dm-vdo/indexer/index.c

26  * The index send two kinds of messages to coordinate between zones: chapter close messages for the
29  * The chapter writer is responsible for committing chapters of records to storage. Since zones can
30  * get different numbers of records, some zones may fall behind others. Each time a zone fills up
32  * and also informs all other zones that it has closed the chapter. Each other zone will then close
33  * the chapter immediately, regardless of how full it is, in order to minimize skew between zones.
37  * Although that chapter is invalid for zones that have moved on, the existence of the open chapter
38  * means that those zones will never ask the volume index about it. No zone is allowed to get more
40  * chapter before the previous one has been closed by all zones, it is forced to wait.
42  * The sparse cache relies on having the same set of chapter indexes available to all zones. When a
68 	/* The number of zones which have submitted a chapter for writing */
134 	zone = index->zones[request->zone_number];
214  * writing until all zones have closed it.
308 	struct index_zone *zone = request->index->zones[request->zone_number];
582 	struct index_zone *zone = index->zones[request->zone_number];
687 				 * We've been told to stop, and all of the zones are in the same
1149 	index->zones[zone_number] = zone;
1179 	result = vdo_allocate(index->zone_count, struct index_zone *, "zones",
1180 			      &index->zones);
1253 		zone = index->zones[z];
1289 	if (index->zones != NULL) {
1291 			free_index_zone(index->zones[i]);
1292 		vdo_free(index->zones);