merge tokudb revision 7776-7883 to toudb.1032b. closes #1260

git-svn-id: file:///svn/toku/tokudb.1032b@7888 c7de825b-a66e-492c-adef-691d508d4ae1

cxx/tests/db_load.cpp

@@ -2,7 +2,7 @@
 #include <assert.h>
 #include <db_cxx.h>
 
-static void hexdump(Dbt *d) {
+static inline void hexdump(Dbt *d) {
     unsigned char *cp = (unsigned char *) d->get_data();
     int n = d->get_size();
     printf(" ");

newbrt/brt-internal.h

@@ -11,10 +11,9 @@
 #include "list.h"
 #include "mempool.h"
 #include "kv-pair.h"
-#include "leafentry.h"
-
 typedef void *OMTVALUE;
 #include "omt.h"
+#include "leafentry.h"
 
 #ifndef BRT_FANOUT
 #define BRT_FANOUT 16
@@ -263,7 +262,13 @@ int toku_cmd_leafval_heaviside (OMTVALUE leafentry, void *extra);
 int toku_brt_root_put_cmd(BRT brt, BRT_CMD cmd, TOKULOGGER logger);
 int toku_cachefile_root_put_cmd (CACHEFILE cf, BRT_CMD cmd, TOKULOGGER logger);
 
-void *mempool_malloc_from_omt(OMT omt, struct mempool *mp, size_t size);
+void *mempool_malloc_from_omt(OMT omt, struct mempool *mp, size_t size, void **maybe_free);
+// Effect: Allocate a new object of size SIZE in MP.  If MP runs out of space, allocate new a new mempool space, and copy all the items
+//  from the OMT (which items refer to items in the old mempool) into the new mempool.
+//  If MAYBE_FREE is NULL then free the old mempool's space.
+//  Otherwise, store the old mempool's space in maybe_free.
+
+void mempool_release(struct mempool *); // release anything that was not released when the ..._norelease function was called.
 
 void toku_verify_all_in_mempool(BRTNODE node);
 
@@ -281,6 +286,7 @@ enum brt_layout_version_e {
 
 void toku_brtheader_free (struct brt_header *h);
 int toku_brtheader_close (CACHEFILE cachefile, void *header_v);
+int toku_brtheader_checkpoint (CACHEFILE cachefile, void *header_v);
 
 #define BLOCK_ALLOCATOR_ALIGNMENT 4096
 // How much must be reserved at the beginning for the block?

newbrt/brt-serialize.c

@@ -271,7 +271,7 @@ void toku_serialize_brtnode_to (int fd, BLOCKNUM blocknum, BRTNODE node, struct
 	printf("%s:%d w.done=%u calculated_size=%u\n", __FILE__, __LINE__, w.ndone, calculated_size);
     assert(calculated_size==w.ndone);
 
-    // The uncompressed part of the header is
+    // The uncompressed part of the block header is
     //   tokuleaf(8),
     //   version(4),
     //   lsn(8),

newbrt/brt-test-helpers.c

@@ -71,12 +71,8 @@ int toku_testsetup_insert_to_leaf (BRT brt, BLOCKNUM blocknum, char *key, int ke
     assert(node->height==0);
 
     u_int32_t lesize, disksize;
-    LEAFENTRY tmp_leafentry;
-    r = le_committed(keylen, key, vallen, val, &lesize, &disksize, &tmp_leafentry);
-
-    LEAFENTRY leafentry = mempool_malloc_from_omt(node->u.l.buffer, &node->u.l.buffer_mempool, lesize);
-    memcpy(leafentry, tmp_leafentry, lesize);
-    toku_free(tmp_leafentry);
+    LEAFENTRY leafentry;
+    r = le_committed(keylen, key, vallen, val, &lesize, &disksize, &leafentry, node->u.l.buffer, &node->u.l.buffer_mempool, 0);
 
