xref: /illumos-gate/usr/src/uts/common/fs/nfs/nfs4_client_debug.c (revision d8d810637042c2a73a269f125dee75e8ed1421d1)
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  *	Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23  *	Use is subject to license terms.
24  */
25 
26 #include <sys/cred.h>
27 #include <sys/kstat.h>
28 #include <sys/list.h>
29 #include <sys/systm.h>
30 #include <sys/vfs.h>
31 #include <sys/vnode.h>
32 #include <sys/cmn_err.h>
33 
34 #include <nfs/nfs4_clnt.h>
35 #include <nfs/rnode4.h>
36 
37 /*
38  * Recovery kstats
39  */
40 typedef struct rkstat {
41 	kstat_named_t	badhandle;
42 	kstat_named_t	badowner;
43 	kstat_named_t	clientid;
44 	kstat_named_t	dead_file;
45 	kstat_named_t	delay;
46 	kstat_named_t	fail_relock;
47 	kstat_named_t	file_diff;
48 	kstat_named_t	no_grace;
49 	kstat_named_t	not_responding;
50 	kstat_named_t	opens_changed;
51 	kstat_named_t	siglost;
52 	kstat_named_t	unexp_action;
53 	kstat_named_t	unexp_errno;
54 	kstat_named_t	unexp_status;
55 	kstat_named_t	wrongsec;
56 	kstat_named_t	lost_state_bad_op;
57 } rkstat_t;
58 
59 static rkstat_t rkstat_template = {
60 	{ "badhandle",		KSTAT_DATA_ULONG },
61 	{ "badowner",		KSTAT_DATA_ULONG },
62 	{ "clientid",		KSTAT_DATA_ULONG },
63 	{ "dead_file",		KSTAT_DATA_ULONG },
64 	{ "delay",		KSTAT_DATA_ULONG },
65 	{ "fail_relock",	KSTAT_DATA_ULONG },
66 	{ "file_diff",		KSTAT_DATA_ULONG },
67 	{ "no_grace",		KSTAT_DATA_ULONG },
68 	{ "not_responding",	KSTAT_DATA_ULONG },
69 	{ "opens_changed",	KSTAT_DATA_ULONG },
70 	{ "siglost",		KSTAT_DATA_ULONG },
71 	{ "unexp_action",	KSTAT_DATA_ULONG },
72 	{ "unexp_errno",	KSTAT_DATA_ULONG },
73 	{ "unexp_status",	KSTAT_DATA_ULONG },
74 	{ "wrongsec",		KSTAT_DATA_ULONG },
75 	{ "bad_op",		KSTAT_DATA_ULONG },
76 };
77 
78 /* maximum number of messages allowed on the mi's mi_msg_list */
79 int nfs4_msg_max = NFS4_MSG_MAX;
80 #define	DEFAULT_LEASE	180
81 
82 /*
83  * Sets the appropiate fields of "ep", given "id" and various parameters.
84  * Assumes that ep's fields have been initialized to zero/null, except for
85  * re_type and mount point info, which are already set.
86  */
87 static void
88 set_event(nfs4_event_type_t id, nfs4_revent_t *ep, mntinfo4_t *mi,
89     rnode4_t *rp1, rnode4_t *rp2, uint_t count, pid_t pid, nfsstat4 nfs4_error,
90     char *server1, char *why, nfs4_tag_type_t tag1, nfs4_tag_type_t tag2,
91     seqid4 seqid1, seqid4 seqid2)
92 {
93 	int len;
94 
95 	switch (id) {
96 	case RE_BAD_SEQID:
97 		ep->re_mi = mi;
98 
99 		/* bad seqid'd file <path/component name> */
100 		if (rp1 && rp1->r_svnode.sv_name)
101 			ep->re_char1 = fn_path(rp1->r_svnode.sv_name);
102 		else
103 			ep->re_char1 = NULL;
104 		ep->re_rp1 = rp1;
105 
106 		/* for LOCK/LOCKU */
107 		ep->re_pid = pid;
108 
109 		ep->re_stat4 = nfs4_error;
110 		ep->re_tag1 = tag1;
111 		ep->re_tag2 = tag2;
112 		ep->re_seqid1 = seqid1;
113 		ep->re_seqid2 = seqid2;
114 		break;
115 	case RE_BADHANDLE:
116 		ASSERT(rp1 != NULL);
117 
118 		/* dead file <path/component name> */
119 		if (rp1->r_svnode.sv_name)
120 			ep->re_char1 = fn_path(rp1->r_svnode.sv_name);
121 		else
122 			ep->re_char1 = NULL;
123 		ep->re_rp1 = rp1;
124 		break;
125 	case RE_CLIENTID:
126 		ep->re_mi = mi;
127 
128 		/* the error we failed with */
129 		ep->re_uint = count;
130 		ep->re_stat4 = nfs4_error;
131 		break;
132 	case RE_DEAD_FILE:
133 		ASSERT(rp1 != NULL);
134 
135 		/* dead file <path/component name> */
136 		if (rp1->r_svnode.sv_name)
137 			ep->re_char1 = fn_path(rp1->r_svnode.sv_name);
138 		else
139 			ep->re_char1 = NULL;
140 		ep->re_rp1 = rp1;
141 
142 		/* why the file got killed */
143 		if (why) {
144 			len = strlen(why);
145 			ep->re_char2 = kmem_alloc(len + 1, KM_SLEEP);
146 			bcopy(why, ep->re_char2, len);
147 			ep->re_char2[len] = '\0';
148 		} else
149 			ep->re_char2 = NULL;
150 
151 		ep->re_stat4 = nfs4_error;
152 		break;
153 	case RE_END:
154 		/* first rnode */
155 		if (rp1 && rp1->r_svnode.sv_name)
156 			ep->re_char1 = fn_path(rp1->r_svnode.sv_name);
157 		else
158 			ep->re_char1 = NULL;
159 		ep->re_rp1 = rp1;
160 
161 		/* second rnode */
162 		if (rp2 && rp2->r_svnode.sv_name)
163 			ep->re_char2 = fn_path(rp2->r_svnode.sv_name);
164 		else
165 			ep->re_char2 = NULL;
166 		ep->re_rp2 = rp2;
167 
168 		ep->re_mi = mi;
169 		break;
170 	case RE_FAIL_RELOCK:
171 		ASSERT(rp1 != NULL);
172 
173 		/* error on fail relock */
174 		ep->re_uint = count;
175 
176 		/* process that failed */
177 		ep->re_pid = pid;
178 
179 		/* nfs4 error */
180 		ep->re_stat4 = nfs4_error;
181 
182 		/* file <path/component name> */
183 		if (rp1->r_svnode.sv_name)
184 			ep->re_char1 = fn_path(rp1->r_svnode.sv_name);
185 		else
186 			ep->re_char1 = NULL;
187 		ep->re_rp1 = rp1;
188 		break;
189 	case RE_FAIL_REMAP_LEN:
190 		/* length of returned filehandle */
191 		ep->re_uint = count;
192 		break;
193 	case RE_FAIL_REMAP_OP:
194 		break;
195 	case RE_FAILOVER:
196 		/* server we're failing over to (if not picking original) */
197 		if (server1 != NULL) {
198 			len = strlen(server1);
199 			ep->re_char1 = kmem_alloc(len + 1, KM_SLEEP);
200 			bcopy(server1, ep->re_char1, len);
201 			ep->re_char1[len] = '\0';
202 		} else {
203 			ep->re_char1 = NULL;
204 		}
205 		break;
206 	case RE_FILE_DIFF:
207 		ASSERT(rp1 != NULL);
208 
209 		/* dead file <path/component name> */
210 		if (rp1->r_svnode.sv_name)
