers may be obtained by running "
90 	    "the \n"
91 	    "::dtrace_state dcmd.)   When data is present on multiple CPUs, "
92 	    "data are\n"
93 	    "presented in CPU order, with records within each CPU ordered "
94 	    "oldest to \n"
95 	    "youngest.  Options:\n\n"
96 	    "-c cpu     Only provide output for specified CPU.\n");
97 }
98 
99 static int
100 dtracemdb_eprobe(dtrace_state_t *state, dtrace_eprobedesc_t *epd)
101 {
102 	dtrace_epid_t epid = epd->dtepd_epid;
103 	dtrace_probe_t probe;
104 	dtrace_ecb_t ecb;
105 	uintptr_t addr, paddr, ap;
106 	dtrace_action_t act;
107 	int nactions, nrecs;
108 
109 	addr = (uintptr_t)state->dts_ecbs +
110 	    (epid - 1) * sizeof (dtrace_ecb_t *);
111 
112 	if (mdb_vread(&addr, sizeof (addr), addr) == -1) {
113 		mdb_warn("failed to read ecb for epid %d", epid);
114 		return (-1);
115 	}
116 
117 	if (addr == 0) {
118 		mdb_warn("epid %d doesn't match an ecb\n", epid);
119 		return (-1);
120 	}
121 
122 	if (mdb_vread(&ecb, sizeof (ecb), addr) == -1) {
123 		mdb_warn("failed to read ecb at %p", addr);
124 		return (-1);
125 	}
126 
127 	paddr = (uintptr_t)ecb.dte_probe;
128 
129 	if (mdb_vread(&probe, sizeof (probe), paddr) == -1) {
130 		mdb_warn("failed to read probe for ecb %p", addr);
131 		return (-1);
132 	}
133 
134 	/*
135 	 * This is a little painful:  in order to find the number of actions,
136 	 * we need to first walk through them.
137 	 */
138 	for (ap = (uintptr_t)ecb.dte_action, nactions = 0; ap != 0; ) {
139 		if (mdb_vread(&act, sizeof (act), ap) == -1) {
140 			mdb_warn("failed to read action %p on ecb %p",
141 			    ap, addr);
142 			return (-1);
143 		}
144 
145 		if (!DTRACEACT_ISAGG(act.dta_kind) && !act.dta_intuple)
146 			nactions++;
147 
148 		ap = (uintptr_t)act.dta_next;
149 	}
150 
151 	nrecs = epd->dtepd_nrecs;
152 	epd->dtepd_nrecs = nactions;
153 	epd->dtepd_probeid = probe.dtpr_id;
154 	epd->dtepd_uarg = ecb.dte_uarg;
155 	epd->dtepd_size = ecb.dte_size;
156 
157 	for (ap = (uintptr_t)ecb.dte_action, nactions = 0; ap != 0; ) {
158 		if (mdb_vread(&act, sizeof (act), ap) == -1) {
159 			mdb_warn("failed to read action %p on ecb %p",
160 			    ap, addr);
161 			return (-1);
162 		}
163 
164 		if (!DTRACEACT_ISAGG(act.dta_kind) && !act.dta_intuple) {
165 			if (nrecs-- == 0)
166 				break;
167 
168 			epd->dtepd_rec[nactions++] = act.dta_rec;
169 		}
170 
171 		ap = (uintptr_t)act.dta_next;
172 	}
173 
174 	return (0);
175 }
176 
177 /*ARGSUSED*/
178 static int
179 dtracemdb_probe(dtrace_state_t *state, dtrace_probedesc_t *pd)
180 {
181 	uintptr_t base, addr, paddr, praddr;
182 	int nprobes, i;
183 	dtrace_probe_t probe;
184 	dtrace_provider_t prov;
185 
186 	if (pd->dtpd_id == DTRACE_IDNONE)
187 		pd->dtpd_id++;
188 
189 	if (mdb_readvar(&base, "dtrace_probes") == -1) {
190 		mdb_warn("failed to read 'dtrace_probes'");
191 		return (-1);
192 	}
193 
194 	if (mdb_readvar(&nprobes, "dtrace_nprobes") == -1) {
195 		mdb_warn("failed to read 'dtrace_nprobes'");
196 		return (-1);
197 	}
198 
199 	for (i = pd->dtpd_id; i <= nprobes; i++) {
200 		addr = base + (i - 1) * sizeof (dtrace_probe_t *);
201 
202 		if (mdb_vread(&paddr, sizeof (paddr), addr) == -1) {
203 			mdb_warn("couldn't read probe pointer at %p", addr);
204 			return (-1);
205 		}
206 
207 		if (paddr != 0)
208 			break;
209 	}
210 
211 	if (paddr == 0) {
212 		errno = ESRCH;
213 		return (-1);
214 	}
215 
216 	if (mdb_vread(&probe, sizeof (probe), paddr) == -1) {
217 		mdb_warn("couldn't read probe at %p", paddr);
218 		return (-1);
219 	}
220 
221 	pd->dtpd_id = probe.dtpr_id;
222 
223 	if (mdb_vread(pd->dtpd_name, DTRACE_NAMELEN,
224 	    (uintptr_t)probe.dtpr_name) == -1) {
225 		mdb_warn("failed to read probe name for probe %p", paddr);
226 		return (-1);
227 	}
228 
229 	if (mdb_vread(pd->dtpd_func, DTRACE_FUNCNAMELEN,
230 	    (uintptr_t)probe.dtpr_func) == -1) {
231 		mdb_warn("failed to read function name for probe %p", paddr);
232 		return (-1);
233 	}
234 
235 	if (mdb_vread(pd->dtpd_mod, DTRACE_MODNAMELEN,
236 	    (uintptr_t)probe.dtpr_mod) == -1) {
237 		mdb_warn("failed to read module name for probe %p", paddr);
238 		return (-1);
239 	}
240 
241 	praddr = (uintptr_t)probe.dtpr_provider;
242 
243 	if (mdb_vread(&prov, sizeof (prov), praddr) == -1) {
244 		mdb_warn("failed to read provider for probe %p", paddr);
245 		return (-1);
246 	}
247 
248 	if (mdb_vread(pd->dtpd_provider, DTRACE_PROVNAMELEN,
249 	    (uintptr_t)prov.dtpv_name) == -1) {
250 		mdb_warn("failed to read provider name for probe %p", paddr);
251 		return (-1);
252 	}
253 
254 	return (0);
255 }
256 
257 /*ARGSUSED*/
258 static int
259 dtracemdb_aggdesc(dtrace_state_t *state, dtrace_aggdesc_t *agd)
260 {
261 	dtrace_aggid_t aggid = agd->dtagd_id;
262 	dtrace_aggregation_t agg;
263 	dtrace_ecb_t ecb;
264 	uintptr_t addr, eaddr, ap, last;
265 	dtrace_action_t act;
266 	dtrace_recdesc_t *lrec;
267 	int nactions, nrecs;
268 
269 	addr = (uintptr_t)state->dts_aggregations +
270 	    (aggid - 1) * sizeof (dtrace_aggregation_t *);
271 
272 	if (mdb_vread(&addr, sizeof (addr), addr) == -1) {
273 		mdb_warn("failed to read aggregation for aggid %d", aggid);
274 		return (-1);
275 	}
276 
277 	if (addr == 0) {
278 		mdb_warn("aggid %d doesn't match an aggregation\n", aggid);
279 		return (-1);
280 	}
281 
282 	if (mdb_vread(&agg, sizeof (agg), addr) == -1) {
283 		mdb_warn("failed to read aggregation at %p", addr);
284 		return (-1);
285 	}
286 
287 	eaddr = (uintptr_t)agg.dtag_ecb;
288 
289 	if (mdb_vread(&ecb, sizeof (ecb), eaddr) == -1) {
290 		mdb_warn("failed to read ecb for aggregation %p", addr);
291 		return (-1);
292 	}
293 
294 	last = (uintptr_t)addr + offsetof(dtrace_aggregation_t, dtag_action);
295 
296 	/*
297 	 * This is a little painful:  in order to find the number of actions,
298 	 * we need to first walk through them.
299 	 */
300 	ap = (uintptr_t)agg.dtag_first;
301 	nactions = 0;
302 
303 	for (;;) {
304 		if (mdb_vread(&act, sizeof (act), ap) == -1) {
305 			mdb_warn("failed to read action %p on aggregation %p",
306 			    ap, addr);
307 			return (-1);
308 		}
309 
310 		nactions++;
311 
312 		if (ap == last)
313 			break;
314 
315 		ap = (uintptr_t)act.dta_next;
316 	}
317 
318 	lrec = &act.dta_rec;
319 	agd->dtagd_size = lrec->dtrd_offset + lrec->dtrd_size - agg.dtag_base;
320 
321 	nrecs = agd->dtagd_nrecs;
322 	agd->dtagd_nrecs = nactions;
323 	agd->dtagd_epid = ecb.dte_epid;
324 
325 	ap = (uintptr_t)agg.dtag_first;
326 	nactions = 0;
327 
328 	for (;;) {
329 		dtrace_recdesc_t rec;
330 
331 		if (mdb_vread(&act, sizeof (act), ap) == -1) {
332 			mdb_warn("failed to read action %p on aggregation %p",
333 			    ap, addr);
334 			return (-1);
335 		}
336 
337 		if (nrecs-- == 0)
338 			break;
339 
340 		rec = act.dta_rec;
341 		rec.dtrd_offset -= agg.dtag_base;
342 		rec.dtrd_uarg = 0;
343 		agd->dtagd_rec[nactions++] = rec;
344 
345 		if (ap == last)
346 			break;
347 
348 		ap = (uintptr_t)act.dta_next;
349 	}
350 
351 	return (0);
352 }
353 
354 static int
355 dtracemdb_bufsnap(dtrace_buffer_t *which, dtrace_bufdesc_t *desc)
356 {
357 	uintptr_t addr;
358 	size_t bufsize;
359 	dtrace_buffer_t buf;
360 	caddr_t data = desc->dtbd_data;
361 	processorid_t max_cpuid, cpu = desc->dtbd_cpu;
362 
363 	if (mdb_readvar(&max_cpuid, "max_cpuid") == -1) {
364 		mdb_warn("failed to read 'max_cpuid'");
365 		errno = EIO;
366 		return (-1);
367 	}
368 
369 	if (cpu < 0 || cpu > max_cpuid) {
370 		errno = EINVAL;
371 		return (-1);
372 	}
373 
374 	addr = (uintptr_t)which + cpu * sizeof (dtrace_buffer_t);
375 
376 	if (mdb_vread(&buf, sizeof (buf), addr) == -1) {
377 		mdb_warn("failed to read buffer description at %p", addr);
378 		errno = EIO;
379 		return (-1);
380 	}
381 
382 	if (buf.dtb_tomax == NULL) {
383 		errno = ENOENT;
384 		return (-1);
385 	}
386 
387 	if (buf.dtb_flags & DTRACEBUF_WRAPPED) {
388 		bufsize = buf.dtb_size;
389 	} else {
390 		bufsize = buf.dtb_offset;
391 	}
392 
393 	if (mdb_vread(data, bufsize, (uintptr_t)buf.dtb_tomax) == -1) {
394 		mdb_warn("couldn't read buffer for CPU %d", cpu);
395 		errno = EIO;
396 		return (-1);
397 	}
398 
399 	if (buf.dtb_offset > buf.dtb_size) {
400 		mdb_warn("buffer for CPU %d has corrupt offset\n", cpu);
401 		errno = EIO;
402 		return (-1);
403 	}
404 
405 	if (buf.dtb_flags & DTRACEBUF_WRAPPED) {
406 		if (buf.dtb_xamot_offset > buf.dtb_size) {
407 			mdb_warn("ringbuffer for CPU %d has corrupt "
408 			    "wrapped offset\n", cpu);
409 			errno = EIO;
410 			return (-1);
411 		}
412 
413 		/*
414 		 * If the ring buffer has wrapped, it needs to be polished.
415 		 * See the comment in dtrace_buffer_polish() for details.
416 		 */
417 		if (buf.dtb_offset < buf.dtb_xamot_offset) {
418 			bzero(data + buf.dtb_offset,
419 			    buf.dtb_xamot_offset - buf.dtb_offset);
420 		}
421 
422 		if (buf.dtb_offset > buf.dtb_xamot_offset) {
423 			bzero(data + buf.dtb_offset,
424 			    buf.dtb_size - buf.dtb_offset);
425 			bzero(data, buf.dtb_xamot_offset);
426 		}
427 
428 		desc->dtbd_oldest = buf.dtb_xamot_offset;
429 	} else {
430 		desc->dtbd_oldest = 0;
431 	}
432 
433 	desc->dtbd_size = bufsize;
434 	desc->dtbd_drops = buf.dtb_drops;
435 	desc->dtbd_errors = buf.dtb_errors;
436 	desc->dtbd_timestamp = gethrtime();
437 
438 	return (0);
439 }
440 
441 /*
442  * This is essentially identical to its cousin in the kernel -- with the
443  * notable exception that we automatically set DTRACEOPT_GRABANON if this
444  * state is an anonymous enabling.
