Annotation of sys/dev/raidframe/rf_parityloggingdags.c, Revision 1.1.1.1
1.1 nbrk 1: /* $OpenBSD: rf_parityloggingdags.c,v 1.4 2002/12/16 07:01:04 tdeval Exp $ */
2: /* $NetBSD: rf_parityloggingdags.c,v 1.4 2000/01/07 03:41:04 oster Exp $ */
3:
4: /*
5: * Copyright (c) 1995 Carnegie-Mellon University.
6: * All rights reserved.
7: *
8: * Author: William V. Courtright II
9: *
10: * Permission to use, copy, modify and distribute this software and
11: * its documentation is hereby granted, provided that both the copyright
12: * notice and this permission notice appear in all copies of the
13: * software, derivative works or modified versions, and any portions
14: * thereof, and that both notices appear in supporting documentation.
15: *
16: * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
17: * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
18: * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
19: *
20: * Carnegie Mellon requests users of this software to return to
21: *
22: * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
23: * School of Computer Science
24: * Carnegie Mellon University
25: * Pittsburgh PA 15213-3890
26: *
27: * any improvements or extensions that they make and grant Carnegie the
28: * rights to redistribute these changes.
29: */
30:
31: #include "rf_archs.h"
32:
33: #if RF_INCLUDE_PARITYLOGGING > 0
34:
35: /*
36: * DAGs specific to parity logging are created here.
37: */
38:
39: #include "rf_types.h"
40: #include "rf_raid.h"
41: #include "rf_dag.h"
42: #include "rf_dagutils.h"
43: #include "rf_dagfuncs.h"
44: #include "rf_debugMem.h"
45: #include "rf_paritylog.h"
46: #include "rf_memchunk.h"
47: #include "rf_general.h"
48:
49: #include "rf_parityloggingdags.h"
50:
51: /*****************************************************************************
52: *
53: * Creates a DAG to perform a large-write operation:
54: *
55: * / Rod \ / Wnd \
56: * H -- NIL- Rod - NIL - Wnd ------ NIL - T
57: * \ Rod / \ Xor - Lpo /
58: *
59: * The writes are not done until the reads complete because if they were done
60: * in parallel, a failure on one of the reads could leave the parity in an
61: * inconsistent state, so that the retry with a new DAG would produce
62: * erroneous parity.
63: *
64: * Note: This DAG has the nasty property that none of the buffers allocated
65: * for reading old data can be freed until the XOR node fires.
66: * Need to fix this.
67: *
68: * The last two arguments are the number of faults tolerated, and function
69: * for the redundancy calculation. The undo for the redundancy calc is assumed
70: * to be null.
71: *
72: *****************************************************************************/
73:
74: void
75: rf_CommonCreateParityLoggingLargeWriteDAG(RF_Raid_t * raidPtr,
76: RF_AccessStripeMap_t *asmap, RF_DagHeader_t *dag_h, void *bp,
77: RF_RaidAccessFlags_t flags, RF_AllocListElem_t *allocList, int nfaults,
78: int (*redFunc) (RF_DagNode_t *))
79: {
80: RF_DagNode_t *nodes, *wndNodes, *rodNodes = NULL, *syncNode, *xorNode;
81: RF_DagNode_t *lpoNode, *blockNode, *unblockNode, *termNode;
82: int nWndNodes, nRodNodes, i;
83: RF_RaidLayout_t *layoutPtr = &(raidPtr->Layout);
84: RF_AccessStripeMapHeader_t *new_asm_h[2];
85: int nodeNum, asmNum;
86: RF_ReconUnitNum_t which_ru;
87: char *sosBuffer, *eosBuffer;
88: RF_PhysDiskAddr_t *pda;
89: RF_StripeNum_t parityStripeID =
90: rf_RaidAddressToParityStripeID(&(raidPtr->Layout),
91: asmap->raidAddress, &which_ru);
92:
93: if (rf_dagDebug)
94: printf("[Creating parity-logging large-write DAG]\n");
95: RF_ASSERT(nfaults == 1); /* This arch only single fault tolerant. */
96: dag_h->creator = "ParityLoggingLargeWriteDAG";
97:
98: /* Alloc the Wnd nodes, the xor node, and the Lpo node. */
