nbd: allow seamless reconnects
THIS SHOULD NEVER GO UPSTREAM.
All in all this is ok, the only questionable thing is the abuse of the
server flags so we know the client cares about having things retried.
This patch will come back when the netlink interface is in place, so
basically the same just minus the special flag.
Signed-off-by: Josef Bacik <jbacik@fb.com>
diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
index 52c787f..e0d770c 100644
--- a/drivers/block/nbd.c
+++ b/drivers/block/nbd.c
@@ -55,6 +55,7 @@ struct nbd_sock {
#define NBD_DISCONNECT_REQUESTED 1
#define NBD_DISCONNECTED 2
#define NBD_HAS_SOCKS_REF 3
+#define NBD_RECONNECT 4
struct nbd_device {
u32 flags;
@@ -66,6 +67,7 @@ struct nbd_device {
wait_queue_head_t socks_wq;
int num_connections;
+ wait_queue_head_t reconn_wq;
int magic;
struct blk_mq_tag_set tag_set;
@@ -88,6 +90,7 @@ struct nbd_device {
struct nbd_cmd {
struct nbd_device *nbd;
int index;
+ int timeouts;
struct completion send_complete;
};
@@ -129,6 +132,25 @@ static const char *nbdcmd_to_ascii(int cmd)
return "invalid";
}
+static int nbd_disconnected(struct nbd_device *nbd)
+{
+ return test_bit(NBD_DISCONNECTED, &nbd->runtime_flags) ||
+ test_bit(NBD_DISCONNECT_REQUESTED, &nbd->runtime_flags);
+}
+
+static int need_reconnect(struct nbd_device *nbd)
+{
+ if (nbd_disconnected(nbd))
+ return 0;
+ if (test_bit(NBD_RECONNECT, &nbd->runtime_flags))
+ return 1;
+ if (nbd->flags & NBD_FLAG_CAN_RECONNECT) {
+ set_bit(NBD_RECONNECT, &nbd->runtime_flags);
+ return 1;
+ }
+ return 0;
+}
+
static int nbd_socks_get_unless_zero(struct nbd_device *nbd)
{
return atomic_add_unless(&nbd->socks_ref, 1, 0);
@@ -150,6 +172,15 @@ static void nbd_socks_put(struct nbd_device *nbd)
}
}
+static int nbd_clear_reconnect(struct nbd_device *nbd)
+{
+ if (test_and_clear_bit(NBD_RECONNECT, &nbd->runtime_flags)) {
+ wake_up(&nbd->reconn_wq);
+ return 1;
+ }
+ return 0;
+}
+
static int nbd_size_clear(struct nbd_device *nbd, struct block_device *bdev)
{
bd_set_size(bdev, 0);
@@ -209,6 +240,7 @@ static void sock_shutdown(struct nbd_device *nbd)
nsock->dead = true;
}
dev_warn(disk_to_dev(nbd->disk), "shutting down sockets\n");
+ nbd_clear_reconnect(nbd);
}
static enum blk_eh_timer_return nbd_xmit_timeout(struct request *req,
@@ -222,26 +254,12 @@ static enum blk_eh_timer_return nbd_xmit_timeout(struct request *req,
return BLK_EH_HANDLED;
}
- if (nbd->num_connections > 1) {
+ if (nbd->num_connections > 1 && cmd->timeouts++ < 1) {
dev_err_ratelimited(nbd_to_dev(nbd),
- "Connection timed out, retrying\n");
- /*
- * Hooray we have more connections, requeue this IO, the submit
- * path will put it on a real connection.
