xref: /netbsd-src/sys/dev/raidframe/rf_engine.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: rf_engine.c,v 1.39 2006/11/16 01:33:23 christos Exp $	*/
2 /*
3  * Copyright (c) 1995 Carnegie-Mellon University.
4  * All rights reserved.
5  *
6  * Author: William V. Courtright II, Mark Holland, Rachad Youssef
7  *
8  * Permission to use, copy, modify and distribute this software and
9  * its documentation is hereby granted, provided that both the copyright
10  * notice and this permission notice appear in all copies of the
11  * software, derivative works or modified versions, and any portions
12  * thereof, and that both notices appear in supporting documentation.
13  *
14  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
15  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
16  * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
17  *
18  * Carnegie Mellon requests users of this software to return to
19  *
20  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
21  *  School of Computer Science
22  *  Carnegie Mellon University
23  *  Pittsburgh PA 15213-3890
24  *
25  * any improvements or extensions that they make and grant Carnegie the
26  * rights to redistribute these changes.
27  */
28 
29 /****************************************************************************
30  *                                                                          *
31  * engine.c -- code for DAG execution engine                                *
32  *                                                                          *
33  * Modified to work as follows (holland):                                   *
34  *   A user-thread calls into DispatchDAG, which fires off the nodes that   *
35  *   are direct successors to the header node.  DispatchDAG then returns,   *
36  *   and the rest of the I/O continues asynchronously.  As each node        *
37  *   completes, the node execution function calls FinishNode().  FinishNode *
38  *   scans the list of successors to the node and increments the antecedent *
39  *   counts.  Each node that becomes enabled is placed on a central node    *
40  *   queue.  A dedicated dag-execution thread grabs nodes off of this       *
41  *   queue and fires them.                                                  *
42  *                                                                          *
43  *   NULL nodes are never fired.                                            *
44  *                                                                          *
45  *   Terminator nodes are never fired, but rather cause the callback        *
46  *   associated with the DAG to be invoked.                                 *
47  *                                                                          *
48  *   If a node fails, the dag either rolls forward to the completion or     *
49  *   rolls back, undoing previously-completed nodes and fails atomically.   *
50  *   The direction of recovery is determined by the location of the failed  *
51  *   node in the graph.  If the failure occurred before the commit node in   *
52  *   the graph, backward recovery is used.  Otherwise, forward recovery is  *
53  *   used.                                                                  *
54  *                                                                          *
55  ****************************************************************************/
56 
57 #include <sys/cdefs.h>
58 __KERNEL_RCSID(0, "$NetBSD: rf_engine.c,v 1.39 2006/11/16 01:33:23 christos Exp $");
59 