99: nWndNodes = asmap->numStripeUnitsAccessed;
100: RF_CallocAndAdd(nodes, nWndNodes + 6, sizeof(RF_DagNode_t),
101: (RF_DagNode_t *), allocList);
102: i = 0;
103: wndNodes = &nodes[i];
104: i += nWndNodes;
105: xorNode = &nodes[i];
106: i += 1;
107: lpoNode = &nodes[i];
108: i += 1;
109: blockNode = &nodes[i];
110: i += 1;
111: syncNode = &nodes[i];
112: i += 1;
113: unblockNode = &nodes[i];
114: i += 1;
115: termNode = &nodes[i];
116: i += 1;
117:
118: dag_h->numCommitNodes = nWndNodes + 1;
119: dag_h->numCommits = 0;
120: dag_h->numSuccedents = 1;
121:
122: rf_MapUnaccessedPortionOfStripe(raidPtr, layoutPtr, asmap, dag_h,
123: new_asm_h, &nRodNodes, &sosBuffer, &eosBuffer, allocList);
124: if (nRodNodes > 0)
125: RF_CallocAndAdd(rodNodes, nRodNodes, sizeof(RF_DagNode_t),
126: (RF_DagNode_t *), allocList);
127:
128: /* Begin node initialization. */
129: rf_InitNode(blockNode, rf_wait, RF_FALSE, rf_NullNodeFunc,
130: rf_NullNodeUndoFunc, NULL, nRodNodes + 1, 0, 0, 0, dag_h,
131: "Nil", allocList);
132: rf_InitNode(unblockNode, rf_wait, RF_FALSE, rf_NullNodeFunc,
133: rf_NullNodeUndoFunc, NULL, 1, nWndNodes + 1, 0, 0, dag_h,
134: "Nil", allocList);
135: rf_InitNode(syncNode, rf_wait, RF_FALSE, rf_NullNodeFunc,
136: rf_NullNodeUndoFunc, NULL, nWndNodes + 1, nRodNodes + 1,
137: 0, 0, dag_h, "Nil", allocList);
138: rf_InitNode(termNode, rf_wait, RF_FALSE, rf_TerminateFunc,
139: rf_TerminateUndoFunc, NULL, 0, 1, 0, 0, dag_h, "Trm", allocList);
140:
141: /* Initialize the Rod nodes. */
142: for (nodeNum = asmNum = 0; asmNum < 2; asmNum++) {
143: if (new_asm_h[asmNum]) {
144: pda = new_asm_h[asmNum]->stripeMap->physInfo;
145: while (pda) {
146: rf_InitNode(&rodNodes[nodeNum], rf_wait,
147: RF_FALSE, rf_DiskReadFunc,
148: rf_DiskReadUndoFunc, rf_GenericWakeupFunc,
149: 1, 1, 4, 0, dag_h, "Rod", allocList);
150: rodNodes[nodeNum].params[0].p = pda;
151: rodNodes[nodeNum].params[1].p = pda->bufPtr;
152: rodNodes[nodeNum].params[2].v = parityStripeID;
153: rodNodes[nodeNum].params[3].v =
154: RF_CREATE_PARAM3(RF_IO_NORMAL_PRIORITY,
155: 0, 0, which_ru);
156: nodeNum++;
157: pda = pda->next;
158: }
159: }
160: }
161: RF_ASSERT(nodeNum == nRodNodes);
162:
163: /* Initialize the wnd nodes. */
164: pda = asmap->physInfo;
165: for (i = 0; i < nWndNodes; i++) {
166: rf_InitNode(&wndNodes[i], rf_wait, RF_TRUE, rf_DiskWriteFunc,
167: rf_DiskWriteUndoFunc, rf_GenericWakeupFunc, 1, 1, 4, 0,
168: dag_h, "Wnd", allocList);
169: RF_ASSERT(pda != NULL);
170: wndNodes[i].params[0].p = pda;
171: wndNodes[i].params[1].p = pda->bufPtr;
172: wndNodes[i].params[2].v = parityStripeID;
173: wndNodes[i].params[3].v =
174: RF_CREATE_PARAM3(RF_IO_NORMAL_PRIORITY, 0, 0, which_ru);
175: pda = pda->next;
176: }
177:
178: /* Initialize the redundancy node. */
179: rf_InitNode(xorNode, rf_wait, RF_TRUE, redFunc, rf_NullNodeUndoFunc,
180: NULL, 1, 1, 2 * (nWndNodes + nRodNodes) + 1, 1, dag_h,
181: "Xr ", allocList);
182: xorNode->flags |= RF_DAGNODE_FLAG_YIELD;
183: for (i = 0; i < nWndNodes; i++) {
184: /* pda */
185: xorNode->params[2 * i + 0] = wndNodes[i].params[0];
186: /* buf ptr */
187: xorNode->params[2 * i + 1] = wndNodes[i].params[1];
188: }
189: for (i = 0; i < nRodNodes; i++) {
190: xorNode->params[2 * (nWndNodes + i) + 0] =
191: rodNodes[i].params[0]; /* pda */
192: xorNode->params[2 * (nWndNodes + i) + 1] =
193: rodNodes[i].params[1]; /* buf ptr */
194: }
195: /* Xor node needs to get at RAID information. */
196: xorNode->params[2 * (nWndNodes + nRodNodes)].p = raidPtr;
197:
198: /*
199: * Look for an Rod node that reads a complete SU. If none, alloc a
200: * buffer to receive the parity info. Note that we can't use a new
201: * data buffer because it will not have gotten written when the xor
202: * occurs.