211 			ep->re_char1 = fn_path(rp1->r_svnode.sv_name);
212 		else
213 			ep->re_char1 = NULL;
214 		ep->re_rp1 = rp1;
215 		break;
216 	case RE_LOST_STATE:
217 		ep->re_uint = count;		/* op number */
218 		if (rp1 && rp1->r_svnode.sv_name)
219 			ep->re_char1 = fn_path(rp1->r_svnode.sv_name);
220 		else
221 			ep->re_char1 = NULL;
222 		ep->re_rp1 = rp1;
223 		if (rp2 && rp2->r_svnode.sv_name)
224 			ep->re_char2 = fn_path(rp2->r_svnode.sv_name);
225 		else
226 			ep->re_char2 = NULL;
227 		ep->re_rp2 = rp2;
228 		break;
229 	case RE_OPENS_CHANGED:
230 		ep->re_mi = mi;
231 
232 		/* original number of open files */
233 		ep->re_uint = count;
234 		/* new number of open files */
235 		ep->re_pid = pid;
236 		break;
237 	case RE_SIGLOST:
238 	case RE_SIGLOST_NO_DUMP:
239 		ASSERT(rp1 != NULL);
240 
241 		/* file <path/component name> */
242 		if (rp1->r_svnode.sv_name)
243 			ep->re_char1 = fn_path(rp1->r_svnode.sv_name);
244 		else
245 			ep->re_char1 = NULL;
246 		ep->re_rp1 = rp1;
247 		ep->re_pid = pid;
248 		ep->re_uint = count;
249 		ep->re_stat4 = nfs4_error;
250 		break;
251 	case RE_START:
252 		/* file <path/component name> */
253 		if (rp1 && rp1->r_svnode.sv_name)
254 			ep->re_char1 = fn_path(rp1->r_svnode.sv_name);
255 		else
256 			ep->re_char1 = NULL;
257 		ep->re_rp1 = rp1;
258 
259 		/* file <path/component name> */
260 		if (rp2 && rp2->r_svnode.sv_name)
261 			ep->re_char2 = fn_path(rp2->r_svnode.sv_name);
262 		else
263 			ep->re_char2 = NULL;
264 		ep->re_rp2 = rp2;
265 
266 		ep->re_mi = mi;
267 		ep->re_uint = count;
268 		break;
269 	case RE_UNEXPECTED_ACTION:
270 	case RE_UNEXPECTED_ERRNO:
271 		/* the error that is unexpected */
272 		ep->re_uint = count;
273 		break;
274 	case RE_UNEXPECTED_STATUS:
275 		/* nfsstat4 error */
276 		ep->re_stat4 = nfs4_error;
277 		break;
278 	case RE_WRONGSEC:
279 		/* the error we failed with */
280 		ep->re_uint = count;
281 
282 		/* file <path/component name> */
283 		if (rp1 && rp1->r_svnode.sv_name)
284 			ep->re_char1 = fn_path(rp1->r_svnode.sv_name);
285 		else
286 			ep->re_char1 = NULL;
287 		ep->re_rp1 = rp1;
288 
289 		/* file <path/component name> */
290 		if (rp2 && rp2->r_svnode.sv_name)
291 			ep->re_char2 = fn_path(rp2->r_svnode.sv_name);
292 		else
293 			ep->re_char2 = NULL;
294 		ep->re_rp2 = rp2;
295 		break;
296 	case RE_LOST_STATE_BAD_OP:
297 		ep->re_uint = count;	/* the unexpected op */
298 		ep->re_pid = pid;
299 		ep->re_rp1 = rp1;
300 		if (rp1 != NULL && rp1->r_svnode.sv_name != NULL)
301 			ep->re_char1 = fn_path(rp1->r_svnode.sv_name);
302 		ep->re_rp2 = rp2;
303 		if (rp2 != NULL && rp2->r_svnode.sv_name != NULL)
304 			ep->re_char2 = fn_path(rp2->r_svnode.sv_name);
305 		break;
306 	default:
307 		break;
308 	}
309 }
310 
311 /*
312  * Sets the appropiate fields of the 'fact' for this 'id'.
313  */
314 static void
315 set_fact(nfs4_fact_type_t id, nfs4_rfact_t *fp, nfsstat4 stat4,
316     nfs4_recov_t raction, nfs_opnum4 op, bool_t reboot, int error,
317     vnode_t *vp)
318 {
319 	rnode4_t *rp1;
320 
321 	switch (id) {
322 	case RF_BADOWNER:
323 		fp->rf_op = op;
324 		fp->rf_reboot = reboot;
325 		fp->rf_stat4 = stat4;
326 		break;
327 	case RF_RENEW_EXPIRED:
328 		break;
329 	case RF_ERR:
330 		fp->rf_op = op;
331 		fp->rf_reboot = reboot;
332 		fp->rf_stat4 = stat4;
333 		fp->rf_action = raction;
334 		fp->rf_error = error;
335 		break;
336 	case RF_SRV_OK:
337 		break;
338 	case RF_SRV_NOT_RESPOND:
339 		break;
340 	case RF_SRVS_OK:
341 		break;
342 	case RF_SRVS_NOT_RESPOND:
343 		gethrestime(&fp->rf_time);
344 		break;
345 	case RF_DELMAP_CB_ERR:
346 		fp->rf_op = op;
347 		fp->rf_stat4 = stat4;
348 
349 		rp1 = VTOR4(vp);
350 		fp->rf_rp1 = rp1;
351 		if (rp1 && rp1->r_svnode.sv_name)
352 			fp->rf_char1 = fn_path(rp1->r_svnode.sv_name);
353 		else
354 			fp->rf_char1 = NULL;
355 		break;
356 	case RF_SENDQ_FULL:
357 		break;
358 	default:
359 		zcmn_err(getzoneid(), CE_NOTE, "illegal fact %d", id);
360 		break;
361 	}
362 }
363 
364 /*
365  * Returns 1 if the event/fact is of a successful communication
366  * from the server; 0 otherwise.
367  */
368 static int
369 successful_comm(nfs4_debug_msg_t *msgp)
370 {
371 	if (msgp->msg_type == RM_EVENT) {
372 		switch (msgp->rmsg_u.msg_event.re_type) {
373 		case RE_BAD_SEQID:
374 		case RE_BADHANDLE:
375 		case RE_FAIL_REMAP_LEN:
376 		case RE_FAIL_REMAP_OP:
377 		case RE_FILE_DIFF:
378 		case RE_START:
379 		case RE_UNEXPECTED_ACTION:
380 		case RE_UNEXPECTED_ERRNO:
381 		case RE_UNEXPECTED_STATUS:
382 		case RE_WRONGSEC:
383 			return (1);
384 		case RE_CLIENTID:
385 		case RE_DEAD_FILE:
386 		case RE_END:
387 		case RE_FAIL_RELOCK:
388 		case RE_FAILOVER:
389 		case RE_LOST_STATE:
390 		case RE_OPENS_CHANGED:
391 		case RE_SIGLOST:
392 		case RE_SIGLOST_NO_DUMP:
393 		case RE_LOST_STATE_BAD_OP:
394 			return (0);
395 		default:
396 			return (0);
397 		}
398 	} else {
399 		switch (msgp->rmsg_u.msg_fact.rf_type) {
400 		case RF_BADOWNER:
401 		case RF_ERR:
402 		case RF_RENEW_EXPIRED:
403 		case RF_SRV_OK:
404 		case RF_SRVS_OK:
405 		case RF_DELMAP_CB_ERR:
406 			return (1);
407 		case RF_SRV_NOT_RESPOND:
408 		case RF_SRVS_NOT_RESPOND:
409 		case RF_SENDQ_FULL:
410 			return (0);
411 		default:
412 			return (0);
413 		}
414 	}
415 }
416 
417 /*
418  * Iterate backwards through the mi's mi_msg_list to find the earliest
419  * message that we should find relevant facts to investigate.