445  */
446 static dof_hdr_t *
447 dtracemdb_dof_create(dtrace_state_t *state, int isanon)
448 {
449 	dof_hdr_t *dof;
450 	dof_sec_t *sec;
451 	dof_optdesc_t *opt;
452 	int i, len = sizeof (dof_hdr_t) +
453 	    roundup(sizeof (dof_sec_t), sizeof (uint64_t)) +
454 	    sizeof (dof_optdesc_t) * DTRACEOPT_MAX;
455 
456 	dof = mdb_zalloc(len, UM_SLEEP);
457 	dof->dofh_ident[DOF_ID_MAG0] = DOF_MAG_MAG0;
458 	dof->dofh_ident[DOF_ID_MAG1] = DOF_MAG_MAG1;
459 	dof->dofh_ident[DOF_ID_MAG2] = DOF_MAG_MAG2;
460 	dof->dofh_ident[DOF_ID_MAG3] = DOF_MAG_MAG3;
461 
462 	dof->dofh_ident[DOF_ID_MODEL] = DOF_MODEL_NATIVE;
463 	dof->dofh_ident[DOF_ID_ENCODING] = DOF_ENCODE_NATIVE;
464 	dof->dofh_ident[DOF_ID_VERSION] = DOF_VERSION;
465 	dof->dofh_ident[DOF_ID_DIFVERS] = DIF_VERSION;
466 	dof->dofh_ident[DOF_ID_DIFIREG] = DIF_DIR_NREGS;
467 	dof->dofh_ident[DOF_ID_DIFTREG] = DIF_DTR_NREGS;
468 
469 	dof->dofh_flags = 0;
470 	dof->dofh_hdrsize = sizeof (dof_hdr_t);
471 	dof->dofh_secsize = sizeof (dof_sec_t);
472 	dof->dofh_secnum = 1;	/* only DOF_SECT_OPTDESC */
473 	dof->dofh_secoff = sizeof (dof_hdr_t);
474 	dof->dofh_loadsz = len;
475 	dof->dofh_filesz = len;
476 	dof->dofh_pad = 0;
477 
478 	/*
479 	 * Fill in the option section header...
480 	 */
481 	sec = (dof_sec_t *)((uintptr_t)dof + sizeof (dof_hdr_t));
482 	sec->dofs_type = DOF_SECT_OPTDESC;
483 	sec->dofs_align = sizeof (uint64_t);
484 	sec->dofs_flags = DOF_SECF_LOAD;
485 	sec->dofs_entsize = sizeof (dof_optdesc_t);
486 
487 	opt = (dof_optdesc_t *)((uintptr_t)sec +
488 	    roundup(sizeof (dof_sec_t), sizeof (uint64_t)));
489 
490 	sec->dofs_offset = (uintptr_t)opt - (uintptr_t)dof;
491 	sec->dofs_size = sizeof (dof_optdesc_t) * DTRACEOPT_MAX;
492 
493 	for (i = 0; i < DTRACEOPT_MAX; i++) {
494 		opt[i].dofo_option = i;
495 		opt[i].dofo_strtab = DOF_SECIDX_NONE;
496 		opt[i].dofo_value = state->dts_options[i];
497 	}
498 
499 	if (isanon)
500 		opt[DTRACEOPT_GRABANON].dofo_value = 1;
501 
502 	return (dof);
503 }
504 
505 static int
506 dtracemdb_format(dtrace_state_t *state, dtrace_fmtdesc_t *desc)
507 {
508 	uintptr_t addr, faddr;
509 	char c;
510 	int len = 0;
511 
512 	if (desc->dtfd_format == 0 || desc->dtfd_format > state->dts_nformats) {
513 		errno = EINVAL;
514 		return (-1);
515 	}
516 
517 	faddr = (uintptr_t)state->dts_formats +
518 	    (desc->dtfd_format - 1) * sizeof (char *);
519 
520 	if (mdb_vread(&addr, sizeof (addr), faddr) == -1) {
521 		mdb_warn("failed to read format string pointer at %p", faddr);
522 		return (-1);
523 	}
524 
525 	do {
526 		if (mdb_vread(&c, sizeof (c), addr + len++) == -1) {
527 			mdb_warn("failed to read format string at %p", addr);
528 			return (-1);
529 		}
530 	} while (c != '\0');
531 
532 	if (len > desc->dtfd_length) {
533 		desc->dtfd_length = len;
534 		return (0);
535 	}
536 
537 	if (mdb_vread(desc->dtfd_string, len, addr) == -1) {
538 		mdb_warn("failed to reread format string at %p", addr);
539 		return (-1);
540 	}
541 
542 	return (0);
543 }
544 
545 static int
546 dtracemdb_status(dtrace_state_t *state, dtrace_status_t *status)
547 {
548 	dtrace_dstate_t *dstate;
549 	int i, j;
550 	uint64_t nerrs;
551 	uintptr_t addr;
552 	int ncpu;
553 
554 	if (mdb_readvar(&ncpu, "_ncpu") == -1) {
555 		mdb_warn("failed to read '_ncpu'");
556 		return (DCMD_ERR);
557 	}
558 
559 	bzero(status, sizeof (dtrace_status_t));
560 
561 	if (state->dts_activity == DTRACE_ACTIVITY_INACTIVE) {
562 		errno = ENOENT;
563 		return (-1);
564 	}
565 
566 	/*
567 	 * For the MDB backend, we never set dtst_exiting or dtst_filled.  This
568 	 * is by design:  we don't want the library to try to stop tracing,
569 	 * because it doesn't particularly mean anything.
570 	 */
571 	nerrs = state->dts_errors;
572 	dstate = &state->dts_vstate.dtvs_dynvars;
573 
574 	for (i = 0; i < ncpu; i++) {
575 		dtrace_dstate_percpu_t dcpu;
576 		dtrace_buffer_t buf;
577 
578 		addr = (uintptr_t)&dstate->dtds_percpu[i];
579 
580 		if (mdb_vread(&dcpu, sizeof (dcpu), addr) == -1) {
581 			mdb_warn("failed to read per-CPU dstate at %p", addr);
582 			return (-1);
583 		}
584 
585 		status->dtst_dyndrops += dcpu.dtdsc_drops;
586 		status->dtst_dyndrops_dirty += dcpu.dtdsc_dirty_drops;
587 		status->dtst_dyndrops_rinsing += dcpu.dtdsc_rinsing_drops;
588 
589 		addr = (uintptr_t)&state->dts_buffer[i];
590 
591 		if (mdb_vread(&buf, sizeof (buf), addr) == -1) {
592 			mdb_warn("failed to read per-CPU buffer at %p", addr);
593 			return (-1);
594 		}
595 
596 		nerrs += buf.dtb_errors;
597 
598 		for (j = 0; j < state->dts_nspeculations; j++) {
599 			dtrace_speculation_t spec;
600 
601 			addr = (uintptr_t)&state->dts_speculations[j];
602 
603 			if (mdb_vread(&spec, sizeof (spec), addr) == -1) {
604 				mdb_warn("failed to read "
605 				    "speculation at %p", addr);
606 				return (-1);
607 			}
608 
609 			addr = (uintptr_t)&spec.dtsp_buffer[i];
610 
611 			if (mdb_vread(&buf, sizeof (buf), addr) == -1) {
612 				mdb_warn("failed to read "
613 				    "speculative buffer at %p", addr);
614 				return (-1);
615 			}
616 
617 			status->dtst_specdrops += buf.dtb_xamot_drops;
618 		}
619 	}
620 
621 	status->dtst_specdrops_busy = state->dts_speculations_busy;
622 	status->dtst_specdrops_unavail = state->dts_speculations_unavail;
623 	status->dtst_errors = nerrs;
624 
625 	return (0);
626 }
627 
628 typedef struct dtracemdb_data {
629 	dtrace_state_t *dtmd_state;
630 	char *dtmd_symstr;
631 	char *dtmd_modstr;
632 	uintptr_t dtmd_addr;
633 	int dtmd_isanon;
634 } dtracemdb_data_t;
635 
636 static int
637 dtracemdb_ioctl(void *varg, int cmd, void *arg)
638 {
639 	dtracemdb_data_t *data = varg;
640 	dtrace_state_t *state = data->dtmd_state;
641 
642 	switch (cmd) {
643 	case DTRACEIOC_CONF: {
644 		dtrace_conf_t *conf = arg;
645 
646 		bzero(conf, sizeof (conf));
647 		conf->dtc_difversion = DIF_VERSION;
648 		conf->dtc_difintregs = DIF_DIR_NREGS;
649 		conf->dtc_diftupregs = DIF_DTR_NREGS;
650 		conf->dtc_ctfmodel = CTF_MODEL_NATIVE;
651 
652 		return (0);
653 	}
654 
655 	case DTRACEIOC_DOFGET: {
656 		dof_hdr_t *hdr = arg, *dof;
657 
658 		dof = dtracemdb_dof_create(state, data->dtmd_isanon);
659 		bcopy(dof, hdr, MIN(hdr->dofh_loadsz, dof->dofh_loadsz));
660 		mdb_free(dof, dof->dofh_loadsz);
661 
662 		return (0);
663 	}
664 
665 	case DTRACEIOC_BUFSNAP:
666 		return (dtracemdb_bufsnap(state->dts_buffer, arg));
667 
668 	case DTRACEIOC_AGGSNAP:
669 		return (dtracemdb_bufsnap(state->dts_aggbuffer, arg));
670 
671 	case DTRACEIOC_AGGDESC:
672 		return (dtracemdb_aggdesc(state, arg));
673 
674 	case DTRACEIOC_EPROBE:
675 		return (dtracemdb_eprobe(state, arg));
676 
677 	case DTRACEIOC_PROBES:
678 		return (dtracemdb_probe(state, arg));
679 
680 	case DTRACEIOC_FORMAT:
681 		return (dtracemdb_format(state, arg));
682 
683 	case DTRACEIOC_STATUS:
684 		return (dtracemdb_status(state, arg));
685 
686 	case DTRACEIOC_GO:
687 		*(processorid_t *)arg = -1;
688 		return (0);
689 
690 	case DTRACEIOC_ENABLE:
691 		errno = ENOTTY; /* see dt_open.c:dtrace_go() */
692 		return (-1);
693 
694 	case DTRACEIOC_PROVIDER:
695 	case DTRACEIOC_PROBEMATCH:
696 		errno = ESRCH;
697 		return (-1);
698 
699 	default:
700 		mdb_warn("unexpected ioctl 0x%x (%s)\n", cmd,
701 		    cmd == DTRACEIOC_PROVIDER	? "DTRACEIOC_PROVIDER" :
702 		    cmd == DTRACEIOC_PROBES	? "DTRACEIOC_PROBES" :
703 		    cmd == DTRACEIOC_BUFSNAP	? "DTRACEIOC_BUFSNAP" :
704 		    cmd == DTRACEIOC_PROBEMATCH	? "DTRACEIOC_PROBEMATCH" :
705 		    cmd == DTRACEIOC_ENABLE	? "DTRACEIOC_ENABLE" :
706 		    cmd == DTRACEIOC_AGGSNAP	? "DTRACEIOC_AGGSNAP" :
707 		    cmd == DTRACEIOC_EPROBE	? "DTRACEIOC_EPROBE" :
708 		    cmd == DTRACEIOC_PROBEARG	? "DTRACEIOC_PROBEARG" :
709 		    cmd == DTRACEIOC_CONF	? "DTRACEIOC_CONF" :
710 		    cmd == DTRACEIOC_STATUS	? "DTRACEIOC_STATUS" :
711 		    cmd == DTRACEIOC_GO		? "DTRACEIOC_GO" :
712 		    cmd == DTRACEIOC_STOP	? "DTRACEIOC_STOP" :
713 		    cmd == DTRACEIOC_AGGDESC	? "DTRACEIOC_AGGDESC" :
714 		    cmd == DTRACEIOC_FORMAT	? "DTRACEIOC_FORMAT" :
715 		    cmd == DTRACEIOC_DOFGET	? "DTRACEIOC_DOFGET" :
716 		    cmd == DTRACEIOC_REPLICATE	? "DTRACEIOC_REPLICATE" :
717 		    "???");
718 		errno = ENXIO;
719 		return (-1);
720 	}
721 }
722 
723 static int
724 dtracemdb_modctl(uintptr_t addr, const struct modctl *m, dtracemdb_data_t *data)
725 {
726 	struct module mod;
727 
728 	if (m->mod_mp == NULL)
729 		return (WALK_NEXT);
730 
731 	if (mdb_vread(&mod, sizeof (mod), (uintptr_t)m->mod_mp) == -1) {
732 		mdb_warn("couldn't read modctl %p's module", addr);
733 		return (WALK_NEXT);
734 	}
735 
736 	if ((uintptr_t)mod.text > data->dtmd_addr)
737 		return (WALK_NEXT);
738 
739 	if ((uintptr_t)mod.text + mod.text_size <= data->dtmd_addr)
740 		return (WALK_NEXT);
741 
742 	if (mdb_readstr(data->dtmd_modstr, MDB_SYM_NAMLEN,
743 	    (uintptr_t)m->mod_modname) == -1)
744 		return (WALK_ERR);
745 
746 	return (WALK_DONE);
747 }
748 
749 static int
750 dtracemdb_lookup_by_addr(void *varg, GElf_Addr addr, GElf_Sym *symp,
751     dtrace_syminfo_t *sip)
752 {
753 	dtracemdb_data_t *data = varg;
754 
755 	if (data->dtmd_symstr == NULL) {
756 		data->dtmd_symstr = mdb_zalloc(MDB_SYM_NAMLEN,
757 		    UM_SLEEP | UM_GC);
758 	}
759 
760 	if (data->dtmd_modstr == NULL) {
761 		data->dtmd_modstr = mdb_zalloc(MDB_SYM_NAMLEN,
762 		    UM_SLEEP | UM_GC);
763 	}
764 
765 	if (symp != NULL) {
766 		if (mdb_lookup_by_addr(addr, MDB_SYM_FUZZY, data->dtmd_symstr,
767 		    MDB_SYM_NAMLEN, symp) == -1)
768 			return (-1);
769 	}
770 
771 	if (sip != NULL) {
772 		data->dtmd_addr = addr;
773 
774 		(void) strcpy(data->dtmd_modstr, "???");
775 
776 		if (mdb_walk("modctl",
777 		    (mdb_walk_cb_t)dtracemdb_modctl, varg) == -1) {
778 			mdb_warn("couldn't walk 'modctl'");
779 			return (-1);
780 		}
781 
782 		sip->dts_object = data->dtmd_modstr;
783 		sip->dts_id = 0;
784 		sip->dts_name = symp != NULL ? data->dtmd_symstr : NULL;
785 	}
786 
787 	return (0);
788 }
789 
790 /*ARGSUSED*/
791 static int
792 dtracemdb_stat(void *varg, processorid_t cpu)
793 {
794 	GElf_Sym sym;
795 	cpu_t c;
796 	uintptr_t caddr, addr;
797 
798 	if (mdb_lookup_by_name("cpu", &sym) == -1) {
799 		mdb_warn("failed to find symbol for 'cpu'");
800 		return (-1);
801 	}
802 
803 	if (cpu * sizeof (uintptr_t) > sym.st_size)
804 		return (-1);
805 
806 	addr = (uintptr_t)sym.st_value + cpu * sizeof (uintptr_t);
807 
808 	if (mdb_vread(&caddr, sizeof (caddr), addr) == -1) {
809 		mdb_warn("failed to read cpu[%d]", cpu);
810 		return (-1);
811 	}
812 
813 	if (caddr == 0)
814 		return (-1);
815 
816 	if (mdb_vread(&c, sizeof (c), caddr) == -1) {
817 		mdb_warn("failed to read cpu at %p", caddr);
818 		return (-1);
819 	}
820 
821 	if (c.cpu_flags & CPU_POWEROFF) {
822 		return (P_POWEROFF);
823 	} else if (c.cpu_flags & CPU_SPARE) {
824 		return (P_SPARE);
825 	} else if (c.cpu_flags & CPU_FAULTED) {
826 		return (P_FAULTED);
827 	} else if (c.cpu_flags & CPU_DISABLED) {
828 		return (P_DISABLED);
829 	} else if ((c.cpu_flags & (CPU_READY | CPU_OFFLINE)) != CPU_READY) {
830 		return (P_OFFLINE);
831 	} else if (c.cpu_flags & CPU_ENABLE) {
832 		return (P_ONLINE);
833 	} else {
834 		return (P_NOINTR);
835 	}
836 }
837 
838 /*ARGSUSED*/
839 static long
840 dtracemdb_sysconf(void *varg, int name)
841 {
842 	int max_ncpus;
843 	processorid_t max_cpuid;
844 
845 	switch (name) {
846 	case _SC_CPUID_MAX:
847 		if (mdb_readvar(&max_cpuid, "max_cpuid") == -1) {
848 			mdb_warn("failed to read 'max_cpuid'");
849 			return (-1);
850 		}
851 
852 		return (max_cpuid);
853 
854 	case _SC_NPROCESSORS_MAX:
855 		if (mdb_readvar(&max_ncpus, "max_ncpus") == -1) {
856 			mdb_warn("failed to read 'max_ncpus'");
857 			return (-1);
858 		}
859 
860 		return (max_ncpus);
861 
862 	default:
863 		mdb_warn("unexpected sysconf code %d\n", name);
864 		return (-1);
865 	}
866 }
867 
868 const dtrace_vector_t dtrace_mdbops = {
869 	dtracemdb_ioctl,
870 	dtracemdb_lookup_by_addr,
871 	dtracemdb_stat,
872 	dtracemdb_sysconf
873 };
874 
875 typedef struct dtrace_dcmddata {
876 	dtrace_hdl_t *dtdd_dtp;
877 	int dtdd_cpu;
878 	int dtdd_quiet;
879 	int dtdd_flowindent;
880 	int dtdd_heading;
881 	FILE *dtdd_output;
882 } dtrace_dcmddata_t;
883 
884 /*
885  * Helper to grab all the content from a file, spit it into a string, and erase
886  * and reset the file.
