Support configurable file backed scratch

src/Makefile

@@ -166,9 +166,9 @@ REDIS_SERVER_NAME=redis-server
 REDIS_SENTINEL_NAME=redis-sentinel
 REDIS_SERVER_OBJ=adlist.o quicklist.o ae.o anet.o dict.o server.o sds.o zmalloc.o lzf_c.o lzf_d.o pqsort.o zipmap.o sha1.o ziplist.o release.o networking.o util.o object.o db.o replication.o rdb.o t_string.o t_list.o t_set.o t_zset.o t_hash.o config.o aof.o pubsub.o multi.o debug.o sort.o intset.o syncio.o cluster.o crc16.o endianconv.o slowlog.o scripting.o bio.o rio.o rand.o memtest.o crc64.o bitops.o sentinel.o notify.o setproctitle.o blocked.o hyperloglog.o latency.o sparkline.o redis-check-rdb.o redis-check-aof.o geo.o lazyfree.o module.o evict.o expire.o geohash.o geohash_helper.o childinfo.o defrag.o siphash.o rax.o t_stream.o listpack.o localtime.o lolwut.o lolwut5.o acl.o storage.o
 REDIS_CLI_NAME=redis-cli
-REDIS_CLI_OBJ=anet.o adlist.o dict.o redis-cli.o zmalloc.o release.o anet.o ae.o crc64.o siphash.o crc16.o storage.o
+REDIS_CLI_OBJ=anet.o adlist.o dict.o redis-cli.o zmalloc.o release.o anet.o ae.o crc64.o siphash.o crc16.o storage-lite.o
 REDIS_BENCHMARK_NAME=redis-benchmark
-REDIS_BENCHMARK_OBJ=ae.o anet.o redis-benchmark.o adlist.o zmalloc.o redis-benchmark.o storage.o
+REDIS_BENCHMARK_OBJ=ae.o anet.o redis-benchmark.o adlist.o zmalloc.o redis-benchmark.o storage-lite.o
 REDIS_CHECK_RDB_NAME=redis-check-rdb
 REDIS_CHECK_AOF_NAME=redis-check-aof
 

src/config.c

@@ -804,6 +804,8 @@ void loadServerConfigFromString(char *config) {
                 err = sentinelHandleConfiguration(argv+1,argc-1);
                 if (err) goto loaderr;
             }
+        } else if (!strcasecmp(argv[0],"scratch-file-path")) {
+            storage_init(argv[1]);
         } else {
             err = "Bad directive or wrong number of arguments"; goto loaderr;
         }

src/object.c

@@ -39,7 +39,7 @@
 /* ===================== Creation and parsing of objects ==================== */
 
 robj *createObject(int type, void *ptr) {
-    robj *o = salloc_obj(); //zmalloc(sizeof(*o), MALLOC_SHARED);
+    robj *o = zmalloc(sizeof(*o), MALLOC_SHARED);
     o->type = type;
     o->encoding = OBJ_ENCODING_RAW;
     o->ptr = ptr;
@@ -362,10 +362,7 @@ void decrRefCount(robj *o) {
         case OBJ_STREAM: freeStreamObject(o); break;
         default: serverPanic("Unknown object type"); break;
         }
-        if (o->type == OBJ_STRING && o->encoding == OBJ_ENCODING_EMBSTR)
-            zfree(o);
-        else
-            sfree_obj(o);
+        zfree(o);
     } else {
         if (o->refcount <= 0) serverPanic("decrRefCount against refcount <= 0");
         if (o->refcount != OBJ_SHARED_REFCOUNT) o->refcount--;

src/redis-benchmark.c

@@ -653,7 +653,7 @@ int main(int argc, const char **argv) {
 
     client c;
 
-    storage_init();
+    storage_init(NULL);
 
     srandom(time(NULL));
     signal(SIGHUP, SIG_IGN);

src/redis-cli.c

@@ -7176,7 +7176,7 @@ static void intrinsicLatencyMode(void) {
 int main(int argc, char **argv) {
     int firstarg;
 
-    storage_init();
+    storage_init(NULL);
     config.hostip = sdsnew("127.0.0.1");
     config.hostport = 6379;
     config.hostsocket = NULL;

src/server.c

@@ -4740,7 +4740,7 @@ int main(int argc, char **argv) {
     struct timeval tv;
     int j;
 
-    storage_init();
+    storage_init(NULL);
 
 #ifdef REDIS_TEST
     if (argc == 3 && !strcasecmp(argv[1], "test")) {

src/storage-lite.c

@@ -0,0 +1,41 @@
+#include <stdlib.h>
+#include <stdio.h>
+#include <memkind.h>
+#include <sys/ioctl.h>
+#include <linux/fs.h>
+#include <unistd.h>
+#include <inttypes.h>
+#include "storage.h"
+#include <assert.h>
+
+// initialize the memory subsystem. 
+//  NOTE: This may be called twice, first with NULL specifying we should use ram
+//      later, after the configuration file is loaded with a path to where we should
+//      place our temporary file.
+void storage_init(const char *tmpfilePath)
+{
+    assert(tmpfilePath == NULL);
+    (void)tmpfilePath;
+}
+
+void *salloc(size_t cb, enum MALLOC_CLASS class)
+{
+    (void)class;
+    return malloc(cb);
+}
+
+void *scalloc(size_t cb, size_t c, enum MALLOC_CLASS class)
+{
+    (void)class;
+    return calloc(cb, c);
+}
+
+void sfree(void *pv)
+{
+    free(pv);
+}
+
+void *srealloc(void *pv, size_t cb)
+{
+    return realloc(pv, cb);
+}

