</listitem>
</varlistentry>
<varlistentry>
+ <term><option>flush</option></term>
+ <listitem>
+ <para>Optional; boolean.</para>
+ <para>When this option is enabled,
+ <command>nbd-server</command> will inform the client that it
+ supports and desires to be sent flush requests when the
+ elevator layer receives them. Receipt of a flush request
+ will cause an fdatasync() (or, if the sync option is set,
+ an fsync()) on the backend storage. This increases
+ reliability in the case of an unclean shutdown at
+ the expense of a degradation of performance. The default
+ state is disabled. This option will have no effect unless
+ supported by the client.
+ </para>
+ </listitem>
+ </varlistentry>
+ <varlistentry>
+ <term><option>fua</option></term>
+ <listitem>
+ <para>Optional; boolean.</para>
+ <para>When this option is enabled,
+ <command>nbd-server</command> will inform the client that it
+ supports and desires to be sent fua (force unit access) commands
+ when the elevator layer receives them. Receipt of a force unit
+ access command will cause the specified command to be synced
+ to backend storage using sync_file_range() if supported, or
+ fdatasync() otherwise. This increases
+ reliability in the case of an unclean shutdown at
+ the expense of a degradation of performance. The default
+ state is disabled. This option will have no effect unless
+ supported by the client.
+ </para>
+ </listitem>
+ </varlistentry>
+ <varlistentry>
+ <term><option>rotational</option></term>
+ <listitem>
+ <para>Optional; boolean.</para>
+ <para>When this option is enabled,
+ <command>nbd-server</command> will inform the client that it
+ it would prefer it to send requests in elevator order, perhaps
+ because it has a backing store and no local elevator. By
+ default, the client uses QUEUE_FLAG_NONROT, which effectively
+ restricts the function of the elevator to block merges. By
+ specifying this flag on the server, the client will not use
+ QUEUE_FLAG_NONROT, meaning the client elevator will perform
+ normal elevator ordering of I/O requests. Note that even when
+ the backing store is on rotating media, it is not normally
+ necessary to specify this flag, as the server's elevator
+ algorithm will be used. This flag is only required where
+ the server will not be using an elevator algorithm or where
+ the elevator algorithm is effectively neutered (e.g. with
+ the sync option set). This option will have no effect unless
+ supported by the client.
+ </para>
+ </listitem>
+ </varlistentry>
+ <varlistentry>
<term><option>sparse_cow</option></term>
<listitem>
<para>Optional; boolean.</para>
#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 )
#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 )
#endif
#ifndef PACKAGE_VERSION
#define PACKAGE_VERSION ""
#define F_SPARSE 16 /**< flag to tell us copyronwrite should use a sparse file */
#define F_SDP 32 /**< flag to tell us the export should be done using the Socket Direct Protocol for RDMA */
#define F_SYNC 64 /**< Whether to fsync() after a write */
+#define F_FLUSH 128 /**< Whether server wants FLUSH to be sent by the client */
+#define F_FUA 256 /**< Whether server wants FUA to be sent by the client */
+#define F_ROTATIONAL 512 /**< Whether server wants the client to implement the elevator algorithm */
GHashTable *children;
char pidfname[256]; /**< name of our PID file */
char pidftemplate[256]; /**< template to be used for the filename of the PID file */
{ "sparse_cow", FALSE, PARAM_BOOL, &(s.flags), F_SPARSE },
{ "sdp", FALSE, PARAM_BOOL, &(s.flags), F_SDP },
{ "sync", FALSE, PARAM_BOOL, &(s.flags), F_SYNC },
+ { "flush", FALSE, PARAM_BOOL, &(s.flags), F_FLUSH },
+ { "fua", FALSE, PARAM_BOOL, &(s.flags), F_FUA },
+ { "rotational", FALSE, PARAM_BOOL, &(s.flags), F_ROTATIONAL },
{ "listenaddr", FALSE, PARAM_STRING, &(s.listenaddr), 0 },
{ "maxconnections", FALSE, PARAM_INT, &(s.max_connections), 0 },
};
* @param client The client we're serving for
* @return The number of bytes actually written, or -1 in case of an error
**/
-ssize_t rawexpwrite(off_t a, char *buf, size_t len, CLIENT *client) {
+ssize_t rawexpwrite(off_t a, char *buf, size_t len, CLIENT *client, int fua) {
int fhandle;
off_t foffset;
size_t maxbytes;
if(maxbytes && len > maxbytes)
len = maxbytes;
- DEBUG4("(WRITE to fd %d offset %llu len %u), ", fhandle, foffset, len);
+ DEBUG5("(WRITE to fd %d offset %llu len %u fua %d), ", fhandle, foffset, 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
+ sync_file_range(fhandle, foffset, len,
+ SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE |
+ SYNC_FILE_RANGE_WAIT_AFTER);
+#else
+ fdatasync(fhandle);
+#endif
}
return retval;
}
* Call rawexpwrite repeatedly until all data has been written.