203: */
204: for (i = 0; i < nRodNodes; i++)
205: if (((RF_PhysDiskAddr_t *) rodNodes[i].params[0].p)
206: ->numSector == raidPtr->Layout.sectorsPerStripeUnit)
207: break;
208: if (i == nRodNodes) {
209: RF_CallocAndAdd(xorNode->results[0], 1,
210: rf_RaidAddressToByte(raidPtr,
211: raidPtr->Layout.sectorsPerStripeUnit), (void *),
212: allocList);
213: } else {
214: xorNode->results[0] = rodNodes[i].params[1].p;
215: }
216:
217: /* Initialize the Lpo node. */
218: rf_InitNode(lpoNode, rf_wait, RF_FALSE, rf_ParityLogOverwriteFunc,
219: rf_ParityLogOverwriteUndoFunc, rf_GenericWakeupFunc, 1, 1, 2, 0,
220: dag_h, "Lpo", allocList);
221:
222: lpoNode->params[0].p = asmap->parityInfo;
223: lpoNode->params[1].p = xorNode->results[0];
224: /* parityInfo must describe entire parity unit. */
225: RF_ASSERT(asmap->parityInfo->next == NULL);
226:
227: /* Connect nodes to form graph. */
228:
229: /* Connect dag header to block node. */
230: RF_ASSERT(dag_h->numSuccedents == 1);
231: RF_ASSERT(blockNode->numAntecedents == 0);
232: dag_h->succedents[0] = blockNode;
233:
234: /* Connect the block node to the Rod nodes. */
235: RF_ASSERT(blockNode->numSuccedents == nRodNodes + 1);
236: for (i = 0; i < nRodNodes; i++) {
237: RF_ASSERT(rodNodes[i].numAntecedents == 1);
238: blockNode->succedents[i] = &rodNodes[i];
239: rodNodes[i].antecedents[0] = blockNode;
240: rodNodes[i].antType[0] = rf_control;
241: }
242:
243: /* Connect the block node to the sync node. */
244: /* necessary if nRodNodes == 0 */
245: RF_ASSERT(syncNode->numAntecedents == nRodNodes + 1);
246: blockNode->succedents[nRodNodes] = syncNode;
247: syncNode->antecedents[0] = blockNode;
248: syncNode->antType[0] = rf_control;
249:
250: /* Connect the Rod nodes to the syncNode. */
251: for (i = 0; i < nRodNodes; i++) {
252: rodNodes[i].succedents[0] = syncNode;
253: syncNode->antecedents[1 + i] = &rodNodes[i];
254: syncNode->antType[1 + i] = rf_control;
255: }
256:
257: /* Connect the sync node to the xor node. */
258: RF_ASSERT(syncNode->numSuccedents == nWndNodes + 1);
259: RF_ASSERT(xorNode->numAntecedents == 1);
260: syncNode->succedents[0] = xorNode;
261: xorNode->antecedents[0] = syncNode;
262: xorNode->antType[0] = rf_trueData; /* Carry forward from sync. */
263:
264: /* Connect the sync node to the Wnd nodes. */
265: for (i = 0; i < nWndNodes; i++) {
266: RF_ASSERT(wndNodes->numAntecedents == 1);
267: syncNode->succedents[1 + i] = &wndNodes[i];
268: wndNodes[i].antecedents[0] = syncNode;
269: wndNodes[i].antType[0] = rf_control;
270: }
271:
272: /* Connect the xor node to the Lpo node. */
273: RF_ASSERT(xorNode->numSuccedents == 1);
274: RF_ASSERT(lpoNode->numAntecedents == 1);
275: xorNode->succedents[0] = lpoNode;
276: lpoNode->antecedents[0] = xorNode;
277: lpoNode->antType[0] = rf_trueData;
278:
279: /* Connect the Wnd nodes to the unblock node. */
280: RF_ASSERT(unblockNode->numAntecedents == nWndNodes + 1);
281: for (i = 0; i < nWndNodes; i++) {
282: RF_ASSERT(wndNodes->numSuccedents == 1);
283: wndNodes[i].succedents[0] = unblockNode;
284: unblockNode->antecedents[i] = &wndNodes[i];
285: unblockNode->antType[i] = rf_control;
286: }
287:
288: /* Connect the Lpo node to the unblock node. */
289: RF_ASSERT(lpoNode->numSuccedents == 1);
290: lpoNode->succedents[0] = unblockNode;
291: unblockNode->antecedents[nWndNodes] = lpoNode;
292: unblockNode->antType[nWndNodes] = rf_control;
293:
294: /* Connect unblock node to terminator. */
295: RF_ASSERT(unblockNode->numSuccedents == 1);
296: RF_ASSERT(termNode->numAntecedents == 1);