887  */
888 static void
889 print_and_truncate_file(FILE *fp)
890 {
891 	long len;
892 	char *out;
893 
894 	/* flush, find length of file, seek to beginning, initialize buffer */
895 	if (fflush(fp) || (len = ftell(fp)) < 0 ||
896 	    fseek(fp, 0, SEEK_SET) < 0) {
897 		mdb_warn("couldn't prepare DTrace output file: %d\n", errno);
898 		return;
899 	}
900 
901 	out = mdb_alloc(len + 1, UM_SLEEP);
902 	out[len] = '\0';
903 
904 	/* read file into buffer, truncate file, and seek to beginning */
905 	if ((fread(out, len + 1, sizeof (char), fp) == 0 && ferror(fp)) ||
906 	    ftruncate(fileno(fp), 0) < 0 || fseek(fp, 0, SEEK_SET) < 0) {
907 		mdb_warn("couldn't read DTrace output file: %d\n", errno);
908 		mdb_free(out, len + 1);
909 		return;
910 	}
911 
912 	mdb_printf("%s", out);
913 	mdb_free(out, len + 1);
914 }
915 
916 /*ARGSUSED*/
917 static int
918 dtrace_dcmdrec(const dtrace_probedata_t *data,
919     const dtrace_recdesc_t *rec, void *arg)
920 {
921 	dtrace_dcmddata_t *dd = arg;
922 
923 	print_and_truncate_file(dd->dtdd_output);
924 
925 	if (rec == NULL) {
926 		/*
927 		 * We have processed the final record; output the newline if
928 		 * we're not in quiet mode.
929 		 */
930 		if (!dd->dtdd_quiet)
931 			mdb_printf("\n");
932 
933 		return (DTRACE_CONSUME_NEXT);
934 	}
935 
936 	return (DTRACE_CONSUME_THIS);
937 }
938 
939 /*ARGSUSED*/
940 static int
941 dtrace_dcmdprobe(const dtrace_probedata_t *data, void *arg)
942 {
943 	dtrace_probedesc_t *pd = data->dtpda_pdesc;
944 	processorid_t cpu = data->dtpda_cpu;
945 	dtrace_dcmddata_t *dd = arg;
946 	char name[DTRACE_FUNCNAMELEN + DTRACE_NAMELEN + 2];
947 
948 	if (dd->dtdd_cpu != -1UL && dd->dtdd_cpu != cpu)
949 		return (DTRACE_CONSUME_NEXT);
950 
951 	if (dd->dtdd_heading == 0) {
952 		if (!dd->dtdd_flowindent) {
953 			if (!dd->dtdd_quiet) {
954 				mdb_printf("%3s %6s %32s\n",
955 				    "CPU", "ID", "FUNCTION:NAME");
956 			}
957 		} else {
958 			mdb_printf("%3s %-41s\n", "CPU", "FUNCTION");
959 		}
960 		dd->dtdd_heading = 1;
961 	}
962 
963 	if (!dd->dtdd_flowindent) {
964 		if (!dd->dtdd_quiet) {
965 			(void) mdb_snprintf(name, sizeof (name), "%s:%s",
966 			    pd->dtpd_func, pd->dtpd_name);
967 
968 			mdb_printf("%3d %6d %32s ", cpu, pd->dtpd_id, name);
969 		}
970 	} else {
971 		int indent = data->dtpda_indent;
972 
973 		if (data->dtpda_flow == DTRACEFLOW_NONE) {
974 			(void) mdb_snprintf(name, sizeof (name), "%*s%s%s:%s",
975 			    indent, "", data->dtpda_prefix, pd->dtpd_func,
976 			    pd->dtpd_name);
977 		} else {
978 			(void) mdb_snprintf(name, sizeof (name), "%*s%s%s",
979 			    indent, "", data->dtpda_prefix, pd->dtpd_func);
980 		}
981 
982 		mdb_printf("%3d %-41s ", cpu, name);
983 	}
984 
985 	return (DTRACE_CONSUME_THIS);
986 }
987 
988 /*ARGSUSED*/
989 static int
990 dtrace_dcmderr(const dtrace_errdata_t *data, void *arg)
991 {
992 	mdb_warn(data->dteda_msg);
993 	return (DTRACE_HANDLE_OK);
994 }
995 
996 /*ARGSUSED*/
997 static int
998 dtrace_dcmddrop(const dtrace_dropdata_t *data, void *arg)
999 {
1000 	mdb_warn(data->dtdda_msg);
1001 	return (DTRACE_HANDLE_OK);
1002 }
1003 
1004 /*ARGSUSED*/
1005 static int
1006 dtrace_dcmdbuffered(const dtrace_bufdata_t *bufdata, void *arg)
1007 {
1008 	mdb_printf("%s", bufdata->dtbda_buffered);
1009 	return (DTRACE_HANDLE_OK);
1010 }
1011 
1012 /*ARGSUSED*/
1013 int
1014 dtrace(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1015 {
1016 	dtrace_state_t state;
1017 	dtrace_hdl_t *dtp;
1018 	int ncpu, err;
1019 	uintptr_t c = -1UL;
1020 	dtrace_dcmddata_t dd;
1021 	dtrace_optval_t val;
1022 	dtracemdb_data_t md;
1023 	int rval = DCMD_ERR;
1024 	dtrace_anon_t anon;
1025 
1026 	if (!(flags & DCMD_ADDRSPEC))
1027 		return (DCMD_USAGE);
1028 
1029 	if (mdb_getopts(argc, argv, 'c', MDB_OPT_UINTPTR, &c, NULL) != argc)
1030 		return (DCMD_USAGE);
1031 
1032 	if (mdb_readvar(&ncpu, "_ncpu") == -1) {
1033 		mdb_warn("failed to read '_ncpu'");
1034 		return (DCMD_ERR);
1035 	}
1036 
1037 	if (mdb_vread(&state, sizeof (state), addr) == -1) {
1038 		mdb_warn("couldn't read dtrace_state_t at %p", addr);
1039 		return (DCMD_ERR);
1040 	}
1041 
1042 	if (state.dts_anon != NULL) {
1043 		addr = (uintptr_t)state.dts_anon;
1044 
1045 		if (mdb_vread(&state, sizeof (state), addr) == -1) {
1046 			mdb_warn("couldn't read anonymous state at %p", addr);
1047 			return (DCMD_ERR);
1048 		}
1049 	}
1050 
1051 	bzero(&md, sizeof (md));
1052 	md.dtmd_state = &state;
1053 
1054 	if ((dtp = dtrace_vopen(DTRACE_VERSION, DTRACE_O_NOSYS, &err,
1055 	    &dtrace_mdbops, &md)) == NULL) {
1056 		mdb_warn("failed to initialize dtrace: %s\n",
1057 		    dtrace_errmsg(NULL, err));
1058 		return (DCMD_ERR);
1059 	}
1060 
1061 	/*
1062 	 * If this is the anonymous enabling, we need to set a bit indicating
1063 	 * that DTRACEOPT_GRABANON should be set.
