X-Git-Url: http://git.alex.org.uk

diff --git a/nbd-server.c b/nbd-server.c
index 4122cef..3343d9d 100644
--- a/nbd-server.c
+++ b/nbd-server.c
@@ -133,17 +133,9 @@ int dontfork = 0;
 /* Debugging macros */
 //#define DODBG
 #ifdef DODBG
-#define DEBUG( a ) printf( a )
-#define DEBUG2( a,b ) printf( a,b )
-#define DEBUG3( a,b,c ) printf( a,b,c )
-#define DEBUG4( a,b,c,d ) printf( a,b,c,d )
-#define DEBUG5( a,b,c,d,e ) printf( a,b,c,d,e )
+#define DEBUG(...) printf(__VA_ARGS__)
 #else
-#define DEBUG( a )
-#define DEBUG2( a,b ) 
-#define DEBUG3( a,b,c ) 
-#define DEBUG4( a,b,c,d ) 
-#define DEBUG5( a,b,c,d,e ) 
+#define DEBUG(...)
 #endif
 #ifndef PACKAGE_VERSION
 #define PACKAGE_VERSION ""
@@ -213,6 +205,7 @@ typedef struct {
 				  disconnects */
 	gchar* servename;    /**< name of the export as selected by nbd-client */
 	int max_connections; /**< maximum number of opened connections */
+	gchar* transactionlog;/**< filename for transaction log */
 } SERVER;
 
 /**
@@ -239,6 +232,7 @@ typedef struct {
 	u32 difffilelen;     /**< number of pages in difffile */
 	u32 *difmap;	     /**< see comment on the global difmap for this one */
 	gboolean modern;     /**< client was negotiated using modern negotiation protocol */
+	int transactionlogfd;/**< fd for transaction log */
 } CLIENT;
 
 /**
@@ -582,6 +576,8 @@ void remove_server(gpointer s) {
 		g_free(server->prerun);
 	if(server->postrun)
 		g_free(server->postrun);
+	if(server->transactionlog)
+		g_free(server->transactionlog);
 	g_free(server);
 }
 
@@ -621,6 +617,9 @@ SERVER* dup_serve(SERVER *s) {
 
 	if(s->postrun)
 		serve->postrun = g_strdup(s->postrun);
+
+	if(s->transactionlog)
+		serve->transactionlog = g_strdup(s->transactionlog);
 	
 	if(s->servename)
 		serve->servename = g_strdup(s->servename);
@@ -719,6 +718,7 @@ GArray* parse_cfile(gchar* f, GError** e) {
 		{ "virtstyle",	FALSE,	PARAM_STRING,	&(virtstyle),		0 },
 		{ "prerun",	FALSE,	PARAM_STRING,	&(s.prerun),		0 },
 		{ "postrun",	FALSE,	PARAM_STRING,	&(s.postrun),		0 },
+		{ "transactionlog", FALSE, PARAM_STRING, &(s.transactionlog),	0 },
 		{ "readonly",	FALSE,	PARAM_BOOL,	&(s.flags),		F_READONLY },
 		{ "multifile",	FALSE,	PARAM_BOOL,	&(s.flags),		F_MULTIFILE },
 		{ "copyonwrite", FALSE,	PARAM_BOOL,	&(s.flags),		F_COPYONWRITE },
@@ -906,7 +906,7 @@ void sigchld_handler(int s) {
 		if(!i) {
 			msg3(LOG_INFO, "SIGCHLD received for an unknown child with PID %ld", (long)pid);
 		} else {
-			DEBUG2("Removing %d from the list of children", pid);
+			DEBUG("Removing %d from the list of children", pid);
 			g_hash_table_remove(children, &pid);
 		}
 	}
@@ -984,7 +984,7 @@ off_t size_autodetect(int fhandle) {
 	if (es > ((off_t)0)) {
 		return es;
         } else {
-                DEBUG2("lseek failed: %d", errno==EBADF?1:(errno==ESPIPE?2:(errno==EINVAL?3:4)));
+                DEBUG("lseek failed: %d", errno==EBADF?1:(errno==ESPIPE?2:(errno==EINVAL?3:4)));
         }
 
 	err("Could not find size of exported block device: %m");
@@ -1074,14 +1074,45 @@ ssize_t rawexpwrite(off_t a, char *buf, size_t len, CLIENT *client, int fua) {
 	if(maxbytes && len > maxbytes)
 		len = maxbytes;
 
