Work in progress global dedupe changes.

pcompress.h

@@ -186,7 +186,7 @@ struct cmp_data {
 	uint64_t chunksize;
 	uint64_t len_cmp, len_cmp_be;
 	uchar_t checksum[CKSUM_MAX_BYTES];
-	int level, cksum_mt;
+	int level, cksum_mt, out_fd;
 	unsigned int id;
 	compress_func_ptr compress;
 	compress_func_ptr decompress;

rabin/global/db.c

@@ -38,11 +38,6 @@
 
 #include "db.h"
 
-#define	ONE_PB (1125899906842624ULL)
-#define	ONE_TB (1099511627776ULL)
-#define	FOUR_MB (4194304ULL)
-#define	EIGHT_MB (8388608ULL)
-
 /*
  * Hashtable structures for in-memory index.
  */
@@ -103,16 +98,16 @@ static cleanup_indx(index_t *indx)
 }
 
 archive_config_t *
-init_global_db_s(char *path, char *tmppath, uint32_t chunksize, int pct_interval,
+init_global_db_s(char *path, char *tmppath, uint32_t chunksize, uint64_t user_chunk_sz,
-		 compress_algo_t algo, cksum_t ck, cksum_t ck_sim, size_t file_sz,
+		 int pct_interval, compress_algo_t algo, cksum_t ck, cksum_t ck_sim,
-		 size_t memlimit, int nthreads)
+		 size_t file_sz, size_t memlimit, int nthreads)
 {
 	archive_config_t *cfg;
 	int rv;
 	float diff;
 
 	cfg = calloc(1, sizeof (archive_config_t));
-	rv = set_config_s(cfg, algo, ck, ck_sim, chunksize, file_sz, pct_interval);
+	rv = set_config_s(cfg, algo, ck, ck_sim, chunksize, file_sz, user_chunk_sz, pct_interval);
 
 	if (path != NULL) {
 		printf("Disk based index not yet implemented.\n");
@@ -125,7 +120,10 @@ init_global_db_s(char *path, char *tmppath, uint32_t chunksize, int pct_interval
 		index_t *indx;
 
 		// Compute total hashtable entries first
-		intervals = 100 / pct_interval - 1;
+		if (pct_interval == 0)
+			intervals = 1;
+		else
+			intervals = 100 / pct_interval - 1;
 		hash_slots = file_sz / cfg->segment_sz_bytes + 1;
 		hash_slots *= intervals;
 		hash_entry_size = sizeof (hash_entry_t) + cfg->similarity_cksum_sz - 1;
@@ -249,6 +247,7 @@ db_lookup_insert_s(archive_config_t *cfg, uchar_t *sim_cksum, int interval,
 			htab[htab_entry] = htab[htab_entry]->next;
 		} else {
 			ent = (hash_entry_t *)malloc(indx->hash_entry_size);
+			indx->memused += indx->hash_entry_size;
 		}
 		ent->seg_offset = seg_offset;
 		ent->next = 0;

rabin/global/db.h

@@ -33,8 +33,9 @@ extern "C" {
 
 archive_config_t *init_global_db(char *configfile);
 archive_config_t *init_global_db_s(char *path, char *tmppath, uint32_t chunksize,
-			int pct_interval, compress_algo_t algo, cksum_t ck,
+			uint64_t user_chunk_sz, int pct_interval, compress_algo_t algo,
-			cksum_t ck_sim, size_t file_sz, size_t memlimit, int nthreads);
+			cksum_t ck, cksum_t ck_sim, size_t file_sz, size_t memlimit,
+			int nthreads);
 uint64_t db_lookup_insert_s(archive_config_t *cfg, uchar_t *sim_cksum, int interval,
 		   uint64_t seg_offset, int do_insert);
 

rabin/global/dedupe_config.c

@@ -42,6 +42,7 @@
 #define	ONE_TB (1099511627776ULL)
 #define	FOUR_MB (4194304ULL)
 #define	EIGHT_MB (8388608ULL)
+#define	EIGHT_GB (8589934592ULL)
 
 static compress_algo_t
 get_compress_level(compress_algo_t algo)
@@ -287,6 +288,7 @@ read_config(char *configfile, archive_config_t *cfg)
 	 */
 	cfg->chunk_sz_bytes = RAB_BLK_AVG_SZ(cfg->chunk_sz);
 	cfg->directory_levels = 2;
+
 	if (cfg->archive_sz < ONE_TB) {
 		segment_sz_bytes = FOUR_MB;
 		cfg->directory_fanout = 128;
@@ -349,8 +351,9 @@ write_config(char *configfile, archive_config_t *cfg)
 
 int
 set_config_s(archive_config_t *cfg, compress_algo_t algo, cksum_t ck, cksum_t ck_sim,
-	     uint32_t chunksize, size_t file_sz, int pct_interval)
+	     uint32_t chunksize, size_t file_sz, uint64_t user_chunk_sz, int pct_interval)
 {
+
 	cfg->algo = algo;
 	cfg->chunk_cksum_type = ck;
 	cfg->similarity_cksum = ck_sim;
@@ -360,9 +363,14 @@ set_config_s(archive_config_t *cfg, compress_algo_t algo, cksum_t ck, cksum_t ck
 	cfg->chunk_sz = chunksize;
 	cfg->chunk_sz_bytes = RAB_BLK_AVG_SZ(cfg->chunk_sz);
 	cfg->pct_interval = pct_interval;
-
 	cfg->archive_sz = file_sz;
-	if (cfg->archive_sz < ONE_TB) {
+	cfg->dedupe_mode = MODE_SIMILARITY;
+
+	if (cfg->archive_sz <= EIGHT_GB) {
+		cfg->dedupe_mode = MODE_SIMPLE;
+		cfg->segment_sz_bytes = user_chunk_sz;
+
+	} else if (cfg->archive_sz < ONE_TB) {
 		cfg->segment_sz_bytes = FOUR_MB;
 
 	} else {

rabin/global/dedupe_config.h

@@ -44,6 +44,12 @@ extern "C" {
 // 8GB
 #define	MIN_ARCHIVE_SZ (8589934592ULL)
 
+typedef enum {
+	MODE_SIMPLE = 0,
+	MODE_SIMILARITY,
+	MODE_ARCHIVE
+} dedupe_mode_t;
+
 typedef struct {
 	char rootdir[PATH_MAX+1];
 	uint32_t chunk_sz; // Numeric ID: 1 - 4k ... 5 - 64k
@@ -58,9 +64,10 @@ typedef struct {
 	int pct_interval; // Similarity based match intervals in %age.
 			// The items below are computed given the above
 			// components.
+	dedupe_mode_t dedupe_mode;
 
 	uint32_t chunk_sz_bytes; // Average chunk size
-	uint32_t segment_sz_bytes; // Segment size in bytes
+	uint64_t segment_sz_bytes; // Segment size in bytes
 	uint32_t segment_sz; // Number of chunks in one segment
 	uint32_t container_sz; // Number of segments
 	int directory_fanout; // Number of subdirectories in a directory
@@ -80,7 +87,8 @@ typedef struct _segment_entry {
 int read_config(char *configfile, archive_config_t *cfg);
 int write_config(char *configfile, archive_config_t *cfg);
 int set_config_s(archive_config_t *cfg, compress_algo_t algo, cksum_t ck, cksum_t ck_sim,
-		      uint32_t chunksize, size_t file_sz, int pct_interval);
+		uint32_t chunksize, size_t file_sz, uint64_t user_chunk_sz,
+		int pct_interval);
 
 #ifdef	__cplusplus
 }