60 #include <sys/errno.h>
61 
62 #include "rf_threadstuff.h"
63 #include "rf_dag.h"
64 #include "rf_engine.h"
65 #include "rf_etimer.h"
66 #include "rf_general.h"
67 #include "rf_dagutils.h"
68 #include "rf_shutdown.h"
69 #include "rf_raid.h"
70 #include "rf_kintf.h"
71 
72 static void rf_ShutdownEngine(void *);
73 static void DAGExecutionThread(RF_ThreadArg_t arg);
74 static void rf_RaidIOThread(RF_ThreadArg_t arg);
75 
76 /* synchronization primitives for this file.  DO_WAIT should be enclosed in a while loop. */
77 
78 #define DO_LOCK(_r_) \
79 do { \
80 	ks = splbio(); \
81 	RF_LOCK_MUTEX((_r_)->node_queue_mutex); \
82 } while (0)
83 
84 #define DO_UNLOCK(_r_) \
85 do { \
86 	RF_UNLOCK_MUTEX((_r_)->node_queue_mutex); \
87 	splx(ks); \
88 } while (0)
89 
90 #define	DO_WAIT(_r_) \
91 	RF_WAIT_COND((_r_)->node_queue, (_r_)->node_queue_mutex)
92 
93 #define	DO_SIGNAL(_r_) \
94 	RF_BROADCAST_COND((_r_)->node_queue)	/* XXX RF_SIGNAL_COND? */
95 
96 static void
97 rf_ShutdownEngine(void *arg)
98 {
99 	RF_Raid_t *raidPtr;
100 	int ks;
101 
102 	raidPtr = (RF_Raid_t *) arg;
103 
104 	/* Tell the rf_RaidIOThread to shutdown */
105 	simple_lock(&(raidPtr->iodone_lock));
106 
107 	raidPtr->shutdown_raidio = 1;
108 	wakeup(&(raidPtr->iodone));
109 
110 	/* ...and wait for it to tell us it has finished */
111 	while (raidPtr->shutdown_raidio)
112  		ltsleep(&(raidPtr->shutdown_raidio), PRIBIO, "raidshutdown", 0,
113 			&(raidPtr->iodone_lock));
114 
115 	simple_unlock(&(raidPtr->iodone_lock));
116 
117  	/* Now shut down the DAG execution engine. */
118  	DO_LOCK(raidPtr);
119   	raidPtr->shutdown_engine = 1;
120   	DO_SIGNAL(raidPtr);
121  	DO_UNLOCK(raidPtr);
122 
123 }
124 
125 int
126 rf_ConfigureEngine(RF_ShutdownList_t **listp, RF_Raid_t *raidPtr,
127 		   RF_Config_t *cfgPtr)
128 {
129 
130 	rf_mutex_init(&raidPtr->node_queue_mutex);
131 	raidPtr->node_queue = NULL;
132 	raidPtr->dags_in_flight = 0;
133 
134 	/* we create the execution thread only once per system boot. no need
135 	 * to check return code b/c the kernel panics if it can't create the
136 	 * thread. */
137 #if RF_DEBUG_ENGINE
138 	if (rf_engineDebug) {
139 		printf("raid%d: Creating engine thread\n", raidPtr->raidid);
140 	}
141 #endif
142 	if (RF_CREATE_ENGINE_THREAD(raidPtr->engine_thread,
143 				    DAGExecutionThread, raidPtr,
144 				    "raid%d", raidPtr->raidid)) {
145 		printf("raid%d: Unable to create engine thread\n",
146 		       raidPtr->raidid);
147 		return (ENOMEM);
148 	}
149 	if (RF_CREATE_ENGINE_THREAD(raidPtr->engine_helper_thread,
150 				    rf_RaidIOThread, raidPtr,
151 				    "raidio%d", raidPtr->raidid)) {
152 		printf("raid%d: Unable to create raidio thread\n",
153 		       raidPtr->raidid);
154 		return (ENOMEM);
155 	}
156 #if RF_DEBUG_ENGINE
157 	if (rf_engineDebug) {
158 		printf("raid%d: Created engine thread\n", raidPtr->raidid);
159 	}
160 #endif
161 
162 	/* engine thread is now running and waiting for work */
163 #if RF_DEBUG_ENGINE
164 	if (rf_engineDebug) {
165 		printf("raid%d: Engine thread running and waiting for events\n", raidPtr->raidid);
166 	}
167 #endif
168 	rf_ShutdownCreate(listp, rf_ShutdownEngine, raidPtr);
169 
170 	return (0);
171 }
172 
173 static int
174 BranchDone(RF_DagNode_t *node)
175 {
176 	int     i;
177 
178 	/* return true if forward execution is completed for a node and it's
179 	 * succedents */
180 	switch (node->status) {
181 	case rf_wait:
182 		/* should never be called in this state */
183 		RF_PANIC();
184 		break;
185 	case rf_fired:
186 		/* node is currently executing, so we're not done */
187 		return (RF_FALSE);
188 	case rf_good:
189 		/* for each succedent recursively check branch */
190 		for (i = 0; i < node->numSuccedents; i++)
191 			if (!BranchDone(node->succedents[i]))
192 				return RF_FALSE;
193 		return RF_TRUE;	/* node and all succedent branches aren't in
194 				 * fired state */
195 	case rf_bad:
196 		/* succedents can't fire */
197 		return (RF_TRUE);
198 	case rf_recover:
199 		/* should never be called in this state */
200 		RF_PANIC();
201 		break;
202 	case rf_undone:
203 	case rf_panic:
204 		/* XXX need to fix this case */
205 		/* for now, assume that we're done */
206 		return (RF_TRUE);
207 	default:
208 		/* illegal node status */
209 		RF_PANIC();
210 		break;
211 	}
212 }
213 
214 static int
215 NodeReady(RF_DagNode_t *node)
216 {
217 	int     ready;
218 
219 	switch (node->dagHdr->status) {
220 	case rf_enable:
221 	case rf_rollForward:
222 		if ((node->status == rf_wait) &&
223 		    (node->numAntecedents == node->numAntDone))
224 			ready = RF_TRUE;
225 		else
226 			ready = RF_FALSE;
227 		break;
228 	case rf_rollBackward:
229 		RF_ASSERT(node->numSuccDone <= node->numSuccedents);
230 		RF_ASSERT(node->numSuccFired <= node->numSuccedents);
231 		RF_ASSERT(node->numSuccFired <= node->numSuccDone);
232 		if ((node->status == rf_good) &&
233 		    (node->numSuccDone == node->numSuccedents))
234 			ready = RF_TRUE;
235 		else
236 			ready = RF_FALSE;
237 		break;
238 	default:
239 		printf("Execution engine found illegal DAG status in NodeReady\n");
240 		RF_PANIC();
241 		break;
242 	}
243 
244 	return (ready);
245 }
246 
247 
248 
249 /* user context and dag-exec-thread context: Fire a node.  The node's
250  * status field determines which function, do or undo, to be fired.
251  * This routine assumes that the node's status field has alread been
252  * set to "fired" or "recover" to indicate the direction of execution.
253  */
254 static void
255 FireNode(RF_DagNode_t *node)
256 {
257 	switch (node->status) {
258 	case rf_fired:
259 		/* fire the do function of a node */
260 #if RF_DEBUG_ENGINE
261 		if (rf_engineDebug) {
262 			printf("raid%d: Firing node 0x%lx (%s)\n",
263 			       node->dagHdr->raidPtr->raidid,
264 			       (unsigned long) node, node->name);
265 		}
266 #endif
267 		if (node->flags & RF_DAGNODE_FLAG_YIELD) {
268 #if defined(__NetBSD__) && defined(_KERNEL)
269 			/* thread_block(); */
270 			/* printf("Need to block the thread here...\n");  */
271 			/* XXX thread_block is actually mentioned in
272 			 * /usr/include/vm/vm_extern.h */
273 #else
274 			thread_block();
275 #endif
276 		}
277 		(*(node->doFunc)) (node);
278 		break;
279 	case rf_recover:
280 		/* fire the undo function of a node */
281 #if RF_DEBUG_ENGINE
282 		if (rf_engineDebug) {
283 			printf("raid%d: Firing (undo) node 0x%lx (%s)\n",
284 			       node->dagHdr->raidPtr->raidid,
285 			       (unsigned long) node, node->name);
286 		}
287 #endif
288 		if (node->flags & RF_DAGNODE_FLAG_YIELD)
289 #if defined(__NetBSD__) && defined(_KERNEL)
290 			/* thread_block(); */
291 			/* printf("Need to block the thread here...\n"); */
292 			/* XXX thread_block is actually mentioned in
293 			 * /usr/include/vm/vm_extern.h */
294 #else
295 			thread_block();
296 #endif
297 		(*(node->undoFunc)) (node);
298 		break;
299 	default:
300 		RF_PANIC();
301 		break;
302 	}
303 }
304 
305 
306 
307 /* user context:
308  * Attempt to fire each node in a linear array.
309  * The entire list is fired atomically.