* @return 0 on success, nonzero on failure
**/
-int rawexpwrite_fully(off_t a, char *buf, size_t len, CLIENT *client) {
+int rawexpwrite_fully(off_t a, char *buf, size_t len, CLIENT *client, int fua) {
ssize_t ret=0;
- while(len > 0 && (ret=rawexpwrite(a, buf, len, client)) > 0 ) {
+ while(len > 0 && (ret=rawexpwrite(a, buf, len, client, fua)) > 0 ) {
a += ret;
buf += ret;
len -= ret;
* @param client The client we're going to write for.
* @return 0 on success, nonzero on failure
**/
-int expwrite(off_t a, char *buf, size_t len, CLIENT *client) {
+int expwrite(off_t a, char *buf, size_t len, CLIENT *client, int fua) {
char pagebuf[DIFFPAGESIZE];
off_t mapcnt,mapl,maph;
off_t wrlen,rdlen;
off_t offset;
if (!(client->server->flags & F_COPYONWRITE))
- return(rawexpwrite_fully(a, buf, len, client));
+ return(rawexpwrite_fully(a, buf, len, client, fua));
DEBUG3("Asked to write %d bytes at %llu.\n", len, (unsigned long long)a);
mapl=a/DIFFPAGESIZE ; maph=(a+len-1)/DIFFPAGESIZE ;
}
len-=wrlen ; a+=wrlen ; buf+=wrlen ;
}
+ if (client->server->flags & F_SYNC) {
+ fsync(client->difffile);
+ } else if (fua) {
+ /* open question: would it be cheaper to do multiple sync_file_ranges?
+ as we iterate through the above?