1064 	 */
1065 	if (mdb_readvar(&anon, "dtrace_anon") == -1) {
1066 		mdb_warn("failed to read 'dtrace_anon'");
1067 		return (DCMD_ERR);
1068 	}
1069 
1070 	md.dtmd_isanon = ((uintptr_t)anon.dta_state == addr);
1071 
1072 	if (dtrace_go(dtp) != 0) {
1073 		mdb_warn("failed to initialize dtrace: %s\n",
1074 		    dtrace_errmsg(dtp, dtrace_errno(dtp)));
1075 		goto err;
1076 	}
1077 
1078 	bzero(&dd, sizeof (dd));
1079 	dd.dtdd_dtp = dtp;
1080 	dd.dtdd_cpu = c;
1081 
1082 	if (dtrace_getopt(dtp, "flowindent", &val) == -1) {
1083 		mdb_warn("couldn't get 'flowindent' option: %s\n",
1084 		    dtrace_errmsg(dtp, dtrace_errno(dtp)));
1085 		goto err;
1086 	}
1087 
1088 	dd.dtdd_flowindent = (val != DTRACEOPT_UNSET);
1089 
1090 	if (dtrace_getopt(dtp, "quiet", &val) == -1) {
1091 		mdb_warn("couldn't get 'quiet' option: %s\n",
1092 		    dtrace_errmsg(dtp, dtrace_errno(dtp)));
1093 		goto err;
1094 	}
1095 
1096 	dd.dtdd_quiet = (val != DTRACEOPT_UNSET);
1097 
1098 	if (dtrace_handle_err(dtp, dtrace_dcmderr, NULL) == -1) {
1099 		mdb_warn("couldn't add err handler: %s\n",
1100 		    dtrace_errmsg(dtp, dtrace_errno(dtp)));
1101 		goto err;
1102 	}
1103 
1104 	if (dtrace_handle_drop(dtp, dtrace_dcmddrop, NULL) == -1) {
1105 		mdb_warn("couldn't add drop handler: %s\n",
1106 		    dtrace_errmsg(dtp, dtrace_errno(dtp)));
1107 		goto err;
1108 	}
1109 
1110 	if (dtrace_handle_buffered(dtp, dtrace_dcmdbuffered, NULL) == -1) {
1111 		mdb_warn("couldn't add buffered handler: %s\n",
1112 		    dtrace_errmsg(dtp, dtrace_errno(dtp)));
1113 		goto err;
1114 	}
1115 
1116 	if (dtrace_status(dtp) == -1) {
1117 		mdb_warn("couldn't get status: %s\n",
1118 		    dtrace_errmsg(dtp, dtrace_errno(dtp)));
1119 		goto err;
1120 	}
1121 
1122 	if (dtrace_aggregate_snap(dtp) == -1) {
1123 		mdb_warn("couldn't snapshot aggregation: %s\n",
1124 		    dtrace_errmsg(dtp, dtrace_errno(dtp)));
1125 		goto err;
1126 	}
1127 
1128 	if ((dd.dtdd_output = tmpfile()) == NULL) {
1129 		mdb_warn("couldn't open DTrace output file: %d\n", errno);
1130 		goto err;
1131 	}
1132 
1133 	if (dtrace_consume(dtp, dd.dtdd_output,
1134 	    dtrace_dcmdprobe, dtrace_dcmdrec, &dd) == -1) {
1135 		mdb_warn("couldn't consume DTrace buffers: %s\n",
1136 		    dtrace_errmsg(dtp, dtrace_errno(dtp)));
1137 	}
1138 
1139 	if (dtrace_aggregate_print(dtp, NULL, NULL) == -1) {
1140 		mdb_warn("couldn't print aggregation: %s\n",
1141 		    dtrace_errmsg(dtp, dtrace_errno(dtp)));
1142 		goto err;
1143 	}
1144 
1145 	rval = DCMD_OK;
1146 err:
1147 	dtrace_close(dtp);
1148 	fclose(dd.dtdd_output);
1149 	return (rval);
1150 }
1151 
1152 static int
1153 dtrace_errhash_cmp(const void *l, const void *r)
1154 {
1155 	uintptr_t lhs = *((uintptr_t *)l);
1156 	uintptr_t rhs = *((uintptr_t *)r);
1157 	dtrace_errhash_t lerr, rerr;
1158 	char lmsg[256], rmsg[256];
1159 
1160 	(void) mdb_vread(&lerr, sizeof (lerr), lhs);
1161 	(void) mdb_vread(&rerr, sizeof (rerr), rhs);
1162 
1163 	if (lerr.dter_msg == NULL)
1164 		return (-1);
1165 
1166 	if (rerr.dter_msg == NULL)
1167 		return (1);
1168 
1169 	(void) mdb_readstr(lmsg, sizeof (lmsg), (uintptr_t)lerr.dter_msg);
1170 	(void) mdb_readstr(rmsg, sizeof (rmsg), (uintptr_t)rerr.dter_msg);
1171 
1172 	return (strcmp(lmsg, rmsg));
1173 }
1174 
1175 int
1176 dtrace_errhash_init(mdb_walk_state_t *wsp)
1177 {
1178 	GElf_Sym sym;
1179 	uintptr_t *hash, addr;
1180 	int i;
1181 
1182 	if (wsp->walk_addr != 0) {
1183 		mdb_warn("dtrace_errhash walk only supports global walks\n");
1184 		return (WALK_ERR);
1185 	}
1186 
1187 	if (mdb_lookup_by_name("dtrace_errhash", &sym) == -1) {
1188 		mdb_warn("couldn't find 'dtrace_errhash' (non-DEBUG kernel?)");
1189 		return (WALK_ERR);
1190 	}
1191 
1192 	addr = (uintptr_t)sym.st_value;
1193 	hash = mdb_alloc(DTRACE_ERRHASHSZ * sizeof (uintptr_t),
1194 	    UM_SLEEP | UM_GC);
1195 
1196 	for (i = 0; i < DTRACE_ERRHASHSZ; i++)
1197 		hash[i] = addr + i * sizeof (dtrace_errhash_t);
1198 
1199 	qsort(hash, DTRACE_ERRHASHSZ, sizeof (uintptr_t), dtrace_errhash_cmp);
1200 
1201 	wsp->walk_addr = 0;
1202 	wsp->walk_data = hash;
1203 
1204 	return (WALK_NEXT);
1205 }
1206 
1207 int
1208 dtrace_errhash_step(mdb_walk_state_t *wsp)
1209 {
1210 	int ndx = (int)wsp->walk_addr;
1211 	uintptr_t *hash = wsp->walk_data;
1212 	dtrace_errhash_t err;
1213 	uintptr_t addr;
1214 
1215 	if (ndx >= DTRACE_ERRHASHSZ)
1216 		return (WALK_DONE);
1217 
1218 	wsp->walk_addr = ndx + 1;
1219 	addr = hash[ndx];
1220 
1221 	if (mdb_vread(&err, sizeof (err), addr) == -1) {
1222 		mdb_warn("failed to read dtrace_errhash_t at %p", addr);
1223 		return (WALK_DONE);
1224 	}
1225 
1226 	if (err.dter_msg == NULL)
1227 		return (WALK_NEXT);
1228 
1229 	return (wsp->walk_callback(addr, &err, wsp->walk_cbdata));
1230 }
1231 
1232 /*ARGSUSED*/
1233 int
1234 dtrace_errhash(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1235 {
1236 	dtrace_errhash_t err;
1237 	char msg[256];
1238 
1239 	if (!(flags & DCMD_ADDRSPEC)) {
1240 		if (mdb_walk_dcmd("dtrace_errhash", "dtrace_errhash",
1241 		    argc, argv) == -1) {
1242 			mdb_warn("can't walk 'dtrace_errhash'");
1243 			return (DCMD_ERR);
1244 		}
1245 
1246 		return (DCMD_OK);
1247 	}
1248 
1249 	if (DCMD_HDRSPEC(flags))
1250 		mdb_printf("%8s %s\n", "COUNT", "ERROR");
1251 
1252 	if (mdb_vread(&err, sizeof (err), addr) == -1) {
1253 		mdb_warn("failed to read dtrace_errhash_t at %p", addr);
1254 		return (DCMD_ERR);
1255 	}
1256 
1257 	addr = (uintptr_t)err.dter_msg;
1258 
1259 	if (mdb_readstr(msg, sizeof (msg), addr) == -1) {
1260 		mdb_warn("failed to read error msg at %p", addr);
1261 		return (DCMD_ERR);
1262 	}
1263 
1264 	mdb_printf("%8d %s", err.dter_count, msg);
1265 
1266 	/*
1267 	 * Some error messages include a newline -- only print the newline
1268 	 * if the message doesn't have one.
1269 	 */
1270 	if (msg[strlen(msg) - 1] != '\n')
1271 		mdb_printf("\n");
1272 
1273 	return (DCMD_OK);
1274 }
1275 
1276 int
1277 dtrace_helptrace_init(mdb_walk_state_t *wsp)
1278 {
1279 	uint32_t next;
1280 	uintptr_t buffer;
1281 
1282 	if (wsp->walk_addr != 0) {
1283 		mdb_warn("dtrace_helptrace only supports global walks\n");
1284 		return (WALK_ERR);
1285 	}
1286 
1287 	if (mdb_readvar(&buffer, "dtrace_helptrace_buffer") == -1) {
1288 		mdb_warn("couldn't read 'dtrace_helptrace_buffer'");
1289 		return (WALK_ERR);
1290 	}
1291 
1292 	if (buffer == 0) {
1293 		mdb_warn("helper tracing is not enabled\n");
1294 		return (WALK_ERR);
1295 	}
1296 
1297 	if (mdb_readvar(&next, "dtrace_helptrace_next") == -1) {
1298 		mdb_warn("couldn't read 'dtrace_helptrace_next'");
1299 		return (WALK_ERR);
1300 	}
1301 
1302 	wsp->walk_addr = next;
1303 
1304 	return (WALK_NEXT);
1305 }
1306 
1307 int
1308 dtrace_helptrace_step(mdb_walk_state_t *wsp)
1309 {
1310 	uint32_t next, size, nlocals, bufsize;
1311 	uintptr_t buffer, addr;
1312 	dtrace_helptrace_t *ht;
1313 	int rval;
1314 
1315 	if (mdb_readvar(&next, "dtrace_helptrace_next") == -1) {
1316 		mdb_warn("couldn't read 'dtrace_helptrace_next'");
1317 		return (WALK_ERR);
1318 	}
1319 
1320 	if (mdb_readvar(&bufsize, "dtrace_helptrace_bufsize") == -1) {
1321 		mdb_warn("couldn't read 'dtrace_helptrace_bufsize'");
1322 		return (WALK_ERR);
1323 	}
1324 
1325 	if (mdb_readvar(&buffer, "dtrace_helptrace_buffer") == -1) {
1326 		mdb_warn("couldn't read 'dtrace_helptrace_buffer'");
1327 		return (WALK_ERR);
1328 	}
1329 
1330 	if (mdb_readvar(&nlocals, "dtrace_helptrace_nlocals") == -1) {
1331 		mdb_warn("couldn't read 'dtrace_helptrace_nlocals'");
1332 		return (WALK_ERR);
1333 	}
1334 
1335 	size = sizeof (dtrace_helptrace_t) +
1336 	    nlocals * sizeof (uint64_t) - sizeof (uint64_t);
1337 
1338 	if (wsp->walk_addr + size > bufsize) {
1339 		if (next == 0)
1340 			return (WALK_DONE);
1341 
1342 		wsp->walk_addr = 0;
1343 	}
1344 
1345 	addr = buffer + wsp->walk_addr;
1346 	ht = alloca(size);
1347 
1348 	if (mdb_vread(ht, size, addr) == -1) {
1349 		mdb_warn("couldn't read entry at %p", addr);
1350 		return (WALK_ERR);
1351 	}
1352 
1353 	if (ht->dtht_helper != NULL) {
1354 		rval = wsp->walk_callback(addr, ht, wsp->walk_cbdata);
1355 
1356 		if (rval != WALK_NEXT)
1357 			return (rval);
1358 	}
1359 
1360 	if (wsp->walk_addr < next && wsp->walk_addr + size >= next)
1361 		return (WALK_DONE);
1362 
1363 	wsp->walk_addr += size;
1364 	return (WALK_NEXT);
1365 }
1366 
1367 int
1368 dtrace_helptrace(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1369 {
1370 	dtrace_helptrace_t help;
1371 	dtrace_helper_action_t helper;
1372 	char where[30];
1373 	uint_t opt_v = FALSE;
1374 	uintptr_t haddr;
1375 
1376 	if (!(flags & DCMD_ADDRSPEC)) {
1377 		if (mdb_walk_dcmd("dtrace_helptrace", "dtrace_helptrace",
1378 		    argc, argv) == -1) {
1379 			mdb_warn("can't walk 'dtrace_helptrace'");
1380 			return (DCMD_ERR);
1381 		}
1382 
1383 		return (DCMD_OK);
1384 	}
1385 
1386 	if (mdb_getopts(argc, argv, 'v',
1387 	    MDB_OPT_SETBITS, TRUE, &opt_v, NULL) != argc)
1388 		return (DCMD_USAGE);
1389 
1390 	if (DCMD_HDRSPEC(flags)) {
1391 		mdb_printf(" %?s %?s %12s %s\n",
1392 		    "ADDR", "HELPER", "WHERE", "DIFO");
1393 	}
1394 
1395 	if (mdb_vread(&help, sizeof (help), addr) == -1) {
1396 		mdb_warn("failed to read dtrace_helptrace_t at %p", addr);
1397 		return (DCMD_ERR);
1398 	}
1399 
1400 	switch (help.dtht_where) {
1401 	case 0:
1402 		(void) mdb_snprintf(where, sizeof (where), "predicate");
1403 		break;
1404 
1405 	case DTRACE_HELPTRACE_NEXT:
1406 		(void) mdb_snprintf(where, sizeof (where), "next");
1407 		break;
1408 
1409 	case DTRACE_HELPTRACE_DONE:
1410 		(void) mdb_snprintf(where, sizeof (where), "done");
1411 		break;
1412 
1413 	case DTRACE_HELPTRACE_ERR:
1414 		(void) mdb_snprintf(where, sizeof (where), "err");
1415 		break;
1416 
1417 	default:
1418 		(void) mdb_snprintf(where, sizeof (where),
1419 		    "action #%d", help.dtht_where);
1420 		break;
1421 	}
1422 
1423 	mdb_printf(" %?p %?p %12s ", addr, help.dtht_helper, where);
1424 
1425 	haddr = (uintptr_t)help.dtht_helper;
1426 
1427 	if (mdb_vread(&helper, sizeof (helper), haddr) == -1) {
1428 		/*
1429 		 * We're not going to warn in this case -- we're just not going
1430 		 * to print anything exciting.
1431 		 */
1432 		mdb_printf("???\n");
1433 	} else {
1434 		switch (help.dtht_where) {
1435 		case 0:
1436 			mdb_printf("%p\n", helper.dtha_predicate);
1437 			break;
1438 
1439 		case DTRACE_HELPTRACE_NEXT:
1440 		case DTRACE_HELPTRACE_DONE:
1441 		case DTRACE_HELPTRACE_ERR:
1442 			mdb_printf("-\n");
1443 			break;
1444 
1445 		default:
1446 			haddr = (uintptr_t)helper.dtha_actions +
1447 			    (help.dtht_where - 1) * sizeof (uintptr_t);
1448 
1449 			if (mdb_vread(&haddr, sizeof (haddr), haddr) == -1) {
1450 				mdb_printf("???\n");
1451 			} else {
1452 				mdb_printf("%p\n", haddr);
1453 			}
1454 		}
1455 	}
1456 
1457 	if (opt_v) {
1458 		int i;
1459 
1460 		if (help.dtht_where == DTRACE_HELPTRACE_ERR) {
1461 			int f = help.dtht_fault;
1462 
1463 			mdb_printf("%?s| %?s %10s |\n", "", "", "");
1464 			mdb_printf("%?s| %?s %10s +->  fault: %s\n", "", "", "",
1465 			    f == DTRACEFLT_BADADDR ? "BADADDR" :
1466 			    f == DTRACEFLT_BADALIGN ? "BADALIGN" :
1467 			    f == DTRACEFLT_ILLOP ? "ILLOP" :
1468 			    f == DTRACEFLT_DIVZERO ? "DIVZERO" :
1469 			    f == DTRACEFLT_NOSCRATCH ? "NOSCRATCH" :
1470 			    f == DTRACEFLT_KPRIV ? "KPRIV" :
1471 			    f == DTRACEFLT_UPRIV ? "UPRIV" :
1472 			    f == DTRACEFLT_TUPOFLOW ? "TUPOFLOW" :
1473 			    f == DTRACEFLT_BADSTACK ? "BADSTACK" :
1474 			    "DTRACEFLT_UNKNOWN");
1475 			mdb_printf("%?s| %?s %12s     addr: 0x%x\n", "", "", "",
1476 			    help.dtht_illval);
1477 			mdb_printf("%?s| %?s %12s   offset: %d\n", "", "", "",
1478 			    help.dtht_fltoffs);
1479 		}
1480 
1481 		mdb_printf("%?s|\n%?s+--> %?s %4s %s\n", "", "",
1482 		    "ADDR", "NDX", "VALUE");
1483 		addr += sizeof (help) - sizeof (uint64_t);
1484 
1485 		for (i = 0; i < help.dtht_nlocals; i++) {
1486 			uint64_t val;
1487 
1488 			if (mdb_vread(&val, sizeof (val), addr) == -1) {
1489 				mdb_warn("couldn't read local at %p", addr);
1490 				continue;
1491 			}
1492 
1493 			mdb_printf("%?s     %?p %4d %p\n", "", addr, i, val);
1494 			addr += sizeof (uint64_t);
1495 		}
1496 
1497 		mdb_printf("\n");
1498 	}
1499 
1500 	return (DCMD_OK);
1501 }
1502 
1503 /*ARGSUSED*/
1504 static int
1505 dtrace_state_walk(uintptr_t addr, const vmem_seg_t *seg, minor_t *highest)
1506 {
1507 	if (seg->vs_end > *highest)
1508 		*highest = seg->vs_end;
1509 
1510 	return (WALK_NEXT);
1511 }
1512 
1513 typedef struct dtrace_state_walk {
1514 	uintptr_t dtsw_softstate;
1515 	minor_t dtsw_max;
1516 	minor_t dtsw_current;
1517 } dtrace_state_walk_t;
1518 
1519 int
1520 dtrace_state_init(mdb_walk_state_t *wsp)
1521 {
1522 	uintptr_t dtrace_minor;
1523 	minor_t max = 0;
1524 	dtrace_state_walk_t *dw;
1525 
1526 	if (wsp->walk_addr != 0) {
1527 		mdb_warn("dtrace_state only supports global walks\n");
1528 		return (WALK_ERR);
1529 	}
1530 
1531 	/*
1532 	 * Find the dtrace_minor vmem arena and walk it to get the maximum
1533 	 * minor number.