310  */
311 static void
312 FireNodeArray(int numNodes, RF_DagNode_t **nodeList)
313 {
314 	RF_DagStatus_t dstat;
315 	RF_DagNode_t *node;
316 	int     i, j;
317 
318 	/* first, mark all nodes which are ready to be fired */
319 	for (i = 0; i < numNodes; i++) {
320 		node = nodeList[i];
321 		dstat = node->dagHdr->status;
322 		RF_ASSERT((node->status == rf_wait) ||
323 			  (node->status == rf_good));
324 		if (NodeReady(node)) {
325 			if ((dstat == rf_enable) ||
326 			    (dstat == rf_rollForward)) {
327 				RF_ASSERT(node->status == rf_wait);
328 				if (node->commitNode)
329 					node->dagHdr->numCommits++;
330 				node->status = rf_fired;
331 				for (j = 0; j < node->numAntecedents; j++)
332 					node->antecedents[j]->numSuccFired++;
333 			} else {
334 				RF_ASSERT(dstat == rf_rollBackward);
335 				RF_ASSERT(node->status == rf_good);
336 				/* only one commit node per graph */
337 				RF_ASSERT(node->commitNode == RF_FALSE);
338 				node->status = rf_recover;
339 			}
340 		}
341 	}
342 	/* now, fire the nodes */
343 	for (i = 0; i < numNodes; i++) {
344 		if ((nodeList[i]->status == rf_fired) ||
345 		    (nodeList[i]->status == rf_recover))
346 			FireNode(nodeList[i]);
347 	}
348 }
349 
350 
351 /* user context:
352  * Attempt to fire each node in a linked list.
353  * The entire list is fired atomically.
354  */
355 static void
356 FireNodeList(RF_DagNode_t *nodeList)
357 {
358 	RF_DagNode_t *node, *next;
359 	RF_DagStatus_t dstat;
360 	int     j;
361 
362 	if (nodeList) {
363 		/* first, mark all nodes which are ready to be fired */
364 		for (node = nodeList; node; node = next) {
365 			next = node->next;
366 			dstat = node->dagHdr->status;
367 			RF_ASSERT((node->status == rf_wait) ||
368 				  (node->status == rf_good));
369 			if (NodeReady(node)) {
370 				if ((dstat == rf_enable) ||
371 				    (dstat == rf_rollForward)) {
372 					RF_ASSERT(node->status == rf_wait);
373 					if (node->commitNode)
374 						node->dagHdr->numCommits++;
375 					node->status = rf_fired;
376 					for (j = 0; j < node->numAntecedents; j++)
377 						node->antecedents[j]->numSuccFired++;
378 				} else {
379 					RF_ASSERT(dstat == rf_rollBackward);
380 					RF_ASSERT(node->status == rf_good);
381 					/* only one commit node per graph */
382 					RF_ASSERT(node->commitNode == RF_FALSE);
383 					node->status = rf_recover;
384 				}
385 			}
386 		}
387 		/* now, fire the nodes */
388 		for (node = nodeList; node; node = next) {
389 			next = node->next;
390 			if ((node->status == rf_fired) ||
391 			    (node->status == rf_recover))
392 				FireNode(node);
393 		}
394 	}
395 }
396 /* interrupt context:
397  * for each succedent
398  *    propagate required results from node to succedent
399  *    increment succedent's numAntDone
400  *    place newly-enable nodes on node queue for firing
401  *
402  * To save context switches, we don't place NIL nodes on the node queue,
403  * but rather just process them as if they had fired.  Note that NIL nodes
404  * that are the direct successors of the header will actually get fired by
405  * DispatchDAG, which is fine because no context switches are involved.
406  *
407  * Important:  when running at user level, this can be called by any
408  * disk thread, and so the increment and check of the antecedent count
409  * must be locked.  I used the node queue mutex and locked down the
410  * entire function, but this is certainly overkill.