297: RF_ASSERT(termNode->numSuccedents == 0);
298: unblockNode->succedents[0] = termNode;
299: termNode->antecedents[0] = unblockNode;
300: termNode->antType[0] = rf_control;
301: }
302:
303:
304: /*****************************************************************************
305: *
306: * Creates a DAG to perform a small-write operation (either raid 5 or pq),
307: * which is as follows:
308: *
309: * Header
310: * |
311: * Block
312: * / | ... \ \
313: * / | \ \
314: * Rod Rod Rod Rop
315: * | \ /| \ / | \/ |
316: * | | | /\ |
317: * Wnd Wnd Wnd X
318: * | \ / |
319: * | \ / |
320: * \ \ / Lpo
321: * \ \ / /
322: * +-> Unblock <-+
323: * |
324: * T
325: *
326: *
327: * R = Read, W = Write, X = Xor, o = old, n = new, d = data, p = parity.
328: * When the access spans a stripe unit boundary and is less than one SU in
329: * size, there will be two Rop -- X -- Wnp branches. I call this the
330: * "double-XOR" case.
331: * The second output from each Rod node goes to the X node. In the double-XOR
332: * case, there are exactly 2 Rod nodes, and each sends one output to one X
333: * node.
334: * There is one Rod -- Wnd -- T branch for each stripe unit being updated.
335: *
336: * The block and unblock nodes are unused. See comment above
337: * CreateFaultFreeReadDAG.
338: *
339: * Note: This DAG ignores all the optimizations related to making the RMWs
340: * atomic.
341: * It also has the nasty property that none of the buffers allocated
342: * for reading old data & parity can be freed until the XOR node fires.
343: * Need to fix this.
344: *
345: * A null qfuncs indicates single fault tolerant.
346: *****************************************************************************/
347:
348: void
349: rf_CommonCreateParityLoggingSmallWriteDAG(RF_Raid_t *raidPtr,
350: RF_AccessStripeMap_t *asmap, RF_DagHeader_t *dag_h, void *bp,
351: RF_RaidAccessFlags_t flags, RF_AllocListElem_t *allocList,
352: RF_RedFuncs_t *pfuncs, RF_RedFuncs_t *qfuncs)
353: {
354: RF_DagNode_t *xorNodes, *blockNode, *unblockNode, *nodes;
355: RF_DagNode_t *readDataNodes, *readParityNodes;
356: RF_DagNode_t *writeDataNodes, *lpuNodes;
357: RF_DagNode_t *unlockDataNodes = NULL, *termNode;
358: RF_PhysDiskAddr_t *pda = asmap->physInfo;
359: int numDataNodes = asmap->numStripeUnitsAccessed;
360: int numParityNodes = (asmap->parityInfo->next) ? 2 : 1;
361: int i, j, nNodes, totalNumNodes;
362: RF_ReconUnitNum_t which_ru;
363: int (*func) (RF_DagNode_t * node), (*undoFunc) (RF_DagNode_t * node);
364: int (*qfunc) (RF_DagNode_t * node);
365: char*name, *qname;
366: RF_StripeNum_t parityStripeID =
367: rf_RaidAddressToParityStripeID(&(raidPtr->Layout),
368: asmap->raidAddress, &which_ru);
369: long nfaults = qfuncs ? 2 : 1;
370: int lu_flag = (rf_enableAtomicRMW) ? 1 : 0; /* Lock/unlock flag. */
371:
372: if (rf_dagDebug)
373: printf("[Creating parity-logging small-write DAG]\n");
374: RF_ASSERT(numDataNodes > 0);
375: RF_ASSERT(nfaults == 1);
376: dag_h->creator = "ParityLoggingSmallWriteDAG";
377:
378: /*
379: * DAG creation occurs in three steps:
380: * 1. Count the number of nodes in the DAG.
381: * 2. Create the nodes.
382: * 3. Initialize the nodes.
383: * 4. Connect the nodes.
384: */
385:
386: /* Step 1. Compute number of nodes in the graph. */
387:
388: /*
389: * Number of nodes: a read and write for each data unit, a redundancy
390: * computation node for each parity node, a read and Lpu for each
391: * parity unit, a block and unblock node (2), a terminator node if
392: * atomic RMW, an unlock node for each data and redundancy unit.