420  */
421 static nfs4_debug_msg_t *
422 find_beginning(nfs4_debug_msg_t *first_msg, mntinfo4_t *mi)
423 {
424 	nfs4_debug_msg_t	*oldest_msg, *cur_msg;
425 	time_t			lease;
426 
427 	ASSERT(mutex_owned(&mi->mi_msg_list_lock));
428 	if (mi->mi_lease_period > 0)
429 		lease = 2 * mi->mi_lease_period;
430 	else
431 		lease = DEFAULT_LEASE;
432 
433 	oldest_msg = first_msg;
434 	cur_msg = list_prev(&mi->mi_msg_list, first_msg);
435 	while (cur_msg &&
436 	    first_msg->msg_time.tv_sec - cur_msg->msg_time.tv_sec < lease) {
437 		oldest_msg = cur_msg;
438 		if ((cur_msg->msg_type == RM_FACT) &&
439 		    (cur_msg->rmsg_u.msg_fact.rf_type == RF_SRV_OK)) {
440 			/* find where we lost contact with the server */
441 			while (cur_msg) {
442 				if ((cur_msg->msg_type == RM_FACT) &&
443 				    (cur_msg->rmsg_u.msg_fact.rf_type ==
444 				    RF_SRV_NOT_RESPOND))
445 					break;
446 				oldest_msg = cur_msg;
447 				cur_msg = list_prev(&mi->mi_msg_list, cur_msg);
448 			}
449 			/*
450 			 * Find the first successful message before
451 			 * we lost contact with the server.
452 			 */
453 			if (cur_msg) {
454 				cur_msg = list_prev(&mi->mi_msg_list, cur_msg);
455 				while (cur_msg && !successful_comm(cur_msg)) {
456 					oldest_msg = cur_msg;
457 					cur_msg = list_prev(&mi->mi_msg_list,
458 					    cur_msg);
459 				}
460 			}
461 			/*
462 			 * If we're not at the dummy head pointer,
463 			 * set the oldest and current message.
464 			 */
465 			if (cur_msg) {
466 				first_msg = cur_msg;
467 				oldest_msg = cur_msg;
468 				cur_msg = list_prev(&mi->mi_msg_list, cur_msg);
469 			}
470 		} else
471 			cur_msg = list_prev(&mi->mi_msg_list, cur_msg);
472 	}
473 
474 	return (oldest_msg);
475 }
476 
477 /*
478  * Returns 1 if facts have been found; 0 otherwise.
479  */
480 static int
481 get_facts(nfs4_debug_msg_t *msgp, nfs4_rfact_t *ret_fp, char **mnt_pt,
482 mntinfo4_t *mi)
483 {
484 	nfs4_debug_msg_t	*cur_msg, *oldest_msg;
485 	nfs4_rfact_t		*cur_fp;
486 	int			found_a_fact = 0;
487 	int			len;
488 
489 	cur_msg = msgp;
490 
491 	/* find the oldest msg to search backwards to */
492 	oldest_msg = find_beginning(cur_msg, mi);
493 	ASSERT(oldest_msg != NULL);
494 
495 	/*
496 	 * Create a fact sheet by searching from our current message
497 	 * backwards to the 'oldest_msg', recording facts along the way
498 	 * until we found facts that have been inspected by another time.
499 	 */
500 	while (cur_msg && cur_msg != list_prev(&mi->mi_msg_list, oldest_msg)) {
501 		if (cur_msg->msg_type != RM_FACT) {
502 			cur_msg = list_prev(&mi->mi_msg_list, cur_msg);
503 			continue;
504 		}
505 
506 		cur_fp = &cur_msg->rmsg_u.msg_fact;
507 		/*
508 		 * If this fact has already been looked at, then so
509 		 * have all preceding facts.  Return Now.
510 		 */
511 		if (cur_fp->rf_status == RFS_INSPECT)
512 			return (found_a_fact);
513 
514 		cur_fp->rf_status = RFS_INSPECT;
515 		found_a_fact = 1;
516 		switch (cur_fp->rf_type) {
517 		case RF_BADOWNER:
518 			break;
519 		case RF_ERR:
520 			/*
521 			 * Don't want to overwrite a fact that was
522 			 * previously found during our current search.
523 			 */
524 			if (!ret_fp->rf_reboot)
525 				ret_fp->rf_reboot = cur_fp->rf_reboot;
526 			if (!ret_fp->rf_stat4)
527 				ret_fp->rf_stat4 = cur_fp->rf_stat4;
528 			if (!ret_fp->rf_action)
529 				ret_fp->rf_action = cur_fp->rf_action;
530 			break;
531 		case RF_RENEW_EXPIRED:
532 			if (cur_msg->msg_mntpt && !(*mnt_pt)) {
533 				len = strlen(cur_msg->msg_mntpt) + 1;
534 				*mnt_pt = kmem_alloc(len, KM_SLEEP);
535 				bcopy(cur_msg->msg_mntpt, *mnt_pt, len);
536 			}
537 			break;
538 		case RF_SRV_OK:
539 			break;
540 		case RF_SRV_NOT_RESPOND:
541 			/*
542 			 * Okay to overwrite this fact as
543 			 * we want the earliest time.
544 			 */
545 			ret_fp->rf_time = cur_fp->rf_time;
546 			break;
547 		case RF_SRVS_OK:
548 			break;
549 		case RF_SRVS_NOT_RESPOND:
550 			break;
551 		case RF_DELMAP_CB_ERR:
552 			break;
553 		case RF_SENDQ_FULL:
554 			break;
555 		default:
556 			zcmn_err(getzoneid(), CE_NOTE,
557 			    "get facts: illegal fact %d", cur_fp->rf_type);
558 			break;
559 		}
560 		cur_msg = list_prev(&mi->mi_msg_list, cur_msg);
561 	}
562 
563 	return (found_a_fact);
564 }
565 
566 /*
567  * Returns 1 if this fact is identical to the last fact recorded
568  * (only checks for a match within the last 2 lease periods).
569  */
570 static int
571 facts_same(nfs4_debug_msg_t *cur_msg, nfs4_debug_msg_t *new_msg,
572     mntinfo4_t *mi)
573 {
574 	nfs4_rfact_t	*fp1, *fp2;
575 	int		lease, len;
576 
577 	ASSERT(mutex_owned(&mi->mi_msg_list_lock));
578 	if (mi->mi_lease_period > 0)
579 		lease = 2 * mi->mi_lease_period;
580 	else
581 		lease = DEFAULT_LEASE;
582 
583 	fp2 = &new_msg->rmsg_u.msg_fact;
584 
585 	while (cur_msg &&
586 	    new_msg->msg_time.tv_sec - cur_msg->msg_time.tv_sec < lease) {
587 		if (cur_msg->msg_type != RM_FACT) {
588 			cur_msg = list_prev(&mi->mi_msg_list, cur_msg);
589 			continue;
590 		}
591 		fp1 = &cur_msg->rmsg_u.msg_fact;
592 		if (fp1->rf_type != fp2->rf_type)
593 			return (0);
594 
595 		/* now actually compare the facts */
596 		if (fp1->rf_action != fp2->rf_action)
597 			return (0);
598 		if (fp1->rf_stat4 != fp2->rf_stat4)
599 			return (0);
600 		if (fp1->rf_reboot != fp2->rf_reboot)
601 			return (0);
602 		if (fp1->rf_op != fp2->rf_op)
603 			return (0);
604 		if (fp1->rf_time.tv_sec != fp2->rf_time.tv_sec)
605 			return (0);
606 		if (fp1->rf_error != fp2->rf_error)
607 			return (0);
608 		if (fp1->rf_rp1 != fp2->rf_rp1)
609 			return (0);
610 		if (cur_msg->msg_srv != NULL) {
611 			if (new_msg->msg_srv == NULL)
612 				return (0);
613 			len = strlen(cur_msg->msg_srv);
614 			if (strncmp(cur_msg->msg_srv, new_msg->msg_srv,
615 			    len) != 0)
616 				return (0);
617 		} else if (new_msg->msg_srv != NULL) {
618 			return (0);
619 		}
620 		if (cur_msg->msg_mntpt != NULL) {
621 			if (new_msg->msg_mntpt == NULL)
622 				return (0);
623 			len = strlen(cur_msg->msg_mntpt);
624 			if (strncmp(cur_msg->msg_mntpt, new_msg->msg_mntpt,
625 			    len) != 0)
626 				return (0);
627 		} else if (new_msg->msg_mntpt != NULL) {
628 			return (0);
629 		}
630 		if (fp1->rf_char1 != NULL) {
631 			if (fp2->rf_char1 == NULL)
632 				return (0);
633 			len = strlen(fp1->rf_char1);
634 			if (strncmp(fp1->rf_char1, fp2->rf_char1, len) != 0)
635 				return (0);
636 		} else if (fp2->rf_char1 != NULL) {
637 			return (0);
638 		}
639 		return (1);
640 	}
641 
642 	return (0);
643 }
644 
645 /*
646  * Returns 1 if these two messages are identical; 0 otherwise.