-	DEBUG5("(WRITE to fd %d offset %llu len %u fua %d), ", fhandle, foffset, len, fua);
+	DEBUG("(WRITE to fd %d offset %llu len %u fua %d), ", fhandle, (long long unsigned)foffset, (unsigned int)len, fua);
 
 	myseek(fhandle, foffset);
 	retval = write(fhandle, buf, len);
 	if(client->server->flags & F_SYNC) {
 		fsync(fhandle);
 	} else if (fua) {
-#ifdef USE_SYNC_FILE_RANGE
+
+	  /* This is where we would do the following
+	   *   #ifdef USE_SYNC_FILE_RANGE
+	   * However, we don't, for the reasons set out below
+	   * by Christoph Hellwig <hch@infradead.org>
+	   *
+	   * [BEGINS] 
+	   * fdatasync is equivalent to fsync except that it does not flush
+	   * non-essential metadata (basically just timestamps in practice), but it
+	   * does flush metadata requried to find the data again, e.g. allocation
+	   * information and extent maps.  sync_file_range does nothing but flush
+	   * out pagecache content - it means you basically won't get your data
+	   * back in case of a crash if you either:
+	   * 
+	   *  a) have a volatile write cache in your disk (e.g. any normal SATA disk)
+	   *  b) are using a sparse file on a filesystem
+	   *  c) are using a fallocate-preallocated file on a filesystem
+	   *  d) use any file on a COW filesystem like btrfs
+	   * 
+	   * e.g. it only does anything useful for you if you do not have a volatile
+	   * write cache, and either use a raw block device node, or just overwrite
+	   * an already fully allocated (and not preallocated) file on a non-COW
+	   * filesystem.
+	   * [ENDS]
+	   *
+	   * What we should do is open a second FD with O_DSYNC set, then write to
+	   * that when appropriate. However, with a Linux client, every REQ_FUA
+	   * immediately follows a REQ_FLUSH, so fdatasync does not cause performance
+	   * problems.
+	   *
+	   */
+#if 0
 		sync_file_range(fhandle, foffset, len,
 				SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE |
 				SYNC_FILE_RANGE_WAIT_AFTER);
@@ -1128,7 +1159,7 @@ ssize_t rawexpread(off_t a, char *buf, size_t len, CLIENT *client) {
 	if(maxbytes && len > maxbytes)
 		len = maxbytes;
 
-	DEBUG4("(READ from fd %d offset %llu len %u), ", fhandle, foffset, len);
+	DEBUG("(READ from fd %d offset %llu len %u), ", fhandle, (long long unsigned int)foffset, (unsigned int)len);
 
 	myseek(fhandle, foffset);
 	return read(fhandle, buf, len);
@@ -1165,7 +1196,7 @@ int expread(off_t a, char *buf, size_t len, CLIENT *client) {
 
 	if (!(client->server->flags & F_COPYONWRITE))
 		return(rawexpread_fully(a, buf, len, client));
-	DEBUG3("Asked to read %d bytes at %llu.\n", len, (unsigned long long)a);
+	DEBUG("Asked to read %u bytes at %llu.\n", (unsigned int)len, (unsigned long long)a);
 
 	mapl=a/DIFFPAGESIZE; maph=(a+len-1)/DIFFPAGESIZE;
 