411  */
412 static void
413 PropagateResults(RF_DagNode_t *node, int context)
414 {
415 	RF_DagNode_t *s, *a;
416 	RF_Raid_t *raidPtr;
417 	int     i, ks;
418 	RF_DagNode_t *finishlist = NULL;	/* a list of NIL nodes to be
419 						 * finished */
420 	RF_DagNode_t *skiplist = NULL;	/* list of nodes with failed truedata
421 					 * antecedents */
422 	RF_DagNode_t *firelist = NULL;	/* a list of nodes to be fired */
423 	RF_DagNode_t *q = NULL, *qh = NULL, *next;
424 	int     j, skipNode;
425 
426 	raidPtr = node->dagHdr->raidPtr;
427 
428 	DO_LOCK(raidPtr);
429 
430 	/* debug - validate fire counts */
431 	for (i = 0; i < node->numAntecedents; i++) {
432 		a = *(node->antecedents + i);
433 		RF_ASSERT(a->numSuccFired >= a->numSuccDone);
434 		RF_ASSERT(a->numSuccFired <= a->numSuccedents);
435 		a->numSuccDone++;
436 	}
437 
438 	switch (node->dagHdr->status) {
439 	case rf_enable:
440 	case rf_rollForward:
441 		for (i = 0; i < node->numSuccedents; i++) {
442 			s = *(node->succedents + i);
443 			RF_ASSERT(s->status == rf_wait);
444 			(s->numAntDone)++;
445 			if (s->numAntDone == s->numAntecedents) {
446 				/* look for NIL nodes */
447 				if (s->doFunc == rf_NullNodeFunc) {
448 					/* don't fire NIL nodes, just process
449 					 * them */
450 					s->next = finishlist;
451 					finishlist = s;
452 				} else {
453 					/* look to see if the node is to be
454 					 * skipped */
455 					skipNode = RF_FALSE;
456 					for (j = 0; j < s->numAntecedents; j++)
457 						if ((s->antType[j] == rf_trueData) && (s->antecedents[j]->status == rf_bad))
458 							skipNode = RF_TRUE;
459 					if (skipNode) {
460 						/* this node has one or more
461 						 * failed true data
462 						 * dependencies, so skip it */
463 						s->next = skiplist;
464 						skiplist = s;
465 					} else
466 						/* add s to list of nodes (q)
467 						 * to execute */
468 						if (context != RF_INTR_CONTEXT) {
469 							/* we only have to
470 							 * enqueue if we're at
471 							 * intr context */
472 							/* put node on
473                                                            a list to
474                                                            be fired
475                                                            after we
476                                                            unlock */
477 							s->next = firelist;
478 							firelist = s;
479 						} else {
480 							/* enqueue the
481 							   node for
482 							   the dag
483 							   exec thread
484 							   to fire */
485 							RF_ASSERT(NodeReady(s));
486 							if (q) {
487 								q->next = s;
488 								q = s;
489 							} else {
490 								qh = q = s;
491 								qh->next = NULL;
492 							}
493 						}
494 				}
495 			}
496 		}
497 
498 		if (q) {
499 			/* xfer our local list of nodes to the node queue */
500 			q->next = raidPtr->node_queue;
501 			raidPtr->node_queue = qh;
502 			DO_SIGNAL(raidPtr);
503 		}
504 		DO_UNLOCK(raidPtr);
505 
506 		for (; skiplist; skiplist = next) {
507 			next = skiplist->next;
508 			skiplist->status = rf_skipped;
509 			for (i = 0; i < skiplist->numAntecedents; i++) {
510 				skiplist->antecedents[i]->numSuccFired++;
511 			}
512 			if (skiplist->commitNode) {
513 				skiplist->dagHdr->numCommits++;
514 			}
515 			rf_FinishNode(skiplist, context);
516 		}
517 		for (; finishlist; finishlist = next) {
518 			/* NIL nodes: no need to fire them */
519 			next = finishlist->next;
520 			finishlist->status = rf_good;
521 			for (i = 0; i < finishlist->numAntecedents; i++) {
522 				finishlist->antecedents[i]->numSuccFired++;
523 			}
524 			if (finishlist->commitNode)
525 				finishlist->dagHdr->numCommits++;
526 			/*
527 		         * Okay, here we're calling rf_FinishNode() on
528 		         * nodes that have the null function as their
529 		         * work proc. Such a node could be the
530 		         * terminal node in a DAG. If so, it will
531 		         * cause the DAG to complete, which will in
532 		         * turn free memory used by the DAG, which
533 		         * includes the node in question. Thus, we