+ */
+ fdatasync(client->difffile);
+ }
+ return 0;
+}
+
+int expflush(CLIENT *client) {
+ int fhandle;
+ off_t foffset;
+ size_t maxbytes;
+ gint i;
+
+ if (client->server->flags & F_COPYONWRITE) {
+ return fsync(client->difffile);
+ }
+
+ for (i = 0; i < client->export->len; i++) {
+ FILE_INFO fi = g_array_index(client->export, FILE_INFO, i);
+ if (fsync(fi.fhandle) < 0)
+ return -1;
+ }
+
return 0;
}
err("Negotiation failed: %m");
if (client->server->flags & F_READONLY)
flags |= NBD_FLAG_READ_ONLY;
+ if (client->server->flags & F_FLUSH)
+ flags |= NBD_FLAG_SEND_FLUSH;
+ if (client->server->flags & F_FUA)
+ flags |= NBD_FLAG_SEND_FUA;
+ if (client->server->flags & F_ROTATIONAL)
+ flags |= NBD_FLAG_ROTATIONAL;
if (!client->modern) {
/* oldstyle */
flags = htonl(flags);
size_t len;
size_t currlen;
size_t writelen;
+ uint16_t command;
#ifdef DODBG
i++;
printf("%d: ", i);
readit(client->net, &request, sizeof(request));
request.from = ntohll(request.from);
request.type = ntohl(request.type);
+ command = request.type & NBD_CMD_MASK_COMMAND;
- if (request.type==NBD_CMD_DISC) {
+ if (command==NBD_CMD_DISC) {
msg2(LOG_INFO, "Disconnect request received.");
if (client->server->flags & F_COPYONWRITE) {
if (client->difmap) g_free(client->difmap) ;
currlen = len;
}
#ifdef DODBG
- printf("%s from %llu (%llu) len %d, ", request.type ? "WRITE" :
+ printf("%s from %llu (%llu) len %d, ", command ? "WRITE" :
"READ", (unsigned long long)request.from,
(unsigned long long)request.from / 512, len);
#endif
continue;
}
- if (request.type==NBD_CMD_WRITE) {
+ if (command==NBD_CMD_WRITE) {
DEBUG("wr: net->buf, ");
while(len > 0) {
readit(client->net, buf, currlen);
ERROR(client, reply, EPERM);
continue;
}
- if (expwrite(request.from, buf, len, client)) {
+ if (expwrite(request.from, buf, len, client,
+ request.type & NBD_CMD_FLAG_FUA)) {
DEBUG("Write failed: %m" );
ERROR(client, reply, errno);
continue;
}
continue;
}
- /* READ */
-
- DEBUG("exp->buf, ");
- memcpy(buf, &reply, sizeof(struct nbd_reply));
- p = buf + sizeof(struct nbd_reply);
- writelen = currlen + sizeof(struct nbd_reply);
- while(len > 0) {
- if (expread(request.from, p, currlen, client)) {
- DEBUG("Read failed: %m");
+
+ if (command==NBD_CMD_FLUSH) {
+ DEBUG("fl: ");
+ if (expflush(client)) {
+ DEBUG("Flush failed: %m");
ERROR(client, reply, errno);
continue;
}
+ SEND(client->net, reply);
+ DEBUG("OK!\n");
+ continue;
+ }
- DEBUG("buf->net, ");
- writeit(client->net, buf, writelen);
- len -= currlen;
- currlen = (len < BUFSIZE) ? len : BUFSIZE;
- p = buf;
- writelen = currlen;
+ if (command==NBD_CMD_READ) {
+ DEBUG("exp->buf, ");
+ memcpy(buf, &reply, sizeof(struct nbd_reply));
+ p = buf + sizeof(struct nbd_reply);
+ writelen = currlen + sizeof(struct nbd_reply);
+ while(len > 0) {
+ if (expread(request.from, p, currlen, client)) {
+ DEBUG("Read failed: %m");
+ ERROR(client, reply, errno);
+ continue;
+ }
+
+ DEBUG("buf->net, ");
+ writeit(client->net, buf, writelen);
+ len -= currlen;
+ currlen = (len < BUFSIZE) ? len : BUFSIZE;
+ p = buf;
+ writelen = currlen;
+ }
+ DEBUG("OK!\n");
+ continue;
}
- DEBUG("OK!\n");
+
+ DEBUG ("Ignoring unknown command\n");
}
return 0;
}
# Yes, that's POSIX sh, not bash!
tmpnam=`mktemp`
+conffile=${tmpnam}.conf
+pidfile=${tmpnam}.pid
ulimit -c unlimited
case $1 in
*/cmd)
# Test with export specified on command line
- ./nbd-server -C /dev/null -p `pwd`/nbd-server.pid 11111 $tmpnam &
+ ./nbd-server -C /dev/null -p ${pidfile} 11111 $tmpnam &
# -p only works if nbd-server wasn't compiled with -DNOFORK or
# -DNODAEMON, which I sometimes do for testing and debugging.
PID=$!