     OMTVALUE storeddatav;
     u_int32_t idx;

newbrt/cachetable.c

@@ -125,6 +125,7 @@ struct cachefile {
 
     void *userdata;
     int (*close_userdata)(CACHEFILE cf, void *userdata); // when closing the last reference to a cachefile, first call this function.
+    int (*checkpoint_userdata)(CACHEFILE cf, void *userdata); // when checkpointing a cachefile, call this function.
 };
 
 int toku_create_cachetable(CACHETABLE *result, long size_limit, LSN initial_lsn, TOKULOGGER logger) {
@@ -229,6 +230,7 @@ int toku_cachetable_openfd (CACHEFILE *cf, CACHETABLE t, int fd, const char *fna
 
 	newcf->userdata = 0;
 	newcf->close_userdata = 0;
+	newcf->checkpoint_userdata = 0;
 
 	*cf = newcf;
 	return 0;
@@ -252,6 +254,7 @@ int toku_cachefile_set_fd (CACHEFILE cf, int fd, const char *fname) {
         return r;
     }
     cf->close_userdata = NULL;
+    cf->checkpoint_userdata = NULL;
     cf->userdata = NULL;
 
     close(cf->fd);
@@ -313,6 +316,7 @@ int toku_cachefile_close (CACHEFILE *cfp, TOKULOGGER logger) {
 	    return r;
 	}
 	cf->close_userdata = NULL;
+	cf->checkpoint_userdata = NULL;
 	cf->userdata = NULL;
         cf->cachetable->cachefiles = remove_cf_from_list(cf, cf->cachetable->cachefiles);
         cachetable_unlock(ct);
@@ -1081,6 +1085,16 @@ int toku_cachetable_checkpoint (CACHETABLE ct) {
             cachetable_complete_write_pair(ct, p, FALSE);
         }
 
+	{
+	    CACHEFILE cf;
+	    for (cf = ct->cachefiles; cf; cf=cf->next) {
+		if (cf->checkpoint_userdata) {
+		    int r = cf->checkpoint_userdata(cf, cf->userdata);
+		    assert(r==0);
+		}
+	    }
+	}
+
         ct->checkpointing = 0; // clear the checkpoint in progress flag
     }
 
@@ -1219,9 +1233,15 @@ int toku_cachetable_get_key_state (CACHETABLE ct, CACHEKEY key, CACHEFILE cf, vo
     return r;
 }
 
-void toku_cachefile_set_userdata (CACHEFILE cf, void *userdata, int (*close_userdata)(CACHEFILE, void*)) {
+void
+toku_cachefile_set_userdata (CACHEFILE cf,
+			     void *userdata,
+			     int (*close_userdata)(CACHEFILE, void*),
+			     int (*checkpoint_userdata)(CACHEFILE, void*))
+{
     cf->userdata = userdata;
     cf->close_userdata = close_userdata;
+    cf->checkpoint_userdata = checkpoint_userdata;
 }
 
 void *toku_cachefile_get_userdata(CACHEFILE cf) {

newbrt/cachetable.h

@@ -61,8 +61,9 @@ typedef void (*CACHETABLE_FLUSH_CALLBACK)(CACHEFILE, CACHEKEY key, void *value,
 // associated with the key are returned.
 typedef int (*CACHETABLE_FETCH_CALLBACK)(CACHEFILE, CACHEKEY key, u_int32_t fullhash, void **value, long *sizep, void *extraargs, LSN *written_lsn);
 
-void toku_cachefile_set_userdata(CACHEFILE cf, void *userdata, int (*close_userdata)(CACHEFILE, void*));
-// Effect: Store some cachefile-specific user data.  When the last reference to a cachefile is closed, we call close_userdata.
+void toku_cachefile_set_userdata(CACHEFILE cf, void *userdata, int (*close_userdata)(CACHEFILE, void*), int (*checkpoint_userdata)(CACHEFILE, void*));
+// Effect: Store some cachefile-specific user data.  When the last reference to a cachefile is closed, we call close_userdata().
+// When the cachefile needs to be checkpointed, we call checkpoint_userdata().
 // If userdata is already non-NULL, then we simply overwrite it.
 void *toku_cachefile_get_userdata(CACHEFILE);
 // Effect: Get the user dataa.

newbrt/leafentry.c

@@ -6,9 +6,20 @@ u_int32_t toku_le_crc(LEAFENTRY v) {
     return x1764_memory(v, leafentry_memsize(v));
 }
 
