Add global index cleanup function.

Fix location of sem_wait().
More comments.

main.c

@@ -551,13 +551,6 @@ redo:
 		memcpy(ubuf, cmpbuf, dedupe_index_sz);
 
 	} else {
-		/*
-		 * This chunk was not deduplicated, however we still need to down the
-		 * semaphore in order to maintain proper thread coordination.
-		 */
-		if (enable_rabin_global) {
-			sem_wait(tdat->rctx->index_sem);
-		}
 		if (HDR & COMPRESSED) {
 			if (HDR & CHUNK_FLAG_PREPROC) {
 				rv = preproc_decompress(tdat->decompress, cseg, tdat->len_cmp,
@@ -606,6 +599,16 @@ redo:
 		tdat->uncompressed_chunk = tdat->compressed_chunk;
 		tdat->compressed_chunk = tmp;
 		tdat->cmp_seg = tdat->uncompressed_chunk;
+	} else {
+		/*
+		 * This chunk was not deduplicated, however we still need to down the
+		 * semaphore in order to maintain proper thread coordination. We do this after
+		 * decompression to achieve better concurrency. Decompression does not need
+		 * to wait for the previous thread's dedupe recovery to complete.
+		 */
+		if (enable_rabin_global) {
+			sem_wait(tdat->rctx->index_sem);
+		}
 	}
 
 	if (!encrypt_type) {
@@ -2575,7 +2578,7 @@ main(int argc, char *argv[])
 	}
 
 	if (enable_rabin_global && enable_delta_encode) {
-		fprintf(stderr, "Global Deduplication does not support Delta Compression yet.\n");
+		fprintf(stderr, "Global Deduplication does not support Delta Compression.\n");
 		exit(1);
 	}
 

rabin/global/db.c

@@ -102,6 +102,7 @@ init_global_db_s(char *path, char *tmppath, uint32_t chunksize, uint64_t user_ch
 
 	if (cfg->dedupe_mode == MODE_SIMPLE) {
 		pct_interval = 0;
+		cfg->pct_interval = 0;
 	}
 
 	if (path != NULL) {
@@ -153,6 +154,7 @@ init_global_db_s(char *path, char *tmppath, uint32_t chunksize, uint64_t user_ch
 		indx->hash_entry_size = hash_entry_size;
 		indx->intervals = intervals;
 		indx->hash_slots = hash_slots / intervals;
+		cfg->nthreads = nthreads;
 
 		for (i = 0; i < intervals; i++) {
 			indx->list[i].tab = (hash_entry_t **)calloc(hash_slots / intervals,
@@ -264,3 +266,21 @@ db_lookup_insert_s(archive_config_t *cfg, uchar_t *sim_cksum, int interval,
 	}
 	return (NULL);
 }
+
+void
+destroy_global_db_s(archive_config_t *cfg)
+{
+	int i;
+	index_t *indx = (index_t *)(cfg->dbdata);
+
+	cleanup_indx(indx);
+	if (cfg->pct_interval > 0) {
+		for (i = 0; i < cfg->nthreads; i++) {
+			close(cfg->seg_fd_r[i]);
+		}
+		free(cfg->seg_fd_r);
+		close(cfg->seg_fd_w);
+		unlink(cfg->rootdir);
+	}
+}
+

rabin/global/db.h

@@ -48,6 +48,7 @@ archive_config_t *init_global_db_s(char *path, char *tmppath, uint32_t chunksize
 			int nthreads);
 hash_entry_t *db_lookup_insert_s(archive_config_t *cfg, uchar_t *sim_cksum, int interval,
 		   uint64_t item_offset, uint32_t item_size, int do_insert);
+void destroy_global_db_s(archive_config_t *cfg);
 
 #ifdef	__cplusplus
 }

rabin/global/dedupe_config.h

@@ -73,8 +73,10 @@ typedef struct {
 	int directory_fanout; // Number of subdirectories in a directory
 	int directory_levels; // Levels of nested directories
 	int num_containers; // Number of containers in a directory
+	int nthreads; // Number of threads processing data segments in parallel
 	int seg_fd_w; 
-	int *seg_fd_r;
+	int *seg_fd_r; // One read-only fd per thread for mapping in portions of the
+		       // segment metadata cache.
 	void *dbdata;
 } archive_config_t;
 

rabin/rabin_dedup.c

@@ -320,6 +320,16 @@ destroy_dedupe_context(dedupe_context_t *ctx)
 #ifndef SSE_MODE
 		if (ctx->current_window_data) slab_free(NULL, ctx->current_window_data);
 #endif
+
+		pthread_mutex_lock(&init_lock);
+		if (arc) {
+			if (arc->dedupe_mode == MODE_SIMPLE) {
+				destroy_global_db_s(arc);
+			}
+		}
+		arc = NULL;
+		pthread_mutex_unlock(&init_lock);
+
 		if (ctx->blocks) {
 			for (i=0; i<ctx->blknum && ctx->blocks[i] != NULL; i++) {
 				slab_free(NULL, ctx->blocks[i]);
@@ -767,6 +777,10 @@ process_blocks:
 			goto dedupe_done;
 		}
 
+		/*
+		 * Subsequent processing below is for per-segment Deduplication.
+		 */
+
 		/*
 		 * Compute hash signature for each block. We do this in a separate loop to 
 		 * have a fast linear scan through the buffer.
@@ -1120,6 +1134,11 @@ dedupe_decompress(dedupe_context_t *ctx, uchar_t *buf, uint64_t *size)
 		return;
 	}
 
+	/*
+	 * Handling for per-segment Deduplication.
+	 * First pass re-create the rabin block array from the index metadata.
+	 * Second pass copy over blocks to the target buffer to re-create the original segment.
+	 */
 	slab_cache_add(sizeof (rabin_blockentry_t));
 	for (blk = 0; blk < blknum; blk++) {
 		if (ctx->blocks[blk] == 0)