Addresses #1765, #1730, #1733, Merge 12079:12119 from 2.0.0 into main

git-svn-id: file:///svn/toku/tokudb@12121 c7de825b-a66e-492c-adef-691d508d4ae1

newbrt/cachetable.c

@@ -183,6 +183,7 @@ struct cachefile {
 			 * because then we'd have to figure out if the transaction was already counted.  If we simply use a count for
 			 * every record in the transaction, we'll be ok.  Hence we use a 64-bit counter to make sure we don't run out.
 			 */
+    BOOL is_closing;    /* TRUE if a cachefile is being close/has been closed. */
     int fd;       /* Bug: If a file is opened read-only, then it is stuck in read-only.  If it is opened read-write, then subsequent writers can write to it too. */
     CACHETABLE cachetable;
     struct fileid fileid;
@@ -299,7 +300,7 @@ int toku_cachetable_openfd (CACHEFILE *cfptr, CACHETABLE ct, int fd, const char
 	if (memcmp(&extant->fileid, &fileid, sizeof(fileid))==0) {
             //File is already open (and in cachetable as extant)
             cachefile_refup(extant);
-	    if (extant->refcount==1) {
+	    if (extant->is_closing) {
 		// if another thread is closing this file, wait until the close is fully complete
 		r = toku_pthread_cond_wait(&extant->openfd_wait, ct->mutex);
 		assert(r == 0);
@@ -430,6 +431,8 @@ int toku_cachefile_close (CACHEFILE *cfp, TOKULOGGER logger, char **error_string
         //Checkpoint holds a reference, close should be impossible if still in use by a checkpoint.
         assert(!cf->next_in_checkpoint);
         assert(!cf->for_checkpoint);
+        assert(!cf->is_closing);
+        cf->is_closing = TRUE; //Mark this cachefile so that no one will re-use it.
 	int r;
 	// cachetable_flush_cachefile() may release and retake cachetable_lock,
 	// allowing another thread to get into toku_cachetable_openfd()
@@ -439,6 +442,8 @@ int toku_cachefile_close (CACHEFILE *cfp, TOKULOGGER logger, char **error_string
 	    if (cf->refcount > 0) {
                 int rs;
 		assert(cf->refcount == 1);       // toku_cachetable_openfd() is single-threaded
+                assert(!cf->next_in_checkpoint); //checkpoint cannot run on a closing file
+                assert(!cf->for_checkpoint);     //checkpoint cannot run on a closing file
 		rs = toku_pthread_cond_signal(&cf->openfd_wait); assert(rs == 0);
 	    }
 	    // we can destroy the condition variable because if there was another thread waiting, it was already signalled
@@ -1556,7 +1561,8 @@ toku_cachetable_begin_checkpoint (CACHETABLE ct, TOKULOGGER logger) {
             CACHEFILE cf;
             assert(ct->cachefiles_in_checkpoint==NULL);
             for (cf = ct->cachefiles; cf; cf=cf->next) {
-                if (cf->refcount>0) {
+                assert(!cf->is_closing); //Closing requires ydb lock (or in checkpoint).  Cannot happen.
+                assert(cf->refcount>0);  //Must have a reference if not closing.
                 //Incremement reference count of cachefile because we're using it for the checkpoint.
                 //This will prevent closing during the checkpoint.
                 cachefile_refup(cf);
@@ -1565,7 +1571,6 @@ toku_cachetable_begin_checkpoint (CACHETABLE ct, TOKULOGGER logger) {
                 cf->for_checkpoint           = TRUE;
             }
         }
-        }
 
 	if (logger) {
 	    // The checkpoint must be performed after the lock is acquired.

newbrt/roll.c

@@ -289,8 +289,12 @@ toku_rollback_rollinclude (BYTESTRING bs,
 }
 
 int
-toku_rollback_tablelock_on_empty_table (FILENUM filenum, TOKUTXN txn, YIELDF UU(yield), void* UU(yield_v))
+toku_rollback_tablelock_on_empty_table (FILENUM filenum,
+                                        TOKUTXN txn,
+                                        YIELDF yield,
+                                        void* yield_v)
 {
+    yield(toku_checkpoint_safe_client_lock, yield_v);
     // on rollback we have to make the file be empty, since we locked an empty table, and then may have done things to it.
 
     CACHEFILE cf;
@@ -307,6 +311,7 @@ toku_rollback_tablelock_on_empty_table (FILENUM filenum, TOKUTXN txn, YIELDF UU(
 	r = toku_brt_truncate(brt);
 	assert(r==0);
     }
+    toku_checkpoint_safe_client_unlock();
 
     return r; 
 }

newbrt/threadpool.c

@@ -5,9 +5,9 @@
 
 #include <stdio.h>
 #include <errno.h>
-#include <string.h>
 
 #include <toku_portability.h>
+#include <string.h>
 #include "toku_pthread.h"
 #include "toku_assert.h"
 #include "memory.h"

src/tests/Makefile

@@ -98,6 +98,7 @@ BDB_DONTRUN_TESTS = \
 	root_fifo_41 \
 	checkpoint_1 \
 	checkpoint_stress \
+	checkpoint_truncate_1 \
 #\ ends prev line
 