647  */
648 static int
649 events_same(nfs4_debug_msg_t *cur_msg, nfs4_debug_msg_t *new_msg,
650     mntinfo4_t *mi)
651 {
652 	nfs4_revent_t	*ep1, *ep2;
653 	int		len;
654 
655 	/* find the last event, bypassing all facts */
656 	while (cur_msg && cur_msg->msg_type != RM_EVENT)
657 		cur_msg = list_prev(&mi->mi_msg_list, cur_msg);
658 
659 	if (!cur_msg)
660 		return (0);
661 
662 	if (cur_msg->msg_type != RM_EVENT)
663 		return (0);
664 
665 	ep1 = &cur_msg->rmsg_u.msg_event;
666 	ep2 = &new_msg->rmsg_u.msg_event;
667 	if (ep1->re_type != ep2->re_type)
668 		return (0);
669 
670 	/*
671 	 * Since we zalloc the buffer, then the two nfs4_debug_msg's
672 	 * must match up even if all the fields weren't filled in
673 	 * the first place.
674 	 */
675 	if (ep1->re_mi != ep2->re_mi)
676 		return (0);
677 	if (ep1->re_uint != ep2->re_uint)
678 		return (0);
679 	if (ep1->re_stat4 != ep2->re_stat4)
680 		return (0);
681 	if (ep1->re_pid != ep2->re_pid)
682 		return (0);
683 	if (ep1->re_rp1 != ep2->re_rp1)
684 		return (0);
685 	if (ep1->re_rp2 != ep2->re_rp2)
686 		return (0);
687 	if (ep1->re_tag1 != ep2->re_tag1)
688 		return (0);
689 	if (ep1->re_tag2 != ep2->re_tag2)
690 		return (0);
691 	if (ep1->re_seqid1 != ep2->re_seqid1)
692 		return (0);
693 	if (ep1->re_seqid2 != ep2->re_seqid2)
694 		return (0);
695 
696 	if (cur_msg->msg_srv != NULL) {
697 		if (new_msg->msg_srv == NULL)
698 			return (0);
699 		len = strlen(cur_msg->msg_srv);
700 		if (strncmp(cur_msg->msg_srv, new_msg->msg_srv, len) != 0)
701 			return (0);
702 	} else if (new_msg->msg_srv != NULL) {
703 		return (0);
704 	}
705 
706 	if (ep1->re_char1 != NULL) {
707 		if (ep2->re_char1 == NULL)
708 			return (0);
709 		len = strlen(ep1->re_char1);
710 		if (strncmp(ep1->re_char1, ep2->re_char1, len) != 0)
711 			return (0);
712 	} else if (ep2->re_char1 != NULL) {
713 		return (0);
714 	}
715 
716 	if (ep1->re_char2 != NULL) {
717 		if (ep2->re_char2 == NULL)
718 			return (0);
719 		len = strlen(ep1->re_char2);
720 		if (strncmp(ep1->re_char2, ep2->re_char2, len) != 0)
721 			return (0);
722 	} else if (ep2->re_char2 != NULL) {
723 		return (0);
724 	}
725 
726 	if (cur_msg->msg_mntpt != NULL) {
727 		if (new_msg->msg_mntpt == NULL)
728 			return (0);
729 		len = strlen(cur_msg->msg_mntpt);
730 		if (strncmp(cur_msg->msg_mntpt, cur_msg->msg_mntpt, len) != 0)
731 			return (0);
732 	} else if (new_msg->msg_mntpt != NULL) {
733 		return (0);
734 	}
735 
736 	return (1);
737 }
738 
739 /*
740  * Free up a recovery event.
741  */
742 static void
743 free_event(nfs4_revent_t *ep)
744 {
745 	int	len;
746 
747 	if (ep->re_char1) {
748 		len = strlen(ep->re_char1) + 1;
749 		kmem_free(ep->re_char1, len);
750 	}
751 	if (ep->re_char2) {
752 		len = strlen(ep->re_char2) + 1;
753 		kmem_free(ep->re_char2, len);
754 	}
755 }
756 
757 /*
758  * Free up a recovery fact.
759  */
760 static void
761 free_fact(nfs4_rfact_t *fp)
762 {
763 	int	len;
764 
765 	if (fp->rf_char1) {
766 		len = strlen(fp->rf_char1) + 1;
767 		kmem_free(fp->rf_char1, len);
768 	}
769 }
770 
771 /*
772  * Free up the message.
773  */
774 void
775 nfs4_free_msg(nfs4_debug_msg_t *msg)
776 {
777 	int len;
778 
779 	if (msg->msg_type == RM_EVENT)
780 		free_event(&msg->rmsg_u.msg_event);
781 	else
782 		free_fact(&msg->rmsg_u.msg_fact);
783 
784 	if (msg->msg_srv) {
785 		len = strlen(msg->msg_srv) + 1;
786 		kmem_free(msg->msg_srv, len);
787 	}
788 
789 	if (msg->msg_mntpt) {
790 		len = strlen(msg->msg_mntpt) + 1;
791 		kmem_free(msg->msg_mntpt, len);
792 	}
793 
794 	/* free up the data structure itself */
795 	kmem_free(msg, sizeof (*msg));
796 }
797 
798 /*
799  * Prints out the interesting facts for recovery events:
800  * -DEAD_FILE
801  * -SIGLOST(_NO_DUMP)
802  */
803 static void
804 print_facts(nfs4_debug_msg_t *msg, mntinfo4_t *mi)
805 {
806 	nfs4_rfact_t *fp;
807 	char *mount_pt;
808 	int len;
809 
810 	if (msg->rmsg_u.msg_event.re_type != RE_DEAD_FILE &&
811 	    msg->rmsg_u.msg_event.re_type != RE_SIGLOST &&
812 	    msg->rmsg_u.msg_event.re_type != RE_SIGLOST_NO_DUMP)
813 		return;
814 
815 	fp = kmem_zalloc(sizeof (*fp), KM_SLEEP);
816 	mount_pt = NULL;
817 
818 	if (get_facts(msg, fp, &mount_pt, mi)) {
819 		char	time[256];
820 
821 
822 		if (fp->rf_time.tv_sec)
823 			(void) snprintf(time, 256, "%ld",
824 			    (gethrestime_sec() - fp->rf_time.tv_sec)/60);
825 		zcmn_err(mi->mi_zone->zone_id, CE_NOTE,
826 		    "!NFS4 FACT SHEET: %s%s %s%s %s %s%s%s %s%s",
827 		    fp->rf_action ? "\n Action: " : "",
828 		    fp->rf_action ? nfs4_recov_action_to_str(fp->rf_action) :
829 		    "",
830 		    fp->rf_stat4 ? "\n NFS4 error: " : "",
831 		    fp->rf_stat4 ? nfs4_stat_to_str(fp->rf_stat4) : "",
832 		    fp->rf_reboot ? "\n Suspected server reboot. " : "",
833 		    fp->rf_time.tv_sec ? "\n Server was down for " : "",
834 		    fp->rf_time.tv_sec ? time : "",
835 		    fp->rf_time.tv_sec ? " minutes." : "",
836 		    mount_pt ? " \n Client's lease expired on mount " : "",
837 		    mount_pt ? mount_pt : "");
838 	}
839 
840 	if (mount_pt) {
841 		len = strlen(mount_pt) + 1;
842 		kmem_free(mount_pt, len);
843 	}
844 
845 	/* free the fact struct itself */
846 	if (fp)
847 		kmem_free(fp, sizeof (*fp));
848 }
849 
850 /*
851  * Print an event message to /var/adm/messages
852  * The last argument to this fuction dictates the repeat status
853  * of the event. If set to 1, it means that we are dumping this
854  * event and it will _never_ be printed after this time. Else if
855  * set to 0 it will be printed again.