1534 	 */
1535 	if (mdb_readvar(&dtrace_minor, "dtrace_minor") == -1) {
1536 		mdb_warn("failed to read 'dtrace_minor'");
1537 		return (WALK_ERR);
1538 	}
1539 
1540 	if (mdb_pwalk("vmem_alloc", (mdb_walk_cb_t)dtrace_state_walk,
1541 	    &max, dtrace_minor) == -1) {
1542 		mdb_warn("couldn't walk 'vmem_alloc'");
1543 		return (WALK_ERR);
1544 	}
1545 
1546 	dw = mdb_zalloc(sizeof (dtrace_state_walk_t), UM_SLEEP | UM_GC);
1547 	dw->dtsw_current = 0;
1548 	dw->dtsw_max = max;
1549 
1550 	if (mdb_readvar(&dw->dtsw_softstate, "dtrace_softstate") == -1) {
1551 		mdb_warn("failed to read 'dtrace_softstate'");
1552 		return (DCMD_ERR);
1553 	}
1554 
1555 	wsp->walk_data = dw;
1556 
1557 	return (WALK_NEXT);
1558 }
1559 
1560 int
1561 dtrace_state_step(mdb_walk_state_t *wsp)
1562 {
1563 	dtrace_state_walk_t *dw = wsp->walk_data;
1564 	uintptr_t statep;
1565 	dtrace_state_t state;
1566 	int rval;
1567 
1568 	while (mdb_get_soft_state_byaddr(dw->dtsw_softstate, dw->dtsw_current,
1569 	    &statep, NULL, 0) == -1) {
1570 		if (dw->dtsw_current >= dw->dtsw_max)
1571 			return (WALK_DONE);
1572 
1573 		dw->dtsw_current++;
1574 	}
1575 
1576 	if (mdb_vread(&state, sizeof (state), statep) == -1) {
1577 		mdb_warn("couldn't read dtrace_state_t at %p", statep);
1578 		return (WALK_NEXT);
1579 	}
1580 
1581 	rval = wsp->walk_callback(statep, &state, wsp->walk_cbdata);
1582 	dw->dtsw_current++;
1583 
1584 	return (rval);
1585 }
1586 
1587 typedef struct dtrace_state_data {
1588 	int dtsd_major;
1589 	uintptr_t dtsd_proc;
1590 	uintptr_t dtsd_softstate;
1591 	uintptr_t dtsd_state;
1592 } dtrace_state_data_t;
1593 
1594 static int
1595 dtrace_state_file(uintptr_t addr, struct file *f, dtrace_state_data_t *data)
1596 {
1597 	vnode_t vnode;
1598 	proc_t proc;
1599 	minor_t minor;
1600 	uintptr_t statep;
1601 
1602 	if (mdb_vread(&vnode, sizeof (vnode), (uintptr_t)f->f_vnode) == -1) {
1603 		mdb_warn("couldn't read vnode at %p", (uintptr_t)f->f_vnode);
1604 		return (WALK_NEXT);
1605 	}
1606 
1607 	if (getmajor(vnode.v_rdev) != data->dtsd_major)
1608 		return (WALK_NEXT);
1609 
1610 	minor = getminor(vnode.v_rdev);
1611 
1612 	if (mdb_vread(&proc, sizeof (proc), data->dtsd_proc) == -1) {
1613 		mdb_warn("failed to read proc at %p", data->dtsd_proc);
1614 		return (WALK_NEXT);
1615 	}
1616 
1617 	if (mdb_get_soft_state_byaddr(data->dtsd_softstate, minor,
1618 	    &statep, NULL, 0) == -1) {
1619 		mdb_warn("failed to read softstate for minor %d", minor);
1620 		return (WALK_NEXT);
1621 	}
1622 
1623 	if (statep != data->dtsd_state)
1624 		return (WALK_NEXT);
1625 
1626 	mdb_printf("%?p %5d %?p %-*s %?p\n", statep, minor,
1627 	    data->dtsd_proc, MAXCOMLEN, proc.p_user.u_comm, addr);
1628 
1629 	return (WALK_NEXT);
1630 }
1631 
1632 /*ARGSUSED*/
1633 static int
1634 dtrace_state_proc(uintptr_t addr, void *ignored, dtrace_state_data_t *data)
1635 {
1636 	data->dtsd_proc = addr;
1637 
1638 	if (mdb_pwalk("file",
1639 	    (mdb_walk_cb_t)dtrace_state_file, data, addr) == -1) {
1640 		mdb_warn("couldn't walk 'file' for proc %p", addr);
1641 		return (WALK_ERR);
1642 	}
1643 
1644 	return (WALK_NEXT);
1645 }
1646 
1647 void
1648 dtrace_state_help(void)
1649 {
1650 	mdb_printf("Given a dtrace_state_t structure, displays all "
1651 	    /*CSTYLED*/
1652 	    "consumers, or \"<anonymous>\"\nif the consumer is anonymous.  If "
1653 	    "no state structure is provided, iterates\nover all state "
1654 	    "structures.\n\n"
1655 	    "Addresses in ADDR column may be provided to ::dtrace to obtain\n"
1656 	    "dtrace(1M)-like output for in-kernel DTrace data.\n");
1657 }
1658 
1659 int
1660 dtrace_state(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
1661 {
1662 	uintptr_t devi;
1663 	struct dev_info info;
1664 	dtrace_state_data_t data;
1665 	dtrace_anon_t anon;
1666 	dtrace_state_t state;
1667 
1668 	if (!(flags & DCMD_ADDRSPEC)) {
1669 		if (mdb_walk_dcmd("dtrace_state",
1670 		    "dtrace_state", argc, argv) == -1) {
1671 			mdb_warn("can't walk dtrace_state");
1672 			return (DCMD_ERR);
1673 		}
1674 		return (DCMD_OK);
1675 	}
1676 
1677 	if (DCMD_HDRSPEC(flags)) {
1678 		mdb_printf("%?s %5s %?s %-*s %?s\n", "ADDR", "MINOR", "PROC",
1679 		    MAXCOMLEN, "NAME", "FILE");
1680 	}
1681 
1682 	/*
1683 	 * First determine if this is anonymous state.
1684 	 */
1685 	if (mdb_readvar(&anon, "dtrace_anon") == -1) {
1686 		mdb_warn("failed to read 'dtrace_anon'");
1687 		return (DCMD_ERR);
1688 	}
1689 
1690 	if ((uintptr_t)anon.dta_state == addr) {
1691 		if (mdb_vread(&state, sizeof (state), addr) == -1) {
1692 			mdb_warn("failed to read anon at %p", addr);
1693 			return (DCMD_ERR);
1694 		}
1695 
1696 		mdb_printf("%?p %5d %?s %-*s %?s\n", addr,
1697 		    getminor(state.dts_dev), "-", MAXCOMLEN,
1698 		    "<anonymous>", "-");
1699 
1700 		return (DCMD_OK);
1701 	}
1702 
1703 	if (mdb_readvar(&devi, "dtrace_devi") == -1) {
1704 		mdb_warn("failed to read 'dtrace_devi'");
1705 		return (DCMD_ERR);
1706 	}
1707 
1708 	if (mdb_vread(&info, sizeof (struct dev_info), devi) == -1) {
1709 		mdb_warn("failed to read 'dev_info'");
1710 		return (DCMD_ERR);
1711 	}
1712 
1713 	data.dtsd_major = info.devi_major;
1714 
1715 	if (mdb_readvar(&data.dtsd_softstate, "dtrace_softstate") == -1) {
1716 		mdb_warn("failed to read 'dtrace_softstate'");
1717 		return (DCMD_ERR);
1718 	}
1719 
1720 	data.dtsd_state = addr;
1721 
1722 	/*
1723 	 * Walk through all processes and all open files looking for this
1724 	 * state.  It must be open somewhere...
1725 	 */
1726 	if (mdb_walk("proc", (mdb_walk_cb_t)dtrace_state_proc, &data) == -1) {
1727 		mdb_warn("couldn't walk 'proc'");
1728 		return (DCMD_ERR);
1729 	}
1730 
1731 	return (DCMD_OK);
1732 }
1733 
1734 typedef struct dtrace_aggkey_data {
1735 	uintptr_t *dtakd_hash;
1736 	uintptr_t dtakd_hashsize;
1737 	uintptr_t dtakd_next;
1738 	uintptr_t dtakd_ndx;
1739 } dtrace_aggkey_data_t;
1740 
1741 int
1742 dtrace_aggkey_init(mdb_walk_state_t *wsp)
1743 {
1744 	dtrace_buffer_t buf;
1745 	uintptr_t addr;
1746 	dtrace_aggbuffer_t agb;
1747 	dtrace_aggkey_data_t *data;
1748 	size_t hsize;
1749 
1750 	if ((addr = wsp->walk_addr) == 0) {
1751 		mdb_warn("dtrace_aggkey walk needs aggregation buffer\n");
1752 		return (WALK_ERR);
1753 	}
1754 
1755 	if (mdb_vread(&buf, sizeof (buf), addr) == -1) {
1756 		mdb_warn("failed to read aggregation buffer at %p", addr);
1757 		return (WALK_ERR);
1758 	}
1759 
1760 	addr = (uintptr_t)buf.dtb_tomax +
1761 	    buf.dtb_size - sizeof (dtrace_aggbuffer_t);
1762 
1763 	if (mdb_vread(&agb, sizeof (agb), addr) == -1) {
1764 		mdb_warn("failed to read dtrace_aggbuffer_t at %p", addr);
1765 		return (WALK_ERR);
1766 	}
1767 
1768 	data = mdb_zalloc(sizeof (dtrace_aggkey_data_t), UM_SLEEP);
1769 
1770 	data->dtakd_hashsize = agb.dtagb_hashsize;
1771 	hsize = agb.dtagb_hashsize * sizeof (dtrace_aggkey_t *);
1772 	data->dtakd_hash = mdb_alloc(hsize, UM_SLEEP);
1773 
1774 	if (mdb_vread(data->dtakd_hash, hsize,
1775 	    (uintptr_t)agb.dtagb_hash) == -1) {
1776 		mdb_warn("failed to read hash at %p",
1777 		    (uintptr_t)agb.dtagb_hash);
1778 		mdb_free(data->dtakd_hash, hsize);
1779 		mdb_free(data, sizeof (dtrace_aggkey_data_t));
1780 		return (WALK_ERR);
1781 	}
1782 
1783 	wsp->walk_data = data;
1784 	return (WALK_NEXT);
1785 }
1786 
1787 int
1788 dtrace_aggkey_step(mdb_walk_state_t *wsp)
1789 {
1790 	dtrace_aggkey_data_t *data = wsp->walk_data;
1791 	dtrace_aggkey_t key;
1792 	uintptr_t addr;
1793 
1794 	while ((addr = data->dtakd_next) == 0) {
1795 		if (data->dtakd_ndx == data->dtakd_hashsize)
1796 			return (WALK_DONE);
1797 
1798 		data->dtakd_next = data->dtakd_hash[data->dtakd_ndx++];
1799 	}
1800 
1801 	if (mdb_vread(&key, sizeof (key), addr) == -1) {
1802 		mdb_warn("failed to read dtrace_aggkey_t at %p", addr);
1803 		return (WALK_ERR);
1804 	}
1805 
1806 	data->dtakd_next = (uintptr_t)key.dtak_next;
1807 
1808 	return (wsp->walk_callback(addr, &key, wsp->walk_cbdata));
1809 }
1810 
1811 void
1812 dtrace_aggkey_fini(mdb_walk_state_t *wsp)
1813 {
1814 	dtrace_aggkey_data_t *data = wsp->walk_data;
1815 	size_t hsize;
1816 
1817 	hsize = data->dtakd_hashsize * sizeof (dtrace_aggkey_t *);
1818 	mdb_free(data->dtakd_hash, hsize);
1819 	mdb_free(data, sizeof (dtrace_aggkey_data_t));
1820 }
1821 
1822 typedef struct dtrace_dynvar_data {
1823 	dtrace_dynhash_t *dtdvd_hash;
1824 	uintptr_t dtdvd_hashsize;
1825 	uintptr_t dtdvd_next;
1826 	uintptr_t dtdvd_ndx;
1827 	uintptr_t dtdvd_sink;
1828 } dtrace_dynvar_data_t;
1829 
1830 int
1831 dtrace_dynvar_init(mdb_walk_state_t *wsp)
1832 {
1833 	uintptr_t addr;
1834 	dtrace_dstate_t dstate;
1835 	dtrace_dynvar_data_t *data;
1836 	size_t hsize;
1837 	GElf_Sym sym;
1838 
1839 	if ((addr = wsp->walk_addr) == 0) {
1840 		mdb_warn("dtrace_dynvar walk needs dtrace_dstate_t\n");
1841 		return (WALK_ERR);
1842 	}
1843 
1844 	if (mdb_vread(&dstate, sizeof (dstate), addr) == -1) {
1845 		mdb_warn("failed to read dynamic state at %p", addr);
1846 		return (WALK_ERR);
1847 	}
1848 
1849 	if (mdb_lookup_by_name("dtrace_dynhash_sink", &sym) == -1) {
1850 		mdb_warn("couldn't find 'dtrace_dynhash_sink'");
1851 		return (WALK_ERR);
1852 	}
1853 
1854 	data = mdb_zalloc(sizeof (dtrace_dynvar_data_t), UM_SLEEP);
1855 
1856 	data->dtdvd_hashsize = dstate.dtds_hashsize;
1857 	hsize = dstate.dtds_hashsize * sizeof (dtrace_dynhash_t);
1858 	data->dtdvd_hash = mdb_alloc(hsize, UM_SLEEP);
1859 	data->dtdvd_sink = (uintptr_t)sym.st_value;
1860 
1861 	if (mdb_vread(data->dtdvd_hash, hsize,
1862 	    (uintptr_t)dstate.dtds_hash) == -1) {
1863 		mdb_warn("failed to read hash at %p",
1864 		    (uintptr_t)dstate.dtds_hash);
1865 		mdb_free(data->dtdvd_hash, hsize);
1866 		mdb_free(data, sizeof (dtrace_dynvar_data_t));
1867 		return (WALK_ERR);
1868 	}
1869 
1870 	data->dtdvd_next = (uintptr_t)data->dtdvd_hash[0].dtdh_chain;
1871 
1872 	wsp->walk_data = data;
1873 	return (WALK_NEXT);
1874 }
1875 
1876 int
1877 dtrace_dynvar_step(mdb_walk_state_t *wsp)
1878 {
1879 	dtrace_dynvar_data_t *data = wsp->walk_data;
1880 	dtrace_dynvar_t dynvar, *dvar;
1881 	size_t dvarsize;
1882 	uintptr_t addr;
1883 	int nkeys;
1884 
1885 	while ((addr = data->dtdvd_next) == data->dtdvd_sink) {
1886 		if (data->dtdvd_ndx == data->dtdvd_hashsize)
1887 			return (WALK_DONE);
1888 
1889 		data->dtdvd_next =
1890 		    (uintptr_t)data->dtdvd_hash[data->dtdvd_ndx++].dtdh_chain;
1891 	}
1892 
1893 	if (mdb_vread(&dynvar, sizeof (dynvar), addr) == -1) {
1894 		mdb_warn("failed to read dtrace_dynvar_t at %p", addr);
1895 		return (WALK_ERR);
1896 	}
1897 
1898 	/*
1899 	 * Now we need to allocate the correct size.