@@ -1175,12 +1206,12 @@ int expread(off_t a, char *buf, size_t len, CLIENT *client) {
 		rdlen=(0<DIFFPAGESIZE-offset && len<(size_t)(DIFFPAGESIZE-offset)) ?
 			len : (size_t)DIFFPAGESIZE-offset;
 		if (client->difmap[mapcnt]!=(u32)(-1)) { /* the block is already there */
-			DEBUG3("Page %llu is at %lu\n", (unsigned long long)mapcnt,
+			DEBUG("Page %llu is at %lu\n", (unsigned long long)mapcnt,
 			       (unsigned long)(client->difmap[mapcnt]));
 			myseek(client->difffile, client->difmap[mapcnt]*DIFFPAGESIZE+offset);
 			if (read(client->difffile, buf, rdlen) != rdlen) return -1;
 		} else { /* the block is not there */
-			DEBUG2("Page %llu is not here, we read the original one\n",
+			DEBUG("Page %llu is not here, we read the original one\n",
 			       (unsigned long long)mapcnt);
 			if(rawexpread_fully(a, buf, rdlen, client)) return -1;
 		}
@@ -1209,7 +1240,7 @@ int expwrite(off_t a, char *buf, size_t len, CLIENT *client, int fua) {
 
 	if (!(client->server->flags & F_COPYONWRITE))
 		return(rawexpwrite_fully(a, buf, len, client, fua)); 
-	DEBUG3("Asked to write %d bytes at %llu.\n", len, (unsigned long long)a);
+	DEBUG("Asked to write %u bytes at %llu.\n", (unsigned int)len, (unsigned long long)a);
 
 	mapl=a/DIFFPAGESIZE ; maph=(a+len-1)/DIFFPAGESIZE ;
 
@@ -1220,7 +1251,7 @@ int expwrite(off_t a, char *buf, size_t len, CLIENT *client, int fua) {
 			len : (size_t)DIFFPAGESIZE-offset;
 
 		if (client->difmap[mapcnt]!=(u32)(-1)) { /* the block is already there */
-			DEBUG3("Page %llu is at %lu\n", (unsigned long long)mapcnt,
+			DEBUG("Page %llu is at %lu\n", (unsigned long long)mapcnt,
 			       (unsigned long)(client->difmap[mapcnt])) ;
 			myseek(client->difffile,
 					client->difmap[mapcnt]*DIFFPAGESIZE+offset);
@@ -1228,7 +1259,7 @@ int expwrite(off_t a, char *buf, size_t len, CLIENT *client, int fua) {
 		} else { /* the block is not there */
 			myseek(client->difffile,client->difffilelen*DIFFPAGESIZE) ;
 			client->difmap[mapcnt]=(client->server->flags&F_SPARSE)?mapcnt:client->difffilelen++;
-			DEBUG3("Page %llu is not here, we put it at %lu\n",
+			DEBUG("Page %llu is not here, we put it at %lu\n",
 			       (unsigned long long)mapcnt,
 			       (unsigned long)(client->difmap[mapcnt]));
 			rdlen=DIFFPAGESIZE ;
@@ -1253,9 +1284,6 @@ int expwrite(off_t a, char *buf, size_t len, CLIENT *client, int fua) {
 }
 
 int expflush(CLIENT *client) {
-	int fhandle;
-	off_t foffset;
-	size_t maxbytes;
 	gint i;
 
         if (client->server->flags & F_COPYONWRITE) {
@@ -1347,6 +1375,7 @@ CLIENT* negotiate(int net, CLIENT *client, GArray* servers) {
 				client->exportsize = OFFT_MAX;
 				client->net = net;
 				client->modern = TRUE;
+				client->transactionlogfd = -1;
 				free(name);
 				return client;
 			}
@@ -1386,7 +1415,9 @@ CLIENT* negotiate(int net, CLIENT *client, GArray* servers) {
 }
 