- */
- if (nbd->socks && nbd->num_connections > 1) {
- if (cmd->index < nbd->num_connections) {
- struct nbd_sock *nsock =
- nbd->socks[cmd->index];
- mutex_lock(&nsock->tx_lock);
- nsock->dead = true;
- kernel_sock_shutdown(nsock->sock, SHUT_RDWR);
- mutex_unlock(&nsock->tx_lock);
- }
- blk_mq_requeue_request(req, true);
- nbd_socks_put(nbd);
- return BLK_EH_RESET_TIMER;
- }
+ "Connection timed out, retrying %p\n", cmd);
+ blk_mq_requeue_request(req, true);
+ nbd_socks_put(nbd);
+ return BLK_EH_NOT_HANDLED;
} else {
dev_err_ratelimited(nbd_to_dev(nbd),
"Connection timed out\n");
@@ -418,12 +436,6 @@ static inline int sock_recv_bvec(struct nbd_device *nbd, int index,
return result;
}
-static int nbd_disconnected(struct nbd_device *nbd)
-{
- return test_bit(NBD_DISCONNECTED, &nbd->runtime_flags) ||
- test_bit(NBD_DISCONNECT_REQUESTED, &nbd->runtime_flags);
-}
-
/* NULL returned = something went wrong, inform userspace */
static struct nbd_cmd *nbd_read_stat(struct nbd_device *nbd, int index)
{
@@ -611,11 +623,30 @@ static int find_fallback(struct nbd_device *nbd, int index)
return new_index;
}
+static int wait_reconnect(struct nbd_cmd *cmd)
+{
+ struct nbd_device *nbd = cmd->nbd;
+ long timeout = nbd->tag_set.timeout;
+ int ret;
+
+ do {
+ printk(KERN_ERR "waiting on reconnect for %p, timeouts %d\n", cmd, cmd->timeouts);
+ ret = wait_event_timeout(nbd->reconn_wq,
+ !test_bit(NBD_RECONNECT,
+ &nbd->runtime_flags),
+ timeout);
+ printk(KERN_ERR "finished waiting to reconnect for %p, ret %d\n", cmd, ret);
+ } while (!ret && cmd->timeouts++ < 1);
+
+ return test_bit(NBD_RECONNECT, &nbd->runtime_flags);
+}
+
static int nbd_handle_cmd(struct nbd_cmd *cmd, int index)
{
struct request *req = blk_mq_rq_from_pdu(cmd);
struct nbd_device *nbd = cmd->nbd;
struct nbd_sock *nsock;
+ int orig_index = index;
int ret;
if (!nbd_socks_get_unless_zero(nbd)) {
@@ -636,6 +667,10 @@ static int nbd_handle_cmd(struct nbd_cmd *cmd, int index)
if (nsock->dead) {
index = find_fallback(nbd, index);
if (index < 0) {
+ if (need_reconnect(nbd) && !wait_reconnect(cmd)) {
+ index = orig_index;
+ goto again;
+ }
nbd_socks_put(nbd);
return -EIO;
}
@@ -732,6 +767,28 @@ static int nbd_add_socket(struct nbd_device *nbd, struct block_device *bdev,
goto out;
}
+ if (unlikely(test_bit(NBD_RECONNECT, &nbd->runtime_flags))) {
+ int i, found = 0;
+
+ for (i = 0; i < nbd->num_connections; i++) {
+ if (nbd->socks[i]->dead) {
+ printk(KERN_ERR "JOSEF: reconnecting sock %d\n", i);
+ nbd->socks[i]->fallback_index = -1;
+ nbd->socks[i]->sock = sock;
+ nbd->socks[i]->dead = false;
+ found = 1;
+ break;
+ }
+ }
+
+ if (!found)
+ dev_err(disk_to_dev(nbd->disk),
+ "No dead connections to reconnect\n");
+ else
+ err = 0;
+ goto out;
+ }
+
err = -ENOMEM;
socks = krealloc(nbd->socks, (nbd->num_connections + 1) *
sizeof(struct nbd_sock *), GFP_KERNEL);
@@ -814,13 +871,10 @@ static int nbd_disconnect(struct nbd_device *nbd, struct block_device *bdev)
if (!nbd_socks_get_unless_zero(nbd))
return -EINVAL;
- mutex_unlock(&nbd->config_lock);
- fsync_bdev(bdev);
- mutex_lock(&nbd->config_lock);
-
if (!test_and_set_bit(NBD_DISCONNECT_REQUESTED,