856  */
857 static void
858 queue_print_event(nfs4_debug_msg_t *msg, mntinfo4_t *mi, int dump)
859 {
860 	nfs4_revent_t		*ep;
861 	zoneid_t		zoneid;
862 
863 	ep = &msg->rmsg_u.msg_event;
864 	zoneid = mi->mi_zone->zone_id;
865 
866 	switch (ep->re_type) {
867 	case RE_BAD_SEQID:
868 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
869 		    "Operation %s for file %s (rnode_pt 0x%p), pid %d using "
870 		    "seqid %d got %s.  Last good seqid was %d for "
871 		    "operation %s.",
872 		    msg->msg_srv, msg->msg_mntpt,
873 		    nfs4_ctags[ep->re_tag1].ct_str, ep->re_char1,
874 		    (void *)ep->re_rp1, ep->re_pid, ep->re_seqid1,
875 		    nfs4_stat_to_str(ep->re_stat4), ep->re_seqid2,
876 		    nfs4_ctags[ep->re_tag2].ct_str);
877 		break;
878 	case RE_BADHANDLE:
879 		ASSERT(ep->re_rp1 != NULL);
880 		if (ep->re_char1 != NULL) {
881 			zcmn_err(zoneid, CE_NOTE,
882 			    "![NFS4][Server: %s][Mntpt: %s]"
883 			    "server %s said filehandle was "
884 			    "invalid for file: %s (rnode_pt 0x%p) on mount %s",
885 			    msg->msg_srv, msg->msg_mntpt, msg->msg_srv,
886 			    ep->re_char1, (void *)ep->re_rp1, msg->msg_mntpt);
887 		} else {
888 			zcmn_err(zoneid, CE_NOTE,
889 			    "![NFS4][Server: %s][Mntpt: %s]"
890 			    "server %s said filehandle was "
891 			    "invalid for file: (rnode_pt 0x%p) on mount %s"
892 			    " for fh:", msg->msg_srv, msg->msg_mntpt,
893 			    msg->msg_srv, (void *)ep->re_rp1, msg->msg_mntpt);
894 			sfh4_printfhandle(ep->re_rp1->r_fh);
895 		}
896 		break;
897 	case RE_CLIENTID:
898 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
899 		    "Can't recover clientid on mount point %s "
900 		    "(mi 0x%p) due to error %d (%s), for server %s.  Marking "
901 		    "file system as unusable.",
902 		    msg->msg_srv, msg->msg_mntpt, msg->msg_mntpt,
903 		    (void *)ep->re_mi, ep->re_uint,
904 		    nfs4_stat_to_str(ep->re_stat4),
905 		    msg->msg_srv);
906 		break;
907 	case RE_DEAD_FILE:
908 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
909 		    "File %s (rnode_pt: %p) was closed due to NFS "
910 		    "recovery error on server %s(%s %s)", msg->msg_srv,
911 		    msg->msg_mntpt, ep->re_char1, (void *)ep->re_rp1,
912 		    msg->msg_srv, ep->re_char2 ? ep->re_char2 : "",
913 		    ep->re_stat4 ? nfs4_stat_to_str(ep->re_stat4) : "");
914 		break;
915 	case RE_END:
916 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
917 		    "NFS Recovery done for mount %s (mi 0x%p) "
918 		    "on server %s, rnode_pt1 %s (0x%p), "
919 		    "rnode_pt2 %s (0x%p)", msg->msg_srv, msg->msg_mntpt,
920 		    msg->msg_mntpt, (void *)ep->re_mi, msg->msg_srv,
921 		    ep->re_char1, (void *)ep->re_rp1, ep->re_char2,
922 		    (void *)ep->re_rp2);
923 		break;
924 	case RE_FAIL_RELOCK:
925 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
926 		    "Couldn't reclaim lock for pid %d for "
927 		    "file %s (rnode_pt 0x%p) on (server %s): error %d",
928 		    msg->msg_srv, msg->msg_mntpt, ep->re_pid, ep->re_char1,
929 		    (void *)ep->re_rp1, msg->msg_srv,
930 		    ep->re_uint ? ep->re_uint : ep->re_stat4);
931 		break;
932 	case RE_FAIL_REMAP_LEN:
933 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
934 		    "remap_lookup: server %s returned bad "
935 		    "fhandle length (%d)", msg->msg_srv, msg->msg_mntpt,
936 		    msg->msg_srv, ep->re_uint);
937 		break;
938 	case RE_FAIL_REMAP_OP:
939 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
940 		    "remap_lookup: didn't get expected OP_GETFH"
941 		    " for server %s", msg->msg_srv, msg->msg_mntpt,
942 		    msg->msg_srv);
943 		break;
944 	case RE_FAILOVER:
945 		if (ep->re_char1)
946 			zcmn_err(zoneid, CE_NOTE,
947 			    "![NFS4][Server: %s][Mntpt: %s]"
948 			    "failing over from %s to %s", msg->msg_srv,
949 			    msg->msg_mntpt, msg->msg_srv, ep->re_char1);
950 		else
951 			zcmn_err(zoneid, CE_NOTE,
952 			    "![NFS4][Server: %s][Mntpt: %s]"
953 			    "NFS4: failing over: selecting "
954 			    "original server %s", msg->msg_srv, msg->msg_mntpt,
955 			    msg->msg_srv);
956 		break;
957 	case RE_FILE_DIFF:
958 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
959 		    "File %s (rnode_pt: %p) on server %s was closed "
960 		    "and failed attempted failover since its is different than "
961 		    "the original file", msg->msg_srv, msg->msg_mntpt,
962 		    ep->re_char1, (void *)ep->re_rp1, msg->msg_srv);
963 		break;
964 	case RE_LOST_STATE:
965 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
966 		    "Lost %s request for fs %s, file %s (rnode_pt: 0x%p), "
967 		    "dir %s (0x%p) for server %s", msg->msg_srv, msg->msg_mntpt,
968 		    nfs4_op_to_str(ep->re_uint), msg->msg_mntpt,
969 		    ep->re_char1, (void *)ep->re_rp1, ep->re_char2,
970 		    (void *)ep->re_rp2, msg->msg_srv);
971 		break;
972 	case RE_OPENS_CHANGED:
973 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
974 		    "The number of open files to reopen changed "
975 		    "for mount %s mi 0x%p (old %d, new %d) on server %s",
976 		    msg->msg_srv, msg->msg_mntpt, msg->msg_mntpt,
977 		    (void *)ep->re_mi, ep->re_uint, ep->re_pid, msg->msg_srv);
978 		break;
979 	case RE_SIGLOST:
980 	case RE_SIGLOST_NO_DUMP:
981 		if (ep->re_uint)
982 			zcmn_err(zoneid, CE_NOTE,
983 			    "![NFS4][Server: %s][Mntpt: %s]"
984 			    "Process %d lost its locks on "
985 			    "file %s (rnode_pt: %p) due to NFS recovery error "
986 			    "(%d) on server %s.", msg->msg_srv, msg->msg_mntpt,
987 			    ep->re_pid, ep->re_char1, (void *)ep->re_rp1,
988 			    ep->re_uint, msg->msg_srv);
989 		else
990 			zcmn_err(zoneid, CE_NOTE,
991 			    "![NFS4][Server: %s][Mntpt: %s]"
992 			    "Process %d lost its locks on "
993 			    "file %s (rnode_pt: %p) due to NFS recovery error "
994 			    "(%s) on server %s.", msg->msg_srv, msg->msg_mntpt,
995 			    ep->re_pid, ep->re_char1, (void *)ep->re_rp1,