 # checkpoint tests depend on this header file, 

src/tests/checkpoint_stress.c

@@ -112,60 +112,6 @@ thin_out(DB* db, int iter) {
 }
 
 
-// assert that correct values are in expected rows
-static void
-verify_sequential_rows(DB* compare_db, int64_t firstkey, int64_t numkeys) {
-    //This does not lock the dbs/grab table locks.
-    //This means that you CANNOT CALL THIS while another thread is modifying the db.
-    //You CAN call it while a txn is open however.
-    int rval = 0;
-    DB_TXN *compare_txn;
-    int r, r1;
-
-    assert(numkeys >= 1);
-    r = env->txn_begin(env, NULL, &compare_txn, DB_READ_UNCOMMITTED);
-        CKERR(r);
-    DBC *c1;
-
-    r = compare_db->cursor(compare_db, compare_txn, &c1, 0);
-        CKERR(r);
-
-
-    DBT key1, val1;
-    DBT key2, val2;
-
-    int64_t k, v;
-
-    dbt_init_realloc(&key1);
-    dbt_init_realloc(&val1);
-
-    dbt_init(&key2, &k, sizeof(k));
-    dbt_init(&val2, &v, sizeof(v));
-
-    k = firstkey;
-    v = generate_val(k);
-    r1 = c1->c_get(c1, &key2, &val2, DB_GET_BOTH);
-    CKERR(r1);
-
-    int64_t i;
-    for (i = 1; i<numkeys; i++) {
-	k = i + firstkey;
-	v = generate_val(k);
-        r1 = c1->c_get(c1, &key1, &val1, DB_NEXT);
-	assert(r1==0);
-	rval = verify_identical_dbts(&key1, &key2) |
-	    verify_identical_dbts(&val1, &val2);
-	assert(rval == 0);
-    }
-    // now verify that there are no rows after the last expected 
-    r1 = c1->c_get(c1, &key1, &val1, DB_NEXT);
-    assert(r1 == DB_NOTFOUND);
-
-    c1->c_close(c1);
-    if (key1.data) toku_free(key1.data);
-    if (val1.data) toku_free(val1.data);
-    compare_txn->commit(compare_txn, 0);
-}
 
 static void
 drop_dead(void) {
@@ -268,7 +214,7 @@ run_test (int iter, int die) {
     cachebytes = K256 * (iter + 1) - (128 * 1024);
     if (cachebytes > max_windows_cachesize)
         cachebytes = 0;
-    if (iter & 1) cachebytes = 0;       // every other iteration use default cachesize
+    if (iter & 2) cachebytes = 0;       // use default cachesize half the time
 
     if (verbose)
 	printf("checkpoint_stress: iter = %d, cachesize (bytes) = 0x%08"PRIx64"\n", iter, cachebytes);

src/tests/checkpoint_test.h

@@ -246,6 +246,15 @@ db_replace(DICTIONARY d, DB_TXN *open_txn) {
     *d = temp;
 }
 
+static void UU()
+db_truncate(DB* db, DB_TXN *txn) {
+    u_int32_t row_count;
+    u_int32_t flags = 0;
+    int r = db->truncate(db, txn, &row_count, flags);
+    assert(row_count == 0);
+    CKERR(r);
+}
+
 static void UU()
 insert_random(DB *db1, DB *db2, DB_TXN *txn) {
     int64_t k = random64();
@@ -355,6 +364,61 @@ insert_n_fixed(DB *db1, DB *db2, DB_TXN *txn, int firstkey, int n) {
 }
 
 
+// assert that correct values are in expected rows
+static void  UU()
+verify_sequential_rows(DB* compare_db, int64_t firstkey, int64_t numkeys) {
+    //This does not lock the dbs/grab table locks.
+    //This means that you CANNOT CALL THIS while another thread is modifying the db.
+    //You CAN call it while a txn is open however.
+    int rval = 0;
+    DB_TXN *compare_txn;
+    int r, r1;
+
+    assert(numkeys >= 1);
+    r = env->txn_begin(env, NULL, &compare_txn, DB_READ_UNCOMMITTED);
+        CKERR(r);
+    DBC *c1;
+
+    r = compare_db->cursor(compare_db, compare_txn, &c1, 0);
+        CKERR(r);
+
+
+    DBT key1, val1;
+    DBT key2, val2;
+
+    int64_t k, v;
+
+    dbt_init_realloc(&key1);
+    dbt_init_realloc(&val1);
+
+    dbt_init(&key2, &k, sizeof(k));
+    dbt_init(&val2, &v, sizeof(v));
+
+    k = firstkey;
+    v = generate_val(k);
+    r1 = c1->c_get(c1, &key2, &val2, DB_GET_BOTH);
+    CKERR(r1);
+
+    int64_t i;
+    for (i = 1; i<numkeys; i++) {
+	k = i + firstkey;
+	v = generate_val(k);
+        r1 = c1->c_get(c1, &key1, &val1, DB_NEXT);
+	assert(r1==0);
+	rval = verify_identical_dbts(&key1, &key2) |
+	    verify_identical_dbts(&val1, &val2);
+	assert(rval == 0);
+    }
+    // now verify that there are no rows after the last expected 
+    r1 = c1->c_get(c1, &key1, &val1, DB_NEXT);
+    assert(r1 == DB_NOTFOUND);
+
+    c1->c_close(c1);
+    if (key1.data) toku_free(key1.data);
+    if (val1.data) toku_free(val1.data);
+    compare_txn->commit(compare_txn, 0);
+}
+
 