393: */
394: totalNumNodes = (2 * numDataNodes) + numParityNodes +
395: (2 * numParityNodes) + 3;
396: if (lu_flag)
397: totalNumNodes += numDataNodes;
398:
399: nNodes = numDataNodes + numParityNodes;
400:
401: dag_h->numCommitNodes = numDataNodes + numParityNodes;
402: dag_h->numCommits = 0;
403: dag_h->numSuccedents = 1;
404:
405: /* Step 2. Create the nodes. */
406: RF_CallocAndAdd(nodes, totalNumNodes, sizeof(RF_DagNode_t),
407: (RF_DagNode_t *), allocList);
408: i = 0;
409: blockNode = &nodes[i];
410: i += 1;
411: unblockNode = &nodes[i];
412: i += 1;
413: readDataNodes = &nodes[i];
414: i += numDataNodes;
415: readParityNodes = &nodes[i];
416: i += numParityNodes;
417: writeDataNodes = &nodes[i];
418: i += numDataNodes;
419: lpuNodes = &nodes[i];
420: i += numParityNodes;
421: xorNodes = &nodes[i];
422: i += numParityNodes;
423: termNode = &nodes[i];
424: i += 1;
425: if (lu_flag) {
426: unlockDataNodes = &nodes[i];
427: i += numDataNodes;
428: }
429: RF_ASSERT(i == totalNumNodes);
430:
431: /* Step 3. Initialize the nodes. */
432: /* Initialize block node (Nil). */
433: rf_InitNode(blockNode, rf_wait, RF_FALSE, rf_NullNodeFunc,
434: rf_NullNodeUndoFunc, NULL, nNodes, 0, 0, 0, dag_h,
435: "Nil", allocList);
436:
437: /* Initialize unblock node (Nil). */
438: rf_InitNode(unblockNode, rf_wait, RF_FALSE, rf_NullNodeFunc,
439: rf_NullNodeUndoFunc, NULL, 1, nNodes, 0, 0, dag_h,
440: "Nil", allocList);
441:
442: /* Initialize terminatory node (Trm). */
443: rf_InitNode(termNode, rf_wait, RF_FALSE, rf_TerminateFunc,
444: rf_TerminateUndoFunc, NULL, 0, 1, 0, 0, dag_h, "Trm", allocList);
445:
446: /* Initialize nodes which read old data (Rod). */
447: for (i = 0; i < numDataNodes; i++) {
448: rf_InitNode(&readDataNodes[i], rf_wait, RF_FALSE,
449: rf_DiskReadFunc, rf_DiskReadUndoFunc, rf_GenericWakeupFunc,
450: nNodes, 1, 4, 0, dag_h, "Rod", allocList);
451: RF_ASSERT(pda != NULL);
452: /* Physical disk addr desc. */
453: readDataNodes[i].params[0].p = pda;
454: readDataNodes[i].params[1].p = rf_AllocBuffer(raidPtr, dag_h,
455: pda, allocList); /* Buffer to hold old data. */
456: readDataNodes[i].params[2].v = parityStripeID;
457: readDataNodes[i].params[3].v =
458: RF_CREATE_PARAM3(RF_IO_NORMAL_PRIORITY, lu_flag,
459: 0, which_ru);
460: pda = pda->next;
461: readDataNodes[i].propList[0] = NULL;
462: readDataNodes[i].propList[1] = NULL;
463: }
464:
465: /* Initialize nodes which read old parity (Rop). */
466: pda = asmap->parityInfo;
467: i = 0;
468: for (i = 0; i < numParityNodes; i++) {
469: RF_ASSERT(pda != NULL);
470: rf_InitNode(&readParityNodes[i], rf_wait, RF_FALSE,
471: rf_DiskReadFunc, rf_DiskReadUndoFunc, rf_GenericWakeupFunc,
472: nNodes, 1, 4, 0, dag_h, "Rop", allocList);
473: readParityNodes[i].params[0].p = pda;
474: readParityNodes[i].params[1].p = rf_AllocBuffer(raidPtr, dag_h,
475: pda, allocList); /* Buffer to hold old parity. */
476: readParityNodes[i].params[2].v = parityStripeID;
477: readParityNodes[i].params[3].v =
478: RF_CREATE_PARAM3(RF_IO_NORMAL_PRIORITY, 0, 0, which_ru);
479: readParityNodes[i].propList[0] = NULL;
480: pda = pda->next;
481: }
482:
483: /* Initialize nodes which write new data (Wnd). */
484: pda = asmap->physInfo;
485: for (i = 0; i < numDataNodes; i++) {
486: RF_ASSERT(pda != NULL);
487: rf_InitNode(&writeDataNodes[i], rf_wait, RF_TRUE,
488: rf_DiskWriteFunc, rf_DiskWriteUndoFunc,
489: rf_GenericWakeupFunc, 1, nNodes, 4, 0, dag_h,