996 			    nfs4_stat_to_str(ep->re_stat4), msg->msg_srv);
997 		break;
998 	case RE_START:
999 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1000 		    "NFS Starting recovery for mount %s "
1001 		    "(mi 0x%p mi_recovflags [0x%x]) on server %s, "
1002 		    "rnode_pt1 %s (0x%p), rnode_pt2 %s (0x%p)", msg->msg_srv,
1003 		    msg->msg_mntpt, msg->msg_mntpt, (void *)ep->re_mi,
1004 		    ep->re_uint, msg->msg_srv, ep->re_char1, (void *)ep->re_rp1,
1005 		    ep->re_char2, (void *)ep->re_rp2);
1006 		break;
1007 	case RE_UNEXPECTED_ACTION:
1008 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1009 		    "NFS recovery: unexpected action (%s) on server %s",
1010 		    msg->msg_srv, msg->msg_mntpt,
1011 		    nfs4_recov_action_to_str(ep->re_uint), msg->msg_srv);
1012 		break;
1013 	case RE_UNEXPECTED_ERRNO:
1014 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1015 		    "NFS recovery: unexpected errno (%d) on server %s",
1016 		    msg->msg_srv, msg->msg_mntpt, ep->re_uint, msg->msg_srv);
1017 		break;
1018 	case RE_UNEXPECTED_STATUS:
1019 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1020 		    "NFS recovery: unexpected NFS status code (%s) "
1021 		    "on server %s", msg->msg_srv, msg->msg_mntpt,
1022 		    nfs4_stat_to_str(ep->re_stat4),
1023 		    msg->msg_srv);
1024 		break;
1025 	case RE_WRONGSEC:
1026 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1027 		    "NFS can't recover from NFS4ERR_WRONGSEC."
1028 		    "  error %d for server %s: rnode_pt1 %s (0x%p)"
1029 		    " rnode_pt2 %s (0x%p)", msg->msg_srv, msg->msg_mntpt,
1030 		    ep->re_uint, msg->msg_srv, ep->re_char1, (void *)ep->re_rp1,
1031 		    ep->re_char2, (void *)ep->re_rp2);
1032 		break;
1033 	case RE_LOST_STATE_BAD_OP:
1034 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1035 		    "NFS lost state with unrecognized op (%d)."
1036 		    "  fs %s, server %s, pid %d, file %s (rnode_pt: 0x%p), "
1037 		    "dir %s (0x%p)", msg->msg_srv, msg->msg_mntpt,
1038 		    ep->re_uint, msg->msg_mntpt, msg->msg_srv, ep->re_pid,
1039 		    ep->re_char1, (void *)ep->re_rp1, ep->re_char2,
1040 		    (void *)ep->re_rp2);
1041 		break;
1042 	default:
1043 		zcmn_err(zoneid, CE_WARN,
1044 		    "!queue_print_event: illegal event %d", ep->re_type);
1045 		break;
1046 	}
1047 
1048 	print_facts(msg, mi);
1049 
1050 	/*
1051 	 * If set this event will not be printed again and is considered
1052 	 * dumped.
1053 	 */
1054 	if (dump)
1055 		msg->msg_status = NFS4_MS_NO_DUMP;
1056 }
1057 
1058 /*
1059  * Print a fact message to /var/adm/messages
1060  */
1061 static void
1062 queue_print_fact(nfs4_debug_msg_t *msg, int dump)
1063 {
1064 	nfs4_rfact_t	*fp;
1065 	zoneid_t	zoneid;
1066 
1067 	fp = &msg->rmsg_u.msg_fact;
1068 	zoneid = getzoneid();
1069 
1070 	switch (fp->rf_type) {
1071 	case RF_BADOWNER:
1072 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1073 		    "NFSMAPID_DOMAIN does not match the server: %s domain\n"
1074 		    "Please check configuration", msg->msg_srv, msg->msg_mntpt,
1075 		    msg->msg_srv);
1076 		break;
1077 	case RF_ERR:
1078 		if (fp->rf_error)
1079 			zcmn_err(zoneid, CE_NOTE,
1080 			    "![NFS4][Server: %s][Mntpt: %s]NFS op %s got "
1081 			    "error %d causing recovery action %s.%s",
1082 			    msg->msg_srv, msg->msg_mntpt,
1083 			    nfs4_op_to_str(fp->rf_op), fp->rf_error,
1084 			    nfs4_recov_action_to_str(fp->rf_action),
1085 			    fp->rf_reboot ?
1086 			    "  Client also suspects that the server rebooted,"
1087 			    " or experienced a network partition." : "");
1088 		else
1089 			zcmn_err(zoneid, CE_NOTE,
1090 			    "![NFS4][Server: %s][Mntpt: %s]NFS op %s got "
1091 			    "error %s causing recovery action %s.%s",
1092 			    msg->msg_srv, msg->msg_mntpt,
1093 			    nfs4_op_to_str(fp->rf_op),
1094 			    nfs4_stat_to_str(fp->rf_stat4),
1095 			    nfs4_recov_action_to_str(fp->rf_action),
1096 			    fp->rf_reboot ?
1097 			    "  Client also suspects that the server rebooted,"
1098 			    " or experienced a network partition." : "");
1099 		break;
1100 	case RF_RENEW_EXPIRED:
1101 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1102 		    "NFS4 renew thread detected client's "
1103 		    "lease has expired. Current open files/locks/IO may fail",
1104 		    msg->msg_srv, msg->msg_mntpt);
1105 		break;
1106 	case RF_SRV_NOT_RESPOND:
1107 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1108 		    "NFS server %s not responding; still trying\n",
1109 		    msg->msg_srv, msg->msg_mntpt, msg->msg_srv);
1110 		break;
1111 	case RF_SRV_OK:
1112 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1113 		    "NFS server %s ok", msg->msg_srv, msg->msg_mntpt,
1114 		    msg->msg_srv);
1115 		break;
1116 	case RF_SRVS_NOT_RESPOND:
1117 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1118 		    "NFS servers %s not responding; still trying", msg->msg_srv,
1119 		    msg->msg_mntpt, msg->msg_srv);
1120 		break;
1121 	case RF_SRVS_OK:
1122 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1123 		    "NFS servers %s ok", msg->msg_srv, msg->msg_mntpt,
1124 		    msg->msg_srv);
1125 		break;
1126 	case RF_DELMAP_CB_ERR:
1127 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1128 		    "NFS op %s got error %s when executing delmap on file %s "
1129 		    "(rnode_pt 0x%p).",
1130 		    msg->msg_srv, msg->msg_mntpt, nfs4_op_to_str(fp->rf_op),
1131 		    nfs4_stat_to_str(fp->rf_stat4), fp->rf_char1,
1132 		    (void *)fp->rf_rp1);
1133 		break;
1134 	case RF_SENDQ_FULL:
1135 		zcmn_err(zoneid, CE_NOTE, "![NFS4][Server: %s][Mntpt: %s]"
1136 		    "send queue to NFS server %s is full; still trying\n",
1137 		    msg->msg_srv, msg->msg_mntpt, msg->msg_srv);
1138 		break;
1139 
1140 	default:
1141 		zcmn_err(zoneid, CE_WARN, "!queue_print_fact: illegal fact %d",
1142 		    fp->rf_type);
1143 	}
1144 
1145 	/*
1146 	 * If set this fact will not be printed again and is considered
1147 	 * dumped.