534 		         * must avoid referencing the node at all
535 		         * after calling rf_FinishNode() on it.  */
536 			rf_FinishNode(finishlist, context);	/* recursive call */
537 		}
538 		/* fire all nodes in firelist */
539 		FireNodeList(firelist);
540 		break;
541 
542 	case rf_rollBackward:
543 		for (i = 0; i < node->numAntecedents; i++) {
544 			a = *(node->antecedents + i);
545 			RF_ASSERT(a->status == rf_good);
546 			RF_ASSERT(a->numSuccDone <= a->numSuccedents);
547 			RF_ASSERT(a->numSuccDone <= a->numSuccFired);
548 
549 			if (a->numSuccDone == a->numSuccFired) {
550 				if (a->undoFunc == rf_NullNodeFunc) {
551 					/* don't fire NIL nodes, just process
552 					 * them */
553 					a->next = finishlist;
554 					finishlist = a;
555 				} else {
556 					if (context != RF_INTR_CONTEXT) {
557 						/* we only have to enqueue if
558 						 * we're at intr context */
559 						/* put node on a list to be
560 						   fired after we unlock */
561 						a->next = firelist;
562 
563 						firelist = a;
564 					} else {
565 						/* enqueue the node for the
566 						   dag exec thread to fire */
567 						RF_ASSERT(NodeReady(a));
568 						if (q) {
569 							q->next = a;
570 							q = a;
571 						} else {
572 							qh = q = a;
573 							qh->next = NULL;
574 						}
575 					}
576 				}
577 			}
578 		}
579 		if (q) {
580 			/* xfer our local list of nodes to the node queue */
581 			q->next = raidPtr->node_queue;
582 			raidPtr->node_queue = qh;
583 			DO_SIGNAL(raidPtr);
584 		}
585 		DO_UNLOCK(raidPtr);
586 		for (; finishlist; finishlist = next) {
587 			/* NIL nodes: no need to fire them */
588 			next = finishlist->next;
589 			finishlist->status = rf_good;
590 			/*
591 		         * Okay, here we're calling rf_FinishNode() on
592 		         * nodes that have the null function as their
593 		         * work proc. Such a node could be the first
594 		         * node in a DAG. If so, it will cause the DAG
595 		         * to complete, which will in turn free memory
596 		         * used by the DAG, which includes the node in
597 		         * question. Thus, we must avoid referencing
598 		         * the node at all after calling
599 		         * rf_FinishNode() on it.  */
600 			rf_FinishNode(finishlist, context);	/* recursive call */
601 		}
602 		/* fire all nodes in firelist */
603 		FireNodeList(firelist);
604 
605 		break;
606 	default:
607 		printf("Engine found illegal DAG status in PropagateResults()\n");
608 		RF_PANIC();
609 		break;
610 	}
611 }
612 
613 
614 
615 /*
616  * Process a fired node which has completed
617  */
618 static void
619 ProcessNode(RF_DagNode_t *node, int context)
620 {
621 	RF_Raid_t *raidPtr;
622 
623 	raidPtr = node->dagHdr->raidPtr;
624 
625 	switch (node->status) {
626 	case rf_good:
627 		/* normal case, don't need to do anything */
628 		break;
629 	case rf_bad:
630 		if ((node->dagHdr->numCommits > 0) ||
631 		    (node->dagHdr->numCommitNodes == 0)) {
632 			/* crossed commit barrier */
633 			node->dagHdr->status = rf_rollForward;
634 #if RF_DEBUG_ENGINE
635 			if (rf_engineDebug) {
636 				printf("raid%d: node (%s) returned fail, rolling forward\n", raidPtr->raidid, node->name);
637 			}
638 #endif
639 		} else {
640 			/* never reached commit barrier */
641 			node->dagHdr->status = rf_rollBackward;
642 #if RF_DEBUG_ENGINE
643 			if (rf_engineDebug) {
644 				printf("raid%d: node (%s) returned fail, rolling backward\n", raidPtr->raidid, node->name);
645 			}
646 #endif
647 		}
648 		break;
649 	case rf_undone:
650 		/* normal rollBackward case, don't need to do anything */
651 		break;
652 	case rf_panic:
653 		/* an undo node failed!!! */
654 		printf("UNDO of a node failed!!!/n");
655 		break;
656 	default:
657 		printf("node finished execution with an illegal status!!!\n");
658 		RF_PANIC();
659 		break;
660 	}
661 
662 	/* enqueue node's succedents (antecedents if rollBackward) for
663 	 * execution */
664 	PropagateResults(node, context);
665 }
666 
667 
668 
669 /* user context or dag-exec-thread context:
670  * This is the first step in post-processing a newly-completed node.