490: "Wnd", allocList);
491: /* Physical disk addr desc. */
492: writeDataNodes[i].params[0].p = pda;
493: /* Buffer holding new data to be written. */
494: writeDataNodes[i].params[1].p = pda->bufPtr;
495: writeDataNodes[i].params[2].v = parityStripeID;
496: writeDataNodes[i].params[3].v =
497: RF_CREATE_PARAM3(RF_IO_NORMAL_PRIORITY, 0, 0, which_ru);
498:
499: if (lu_flag) {
500: /* Initialize node to unlock the disk queue. */
501: rf_InitNode(&unlockDataNodes[i], rf_wait, RF_FALSE,
502: rf_DiskUnlockFunc, rf_DiskUnlockUndoFunc,
503: rf_GenericWakeupFunc, 1, 1, 2, 0, dag_h,
504: "Und", allocList);
505: /* Physical disk addr desc. */
506: unlockDataNodes[i].params[0].p = pda;
507: unlockDataNodes[i].params[1].v =
508: RF_CREATE_PARAM3(RF_IO_NORMAL_PRIORITY, 0,
509: lu_flag, which_ru);
510: }
511: pda = pda->next;
512: }
513:
514:
515: /* Initialize nodes which compute new parity. */
516: /*
517: * We use the simple XOR func in the double-XOR case, and when we're
518: * accessing only a portion of one stripe unit. The distinction
519: * between the two is that the regular XOR func assumes that the
520: * targbuf is a full SU in size, and examines the pda associated with
521: * the buffer to decide where within the buffer to XOR the data,
522: * whereas the simple XOR func just XORs the data into the start of
523: * the buffer.
524: */
525: if ((numParityNodes == 2) || ((numDataNodes == 1) &&
526: (asmap->totalSectorsAccessed <
527: raidPtr->Layout.sectorsPerStripeUnit))) {
528: func = pfuncs->simple;
529: undoFunc = rf_NullNodeUndoFunc;
530: name = pfuncs->SimpleName;
531: if (qfuncs) {
532: qfunc = qfuncs->simple;
533: qname = qfuncs->SimpleName;
534: }
535: } else {
536: func = pfuncs->regular;
537: undoFunc = rf_NullNodeUndoFunc;
538: name = pfuncs->RegularName;
539: if (qfuncs) {
540: qfunc = qfuncs->regular;
541: qname = qfuncs->RegularName;
542: }
543: }
544: /*
545: * Initialize the xor nodes: params are {pda,buf} from {Rod,Wnd,Rop}
546: * nodes, and raidPtr.
547: */
548: if (numParityNodes == 2) { /* Double-XOR case. */
549: for (i = 0; i < numParityNodes; i++) {
550: rf_InitNode(&xorNodes[i], rf_wait, RF_TRUE, func,
551: undoFunc, NULL, 1, nNodes, 7, 1, dag_h, name,
552: allocList); /* No wakeup func for XOR. */
553: xorNodes[i].flags |= RF_DAGNODE_FLAG_YIELD;
554: xorNodes[i].params[0] = readDataNodes[i].params[0];
555: xorNodes[i].params[1] = readDataNodes[i].params[1];
556: xorNodes[i].params[2] = readParityNodes[i].params[0];
557: xorNodes[i].params[3] = readParityNodes[i].params[1];
558: xorNodes[i].params[4] = writeDataNodes[i].params[0];
559: xorNodes[i].params[5] = writeDataNodes[i].params[1];
560: xorNodes[i].params[6].p = raidPtr;
561: /* Use old parity buf as target buf. */
562: xorNodes[i].results[0] = readParityNodes[i].params[1].p;
563: }
564: } else {
565: /* There is only one xor node in this case. */
566: rf_InitNode(&xorNodes[0], rf_wait, RF_TRUE, func, undoFunc,
567: NULL, 1, nNodes,
568: (2 * (numDataNodes + numDataNodes + 1) + 1), 1,
569: dag_h, name, allocList);
570: xorNodes[0].flags |= RF_DAGNODE_FLAG_YIELD;
571: for (i = 0; i < numDataNodes + 1; i++) {
572: /* Set up params related to Rod and Rop nodes. */
573: xorNodes[0].params[2 * i + 0] =
574: readDataNodes[i].params[0]; /* pda */
575: xorNodes[0].params[2 * i + 1] =
576: readDataNodes[i].params[1]; /* Buffer pointer */
577: }
578: for (i = 0; i < numDataNodes; i++) {