1148 	 */
1149 	if (dump)
1150 		msg->msg_status = NFS4_MS_NO_DUMP;
1151 }
1152 
1153 /*
1154  * Returns 1 if the entire queue should be dumped, 0 otherwise.
1155  */
1156 static int
1157 id_to_dump_queue(nfs4_event_type_t id)
1158 {
1159 	switch (id) {
1160 	case RE_DEAD_FILE:
1161 	case RE_SIGLOST:
1162 	case RE_WRONGSEC:
1163 	case RE_CLIENTID:
1164 		return (1);
1165 	default:
1166 		return (0);
1167 	}
1168 }
1169 
1170 /*
1171  * Returns 1 if the event (but not the entire queue) should be printed;
1172  * 0 otherwise.
1173  */
1174 static int
1175 id_to_dump_solo_event(nfs4_event_type_t id)
1176 {
1177 	switch (id) {
1178 	case RE_BAD_SEQID:
1179 	case RE_BADHANDLE:
1180 	case RE_FAIL_REMAP_LEN:
1181 	case RE_FAIL_REMAP_OP:
1182 	case RE_FAILOVER:
1183 	case RE_OPENS_CHANGED:
1184 	case RE_SIGLOST_NO_DUMP:
1185 	case RE_UNEXPECTED_ACTION:
1186 	case RE_UNEXPECTED_ERRNO:
1187 	case RE_UNEXPECTED_STATUS:
1188 	case RE_LOST_STATE_BAD_OP:
1189 		return (1);
1190 	default:
1191 		return (0);
1192 	}
1193 }
1194 
1195 /*
1196  * Returns 1 if the fact (but not the entire queue) should be printed;
1197  * 0 otherwise.
1198  */
1199 static int
1200 id_to_dump_solo_fact(nfs4_fact_type_t id)
1201 {
1202 	switch (id) {
1203 	case RF_SRV_NOT_RESPOND:
1204 	case RF_SRV_OK:
1205 	case RF_SRVS_NOT_RESPOND:
1206 	case RF_SRVS_OK:
1207 	case RF_SENDQ_FULL:
1208 		return (1);
1209 	default:
1210 		return (0);
1211 	}
1212 }
1213 
1214 /*
1215  * Update a kernel stat
1216  */
1217 static void
1218 update_recov_kstats(nfs4_debug_msg_t *msg, mntinfo4_t *mi)
1219 {
1220 	rkstat_t	*rsp;
1221 
1222 	if (!mi->mi_recov_ksp)
1223 		return;
1224 
1225 	rsp = (rkstat_t *)mi->mi_recov_ksp->ks_data;
1226 
1227 	if (msg->msg_type == RM_EVENT) {
1228 		switch (msg->rmsg_u.msg_event.re_type) {
1229 		case RE_BADHANDLE:
1230 			rsp->badhandle.value.ul++;
1231 			break;
1232 		case RE_CLIENTID:
1233 			rsp->clientid.value.ul++;
1234 			break;
1235 		case RE_DEAD_FILE:
1236 			rsp->dead_file.value.ul++;
1237 			break;
1238 		case RE_FAIL_RELOCK:
1239 			rsp->fail_relock.value.ul++;
1240 			break;
1241 		case RE_FILE_DIFF:
1242 			rsp->file_diff.value.ul++;
1243 			break;
1244 		case RE_OPENS_CHANGED:
1245 			rsp->opens_changed.value.ul++;
1246 			break;
1247 		case RE_SIGLOST:
1248 		case RE_SIGLOST_NO_DUMP:
1249 			rsp->siglost.value.ul++;
1250 			break;
1251 		case RE_UNEXPECTED_ACTION:
1252 			rsp->unexp_action.value.ul++;
1253 			break;
1254 		case RE_UNEXPECTED_ERRNO:
1255 			rsp->unexp_errno.value.ul++;
1256 			break;
1257 		case RE_UNEXPECTED_STATUS:
1258 			rsp->unexp_status.value.ul++;
1259 			break;
1260 		case RE_WRONGSEC:
1261 			rsp->wrongsec.value.ul++;
1262 			break;
1263 		case RE_LOST_STATE_BAD_OP:
1264 			rsp->lost_state_bad_op.value.ul++;
1265 			break;
1266 		default:
1267 			break;
1268 		}
1269 	} else if (msg->msg_type == RM_FACT) {
1270 		switch (msg->rmsg_u.msg_fact.rf_type) {
1271 		case RF_BADOWNER:
1272 			rsp->badowner.value.ul++;
1273 			break;
1274 		case RF_SRV_NOT_RESPOND:
1275 			rsp->not_responding.value.ul++;
1276 			break;
1277 		default:
1278 			break;
1279 		}
1280 	}
1281 }
1282 
1283 /*
1284  * Dump the mi's mi_msg_list of recovery messages.
1285  */
1286 static void
1287 dump_queue(mntinfo4_t *mi, nfs4_debug_msg_t *msg)
1288 {
1289 	nfs4_debug_msg_t *tmp_msg;
1290 
1291 	ASSERT(mutex_owned(&mi->mi_msg_list_lock));
1292 
1293 	/* update kstats */
1294 	update_recov_kstats(msg, mi);
1295 
1296 	/*
1297 	 * If we aren't supposed to dump the queue then see if we
1298 	 * should just print this single message, then return.
1299 	 */
1300 	if (!id_to_dump_queue(msg->rmsg_u.msg_event.re_type)) {
1301 		if (id_to_dump_solo_event(msg->rmsg_u.msg_event.re_type))
1302 			queue_print_event(msg, mi, 0);
1303 		return;
1304 	}
1305 
1306 	/*
1307 	 * Write all events/facts in the queue that haven't been
1308 	 * previously written to disk.
1309 	 */
1310 	tmp_msg = list_head(&mi->mi_msg_list);
1311 	while (tmp_msg) {
1312 		if (tmp_msg->msg_status == NFS4_MS_DUMP) {
1313 			if (tmp_msg->msg_type == RM_EVENT)
1314 				queue_print_event(tmp_msg, mi, 1);
1315 			else if (tmp_msg->msg_type == RM_FACT)
1316 				queue_print_fact(tmp_msg, 1);
1317 		}
1318 		tmp_msg = list_next(&mi->mi_msg_list, tmp_msg);
1319 	}
1320 }
1321 
1322 /*
1323  * Places the event into mi's debug recovery message queue.  Some of the
1324  * fields can be overloaded to be a generic value, depending on the event
1325  * type.  These include "count", "why".
1326  */
1327 void
1328 nfs4_queue_event(nfs4_event_type_t id, mntinfo4_t *mi, char *server1,
1329     uint_t count, vnode_t *vp1, vnode_t *vp2, nfsstat4 nfs4_error,
1330     char *why, pid_t pid, nfs4_tag_type_t tag1, nfs4_tag_type_t tag2,
1331     seqid4 seqid1, seqid4 seqid2)
1332 {
1333 	nfs4_debug_msg_t	*msg;
1334 	nfs4_revent_t		*ep;
1335 	char			*cur_srv;
1336 	rnode4_t		*rp1 = NULL, *rp2 = NULL;
1337 	refstr_t		*mntpt;
1338 
1339 	ASSERT(mi != NULL);
1340 	if (vp1)
1341 		rp1 = VTOR4(vp1);
1342 	if (vp2)
1343 		rp2 = VTOR4(vp2);
1344 
1345 	/*
1346 	 * Initialize the message with the relevant server/mount_pt/time
1347 	 * information. Also place the relevent event related info.
