cow friendly HGETALL and variants

[redis.git] / src / vm.c

diff --git a/src/vm.c b/src/vm.c
index a34387521d0cac989b73fe9f7021dd53ba917a3b..ac0d92e332cf091b88c5caeb2b3a9b439869a2a5 100644 (file)
--- a/src/vm.c
+++ b/src/vm.c
@@ -96,7 +96,6 @@ void vmInit(void) {
     server.io_processed = listCreate();
     server.io_ready_clients = listCreate();
     pthread_mutex_init(&server.io_mutex,NULL);
-    pthread_mutex_init(&server.obj_freelist_mutex,NULL);
     pthread_mutex_init(&server.io_swapfile_mutex,NULL);
     server.io_active_threads = 0;
     if (pipe(pipefds) == -1) {
@@ -110,6 +109,11 @@ void vmInit(void) {
     /* LZF requires a lot of stack */
     pthread_attr_init(&server.io_threads_attr);
     pthread_attr_getstacksize(&server.io_threads_attr, &stacksize);
+
+    /* Solaris may report a stacksize of 0, let's set it to 1 otherwise
+     * multiplying it by 2 in the while loop later will not really help ;) */
+    if (!stacksize) stacksize = 1;
+
     while (stacksize < REDIS_THREAD_STACK_SIZE) stacksize *= 2;
     pthread_attr_setstacksize(&server.io_threads_attr, stacksize);
     /* Listen for events in the threaded I/O pipe */
@@ -259,7 +263,7 @@ int vmWriteObjectOnSwap(robj *o, off_t page) {
  * If we can't find enough contiguous empty pages to swap the object on disk
  * NULL is returned. */
 vmpointer *vmSwapObjectBlocking(robj *val) {
-    off_t pages = rdbSavedObjectPages(val,NULL);
+    off_t pages = rdbSavedObjectPages(val);
     off_t page;
     vmpointer *vp;
 
@@ -357,7 +361,7 @@ robj *vmPreviewObject(robj *o) {
 double computeObjectSwappability(robj *o) {
     /* actual age can be >= minage, but not < minage. As we use wrapping
      * 21 bit clocks with minutes resolution for the LRU. */
-    time_t minage = abs(server.lruclock - o->lru);
+    time_t minage = estimateObjectIdleTime(o);
     long asize = 0, elesize;
     robj *ele;
     list *l;
@@ -548,7 +552,15 @@ void freeIOJob(iojob *j) {
 
 /* Every time a thread finished a Job, it writes a byte into the write side
  * of an unix pipe in order to "awake" the main thread, and this function
- * is called. */
+ * is called.
+ *
+ * Note that this is called both by the event loop, when a I/O thread
+ * sends a byte in the notification pipe, and is also directly called from
+ * waitEmptyIOJobsQueue().
+ *
+ * In the latter case we don't want to swap more, so we use the
+ * "privdata" argument setting it to a not NULL value to signal this
+ * condition. */
 void vmThreadedIOCompletedJob(aeEventLoop *el, int fd, void *privdata,
             int mask)
 {
@@ -558,6 +570,8 @@ void vmThreadedIOCompletedJob(aeEventLoop *el, int fd, void *privdata,
     REDIS_NOTUSED(mask);
     REDIS_NOTUSED(privdata);
 
+    if (privdata != NULL) trytoswap = 0; /* check the comments above... */
+
     /* For every byte we read in the read side of the pipe, there is one
      * I/O job completed to process. */
     while((retval = read(fd,buf,1)) == 1) {
@@ -807,9 +821,7 @@ void *IOThreadEntryPoint(void *arg) {
             vmpointer *vp = (vmpointer*)j->id;
             j->val = vmReadObjectFromSwap(j->page,vp->vtype);
         } else if (j->type == REDIS_IOJOB_PREPARE_SWAP) {
-            FILE *fp = fopen("/dev/null","w+");
-            j->pages = rdbSavedObjectPages(j->val,fp);
-            fclose(fp);
+            j->pages = rdbSavedObjectPages(j->val);
         } else if (j->type == REDIS_IOJOB_DO_SWAP) {
             if (vmWriteObjectOnSwap(j->val,j->page) == REDIS_ERR)
                 j->canceled = 1;
@@ -869,7 +881,8 @@ void waitEmptyIOJobsQueue(void) {
         io_processed_len = listLength(server.io_processed);
         unlockThreadedIO();
         if (io_processed_len) {
-            vmThreadedIOCompletedJob(NULL,server.io_ready_pipe_read,NULL,0);
+            vmThreadedIOCompletedJob(NULL,server.io_ready_pipe_read,
+                                                        (void*)0xdeadbeef,0);
             usleep(1000); /* 1 millisecond */
         } else {
             usleep(10000); /* 10 milliseconds */