579: /* Set up params related to Wnd and Wnp nodes. */
580: xorNodes[0].params[2 * (numDataNodes + 1 + i) + 0] =
581: writeDataNodes[i].params[0]; /* pda */
582: xorNodes[0].params[2 * (numDataNodes + 1 + i) + 1] =
583: writeDataNodes[i].params[1]; /* Buffer pointer */
584: }
585: xorNodes[0].params[2 * (numDataNodes + numDataNodes + 1)].p =
586: raidPtr; /* Xor node needs to get at RAID information. */
587: xorNodes[0].results[0] = readParityNodes[0].params[1].p;
588: }
589:
590: /* Initialize the log node(s). */
591: pda = asmap->parityInfo;
592: for (i = 0; i < numParityNodes; i++) {
593: RF_ASSERT(pda);
594: rf_InitNode(&lpuNodes[i], rf_wait, RF_FALSE,
595: rf_ParityLogUpdateFunc, rf_ParityLogUpdateUndoFunc,
596: rf_GenericWakeupFunc, 1, 1, 2, 0, dag_h, "Lpu", allocList);
597: lpuNodes[i].params[0].p = pda; /* PhysDiskAddr of parity. */
598: /* Buffer pointer to parity. */
599: lpuNodes[i].params[1].p = xorNodes[i].results[0];
600: pda = pda->next;
601: }
602:
603:
604: /* Step 4. Connect the nodes. */
605:
606: /* Connect header to block node. */
607: RF_ASSERT(dag_h->numSuccedents == 1);
608: RF_ASSERT(blockNode->numAntecedents == 0);
609: dag_h->succedents[0] = blockNode;
610:
611: /* Connect block node to read old data nodes. */
612: RF_ASSERT(blockNode->numSuccedents == (numDataNodes + numParityNodes));
613: for (i = 0; i < numDataNodes; i++) {
614: blockNode->succedents[i] = &readDataNodes[i];
615: RF_ASSERT(readDataNodes[i].numAntecedents == 1);
616: readDataNodes[i].antecedents[0] = blockNode;
617: readDataNodes[i].antType[0] = rf_control;
618: }
619:
620: /* Connect block node to read old parity nodes. */
621: for (i = 0; i < numParityNodes; i++) {
622: blockNode->succedents[numDataNodes + i] = &readParityNodes[i];
623: RF_ASSERT(readParityNodes[i].numAntecedents == 1);
624: readParityNodes[i].antecedents[0] = blockNode;
625: readParityNodes[i].antType[0] = rf_control;
626: }
627:
628: /* Connect read old data nodes to write new data nodes. */
629: for (i = 0; i < numDataNodes; i++) {
630: RF_ASSERT(readDataNodes[i].numSuccedents ==
631: numDataNodes + numParityNodes);
632: for (j = 0; j < numDataNodes; j++) {
633: RF_ASSERT(writeDataNodes[j].numAntecedents ==
634: numDataNodes + numParityNodes);
635: readDataNodes[i].succedents[j] = &writeDataNodes[j];
636: writeDataNodes[j].antecedents[i] = &readDataNodes[i];
637: if (i == j)
638: writeDataNodes[j].antType[i] = rf_antiData;
639: else
640: writeDataNodes[j].antType[i] = rf_control;
641: }
642: }
643:
644: /* Connect read old data nodes to xor nodes. */
645: for (i = 0; i < numDataNodes; i++)
646: for (j = 0; j < numParityNodes; j++) {
647: RF_ASSERT(xorNodes[j].numAntecedents ==
648: numDataNodes + numParityNodes);
649: readDataNodes[i].succedents[numDataNodes + j] =
650: &xorNodes[j];
651: xorNodes[j].antecedents[i] = &readDataNodes[i];
652: xorNodes[j].antType[i] = rf_trueData;
653: }
654:
655: /* Connect read old parity nodes to write new data nodes. */
656: for (i = 0; i < numParityNodes; i++) {
657: RF_ASSERT(readParityNodes[i].numSuccedents ==
658: numDataNodes + numParityNodes);
659: for (j = 0; j < numDataNodes; j++) {
660: readParityNodes[i].succedents[j] = &writeDataNodes[j];
661: writeDataNodes[j].antecedents[numDataNodes + i] =
662: &readParityNodes[i];
663: writeDataNodes[j].antType[numDataNodes + i] =
664: rf_control;
665: }
666: }
667:
668: /* Connect read old parity nodes to xor nodes. */