1900 	 */
1901 	nkeys = dynvar.dtdv_tuple.dtt_nkeys;
1902 	dvarsize = (uintptr_t)&dynvar.dtdv_tuple.dtt_key[nkeys] -
1903 	    (uintptr_t)&dynvar;
1904 
1905 	dvar = alloca(dvarsize);
1906 
1907 	if (mdb_vread(dvar, dvarsize, addr) == -1) {
1908 		mdb_warn("failed to read dtrace_dynvar_t at %p", addr);
1909 		return (WALK_ERR);
1910 	}
1911 
1912 	data->dtdvd_next = (uintptr_t)dynvar.dtdv_next;
1913 
1914 	return (wsp->walk_callback(addr, dvar, wsp->walk_cbdata));
1915 }
1916 
1917 void
1918 dtrace_dynvar_fini(mdb_walk_state_t *wsp)
1919 {
1920 	dtrace_dynvar_data_t *data = wsp->walk_data;
1921 	size_t hsize;
1922 
1923 	hsize = data->dtdvd_hashsize * sizeof (dtrace_dynvar_t *);
1924 	mdb_free(data->dtdvd_hash, hsize);
1925 	mdb_free(data, sizeof (dtrace_dynvar_data_t));
1926 }
1927 
1928 typedef struct dtrace_hashstat_data {
1929 	size_t *dthsd_counts;
1930 	size_t dthsd_hashsize;
1931 	char *dthsd_data;
1932 	size_t dthsd_size;
1933 	int dthsd_header;
1934 } dtrace_hashstat_data_t;
1935 
1936 typedef void (*dtrace_hashstat_func_t)(dtrace_hashstat_data_t *);
1937 
1938 static void
1939 dtrace_hashstat_additive(dtrace_hashstat_data_t *data)
1940 {
1941 	int i;
1942 	int hval = 0;
1943 
1944 	for (i = 0; i < data->dthsd_size; i++)
1945 		hval += data->dthsd_data[i];
1946 
1947 	data->dthsd_counts[hval % data->dthsd_hashsize]++;
1948 }
1949 
1950 static void
1951 dtrace_hashstat_shifty(dtrace_hashstat_data_t *data)
1952 {
1953 	uint64_t hval = 0;
1954 	int i;
1955 
1956 	if (data->dthsd_size < sizeof (uint64_t)) {
1957 		dtrace_hashstat_additive(data);
1958 		return;
1959 	}
1960 
1961 	for (i = 0; i < data->dthsd_size; i += sizeof (uint64_t)) {
1962 		/* LINTED - alignment */
1963 		uint64_t val = *((uint64_t *)&data->dthsd_data[i]);
1964 
1965 		hval += (val & ((1 << NBBY) - 1)) +
1966 		    ((val >> NBBY) & ((1 << NBBY) - 1)) +
1967 		    ((val >> (NBBY << 1)) & ((1 << NBBY) - 1)) +
1968 		    ((val >> (NBBY << 2)) & ((1 << NBBY) - 1)) +
1969 		    (val & USHRT_MAX) + (val >> (NBBY << 1) & USHRT_MAX);
1970 	}
1971 
1972 	data->dthsd_counts[hval % data->dthsd_hashsize]++;
1973 }
1974 
1975 static void
1976 dtrace_hashstat_knuth(dtrace_hashstat_data_t *data)
1977 {
1978 	int i;
1979 	int hval = data->dthsd_size;
1980 
1981 	for (i = 0; i < data->dthsd_size; i++)
1982 		hval = (hval << 4) ^ (hval >> 28) ^ data->dthsd_data[i];
1983 
1984 	data->dthsd_counts[hval % data->dthsd_hashsize]++;
1985 }
1986 
1987 static void
1988 dtrace_hashstat_oneatatime(dtrace_hashstat_data_t *data)
1989 {
1990 	int i;
1991 	uint32_t hval = 0;
1992 
1993 	for (i = 0; i < data->dthsd_size; i++) {
1994 		hval += data->dthsd_data[i];
1995 		hval += (hval << 10);
1996 		hval ^= (hval >> 6);
1997 	}
1998 
1999 	hval += (hval << 3);
2000 	hval ^= (hval >> 11);
2001 	hval += (hval << 15);
2002 
2003 	data->dthsd_counts[hval % data->dthsd_hashsize]++;
2004 }
2005 
2006 static void
2007 dtrace_hashstat_fnv(dtrace_hashstat_data_t *data)
2008 {
2009 	static const uint32_t prime = 0x01000193;
2010 	uint32_t hval = 0;
2011 	int i;
2012 
2013 	for (i = 0; i < data->dthsd_size; i++) {
2014 		hval *= prime;
2015 		hval ^= data->dthsd_data[i];
2016 	}
2017 
2018 	data->dthsd_counts[hval % data->dthsd_hashsize]++;
2019 }
2020 
2021 static void
2022 dtrace_hashstat_stats(char *name, dtrace_hashstat_data_t *data)
2023 {
2024 	size_t nz = 0, i;
2025 	int longest = 0;
2026 	size_t ttl = 0;
2027 	double sum = 0.0;
2028 	double avg;
2029 	uint_t util, stddev;
2030 
2031 	if (!data->dthsd_header) {
2032 		mdb_printf("%15s %11s %11s %11s %11s %11s\n", "NAME",
2033 		    "HASHSIZE", "%UTIL", "LONGEST", "AVERAGE", "STDDEV");
2034 		data->dthsd_header = 1;
2035 	}
2036 
2037 	for (i = 0; i < data->dthsd_hashsize; i++) {
2038 		if (data->dthsd_counts[i] != 0) {
2039 			nz++;
2040 
2041 			if (data->dthsd_counts[i] > longest)
2042 				longest = data->dthsd_counts[i];
2043 
2044 			ttl += data->dthsd_counts[i];
2045 		}
2046 	}
2047 
2048 	if (nz == 0) {
2049 		mdb_printf("%15s %11d %11s %11s %11s %11s\n", name,
2050 		    data->dthsd_hashsize, "-", "-", "-", "-");
2051 		return;
2052 	}
2053 
2054 	avg = (double)ttl / (double)nz;
2055 
2056 	for (i = 0; i < data->dthsd_hashsize; i++) {
2057 		double delta = (double)data->dthsd_counts[i] - avg;
2058 
2059 		if (data->dthsd_counts[i] == 0)
2060 			continue;
2061 
2062 		sum += delta * delta;
2063 	}
2064 
2065 	util = (nz * 1000) / data->dthsd_hashsize;
2066 	stddev = (uint_t)sqrt(sum / (double)nz) * 10;
2067 
2068 	mdb_printf("%15s %11d %9u.%1u %11d %11d %9u.%1u\n", name,
2069 	    data->dthsd_hashsize, util / 10, util % 10, longest, ttl / nz,
2070 	    stddev / 10, stddev % 10);
2071 }
2072 
2073 static struct dtrace_hashstat {
2074 	char *dths_name;
2075 	dtrace_hashstat_func_t dths_func;
2076 } _dtrace_hashstat[] = {
2077 	{ "<actual>", NULL },
2078 	{ "additive", dtrace_hashstat_additive },
2079 	{ "shifty", dtrace_hashstat_shifty },
2080 	{ "knuth", dtrace_hashstat_knuth },
2081 	{ "one-at-a-time", dtrace_hashstat_oneatatime },
2082 	{ "fnv", dtrace_hashstat_fnv },
2083 	{ NULL, 0 }
2084 };
2085 
2086 typedef struct dtrace_aggstat_data {
2087 	dtrace_hashstat_data_t dtagsd_hash;
2088 	dtrace_hashstat_func_t dtagsd_func;
2089 } dtrace_aggstat_data_t;
2090 
2091 static int
2092 dtrace_aggstat_walk(uintptr_t addr, dtrace_aggkey_t *key,
2093     dtrace_aggstat_data_t *data)
2094 {
2095 	dtrace_hashstat_data_t *hdata = &data->dtagsd_hash;
2096 	size_t size;
2097 
2098 	if (data->dtagsd_func == NULL) {
2099 		size_t bucket = key->dtak_hashval % hdata->dthsd_hashsize;
2100 
2101 		hdata->dthsd_counts[bucket]++;
2102 		return (WALK_NEXT);
2103 	}
2104 
2105 	/*
2106 	 * We need to read the data.
2107 	 */
2108 	size = key->dtak_size - sizeof (dtrace_aggid_t);
2109 	addr = (uintptr_t)key->dtak_data + sizeof (dtrace_aggid_t);
2110 	hdata->dthsd_data = alloca(size);
2111 	hdata->dthsd_size = size;
2112 
2113 	if (mdb_vread(hdata->dthsd_data, size, addr) == -1) {
2114 		mdb_warn("couldn't read data at %p", addr);
2115 		return (WALK_ERR);
2116 	}
2117 
2118 	data->dtagsd_func(hdata);
2119 
2120 	return (WALK_NEXT);
2121 }
2122 
2123 /*ARGSUSED*/
2124 int
2125 dtrace_aggstat(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2126 {
2127 	dtrace_buffer_t buf;
2128 	uintptr_t aaddr;
2129 	dtrace_aggbuffer_t agb;
2130 	size_t hsize, i, actual, prime, evenpow;
2131 	dtrace_aggstat_data_t data;
2132 	dtrace_hashstat_data_t *hdata = &data.dtagsd_hash;
2133 
2134 	bzero(&data, sizeof (data));
2135 
2136 	if (!(flags & DCMD_ADDRSPEC))
2137 		return (DCMD_USAGE);
2138 
2139 	if (mdb_vread(&buf, sizeof (buf), addr) == -1) {
2140 		mdb_warn("failed to read aggregation buffer at %p", addr);
2141 		return (DCMD_ERR);
2142 	}
2143 
2144 	aaddr = (uintptr_t)buf.dtb_tomax +
2145 	    buf.dtb_size - sizeof (dtrace_aggbuffer_t);
2146 
2147 	if (mdb_vread(&agb, sizeof (agb), aaddr) == -1) {
2148 		mdb_warn("failed to read dtrace_aggbuffer_t at %p", aaddr);
2149 		return (DCMD_ERR);
2150 	}
2151 
2152 	hsize = (actual = agb.dtagb_hashsize) * sizeof (size_t);
2153 	hdata->dthsd_counts = mdb_alloc(hsize, UM_SLEEP | UM_GC);
2154 
2155 	/*
2156 	 * Now pick the largest prime smaller than the hash size.  (If the
2157 	 * existing size is prime, we'll pick a smaller prime just for the
2158 	 * hell of it.)