-int le_committed (u_int32_t klen, void* kval, u_int32_t dlen, void* dval, u_int32_t *resultsize, u_int32_t *disksize, LEAFENTRY *result) {
+static void *
+le_malloc(OMT omt, struct mempool *mp, size_t size, void **maybe_free)
+{
+    if (omt)
+	return mempool_malloc_from_omt(omt, mp, size, maybe_free);
+    else
+	return toku_malloc(size);
+}
+
+int
+le_committed (u_int32_t klen, void* kval, u_int32_t dlen, void* dval, u_int32_t *resultsize, u_int32_t *disksize, LEAFENTRY *result,
+	      OMT omt, struct mempool *mp, void **maybe_free) {
     size_t size = 9+klen+dlen;
-    unsigned char *lec=toku_malloc(size);
+    unsigned char *lec=le_malloc(omt, mp, size, maybe_free);
     assert(lec);
     lec[0] = LE_COMMITTED;
     putint(lec+1, klen);
@@ -20,10 +31,12 @@ int le_committed (u_int32_t klen, void* kval, u_int32_t dlen, void* dval, u_int3
     *result=(LEAFENTRY)lec;
     return 0;
 }
+
 int le_both (TXNID xid, u_int32_t klen, void* kval, u_int32_t clen, void* cval, u_int32_t plen, void* pval,
-	     u_int32_t *resultsize, u_int32_t *disksize, LEAFENTRY *result) {
+	     u_int32_t *resultsize, u_int32_t *disksize, LEAFENTRY *result,
+	     OMT omt, struct mempool *mp, void **maybe_free) {
     size_t size = 1+8+4*3+klen+clen+plen;
-    unsigned char *lec=toku_malloc(size);
+    unsigned char *lec=le_malloc(omt, mp, size, maybe_free);
     assert(lec);
     lec[0] = LE_BOTH;
     putint64(lec+1,          xid);
@@ -39,10 +52,13 @@ int le_both (TXNID xid, u_int32_t klen, void* kval, u_int32_t clen, void* cval,
     return 0;
     
 }
-int le_provdel  (TXNID xid, u_int32_t klen, void* kval, u_int32_t dlen, void* dval, 
-		 u_int32_t *memsize, u_int32_t *disksize, LEAFENTRY *result) {
+
+int
+le_provdel (TXNID xid, u_int32_t klen, void* kval, u_int32_t dlen, void* dval,
+	    u_int32_t *memsize, u_int32_t *disksize, LEAFENTRY *result,
+	    OMT omt, struct mempool *mp, void **maybe_free) {
     size_t size = 1 + 8 + 2*4 + klen + dlen;
-    unsigned char *lec=toku_malloc(size);
+    unsigned char *lec= le_malloc(omt, mp, size, maybe_free);
     assert(lec);
     lec[0] = LE_PROVDEL;
     putint64(lec+1,          xid);
@@ -55,9 +71,12 @@ int le_provdel  (TXNID xid, u_int32_t klen, void* kval, u_int32_t dlen, void* dv
     *result=(LEAFENTRY)lec;
     return 0;
 }
-int le_provpair (TXNID xid, u_int32_t klen, void* kval, u_int32_t plen, void* pval, u_int32_t *memsize, u_int32_t *disksize, LEAFENTRY *result) {
+
+int
+le_provpair (TXNID xid, u_int32_t klen, void* kval, u_int32_t plen, void* pval, u_int32_t *memsize, u_int32_t *disksize, LEAFENTRY *result,
+	     OMT omt, struct mempool *mp, void **maybe_free) {
     size_t size = 1 + 8 + 2*4 + klen + plen;
-    unsigned char *lec=toku_malloc(size);
+    unsigned char *lec= le_malloc(omt, mp, size, maybe_free);
     assert(lec);
     lec[0] = LE_PROVPAIR;
     putint64(lec+1,          xid);
@@ -154,7 +173,8 @@ int toku_fread_LEAFENTRY(FILE *f, LEAFENTRY *le, struct x1764 *checksum, u_int32
 	r = toku_fread_BYTESTRING(f, &a, checksum, len);  if (r!=0) return r;
 	r = toku_fread_BYTESTRING(f, &b, checksum, len);  if (r!=0) return r;
 	r = le_committed(a.len, a.data, b.len, b.data,
-			 &memsize, &disksize, le);
+			 &memsize, &disksize, le,
+			 0, 0, 0);
 	toku_free_BYTESTRING(a);
 	toku_free_BYTESTRING(b);
 	return r;
@@ -164,7 +184,8 @@ int toku_fread_LEAFENTRY(FILE *f, LEAFENTRY *le, struct x1764 *checksum, u_int32
 	r = toku_fread_BYTESTRING(f, &b, checksum, len);  if (r!=0) return r;
 	r = toku_fread_BYTESTRING(f, &c, checksum, len);  if (r!=0) return r;
 	r = le_both(xid, a.len, a.data, b.len, b.data, c.len, c.data,
-		    &memsize, &disksize, le);
+		    &memsize, &disksize, le,
+		    0, 0, 0);
 	toku_free_BYTESTRING(a);
 	toku_free_BYTESTRING(b);
 	toku_free_BYTESTRING(c);
@@ -174,7 +195,8 @@ int toku_fread_LEAFENTRY(FILE *f, LEAFENTRY *le, struct x1764 *checksum, u_int32
 	r = toku_fread_BYTESTRING(f, &a, checksum, len);  if (r!=0) return r;
 	r = toku_fread_BYTESTRING(f, &b, checksum, len);  if (r!=0) return r;
 	r = le_provdel(xid, a.len, a.data, b.len, b.data,
-		       &memsize, &disksize, le);
+		       &memsize, &disksize, le,
+		       0, 0, 0);
 	toku_free_BYTESTRING(a);
 	toku_free_BYTESTRING(b);
 	return r;
@@ -183,7 +205,8 @@ int toku_fread_LEAFENTRY(FILE *f, LEAFENTRY *le, struct x1764 *checksum, u_int32
 	r = toku_fread_BYTESTRING(f, &a, checksum, len);  if (r!=0) return r;
 	r = toku_fread_BYTESTRING(f, &b, checksum, len);  if (r!=0) return r;
 	r = le_provpair(xid, a.len, a.data, b.len, b.data,
-			&memsize, &disksize, le);
+			&memsize, &disksize, le,
+			0, 0, 0);
 	toku_free_BYTESTRING(a);
 	toku_free_BYTESTRING(b);
 	return r;