669: for (i = 0; i < numParityNodes; i++)
670: for (j = 0; j < numParityNodes; j++) {
671: readParityNodes[i].succedents[numDataNodes + j] =
672: &xorNodes[j];
673: xorNodes[j].antecedents[numDataNodes + i] =
674: &readParityNodes[i];
675: xorNodes[j].antType[numDataNodes + i] = rf_trueData;
676: }
677:
678: /* Connect xor nodes to write new parity nodes. */
679: for (i = 0; i < numParityNodes; i++) {
680: RF_ASSERT(xorNodes[i].numSuccedents == 1);
681: RF_ASSERT(lpuNodes[i].numAntecedents == 1);
682: xorNodes[i].succedents[0] = &lpuNodes[i];
683: lpuNodes[i].antecedents[0] = &xorNodes[i];
684: lpuNodes[i].antType[0] = rf_trueData;
685: }
686:
687: for (i = 0; i < numDataNodes; i++) {
688: if (lu_flag) {
689: /* Connect write new data nodes to unlock nodes. */
690: RF_ASSERT(writeDataNodes[i].numSuccedents == 1);
691: RF_ASSERT(unlockDataNodes[i].numAntecedents == 1);
692: writeDataNodes[i].succedents[0] = &unlockDataNodes[i];
693: unlockDataNodes[i].antecedents[0] = &writeDataNodes[i];
694: unlockDataNodes[i].antType[0] = rf_control;
695:
696: /* Connect unlock nodes to unblock node. */
697: RF_ASSERT(unlockDataNodes[i].numSuccedents == 1);
698: RF_ASSERT(unblockNode->numAntecedents ==
699: (numDataNodes + (nfaults * numParityNodes)));
700: unlockDataNodes[i].succedents[0] = unblockNode;
701: unblockNode->antecedents[i] = &unlockDataNodes[i];
702: unblockNode->antType[i] = rf_control;
703: } else {
704: /* Connect write new data nodes to unblock node. */
705: RF_ASSERT(writeDataNodes[i].numSuccedents == 1);
706: RF_ASSERT(unblockNode->numAntecedents ==
707: (numDataNodes + (nfaults * numParityNodes)));
708: writeDataNodes[i].succedents[0] = unblockNode;
709: unblockNode->antecedents[i] = &writeDataNodes[i];
710: unblockNode->antType[i] = rf_control;
711: }
712: }
713:
714: /* Connect write new parity nodes to unblock node. */
715: for (i = 0; i < numParityNodes; i++) {
716: RF_ASSERT(lpuNodes[i].numSuccedents == 1);
717: lpuNodes[i].succedents[0] = unblockNode;
718: unblockNode->antecedents[numDataNodes + i] = &lpuNodes[i];
719: unblockNode->antType[numDataNodes + i] = rf_control;
720: }
721:
722: /* Connect unblock node to terminator. */
723: RF_ASSERT(unblockNode->numSuccedents == 1);
724: RF_ASSERT(termNode->numAntecedents == 1);
725: RF_ASSERT(termNode->numSuccedents == 0);
726: unblockNode->succedents[0] = termNode;
727: termNode->antecedents[0] = unblockNode;
728: termNode->antType[0] = rf_control;
729: }
730:
731:
732: void
733: rf_CreateParityLoggingSmallWriteDAG(RF_Raid_t *raidPtr,
734: RF_AccessStripeMap_t *asmap, RF_DagHeader_t *dag_h, void *bp,
735: RF_RaidAccessFlags_t flags, RF_AllocListElem_t *allocList,
736: RF_RedFuncs_t *pfuncs, RF_RedFuncs_t *qfuncs)
737: {
738: dag_h->creator = "ParityLoggingSmallWriteDAG";
739: rf_CommonCreateParityLoggingSmallWriteDAG(raidPtr, asmap, dag_h, bp,
740: flags, allocList, &rf_xorFuncs, NULL);
741: }
742:
743:
744: void
745: rf_CreateParityLoggingLargeWriteDAG(RF_Raid_t *raidPtr,
746: RF_AccessStripeMap_t *asmap, RF_DagHeader_t *dag_h, void *bp,
747: RF_RaidAccessFlags_t flags, RF_AllocListElem_t *allocList, int nfaults,
748: int (*redFunc) (RF_DagNode_t *))
749: {
750: dag_h->creator = "ParityLoggingSmallWriteDAG";
751: rf_CommonCreateParityLoggingLargeWriteDAG(raidPtr, asmap, dag_h, bp,
752: flags, allocList, 1, rf_RegularXorFunc);
753: }
754: #endif /* RF_INCLUDE_PARITYLOGGING > 0 */
CVSweb