2159 	 */
2160 	for (prime = agb.dtagb_hashsize - 1; prime > 7; prime--) {
2161 		size_t limit = prime / 7;
2162 
2163 		for (i = 2; i < limit; i++) {
2164 			if ((prime % i) == 0)
2165 				break;
2166 		}
2167 
2168 		if (i == limit)
2169 			break;
2170 	}
2171 
2172 	/*
2173 	 * And now we want to pick the largest power of two smaller than the
2174 	 * hashsize.
2175 	 */
2176 	for (i = 0; (1 << i) < agb.dtagb_hashsize; i++)
2177 		continue;
2178 
2179 	evenpow = (1 << (i - 1));
2180 
2181 	for (i = 0; _dtrace_hashstat[i].dths_name != NULL; i++) {
2182 		data.dtagsd_func = _dtrace_hashstat[i].dths_func;
2183 
2184 		hdata->dthsd_hashsize = actual;
2185 		hsize = hdata->dthsd_hashsize * sizeof (size_t);
2186 		bzero(hdata->dthsd_counts, hsize);
2187 
2188 		if (mdb_pwalk("dtrace_aggkey",
2189 		    (mdb_walk_cb_t)dtrace_aggstat_walk, &data, addr) == -1) {
2190 			mdb_warn("failed to walk dtrace_aggkey at %p", addr);
2191 			return (DCMD_ERR);
2192 		}
2193 
2194 		dtrace_hashstat_stats(_dtrace_hashstat[i].dths_name, hdata);
2195 
2196 		/*
2197 		 * If we were just printing the actual value, we won't try
2198 		 * any of the sizing experiments.
2199 		 */
2200 		if (data.dtagsd_func == NULL)
2201 			continue;
2202 
2203 		hdata->dthsd_hashsize = prime;
2204 		hsize = hdata->dthsd_hashsize * sizeof (size_t);
2205 		bzero(hdata->dthsd_counts, hsize);
2206 
2207 		if (mdb_pwalk("dtrace_aggkey",
2208 		    (mdb_walk_cb_t)dtrace_aggstat_walk, &data, addr) == -1) {
2209 			mdb_warn("failed to walk dtrace_aggkey at %p", addr);
2210 			return (DCMD_ERR);
2211 		}
2212 
2213 		dtrace_hashstat_stats(_dtrace_hashstat[i].dths_name, hdata);
2214 
2215 		hdata->dthsd_hashsize = evenpow;
2216 		hsize = hdata->dthsd_hashsize * sizeof (size_t);
2217 		bzero(hdata->dthsd_counts, hsize);
2218 
2219 		if (mdb_pwalk("dtrace_aggkey",
2220 		    (mdb_walk_cb_t)dtrace_aggstat_walk, &data, addr) == -1) {
2221 			mdb_warn("failed to walk dtrace_aggkey at %p", addr);
2222 			return (DCMD_ERR);
2223 		}
2224 
2225 		dtrace_hashstat_stats(_dtrace_hashstat[i].dths_name, hdata);
2226 	}
2227 
2228 	return (DCMD_OK);
2229 }
2230 
2231 /*ARGSUSED*/
2232 static int
2233 dtrace_dynstat_walk(uintptr_t addr, dtrace_dynvar_t *dynvar,
2234     dtrace_aggstat_data_t *data)
2235 {
2236 	dtrace_hashstat_data_t *hdata = &data->dtagsd_hash;
2237 	dtrace_tuple_t *tuple = &dynvar->dtdv_tuple;
2238 	dtrace_key_t *key = tuple->dtt_key;
2239 	size_t size = 0, offs = 0;
2240 	int i, nkeys = tuple->dtt_nkeys;
2241 	char *buf;
2242 
2243 	if (data->dtagsd_func == NULL) {
2244 		size_t bucket = dynvar->dtdv_hashval % hdata->dthsd_hashsize;
2245 
2246 		hdata->dthsd_counts[bucket]++;
2247 		return (WALK_NEXT);
2248 	}
2249 
2250 	/*
2251 	 * We want to hand the hashing algorithm a contiguous buffer.  First
2252 	 * run through the tuple and determine the size.
2253 	 */
2254 	for (i = 0; i < nkeys; i++) {
2255 		if (key[i].dttk_size == 0) {
2256 			size += sizeof (uint64_t);
2257 		} else {
2258 			size += key[i].dttk_size;
2259 		}
2260 	}
2261 
2262 	buf = alloca(size);
2263 
2264 	/*
2265 	 * Now go back through the tuple and copy the data into the buffer.
2266 	 */
2267 	for (i = 0; i < nkeys; i++) {
2268 		if (key[i].dttk_size == 0) {
2269 			bcopy(&key[i].dttk_value, &buf[offs],
2270 			    sizeof (uint64_t));
2271 			offs += sizeof (uint64_t);
2272 		} else {
2273 			if (mdb_vread(&buf[offs], key[i].dttk_size,
2274 			    key[i].dttk_value) == -1) {
2275 				mdb_warn("couldn't read tuple data at %p",
2276 				    key[i].dttk_value);
2277 				return (WALK_ERR);
2278 			}
2279 
2280 			offs += key[i].dttk_size;
2281 		}
2282 	}
2283 
2284 	hdata->dthsd_data = buf;
2285 	hdata->dthsd_size = size;
2286 
2287 	data->dtagsd_func(hdata);
2288 
2289 	return (WALK_NEXT);
2290 }
2291 
2292 /*ARGSUSED*/
2293 int
2294 dtrace_dynstat(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2295 {
2296 	dtrace_dstate_t dstate;
2297 	size_t hsize, i, actual, prime;
2298 	dtrace_aggstat_data_t data;
2299 	dtrace_hashstat_data_t *hdata = &data.dtagsd_hash;
2300 
2301 	bzero(&data, sizeof (data));
2302 
2303 	if (!(flags & DCMD_ADDRSPEC))
2304 		return (DCMD_USAGE);
2305 
2306 	if (mdb_vread(&dstate, sizeof (dstate), addr) == -1) {
2307 		mdb_warn("failed to read dynamic variable state at %p", addr);
2308 		return (DCMD_ERR);
2309 	}
2310 
2311 	hsize = (actual = dstate.dtds_hashsize) * sizeof (size_t);
2312 	hdata->dthsd_counts = mdb_alloc(hsize, UM_SLEEP | UM_GC);
2313 
2314 	/*
2315 	 * Now pick the largest prime smaller than the hash size.  (If the
2316 	 * existing size is prime, we'll pick a smaller prime just for the
2317 	 * hell of it.)
2318 	 */
2319 	for (prime = dstate.dtds_hashsize - 1; prime > 7; prime--) {
2320 		size_t limit = prime / 7;
2321 
2322 		for (i = 2; i < limit; i++) {
2323 			if ((prime % i) == 0)
2324 				break;
2325 		}
2326 
2327 		if (i == limit)
2328 			break;
2329 	}
2330 
2331 	for (i = 0; _dtrace_hashstat[i].dths_name != NULL; i++) {
2332 		data.dtagsd_func = _dtrace_hashstat[i].dths_func;
2333 
2334 		hdata->dthsd_hashsize = actual;
2335 		hsize = hdata->dthsd_hashsize * sizeof (size_t);
2336 		bzero(hdata->dthsd_counts, hsize);
2337 
2338 		if (mdb_pwalk("dtrace_dynvar",
2339 		    (mdb_walk_cb_t)dtrace_dynstat_walk, &data, addr) == -1) {
2340 			mdb_warn("failed to walk dtrace_dynvar at %p", addr);
2341 			return (DCMD_ERR);
2342 		}
2343 
2344 		dtrace_hashstat_stats(_dtrace_hashstat[i].dths_name, hdata);
2345 
2346 		/*
2347 		 * If we were just printing the actual value, we won't try
2348 		 * any of the sizing experiments.
2349 		 */
2350 		if (data.dtagsd_func == NULL)
2351 			continue;
2352 
2353 		hdata->dthsd_hashsize = prime;
2354 		hsize = hdata->dthsd_hashsize * sizeof (size_t);
2355 		bzero(hdata->dthsd_counts, hsize);
2356 
2357 		if (mdb_pwalk("dtrace_dynvar",
2358 		    (mdb_walk_cb_t)dtrace_dynstat_walk, &data, addr) == -1) {
2359 			mdb_warn("failed to walk dtrace_aggkey at %p", addr);
2360 			return (DCMD_ERR);
2361 		}
2362 
2363 		dtrace_hashstat_stats(_dtrace_hashstat[i].dths_name, hdata);
2364 	}
2365 
2366 	return (DCMD_OK);
2367 }
2368 
2369 typedef struct dtrace_ecb_walk {
2370 	dtrace_ecb_t **dtew_ecbs;
2371 	int dtew_necbs;
2372 	int dtew_curecb;
2373 } dtrace_ecb_walk_t;
2374 
2375 static int
2376 dtrace_ecb_init(mdb_walk_state_t *wsp)
2377 {
2378 	uintptr_t addr;
2379 	dtrace_state_t state;
2380 	dtrace_ecb_walk_t *ecbwp;
2381 
2382 	if ((addr = wsp->walk_addr) == 0) {
2383 		mdb_warn("dtrace_ecb walk needs dtrace_state_t\n");
2384 		return (WALK_ERR);
2385 	}
2386 
2387 	if (mdb_vread(&state, sizeof (state), addr) == -1) {
2388 		mdb_warn("failed to read dtrace state pointer at %p", addr);
2389 		return (WALK_ERR);
2390 	}
2391 
2392 	ecbwp = mdb_zalloc(sizeof (dtrace_ecb_walk_t), UM_SLEEP | UM_GC);
2393 
2394 	ecbwp->dtew_ecbs = state.dts_ecbs;
2395 	ecbwp->dtew_necbs = state.dts_necbs;
2396 	ecbwp->dtew_curecb = 0;
2397 
2398 	wsp->walk_data = ecbwp;
2399 
2400 	return (WALK_NEXT);
2401 }
2402 
2403 static int
2404 dtrace_ecb_step(mdb_walk_state_t *wsp)
2405 {
2406 	uintptr_t ecbp, addr;
2407 	dtrace_ecb_walk_t *ecbwp = wsp->walk_data;
2408 
2409 	addr = (uintptr_t)ecbwp->dtew_ecbs +
2410 	    ecbwp->dtew_curecb * sizeof (dtrace_ecb_t *);
2411 
2412 	if (ecbwp->dtew_curecb++ == ecbwp->dtew_necbs)
2413 		return (WALK_DONE);
2414 
2415 	if (mdb_vread(&ecbp, sizeof (addr), addr) == -1) {
2416 		mdb_warn("failed to read ecb at entry %d\n",
2417 		    ecbwp->dtew_curecb);
2418 		return (WALK_ERR);
2419 	}
2420 
2421 	if (ecbp == 0)
2422 		return (WALK_NEXT);
2423 
2424 	return (wsp->walk_callback(ecbp, NULL, wsp->walk_cbdata));
2425 }
2426 
2427 static void
2428 dtrace_options_numtostr(uint64_t num, char *buf, size_t len)
2429 {
2430 	uint64_t n = num;
2431 	int index = 0;
2432 	char u;
2433 
2434 	while (n >= 1024) {
2435 		n = (n + (1024 / 2)) / 1024; /* Round up or down */
2436 		index++;
2437 	}
2438 
2439 	u = " KMGTPE"[index];
2440 
2441 	if (index == 0) {
2442 		(void) mdb_snprintf(buf, len, "%llu", (u_longlong_t)n);
2443 	} else if (n < 10 && (num & (num - 1)) != 0) {
2444 		(void) mdb_snprintf(buf, len, "%.2f%c",
2445 		    (double)num / (1ULL << 10 * index), u);
2446 	} else if (n < 100 && (num & (num - 1)) != 0) {
2447 		(void) mdb_snprintf(buf, len, "%.1f%c",
2448 		    (double)num / (1ULL << 10 * index), u);
2449 	} else {
2450 		(void) mdb_snprintf(buf, len, "%llu%c", (u_longlong_t)n, u);
2451 	}
2452 }
2453 
2454 static void
2455 dtrace_options_numtohz(uint64_t num, char *buf, size_t len)
2456 {
2457 	(void) mdb_snprintf(buf, len, "%dhz", NANOSEC/num);
2458 }
2459 
2460 static void
2461 dtrace_options_numtobufpolicy(uint64_t num, char *buf, size_t len)
2462 {
2463 	char *policy = "unknown";
2464 
2465 	switch (num) {
2466 		case DTRACEOPT_BUFPOLICY_RING:
2467 			policy = "ring";
2468 			break;
2469 
2470 		case DTRACEOPT_BUFPOLICY_FILL:
2471 			policy = "fill";
2472 			break;
2473 
2474 		case DTRACEOPT_BUFPOLICY_SWITCH:
2475 			policy = "switch";
2476 			break;
2477 	}
2478 
2479 	(void) mdb_snprintf(buf, len, "%s", policy);
2480 }
2481 
2482 static void
2483 dtrace_options_numtocpu(uint64_t cpu, char *buf, size_t len)
2484 {
2485 	if (cpu == DTRACE_CPUALL)
2486 		(void) mdb_snprintf(buf, len, "%7s", "unbound");
2487 	else
2488 		(void) mdb_snprintf(buf, len, "%d", cpu);
2489 }
2490 
2491 typedef void (*dtrace_options_func_t)(uint64_t, char *, size_t);
2492 
2493 static struct dtrace_options {
2494 	char *dtop_optstr;
2495 	dtrace_options_func_t dtop_func;
2496 } _dtrace_options[] = {
2497 	{ "bufsize", dtrace_options_numtostr },
2498 	{ "bufpolicy", dtrace_options_numtobufpolicy },
2499 	{ "dynvarsize", dtrace_options_numtostr },