1348 	 */
1349 	msg = kmem_zalloc(sizeof (*msg), KM_SLEEP);
1350 	msg->msg_type = RM_EVENT;
1351 	msg->msg_status = NFS4_MS_DUMP;
1352 	ep = &msg->rmsg_u.msg_event;
1353 	ep->re_type = id;
1354 	gethrestime(&msg->msg_time);
1355 
1356 	cur_srv = mi->mi_curr_serv->sv_hostname;
1357 	msg->msg_srv = strdup(cur_srv);
1358 	mntpt = vfs_getmntpoint(mi->mi_vfsp);
1359 	msg->msg_mntpt = strdup(refstr_value(mntpt));
1360 	refstr_rele(mntpt);
1361 
1362 	set_event(id, ep, mi, rp1, rp2, count, pid, nfs4_error, server1,
1363 	    why, tag1, tag2, seqid1, seqid2);
1364 
1365 	mutex_enter(&mi->mi_msg_list_lock);
1366 
1367 	/* if this event is the same as the last event, drop it */
1368 	if (events_same(list_tail(&mi->mi_msg_list), msg, mi)) {
1369 		mutex_exit(&mi->mi_msg_list_lock);
1370 		nfs4_free_msg(msg);
1371 		return;
1372 	}
1373 
1374 	/* queue the message at the end of the list */
1375 	list_insert_tail(&mi->mi_msg_list, msg);
1376 
1377 	dump_queue(mi, msg);
1378 
1379 	if (mi->mi_msg_count == nfs4_msg_max) {
1380 		nfs4_debug_msg_t *rm_msg;
1381 
1382 		/* remove the queue'd message at the front of the list */
1383 		rm_msg = list_head(&mi->mi_msg_list);
1384 		list_remove(&mi->mi_msg_list, rm_msg);
1385 		mutex_exit(&mi->mi_msg_list_lock);
1386 		nfs4_free_msg(rm_msg);
1387 	} else {
1388 		mi->mi_msg_count++;
1389 		mutex_exit(&mi->mi_msg_list_lock);
1390 	}
1391 }
1392 
1393 /*
1394  * Places the fact into mi's debug recovery messages queue.
1395  */
1396 void
1397 nfs4_queue_fact(nfs4_fact_type_t fid, mntinfo4_t *mi, nfsstat4 stat4,
1398     nfs4_recov_t raction, nfs_opnum4 op, bool_t reboot, char *srvname,
1399     int error, vnode_t *vp)
1400 {
1401 	nfs4_debug_msg_t	*msg;
1402 	nfs4_rfact_t		*fp;
1403 	char			*cur_srv;
1404 	refstr_t		*mntpt;
1405 
1406 	/*
1407 	 * Initialize the message with the relevant server/mount_pt/time
1408 	 * information. Also place the relevant fact related info.
1409 	 */
1410 	msg = kmem_zalloc(sizeof (*msg), KM_SLEEP);
1411 	msg->msg_type = RM_FACT;
1412 	msg->msg_status = NFS4_MS_DUMP;
1413 	gethrestime(&msg->msg_time);
1414 
1415 	if (srvname)
1416 		cur_srv = srvname;
1417 	else
1418 		cur_srv = mi->mi_curr_serv->sv_hostname;
1419 
1420 	msg->msg_srv = strdup(cur_srv);
1421 	mntpt = vfs_getmntpoint(mi->mi_vfsp);
1422 	msg->msg_mntpt = strdup(refstr_value(mntpt));
1423 	refstr_rele(mntpt);
1424 
1425 	fp = &msg->rmsg_u.msg_fact;
1426 	fp->rf_type = fid;
1427 	fp->rf_status = RFS_NO_INSPECT;
1428 	set_fact(fid, fp, stat4, raction, op, reboot, error, vp);
1429 
1430 	update_recov_kstats(msg, mi);
1431 
1432 	mutex_enter(&mi->mi_msg_list_lock);
1433 
1434 	/* if this fact is the same as the last fact, drop it */
1435 	if (facts_same(list_tail(&mi->mi_msg_list), msg, mi)) {
1436 		mutex_exit(&mi->mi_msg_list_lock);
1437 		nfs4_free_msg(msg);
1438 		return;
1439 	}
1440 
1441 	/* queue the message at the end of the list */
1442 	list_insert_tail(&mi->mi_msg_list, msg);
1443 
1444 	if (id_to_dump_solo_fact(msg->rmsg_u.msg_fact.rf_type))
1445 		queue_print_fact(msg, 0);
1446 
1447 	if (mi->mi_msg_count == nfs4_msg_max) {
1448 		nfs4_debug_msg_t *rm_msg;
1449 
1450 		/* remove the queue'd message at the front of the list */
1451 		rm_msg = list_head(&mi->mi_msg_list);
1452 		list_remove(&mi->mi_msg_list, rm_msg);
1453 		mutex_exit(&mi->mi_msg_list_lock);
1454 		nfs4_free_msg(rm_msg);
1455 	} else {
1456 		mi->mi_msg_count++;
1457 		mutex_exit(&mi->mi_msg_list_lock);
1458 	}
1459 }
1460 
1461 /*
1462  * Initialize the 'mi_recov_kstat' kstat.
1463  */
1464 void
1465 nfs4_mnt_recov_kstat_init(vfs_t *vfsp)
1466 {
1467 	mntinfo4_t *mi = VFTOMI4(vfsp);
1468 	kstat_t		*ksp;
1469 	zoneid_t	zoneid = mi->mi_zone->zone_id;
1470 
1471 	/*
1472 	 * Create the version specific kstats.
1473 	 *
1474 	 * PSARC 2001/697 Contract Private Interface
1475 	 * All nfs kstats are under SunMC contract
1476 	 * Please refer to the PSARC listed above and contact
1477 	 * SunMC before making any changes!
1478 	 *
1479 	 * Changes must be reviewed by Solaris File Sharing
1480 	 * Changes must be communicated to contract-2001-697@sun.com
1481 	 *
1482 	 */
1483 
1484 	if ((ksp = kstat_create_zone("nfs", getminor(vfsp->vfs_dev),
1485 	    "mi_recov_kstat", "misc", KSTAT_TYPE_NAMED,
1486 	    sizeof (rkstat_t) / sizeof (kstat_named_t),
1487 	    KSTAT_FLAG_WRITABLE, zoneid)) == NULL) {
1488 		mi->mi_recov_ksp = NULL;
1489 		zcmn_err(GLOBAL_ZONEID, CE_NOTE,
1490 		    "!mi_recov_kstat for mi %p failed\n",
1491 		    (void *)mi);
1492 		return;
1493 	}
1494 	if (zoneid != GLOBAL_ZONEID)
1495 		kstat_zone_add(ksp, GLOBAL_ZONEID);
1496 	mi->mi_recov_ksp = ksp;
1497 	bcopy(&rkstat_template, ksp->ks_data, sizeof (rkstat_t));
1498 	kstat_install(ksp);
1499 }
1500 
1501 /*
1502  * Increment the "delay" kstat.
1503  */
1504 void
1505 nfs4_mi_kstat_inc_delay(mntinfo4_t *mi)
1506 {
1507 	rkstat_t    *rsp;
1508 
1509 	if (!mi->mi_recov_ksp)
1510 		return;
1511 
1512 	rsp = (rkstat_t *)mi->mi_recov_ksp->ks_data;
1513 	rsp->delay.value.ul++;
1514 }
1515 
1516 /*
1517  * Increment the "no_grace" kstat.
1518  */
1519 void
1520 nfs4_mi_kstat_inc_no_grace(mntinfo4_t *mi)
1521 {
1522 	rkstat_t	*rsp;
1523 
1524 	if (!mi->mi_recov_ksp)
1525 		return;
1526 
1527 	rsp = (rkstat_t *)mi->mi_recov_ksp->ks_data;
1528 	rsp->no_grace.value.ul++;
1529 }
1530