 
 static void UU()

src/tests/checkpoint_truncate_1.c

+/* -*- mode: C; c-basic-offset: 4 -*- */
+#ident "Copyright (c) 2007 Tokutek Inc.  All rights reserved."
+#ident "$Id$"
+
+#include <db.h>
+#include <sys/stat.h>
+#include "toku_pthread.h"
+#include "test.h"
+#include "checkpoint_test.h"
+
+
+// Purpose of this test is to verify that truncate performed during checkpoint does not cause crash.
+// Method:
+//   write two dictionaries, control and test
+//   take a checkpoint
+//     during the checkpoint (in the callback) trigger a second thread that will truncate test dictionary
+//   verify contents of control dictionary
+//
+// Potential improvements if necessary:
+//   make test more strenuous by checkpointing and truncating many dictionaries
+//   add random delays to alter timing per iteration
+//   verify that test dictionary does not exist on disk
+
+
+// Only useful for single threaded testing, 
+// but can be accessed from checkpoint_callback.
+static DICTIONARY test_dictionary = NULL;
+
+static toku_pthread_t thread;
+
+static int iter = 0;
+
+static void
+checkpoint_truncate_test(u_int32_t flags, u_int32_t n) {
+    void * ignore;
+
+    if (verbose) { 
+        printf("%s(%s):%d, n=0x%03x, checkpoint=%01x, flags=0x%05x\n", 
+               __FILE__, __FUNCTION__, __LINE__, 
+               n, 1, flags); 
+	printf("Verify that truncate done during checkpoint does not crash, iter = %d\n", iter);
+        fflush(stdout); 
+    }
+    dir_create();
+    env_startup(0);
+
+    DICTIONARY_S db_control;
+    init_dictionary(&db_control, flags, "control");
+    DICTIONARY_S db_test;
+    init_dictionary(&db_test, flags, "test");
+    test_dictionary = &db_test;
+
+    db_startup(&db_test, NULL);
+    db_startup(&db_control, NULL);
+    int64_t firstkey = 0;
+    int64_t numkeys = n;
+    insert_n_fixed(db_test.db, db_control.db, NULL, firstkey, numkeys);
+    snapshot(&db_test, TRUE);  // take checkpoint, truncate db_test during checkpoint callback
+    verify_sequential_rows(db_control.db, firstkey, numkeys);
+    pthread_join(thread, &ignore);
+    db_shutdown(&db_control);
+    env_shutdown();
+}
+
+
+// Purpose is to truncate test db while checkpoint is 
+// in progress.
+void *
+truncate_thread(void * extra) {
+    DICTIONARY d = *(DICTIONARY*) extra;
+    char name[MAX_NAME*2];
+    fill_name(d, name, sizeof(name));
+
+    // maybe insert random delay here if necessary
+
+    if (verbose) {
+	printf("truncating %s\n",
+	       name);
+	fflush(stdout);
+    }
+    if (iter & 1)
+	pthread_yield();  // increase probability of collision by having some different timing
+    db_truncate(d->db, NULL);
+    return NULL;
+} 
+
+
+void checkpoint_callback_1(void * extra) {
+    int r = toku_pthread_create(&thread, 0, truncate_thread, extra);
+    CKERR(r);
+}
+
+int
+test_main (int argc, char *argv[]) {
+    int limit = 512;
+    parse_args(argc, argv);
+
+    db_env_set_checkpoint_callback(checkpoint_callback_1, &test_dictionary);
+    for (iter = 0; iter < limit; iter++) {
+	checkpoint_truncate_test(0, 16*1024+1);
+    }
+    db_env_set_checkpoint_callback(NULL, NULL);
+
+    return 0;
+}

src/ydb.c

@@ -3665,7 +3665,12 @@ static const DBT* toku_db_dbt_neg_infty(void) {
 }
 
 static int locked_db_truncate(DB *db, DB_TXN *txn, u_int32_t *row_count, u_int32_t flags) {
-    toku_ydb_lock(); int r = toku_db_truncate(db, txn, row_count, flags); toku_ydb_unlock(); return r;
+    toku_checkpoint_safe_client_lock();
+    toku_ydb_lock();
+    int r = toku_db_truncate(db, txn, row_count, flags);
+    toku_ydb_unlock();
+    toku_checkpoint_safe_client_unlock();
+    return r;
 }
 
 static int toku_db_create(DB ** db, DB_ENV * env, u_int32_t flags) {