2500 	{ "aggsize", dtrace_options_numtostr },
2501 	{ "specsize", dtrace_options_numtostr },
2502 	{ "nspec", dtrace_options_numtostr },
2503 	{ "strsize", dtrace_options_numtostr },
2504 	{ "cleanrate", dtrace_options_numtohz },
2505 	{ "cpu", dtrace_options_numtocpu },
2506 	{ "bufresize", dtrace_options_numtostr },
2507 	{ "grabanon", dtrace_options_numtostr },
2508 	{ "flowindent", dtrace_options_numtostr },
2509 	{ "quiet", dtrace_options_numtostr },
2510 	{ "stackframes", dtrace_options_numtostr },
2511 	{ "ustackframes", dtrace_options_numtostr },
2512 	{ "aggrate", dtrace_options_numtohz },
2513 	{ "switchrate", dtrace_options_numtohz },
2514 	{ "statusrate", dtrace_options_numtohz },
2515 	{ "destructive", dtrace_options_numtostr },
2516 	{ "stackindent", dtrace_options_numtostr },
2517 	{ "rawbytes", dtrace_options_numtostr },
2518 	{ "jstackframes", dtrace_options_numtostr },
2519 	{ "jstackstrsize", dtrace_options_numtostr },
2520 	{ "aggsortkey", dtrace_options_numtostr },
2521 	{ "aggsortrev", dtrace_options_numtostr },
2522 	{ "aggsortpos", dtrace_options_numtostr },
2523 	{ "aggsortkeypos", dtrace_options_numtostr },
2524 	{ "temporal", dtrace_options_numtostr },
2525 	{ "agghist", dtrace_options_numtostr },
2526 	{ "aggpack", dtrace_options_numtostr },
2527 	{ "aggzoom", dtrace_options_numtostr },
2528 	{ "zone", dtrace_options_numtostr }
2529 };
2530 
2531 CTASSERT(ARRAY_SIZE(_dtrace_options) == DTRACEOPT_MAX);
2532 
2533 static void
2534 dtrace_options_help(void)
2535 {
2536 	mdb_printf("Given a dtrace_state_t structure, displays the "
2537 	    "current tunable option\nsettings.\n");
2538 }
2539 
2540 /*ARGSUSED*/
2541 static int
2542 dtrace_options(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2543 {
2544 	dtrace_state_t state;
2545 	int i = 0;
2546 	dtrace_optval_t *options;
2547 	char val[32];
2548 
2549 	if (!(flags & DCMD_ADDRSPEC))
2550 		return (DCMD_USAGE);
2551 
2552 	if (mdb_vread(&state, sizeof (dtrace_state_t), (uintptr_t)addr) == -1) {
2553 		mdb_warn("failed to read state pointer at %p\n", addr);
2554 		return (DCMD_ERR);
2555 	}
2556 
2557 	options = &state.dts_options[0];
2558 
2559 	mdb_printf("%<u>%-25s %s%</u>\n", "OPTION", "VALUE");
2560 	for (i = 0; i < DTRACEOPT_MAX; i++) {
2561 		if (options[i] == DTRACEOPT_UNSET) {
2562 			mdb_printf("%-25s %s\n",
2563 			    _dtrace_options[i].dtop_optstr, "UNSET");
2564 		} else {
2565 			(void) _dtrace_options[i].dtop_func(options[i],
2566 			    val, 32);
2567 			mdb_printf("%-25s %s\n",
2568 			    _dtrace_options[i].dtop_optstr, val);
2569 		}
2570 	}
2571 
2572 	return (DCMD_OK);
2573 }
2574 
2575 static int
2576 pid2state_init(mdb_walk_state_t *wsp)
2577 {
2578 	dtrace_state_data_t *data;
2579 	uintptr_t devi;
2580 	uintptr_t proc;
2581 	struct dev_info info;
2582 	pid_t pid = (pid_t)wsp->walk_addr;
2583 
2584 	if (wsp->walk_addr == 0) {
2585 		mdb_warn("pid2state walk requires PID\n");
2586 		return (WALK_ERR);
2587 	}
2588 
2589 	data = mdb_zalloc(sizeof (dtrace_state_data_t), UM_SLEEP | UM_GC);
2590 
2591 	if (mdb_readvar(&data->dtsd_softstate, "dtrace_softstate") == -1) {
2592 		mdb_warn("failed to read 'dtrace_softstate'");
2593 		return (DCMD_ERR);
2594 	}
2595 
2596 	if ((proc = mdb_pid2proc(pid, NULL)) == 0) {
2597 		mdb_warn("PID 0t%d not found\n", pid);
2598 		return (DCMD_ERR);
2599 	}
2600 
2601 	if (mdb_readvar(&devi, "dtrace_devi") == -1) {
2602 		mdb_warn("failed to read 'dtrace_devi'");
2603 		return (DCMD_ERR);
2604 	}
2605 
2606 	if (mdb_vread(&info, sizeof (struct dev_info), devi) == -1) {
2607 		mdb_warn("failed to read 'dev_info'");
2608 		return (DCMD_ERR);
2609 	}
2610 
2611 	data->dtsd_major = info.devi_major;
2612 	data->dtsd_proc = proc;
2613 
2614 	wsp->walk_data = data;
2615 
2616 	return (WALK_NEXT);
2617 }
2618 
2619 /*ARGSUSED*/
2620 static int
2621 pid2state_file(uintptr_t addr, struct file *f, dtrace_state_data_t *data)
2622 {
2623 	vnode_t vnode;
2624 	minor_t minor;
2625 	uintptr_t statep;
2626 
2627 	/* Get the vnode for this file */
2628 	if (mdb_vread(&vnode, sizeof (vnode), (uintptr_t)f->f_vnode) == -1) {
2629 		mdb_warn("couldn't read vnode at %p", (uintptr_t)f->f_vnode);
2630 		return (WALK_NEXT);
2631 	}
2632 
2633 
2634 	/* Is this the dtrace device? */
2635 	if (getmajor(vnode.v_rdev) != data->dtsd_major)
2636 		return (WALK_NEXT);
2637 
2638 	/* Get the minor number for this device entry */
2639 	minor = getminor(vnode.v_rdev);
2640 
2641 	if (mdb_get_soft_state_byaddr(data->dtsd_softstate, minor,
2642 	    &statep, NULL, 0) == -1) {
2643 		mdb_warn("failed to read softstate for minor %d", minor);
2644 		return (WALK_NEXT);
2645 	}
2646 
2647 	mdb_printf("%p\n", statep);
2648 
2649 	return (WALK_NEXT);
2650 }
2651 
2652 static int
2653 pid2state_step(mdb_walk_state_t *wsp)
2654 {
2655 	dtrace_state_data_t *ds = wsp->walk_data;
2656 
2657 	if (mdb_pwalk("file",
2658 	    (mdb_walk_cb_t)pid2state_file, ds, ds->dtsd_proc) == -1) {
2659 		mdb_warn("couldn't walk 'file' for proc %p", ds->dtsd_proc);
2660 		return (WALK_ERR);
2661 	}
2662 
2663 	return (WALK_DONE);
2664 }
2665 
2666 /*ARGSUSED*/
2667 static int
2668 dtrace_probes_walk(uintptr_t addr, void *ignored, uintptr_t *target)
2669 {
2670 	dtrace_ecb_t ecb;
2671 	dtrace_probe_t probe;
2672 	dtrace_probedesc_t pd;
2673 
2674 	if (addr == 0)
2675 		return (WALK_ERR);
2676 
2677 	if (mdb_vread(&ecb, sizeof (dtrace_ecb_t), addr) == -1) {
2678 		mdb_warn("failed to read ecb %p\n", addr);
2679 		return (WALK_ERR);
2680 	}
2681 
2682 	if (ecb.dte_probe == NULL)
2683 		return (WALK_ERR);
2684 
2685 	if (mdb_vread(&probe, sizeof (dtrace_probe_t),
2686 	    (uintptr_t)ecb.dte_probe) == -1) {
2687 		mdb_warn("failed to read probe %p\n", ecb.dte_probe);
2688 		return (WALK_ERR);
2689 	}
2690 
2691 	pd.dtpd_id = probe.dtpr_id;
2692 	dtracemdb_probe(NULL, &pd);
2693 
2694 	mdb_printf("%5d %10s %17s %33s %s\n", pd.dtpd_id, pd.dtpd_provider,
2695 	    pd.dtpd_mod, pd.dtpd_func, pd.dtpd_name);
2696 
2697 	return (WALK_NEXT);
2698 }
2699 
2700 static void
2701 dtrace_probes_help(void)
2702 {
2703 	mdb_printf("Given a dtrace_state_t structure, displays all "
2704 	    "its active enablings.  If no\nstate structure is provided, "
2705 	    "all available probes are listed.\n");
2706 }
2707 
2708 /*ARGSUSED*/
2709 static int
2710 dtrace_probes(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
2711 {
2712 	dtrace_probedesc_t pd;
2713 	uintptr_t caddr, base, paddr;
2714 	int nprobes, i;
2715 
2716 	mdb_printf("%5s %10s %17s %33s %s\n",
2717 	    "ID", "PROVIDER", "MODULE", "FUNCTION", "NAME");
2718 
2719 	if (!(flags & DCMD_ADDRSPEC)) {
2720 		/*
2721 		 * If no argument is provided just display all available
2722 		 * probes.
2723 		 */
2724 		if (mdb_readvar(&base, "dtrace_probes") == -1) {
2725 			mdb_warn("failed to read 'dtrace_probes'");
2726 			return (-1);
2727 		}
2728 
2729 		if (mdb_readvar(&nprobes, "dtrace_nprobes") == -1) {
2730 			mdb_warn("failed to read 'dtrace_nprobes'");
2731 			return (-1);
2732 		}
2733 
2734 		for (i = 0; i < nprobes; i++) {
2735 			caddr = base + i  * sizeof (dtrace_probe_t *);
2736 
2737 			if (mdb_vread(&paddr, sizeof (paddr), caddr) == -1) {
2738 				mdb_warn("couldn't read probe pointer at %p",
2739 				    caddr);
2740 				continue;
2741 			}
2742 
2743 			if (paddr == 0)
2744 				continue;
2745 
2746 			pd.dtpd_id = i + 1;
2747 			if (dtracemdb_probe(NULL, &pd) == 0) {
2748 				mdb_printf("%5d %10s %17s %33s %s\n",
2749 				    pd.dtpd_id, pd.dtpd_provider,
2750 				    pd.dtpd_mod, pd.dtpd_func, pd.dtpd_name);
2751 			}
2752 		}
2753 	} else {
2754 		if (mdb_pwalk("dtrace_ecb", (mdb_walk_cb_t)dtrace_probes_walk,
2755 		    NULL, addr) == -1) {
2756 			mdb_warn("couldn't walk 'dtrace_ecb'");
2757 			return (DCMD_ERR);
2758 		}
2759 	}
2760 
2761 	return (DCMD_OK);
2762 }
2763 
2764 const mdb_dcmd_t kernel_dcmds[] = {
2765 	{ "id2probe", ":", "translate a dtrace_id_t to a dtrace_probe_t",
2766 	    id2probe },
2767 	{ "dtrace", ":[-c cpu]", "print dtrace(1M)-like output",
2768 	    dtrace, dtrace_help },
2769 	{ "dtrace_errhash", ":", "print DTrace error hash", dtrace_errhash },
2770 	{ "dtrace_helptrace", ":", "print DTrace helper trace",
2771 	    dtrace_helptrace },
2772 	{ "dtrace_state", ":", "print active DTrace consumers", dtrace_state,
2773 	    dtrace_state_help },
2774 	{ "dtrace_aggstat", ":",
2775 	    "print DTrace aggregation hash statistics", dtrace_aggstat },
2776 	{ "dtrace_dynstat", ":",
2777 	    "print DTrace dynamic variable hash statistics", dtrace_dynstat },
2778 	{ "dtrace_options", ":",
2779 	    "print a DTrace consumer's current tuneable options",
2780 	    dtrace_options, dtrace_options_help },
2781 	{ "dtrace_probes", "?", "print a DTrace consumer's enabled probes",
2782 	    dtrace_probes, dtrace_probes_help },
2783 	{ NULL }
2784 };
2785 
2786 const mdb_walker_t kernel_walkers[] = {
2787 	{ "dtrace_errhash", "walk hash of DTrace error messasges",
2788 		dtrace_errhash_init, dtrace_errhash_step },
2789 	{ "dtrace_helptrace", "walk DTrace helper trace entries",
2790 		dtrace_helptrace_init, dtrace_helptrace_step },
2791 	{ "dtrace_state", "walk DTrace per-consumer softstate",
2792 		dtrace_state_init, dtrace_state_step },
2793 	{ "dtrace_aggkey", "walk DTrace aggregation keys",
2794 		dtrace_aggkey_init, dtrace_aggkey_step, dtrace_aggkey_fini },
2795 	{ "dtrace_dynvar", "walk DTrace dynamic variables",
2796 		dtrace_dynvar_init, dtrace_dynvar_step, dtrace_dynvar_fini },
2797 	{ "dtrace_ecb", "walk a DTrace consumer's enabling control blocks",
2798 		dtrace_ecb_init, dtrace_ecb_step },
2799 	{ "pid2state", "walk a processes dtrace_state structures",
2800 	    pid2state_init, pid2state_step },
2801 	{ NULL }
2802 };
2803