newbrt/leafentry.h

@@ -33,13 +33,16 @@
 
 u_int32_t toku_le_crc(LEAFENTRY v);
 
-int le_committed (u_int32_t klen, void* kval, u_int32_t dlen, void* dval, u_int32_t *resultsize, u_int32_t *disksize, LEAFENTRY *result);
+int le_committed (u_int32_t klen, void* kval, u_int32_t dlen, void* dval, u_int32_t *resultsize, u_int32_t *disksize, LEAFENTRY *result,
+		  OMT, struct mempool *, void **maybe_free);
 int le_both (TXNID xid, u_int32_t cklen, void* ckval, u_int32_t cdlen, void* cdval, u_int32_t pdlen, void* pdval,
-	     u_int32_t *memsize, u_int32_t *disksize, LEAFENTRY *result);
-int le_provdel  (TXNID xid, u_int32_t klen, void* kval, u_int32_t dlen, void* dval,
-		 u_int32_t *resultsize, u_int32_t *memsize, LEAFENTRY *result);
-int le_provpair (TXNID xid, u_int32_t klen, void* kval, u_int32_t dlen, void* dval,
-		 u_int32_t *memsize, u_int32_t *disksize, LEAFENTRY *result);
+	     u_int32_t *memsize, u_int32_t *disksize, LEAFENTRY *result,
+	     OMT, struct mempool *, void **maybe_free);
+int le_provdel (TXNID xid, u_int32_t klen, void* kval, u_int32_t dlen, void* dval,
+		u_int32_t *resultsize, u_int32_t *memsize, LEAFENTRY *result,
+		OMT, struct mempool *, void **maybe_free);
+int le_provpair (TXNID xid, u_int32_t klen, void* kval, u_int32_t plen, void* pval, u_int32_t *memsize, u_int32_t *disksize, LEAFENTRY *result,
+		 OMT omt, struct mempool *mp, void **maybe_free);
 
 enum le_state { LE_COMMITTED=1, // A committed pair.
 		LE_BOTH,        // A committed pair and a provisional pair.

