2018-12-11 06:37:47

by Medad Young

Subject: [PATCH] nbd:clear NBD_BOUND flag when NBD connection is closed

From: Medad <[email protected]>

If we do NOT clear NBD_BOUND flag when NBD connection is closed,
then the original NBD device could not be used again.

Signed-off-by: Medad <[email protected]>
---
drivers/block/nbd.c | 11 ++++++++---
1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
index 64278f4..5c88490 100644
--- a/drivers/block/nbd.c
+++ b/drivers/block/nbd.c
@@ -277,10 +277,14 @@ static void sock_shutdown(struct nbd_device *nbd)
struct nbd_config *config = nbd->config;
int i;

- if (config->num_connections == 0)
+ if (config->num_connections == 0) {
+ clear_bit(NBD_BOUND, &config->runtime_flags);
return;
- if (test_and_set_bit(NBD_DISCONNECTED, &config->runtime_flags))
+ }
+ if (test_and_set_bit(NBD_DISCONNECTED, &config->runtime_flags)) {
+ clear_bit(NBD_BOUND, &config->runtime_flags);
return;
+ }

for (i = 0; i < config->num_connections; i++) {
struct nbd_sock *nsock = config->socks[i];
@@ -944,7 +948,7 @@ static int nbd_reconnect_socket(struct nbd_device *nbd, unsigned long arg)
sockfd_put(old);

clear_bit(NBD_DISCONNECTED, &config->runtime_flags);
-
+ clear_bit(NBD_BOUND, &config->runtime_flags);
/* We take the tx_mutex in an error path in the recv_work, so we
* need to queue_work outside of the tx_mutex.
*/
@@ -1020,6 +1024,7 @@ static int nbd_disconnect(struct nbd_device *nbd)
dev_info(disk_to_dev(nbd->disk), "NBD_DISCONNECT\n");
set_bit(NBD_DISCONNECT_REQUESTED, &config->runtime_flags);
send_disconnects(nbd);
+ clear_bit(NBD_BOUND, &config->runtime_flags);
return 0;
}

--
2.7.4

2018-12-13 14:19:32

by Adriana Kobylak

Subject: Re: [PATCH] nbd:clear NBD_BOUND flag when NBD connection is closed

On 2018-12-11 00:17, [email protected] wrote:
> From: Medad <[email protected]>
>
> If we do NOT clear NBD_BOUND flag when NBD connection is closed,
> then the original NBD device could not be used again.
>
> Signed-off-by: Medad <[email protected]>

Tested on Witherspoon (OpenPOWER P9 with AST2500) and verified with
this patch that the NBD device could be disconnected/reconnected.

Tested by: Adriana Kobylak <[email protected]>

> ---
> drivers/block/nbd.c | 11 ++++++++---
> 1 file changed, 8 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
> index 64278f4..5c88490 100644
> --- a/drivers/block/nbd.c
> +++ b/drivers/block/nbd.c
> @@ -277,10 +277,14 @@ static void sock_shutdown(struct nbd_device *nbd)
> struct nbd_config *config = nbd->config;
> int i;
>
> - if (config->num_connections == 0)
> + if (config->num_connections == 0) {
> + clear_bit(NBD_BOUND, &config->runtime_flags);
> return;
> - if (test_and_set_bit(NBD_DISCONNECTED, &config->runtime_flags))
> + }
> + if (test_and_set_bit(NBD_DISCONNECTED, &config->runtime_flags)) {
> + clear_bit(NBD_BOUND, &config->runtime_flags);
> return;
> + }
>
> for (i = 0; i < config->num_connections; i++) {
> struct nbd_sock *nsock = config->socks[i];
> @@ -944,7 +948,7 @@ static int nbd_reconnect_socket(struct nbd_device
> *nbd, unsigned long arg)
> sockfd_put(old);
>
> clear_bit(NBD_DISCONNECTED, &config->runtime_flags);
> -
> + clear_bit(NBD_BOUND, &config->runtime_flags);
> /* We take the tx_mutex in an error path in the recv_work, so we
> * need to queue_work outside of the tx_mutex.
> */
> @@ -1020,6 +1024,7 @@ static int nbd_disconnect(struct nbd_device *nbd)
> dev_info(disk_to_dev(nbd->disk), "NBD_DISCONNECT\n");
> set_bit(NBD_DISCONNECT_REQUESTED, &config->runtime_flags);
> send_disconnects(nbd);
> + clear_bit(NBD_BOUND, &config->runtime_flags);
> return 0;
> }