671  * This routine is called by each node execution function to mark the node
672  * as complete and fire off any successors that have been enabled.
673  */
674 int
675 rf_FinishNode(RF_DagNode_t *node, int context)
676 {
677 	int     retcode = RF_FALSE;
678 	node->dagHdr->numNodesCompleted++;
679 	ProcessNode(node, context);
680 
681 	return (retcode);
682 }
683 
684 
685 /* user context: submit dag for execution, return non-zero if we have
686  * to wait for completion.  if and only if we return non-zero, we'll
687  * cause cbFunc to get invoked with cbArg when the DAG has completed.
688  *
689  * for now we always return 1.  If the DAG does not cause any I/O,
690  * then the callback may get invoked before DispatchDAG returns.
691  * There's code in state 5 of ContinueRaidAccess to handle this.
692  *
693  * All we do here is fire the direct successors of the header node.
694  * The DAG execution thread does the rest of the dag processing.  */
695 int
696 rf_DispatchDAG(RF_DagHeader_t *dag, void (*cbFunc) (void *),
697 	       void *cbArg)
698 {
699 	RF_Raid_t *raidPtr;
700 
701 	raidPtr = dag->raidPtr;
702 #if RF_ACC_TRACE > 0
703 	if (dag->tracerec) {
704 		RF_ETIMER_START(dag->tracerec->timer);
705 	}
706 #endif
707 #if DEBUG
708 #if RF_DEBUG_VALIDATE_DAG
709 	if (rf_engineDebug || rf_validateDAGDebug) {
710 		if (rf_ValidateDAG(dag))
711 			RF_PANIC();
712 	}
713 #endif
714 #endif
715 #if RF_DEBUG_ENGINE
716 	if (rf_engineDebug) {
717 		printf("raid%d: Entering DispatchDAG\n", raidPtr->raidid);
718 	}
719 #endif
720 	raidPtr->dags_in_flight++;	/* debug only:  blow off proper
721 					 * locking */
722 	dag->cbFunc = cbFunc;
723 	dag->cbArg = cbArg;
724 	dag->numNodesCompleted = 0;
725 	dag->status = rf_enable;
726 	FireNodeArray(dag->numSuccedents, dag->succedents);
727 	return (1);
728 }
729 /* dedicated kernel thread: the thread that handles all DAG node
730  * firing.  To minimize locking and unlocking, we grab a copy of the
731  * entire node queue and then set the node queue to NULL before doing
732  * any firing of nodes.  This way we only have to release the lock
733  * once.  Of course, it's probably rare that there's more than one
734  * node in the queue at any one time, but it sometimes happens.
735  */
736 
737 static void
738 DAGExecutionThread(RF_ThreadArg_t arg)
739 {
740 	RF_DagNode_t *nd, *local_nq, *term_nq, *fire_nq;
741 	RF_Raid_t *raidPtr;
742 	int     ks;
743 	int     s;
744 
745 	raidPtr = (RF_Raid_t *) arg;
746 
747 #if RF_DEBUG_ENGINE
748 	if (rf_engineDebug) {
749 		printf("raid%d: Engine thread is running\n", raidPtr->raidid);
750 	}
751 #endif
752 	s = splbio();
753 
754 	DO_LOCK(raidPtr);
755 	while (!raidPtr->shutdown_engine) {
756 
757 		while (raidPtr->node_queue != NULL) {
758 			local_nq = raidPtr->node_queue;
759 			fire_nq = NULL;
760 			term_nq = NULL;
761 			raidPtr->node_queue = NULL;
762 			DO_UNLOCK(raidPtr);
763 
764 			/* first, strip out the terminal nodes */
765 			while (local_nq) {
766 				nd = local_nq;
767 				local_nq = local_nq->next;
768 				switch (nd->dagHdr->status) {
769 				case rf_enable:
770 				case rf_rollForward:
771 					if (nd->numSuccedents == 0) {