newbrt/recover.c

@@ -158,7 +158,7 @@ static void toku_recover_fheader (LSN UU(lsn), TXNID UU(txnid),FILENUM filenum,L
     pair->brt->h = h;
     pair->brt->nodesize = h->nodesize;
     pair->brt->flags    = h->nodesize;
-    toku_cachefile_set_userdata(pair->cf, pair->brt->h, toku_brtheader_close);
+    toku_cachefile_set_userdata(pair->cf, pair->brt->h, toku_brtheader_close, toku_brtheader_checkpoint);
 }
 
 static void
@@ -613,7 +613,7 @@ toku_recover_insertleafentry (LSN lsn, FILENUM filenum, BLOCKNUM blocknum, u_int
     node->log_lsn = lsn;
     {
 	int memsize = leafentry_memsize(newleafentry);
-	void *mem = mempool_malloc_from_omt(node->u.l.buffer, &node->u.l.buffer_mempool, memsize);
+	void *mem = mempool_malloc_from_omt(node->u.l.buffer, &node->u.l.buffer_mempool, memsize, 0);
 	assert(mem);
 	memcpy(mem, newleafentry, memsize);
 	r = toku_omt_insert_at(node->u.l.buffer, mem, idx);

newbrt/tests/test-leafentry.c

 #include "portability.h"
 #include <string.h>
 #include "brttypes.h"
-#include "leafentry.h"
+#include "includes.h"
 
 static void test_leafentry_1 (void) {
     LEAFENTRY l;
     int r;
     u_int32_t msize, dsize;
-    r = le_committed(4, "abc", 3, "xy", &msize, &dsize, &l);
+    r = le_committed(4, "abc", 3, "xy", &msize, &dsize, &l, 0, 0, 0);
     assert(r==0);
     char expect[] = {LE_COMMITTED,
 		     0, 0, 0, 4,
@@ -24,7 +24,7 @@ static void test_leafentry_2 (void) {
     LEAFENTRY l;
     int r;
     u_int32_t msize, dsize;
-    r = le_both(0x0123456789abcdef0LL, 3, "ab", 4, "xyz", 5, "lmno", &msize, &dsize, &l);
+    r = le_both(0x0123456789abcdef0LL, 3, "ab", 4, "xyz", 5, "lmno", &msize, &dsize, &l, 0, 0, 0);
     assert(r==0);
     char expect[] = {LE_BOTH,
 		     0x12, 0x34, 0x56, 0x78, 0x9a, 0xbc, 0xde, 0xf0,
@@ -41,7 +41,7 @@ static void test_leafentry_3 (void) {
     LEAFENTRY l;
     int r;
     u_int32_t msize, dsize;
-    r = le_provdel(0x0123456789abcdef0LL, 3, "ab", 5, "lmno", &msize, &dsize, &l);
+    r = le_provdel(0x0123456789abcdef0LL, 3, "ab", 5, "lmno", &msize, &dsize, &l, 0, 0, 0);
     assert(r==0);
     char expect[] = {LE_PROVDEL,
 		     0x12, 0x34, 0x56, 0x78, 0x9a, 0xbc, 0xde, 0xf0,
@@ -57,7 +57,7 @@ static void test_leafentry_4 (void) {
     LEAFENTRY l;
     int r;
     u_int32_t msize, dsize;
-    r = le_provpair(0x0123456789abcdef0LL, 3, "ab", 5, "lmno", &msize, &dsize, &l);
+    r = le_provpair(0x0123456789abcdef0LL, 3, "ab", 5, "lmno", &msize, &dsize, &l, 0, 0, 0);
     assert(r==0);
     char expect[] = {LE_PROVPAIR,
 		     0x12, 0x34, 0x56, 0x78, 0x9a, 0xbc, 0xde, 0xf0,
@@ -89,7 +89,7 @@ static void test_leafentry_3long (void) {
     LEAFENTRY l;
     int r;
     u_int32_t msize, dsize;
-    r = le_provdel(0x0123456789abcdef0LL, 301, zeros, 1025, zeros, &msize, &dsize, &l);
+    r = le_provdel(0x0123456789abcdef0LL, 301, zeros, 1025, zeros, &msize, &dsize, &l, 0, 0, 0);
     assert(r==0);
     assert(sizeof(expect_3long)==msize);
     assert(msize==dsize);