src/storage.c

@@ -10,7 +10,7 @@
 #include "storage.h"
 
 struct memkind *mkdisk = NULL;
-static char *PMEM_DIR = "/mnt/btrfs_scratch/";
+static const char *PMEM_DIR = NULL;
 
 void handle_prefork();
 void handle_postfork_parent();
@@ -85,9 +85,6 @@ void *pool_alloc(struct alloc_pool *ppool)
     }
 }
 
-#pragma weak serverLog
-void serverLog(int level, const char*fmt, ...){}
-
 void pool_free(struct alloc_pool *ppool, void *pv)
 {
     struct object_page *cur = ppool->pobjpageHead;
@@ -113,18 +110,62 @@ void pool_free(struct alloc_pool *ppool, void *pv)
 struct alloc_pool poolobj;
 struct alloc_pool poolembstrobj;
 
-void storage_init()
+int forkFile()
 {
-    int errv = memkind_create_pmem(PMEM_DIR, 0, &mkdisk);
-    if (errv)
+    int fdT;
+    memkind_tmpfile(PMEM_DIR, &fdT);
+    if (ioctl(fdT, FICLONE, memkind_fd(mkdisk)) == -1)
     {
-        fprintf(stderr, "Memory pool creation failed: %d\n", errv);
-        exit(EXIT_FAILURE);
+        return -1;
     }
-    pool_initialize(&poolobj, sizeof(robj));
-    pool_initialize(&poolembstrobj, EMBSTR_ROBJ_SIZE);
+    return fdT;
+}
 
-    pthread_atfork(handle_prefork, handle_postfork_parent, handle_postfork_child);
+// initialize the memory subsystem. 
+//  NOTE: This may be called twice, first with NULL specifying we should use ram
+//      later, after the configuration file is loaded with a path to where we should
+//      place our temporary file.
+void storage_init(const char *tmpfilePath)
+{
+    if (tmpfilePath == NULL)
+    {
+        serverAssert(mkdisk == NULL);
+        mkdisk = MEMKIND_DEFAULT;
+    }
+    else
+    {
+        // First create the file
+        serverAssert(mkdisk == MEMKIND_DEFAULT);
+        PMEM_DIR = memkind_malloc(MEMKIND_DEFAULT, strlen(tmpfilePath));
+        strcpy((char*)PMEM_DIR, tmpfilePath);
+        int errv = memkind_create_pmem(PMEM_DIR, 0, &mkdisk);
+        if (errv == MEMKIND_ERROR_INVALID)
+        {
+            serverLog(LOG_CRIT, "Memory pool creation failed: %s", strerror(errno));
+            exit(EXIT_FAILURE);
+        }
+        else if (errv)
+        {
+            char msgbuf[1024];
+            memkind_error_message(errv, msgbuf, 1024);
+            serverLog(LOG_CRIT, "Memory pool creation failed: %s", msgbuf);
+            exit(EXIT_FAILURE);
+        }
+
+        // Next test if COW is working
+        int fdTest = forkFile();
+        if (fdTest < 0)
+        {
+            serverLog(LOG_ERR, "Scratch file system does not support Copy on Write.  To fix this scratch-file-path must point to a path on a filesystem which supports copy on write, such as btrfs.");
+            exit(EXIT_FAILURE);
+        }
+        close(fdTest);
+
+        pool_initialize(&poolobj, sizeof(robj));
+        pool_initialize(&poolembstrobj, EMBSTR_ROBJ_SIZE);
+
+        pthread_atfork(handle_prefork, handle_postfork_parent, handle_postfork_child);
+    }
 }
 
 
@@ -184,12 +225,9 @@ void *srealloc(void *pv, size_t cb)
 int fdNew = -1;
 void handle_prefork()
 {
-    memkind_tmpfile(PMEM_DIR, &fdNew);
-    if (ioctl(fdNew, FICLONE, memkind_fd(mkdisk)) == -1)
-    {
-        perror("failed to fork file");
-        exit(EXIT_FAILURE);
-    }
+    fdNew = forkFile();
+    if (fdNew < 0)
+        serverLog(LOG_ERR, "Failed to clone scratch file");
 }
 
 void handle_postfork_parent()

src/storage.h

@@ -9,12 +9,7 @@ enum MALLOC_CLASS
     MALLOC_SHARED,
 };
 
-void storage_init(void);
-
-struct redisObject *salloc_obj();
-void sfree_obj(struct redisObject *obj);
-struct redisObject *salloc_objembstr();
-void sfree_objembstr(struct redisObject *obj);
+void storage_init(const char *tmpfilePath);
 
 void *salloc(size_t cb, enum MALLOC_CLASS class);
 void *scalloc(size_t cb, size_t c, enum MALLOC_CLASS class);