 /** sending macro. */
-#define SEND(net,reply) writeit( net, &reply, sizeof( reply ));
+#define SEND(net,reply) { writeit( net, &reply, sizeof( reply )); \
+	if (client->transactionlogfd != -1) \
+		writeit(client->transactionlogfd, &reply, sizeof(reply)); }
 /** error macro. */
 #define ERROR(client,reply,errcode) { reply.error = htonl(errcode); SEND(client->net,reply); reply.error = 0; }
 /**
@@ -1421,6 +1452,9 @@ int mainloop(CLIENT *client) {
 		printf("%d: ", i);
 #endif
 		readit(client->net, &request, sizeof(request));
+		if (client->transactionlogfd != -1)
+			writeit(client->transactionlogfd, &request, sizeof(request));
+
 		request.from = ntohll(request.from);
 		request.type = ntohl(request.type);
 		command = request.type & NBD_CMD_MASK_COMMAND;
@@ -1447,11 +1481,9 @@ int mainloop(CLIENT *client) {
 		} else {
 			currlen = len;
 		}
-#ifdef DODBG
-		printf("%s from %llu (%llu) len %d, ", command ? "WRITE" :
+		DEBUG("%s from %llu (%llu) len %d, ", command ? "WRITE" :
 				"READ", (unsigned long long)request.from,
-				(unsigned long long)request.from / 512, len);
-#endif
+				(unsigned long long)request.from / 512, (unsigned int)len);
 		memcpy(reply.handle, request.handle, sizeof(reply.handle));
 
 		if ((command==NBD_CMD_WRITE) || (command==NBD_CMD_READ)) {
@@ -1508,6 +1540,8 @@ int mainloop(CLIENT *client) {
 		if (command==NBD_CMD_READ) {
 			DEBUG("exp->buf, ");
 			memcpy(buf, &reply, sizeof(struct nbd_reply));
+			if (client->transactionlogfd != -1)
+				writeit(client->transactionlogfd, &reply, sizeof(reply));
 			p = buf + sizeof(struct nbd_reply);
 			writelen = currlen + sizeof(struct nbd_reply);
 			while(len > 0) {
@@ -1520,6 +1554,7 @@ int mainloop(CLIENT *client) {
 				DEBUG("buf->net, ");
 				writeit(client->net, buf, writelen);
 				len -= currlen;
+				request.from += currlen;
 				currlen = (len < BUFSIZE) ? len : BUFSIZE;
 				p = buf;
 				writelen = currlen;
@@ -1559,7 +1594,7 @@ void setupexport(CLIENT* client) {
 		} else {
 			tmpname=g_strdup(client->exportname);
 		}
-		DEBUG2( "Opening %s\n", tmpname );
+		DEBUG( "Opening %s\n", tmpname );
 		fi.fhandle = open(tmpname, mode);
 		if(fi.fhandle == -1 && mode == O_RDWR) {
 			/* Try again because maybe media was read-only */
@@ -1658,6 +1693,15 @@ int do_run(gchar* command, gchar* file) {
  * @param client a connected client
  **/
 void serveconnection(CLIENT *client) {
+	if (client->server->transactionlog && (client->transactionlogfd == -1))
+	{
+		if (-1 == (client->transactionlogfd = open(client->server->transactionlog,
+							   O_WRONLY | O_CREAT,
+							   S_IRUSR | S_IWUSR)))
+			g_warning("Could not open transaction log %s",
+				  client->server->transactionlog);
+	}
+
 	if(do_run(client->server->prerun, client->exportname)) {
 		exit(EXIT_FAILURE);
 	}
@@ -1671,6 +1715,12 @@ void serveconnection(CLIENT *client) {
 
 	mainloop(client);
 	do_run(client->server->postrun, client->exportname);
+
+	if (-1 != client->transactionlogfd)
+	{
+		close(client->transactionlogfd);
+		client->transactionlogfd = -1;
+	}
 }
 
 /**
@@ -1856,6 +1906,7 @@ int serveloop(GArray* servers) {
 					client->server=serve;
 					client->exportsize=OFFT_MAX;
 					client->net=net;
+					client->transactionlogfd = -1;
 				}
 				set_peername(net, client);
 				if (!authorized_client(client)) {