&nbd->runtime_flags))
send_disconnects(nbd);
+ nbd_clear_reconnect(nbd);
nbd_socks_put(nbd);
return 0;
}
@@ -843,10 +897,14 @@ static int nbd_start_device(struct nbd_device *nbd, struct block_device *bdev)
int num_connections = nbd->num_connections;
int error = 0, i;
- if (nbd->task_recv)
+ if (nbd->task_recv) {
+ nbd_clear_reconnect(nbd);
return -EBUSY;
- if (!nbd->socks)
+ }
+ if (!nbd->socks) {
+ nbd_clear_reconnect(nbd);
return -EINVAL;
+ }
if (num_connections > 1 &&
!(nbd->flags & NBD_FLAG_CAN_MULTI_CONN)) {
dev_err(disk_to_dev(nbd->disk), "server does not support multiple connections per device.\n");
@@ -854,13 +912,13 @@ static int nbd_start_device(struct nbd_device *nbd, struct block_device *bdev)
goto out_err;
}
- blk_mq_update_nr_hw_queues(&nbd->tag_set, nbd->num_connections);
args = kcalloc(num_connections, sizeof(*args), GFP_KERNEL);
if (!args) {
error = -ENOMEM;
goto out_err;
}
nbd->task_recv = current;
+ nbd->task_setup = current;
mutex_unlock(&nbd->config_lock);
nbd_parse_flags(nbd, bdev);
@@ -883,13 +941,25 @@ static int nbd_start_device(struct nbd_device *nbd, struct block_device *bdev)
args[i].index = i;
queue_work(recv_workqueue, &args[i].work);
}
+
+ if (!nbd_clear_reconnect(nbd))
+ blk_mq_update_nr_hw_queues(&nbd->tag_set,
+ nbd->num_connections);
wait_event_interruptible(nbd->recv_wq,
atomic_read(&nbd->recv_threads) == 0);
for (i = 0; i < num_connections; i++)
flush_work(&args[i].work);
nbd_dev_dbg_close(nbd);
- nbd_size_clear(nbd, bdev);
device_remove_file(disk_to_dev(nbd->disk), &pid_attr);
+ if (nbd->flags & NBD_FLAG_CAN_RECONNECT &&
+ !test_bit(NBD_DISCONNECT_REQUESTED, &nbd->runtime_flags)) {
+ set_bit(NBD_RECONNECT, &nbd->runtime_flags);
+ mutex_lock(&nbd->config_lock);
+ nbd->task_recv = NULL;
+ printk(KERN_ERR "JOSEF: TRYING RECONNECT\n");
+ return -EAGAIN;
+ }
+ nbd_size_clear(nbd, bdev);
out_recv:
mutex_lock(&nbd->config_lock);
nbd->task_recv = NULL;
@@ -903,6 +973,7 @@ static int nbd_start_device(struct nbd_device *nbd, struct block_device *bdev)
error = -ETIMEDOUT;
nbd_reset(nbd);
+ printk(KERN_ERR "JOSEF: DOIT RETURNING %d, flags %u\n", error, nbd->flags);
return error;
}
@@ -1110,6 +1181,7 @@ static int nbd_init_request(void *data, struct request *rq,
{
struct nbd_cmd *cmd = blk_mq_rq_to_pdu(rq);
cmd->nbd = data;
+ cmd->timeouts = 0;
return 0;
}
@@ -1203,6 +1275,7 @@ static int nbd_dev_add(int index)
sprintf(disk->disk_name, "nbd%d", index);
init_waitqueue_head(&nbd->recv_wq);
init_waitqueue_head(&nbd->socks_wq);
+ init_waitqueue_head(&nbd->reconn_wq);
nbd_reset(nbd);
add_disk(disk);
return index;
diff --git a/include/uapi/linux/nbd.h b/include/uapi/linux/nbd.h
index c91c642..66e5ec7 100644
--- a/include/uapi/linux/nbd.h
+++ b/include/uapi/linux/nbd.h
@@ -44,7 +44,7 @@ enum {
/* there is a gap here to match userspace */
#define NBD_FLAG_SEND_TRIM (1 << 5) /* send trim/discard */
#define NBD_FLAG_CAN_MULTI_CONN (1 << 8) /* Server supports multiple connections per export. */
-
+#define NBD_FLAG_CAN_RECONNECT (1 << 20)
/* userspace doesn't need the nbd_device structure */
/* These are sent over the network in the request/reply magic fields */