772 						/* end of the dag, add to
773 						 * callback list */
774 						nd->next = term_nq;
775 						term_nq = nd;
776 					} else {
777 						/* not the end, add to the
778 						 * fire queue */
779 						nd->next = fire_nq;
780 						fire_nq = nd;
781 					}
782 					break;
783 				case rf_rollBackward:
784 					if (nd->numAntecedents == 0) {
785 						/* end of the dag, add to the
786 						 * callback list */
787 						nd->next = term_nq;
788 						term_nq = nd;
789 					} else {
790 						/* not the end, add to the
791 						 * fire queue */
792 						nd->next = fire_nq;
793 						fire_nq = nd;
794 					}
795 					break;
796 				default:
797 					RF_PANIC();
798 					break;
799 				}
800 			}
801 
802 			/* execute callback of dags which have reached the
803 			 * terminal node */
804 			while (term_nq) {
805 				nd = term_nq;
806 				term_nq = term_nq->next;
807 				nd->next = NULL;
808 				(nd->dagHdr->cbFunc) (nd->dagHdr->cbArg);
809 				raidPtr->dags_in_flight--;	/* debug only */
810 			}
811 
812 			/* fire remaining nodes */
813 			FireNodeList(fire_nq);
814 
815 			DO_LOCK(raidPtr);
816 		}
817 		while (!raidPtr->shutdown_engine &&
818 		       raidPtr->node_queue == NULL) {
819 			DO_WAIT(raidPtr);
820 		}
821 	}
822 	DO_UNLOCK(raidPtr);
823 
824 	splx(s);
825 	kthread_exit(0);
826 }
827 
828 /*
829  * rf_RaidIOThread() -- When I/O to a component begins, raidstrategy()
830  * puts the I/O on a buf_queue, and then signals raidPtr->iodone.  If
831  * necessary, this function calls raidstart() to initiate the I/O.
832  * When I/O to a component completes, KernelWakeupFunc() puts the
833  * completed request onto raidPtr->iodone TAILQ.  This function looks
834  * after requests on that queue by calling rf_DiskIOComplete() for the
835  * request, and by calling any required CompleteFunc for the request.
836  */
837 
838 static void
839 rf_RaidIOThread(RF_ThreadArg_t arg)
840 {
841 	RF_Raid_t *raidPtr;
842 	RF_DiskQueueData_t *req;
843 	int s;
844 
845 	raidPtr = (RF_Raid_t *) arg;
846 
847 	s = splbio();
848 	simple_lock(&(raidPtr->iodone_lock));
849 
850 	while (!raidPtr->shutdown_raidio) {
851 		/* if there is nothing to do, then snooze. */
852 		if (TAILQ_EMPTY(&(raidPtr->iodone)) &&
853 		    rf_buf_queue_check(raidPtr->raidid)) {
854 			ltsleep(&(raidPtr->iodone), PRIBIO, "raidiow", 0,
855 				&(raidPtr->iodone_lock));
856 		}
857 
858 		/* See what I/Os, if any, have arrived */
859 		while ((req = TAILQ_FIRST(&(raidPtr->iodone))) != NULL) {
860 			TAILQ_REMOVE(&(raidPtr->iodone), req, iodone_entries);
861 			simple_unlock(&(raidPtr->iodone_lock));
862 			rf_DiskIOComplete(req->queue, req, req->error);
863 			(req->CompleteFunc) (req->argument, req->error);
864 			simple_lock(&(raidPtr->iodone_lock));
865 		}
866 
867 		/* process any pending outgoing IO */
868 		simple_unlock(&(raidPtr->iodone_lock));
869 		raidstart(raidPtr);
870 		simple_lock(&(raidPtr->iodone_lock));
871 
872 	}
873 
874 	/* Let rf_ShutdownEngine know that we're done... */
875 	raidPtr->shutdown_raidio = 0;
876 	wakeup(&(raidPtr->shutdown_raidio));
877 
878 	simple_unlock(&(raidPtr->iodone_lock));
879 	splx(s);
880 
881 	kthread_exit(0);
882 }
883