;;
*/cfgsize)
# Test oversized requests
- ./nbd-server -C /dev/null -p `pwd`/nbd-server.pid 11111 $tmpnam &
+ ./nbd-server -C /dev/null -p ${pidfile} 11111 $tmpnam &
# -p only works if nbd-server wasn't compiled with -DNOFORK or
# -DNODAEMON, which I sometimes do for testing and debugging.
PID=$!
;;
*/cfg1)
# Test with export specified in config file
- cat > nbd-server.conf <<EOF
+ cat > ${conffile} <<EOF
[generic]
oldstyle = true
[export]
exportname = $tmpnam
port = 11112
EOF
- ./nbd-server -C nbd-server.conf -p `pwd`/nbd-server.pid &
+ ./nbd-server -C ${conffile} -p ${pidfile} &
PID=$!
sleep 1
./nbd-tester-client 127.0.0.1 11112
*/cfgmulti)
# Test with multiple exports specified in config file, and
# testing more options too
- cat >nbd-server.conf <<EOF
+ cat >${conffile} <<EOF
[generic]
oldstyle = true
[export1]
readonly = true
listenaddr = 127.0.0.1
EOF
- ./nbd-server -C nbd-server.conf -p `pwd`/nbd-server.pid &
+ ./nbd-server -C ${conffile} -p ${pidfile} &
PID=$!
sleep 1
./nbd-tester-client localhost 11113
retval=$?
if [ $retval -ne 0 ]
then
- if [ -f nbd-server.pid ]
+ if [ -f ${pidfile} ]
then
- kill `cat nbd-server.pid`
- rm -f nbd-server.pid
+ kill `cat ${pidfile}`
+ rm -f ${pidfile}
else
kill $PID
fi
if [ -z "$2" ]
then
- rm -f $tmpnam nbd-server.conf
+ rm -f $tmpnam ${conffile}
fi
exit $retval
fi
;;
*/cfgnew)
# Test new-style exports
- cat >nbd-server.conf <<EOF
+ cat >${conffile} <<EOF
[generic]
[export1]
exportname = $tmpnam
EOF
- ./nbd-server -C nbd-server.conf -p `pwd`/nbd-server.pid &
+ ./nbd-server -C ${conffile} -p ${pidfile} &
PID=$!
sleep 1
./nbd-tester-client localhost -N export1
retval=$?
;;
*/write)
- # Test new-style exports
- cat >nbd-server.conf <<EOF
+ # Test writing
+ cat >${conffile} <<EOF
[generic]
[export1]
exportname = $tmpnam
EOF
- ./nbd-server -C nbd-server.conf -p `pwd`/nbd-server.pid &
+ ./nbd-server -C ${conffile} -p ${pidfile} &
PID=$!
sleep 1
./nbd-tester-client localhost -N export1 -w
retval=$?
;;
*/flush)
- # Test new-style exports
- cat >nbd-server.conf <<EOF
+ # Test writes with flush
+ cat >${conffile} <<EOF
[generic]
[export1]
exportname = $tmpnam
+ flush = true
+ fua = true
+ rotational = true
EOF
- ./nbd-server -C nbd-server.conf -p `pwd`/nbd-server.pid &
+ ./nbd-server -C ${conffile} -p ${pidfile} &
PID=$!
sleep 1
./nbd-tester-client localhost -N export1 -w -f
exit 1
;;
esac
-if [ -f nbd-server.pid ]
+if [ -f ${pidfile} ]
then
- kill `cat nbd-server.pid`
- rm -f nbd-server.pid
+ kill `cat ${pidfile}`
+ rm -f ${pidfile}
else
kill $PID
fi
if [ -z "$2" ]
then
- rm -f $tmpnam nbd-server.conf
+ rm -f $tmpnam ${conffile}
fi
if [ $retval -ne 0 ]
then
projects / nbd.git / commitdiff