[PATCH] tcmu: wait for nl reply only if there are listeners

[PATCH] tcmu: wait for nl reply only if there are listeners

[Mike Christie]

genlmsg_multicast_allns used to return -ESRCH even if the message was
successfully sent to a listener. With commit:

commit cb9f7a9a5c96a773bbc9c70660dc600cfff82f82
Author: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Date:   Tue Feb 6 14:48:32 2018 +0100

    netlink: ensure to loop over all netns in genlmsg_multicast_allns()

it now will return success if the message was sent to a listener. With
that patch, tcmu can now immediately fail if -ESRCH is returned because
we know there will be no reply.

Signed-off-by: Mike Christie <mchristi@redhat.com>

---
 drivers/target/target_core_user.c | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/drivers/target/target_core_user.c b/drivers/target/target_core_user.c
index 5831e0e..dccc13c 100644
--- a/drivers/target/target_core_user.c
+++ b/drivers/target/target_core_user.c
@@ -1794,9 +1794,6 @@ static int tcmu_netlink_event_send(struct tcmu_dev *udev,
 
 	ret = genlmsg_multicast_allns(&tcmu_genl_family, skb, 0,
 				      TCMU_MCGRP_CONFIG, GFP_KERNEL);
-	/* We don't care if no one is listening */
-	if (ret = -ESRCH)
-		ret = 0;
 	if (!ret)
 		ret = tcmu_wait_genl_cmd_reply(udev);
 	return ret;
-- 
1.8.3.1

Re: [PATCH] tcmu: wait for nl reply only if there are listeners

[Mike Christie]

Martin, I am self nacking this patch for now.


On 02/01/2019 09:01 PM, Mike Christie wrote:
> genlmsg_multicast_allns used to return -ESRCH even if the message was
> successfully sent to a listener. With commit:
> 
> commit cb9f7a9a5c96a773bbc9c70660dc600cfff82f82
> Author: Nicolas Dichtel <nicolas.dichtel@6wind.com>
> Date:   Tue Feb 6 14:48:32 2018 +0100
> 
>     netlink: ensure to loop over all netns in genlmsg_multicast_allns()
> 
> it now will return success if the message was sent to a listener. With
> that patch, tcmu can now immediately fail if -ESRCH is returned because
> we know there will be no reply.
> 
> Signed-off-by: Mike Christie <mchristi@redhat.com>
> 
> ---
>  drivers/target/target_core_user.c | 3 ---
>  1 file changed, 3 deletions(-)
> 
> diff --git a/drivers/target/target_core_user.c b/drivers/target/target_core_user.c
> index 5831e0e..dccc13c 100644
> --- a/drivers/target/target_core_user.c
> +++ b/drivers/target/target_core_user.c
> @@ -1794,9 +1794,6 @@ static int tcmu_netlink_event_send(struct tcmu_dev *udev,
>  
>  	ret = genlmsg_multicast_allns(&tcmu_genl_family, skb, 0,
>  				      TCMU_MCGRP_CONFIG, GFP_KERNEL);
> -	/* We don't care if no one is listening */
> -	if (ret = -ESRCH)
> -		ret = 0;
>  	if (!ret)
>  		ret = tcmu_wait_genl_cmd_reply(udev);
>  	return ret;
> 

[PATCH] tcmu: wait for nl reply only if there are listeners or during an add

[Cathy Avery]

genlmsg_multicast_allns now returns the correct statuses when a
message is sent to a listener. However in the case of adding a
device we want to wait for the listener otherwise we may miss the
the device during startup.

Signed-off-by: Cathy Avery <cavery@redhat.com>
---
 drivers/target/target_core_user.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/drivers/target/target_core_user.c b/drivers/target/target_core_user.c
index 5831e0eecea1..2d3f8f5403d8 100644
--- a/drivers/target/target_core_user.c
+++ b/drivers/target/target_core_user.c
@@ -1794,11 +1794,12 @@ static int tcmu_netlink_event_send(struct tcmu_dev *udev,
 
 	ret = genlmsg_multicast_allns(&tcmu_genl_family, skb, 0,
 				      TCMU_MCGRP_CONFIG, GFP_KERNEL);
-	/* We don't care if no one is listening */
-	if (ret = -ESRCH)
-		ret = 0;
-	if (!ret)
-		ret = tcmu_wait_genl_cmd_reply(udev);
+
+	/* Wait during an add as the listener may not be up yet */
+	if (ret = 0 ||
+	   (ret = -ESRCH && cmd = TCMU_CMD_ADDED_DEVICE))
+		return tcmu_wait_genl_cmd_reply(udev);
+
 	return ret;
 }
 
-- 
2.19.1

Re: [PATCH] tcmu: wait for nl reply only if there are listeners or during an add

[Mike Christie]

On 02/28/2019 01:28 PM, Cathy Avery wrote:
> genlmsg_multicast_allns now returns the correct statuses when a
> message is sent to a listener. However in the case of adding a
> device we want to wait for the listener otherwise we may miss the
> the device during startup.
> 
> Signed-off-by: Cathy Avery <cavery@redhat.com>
> ---
>  drivers/target/target_core_user.c | 11 ++++++-----
>  1 file changed, 6 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/target/target_core_user.c b/drivers/target/target_core_user.c
> index 5831e0eecea1..2d3f8f5403d8 100644
> --- a/drivers/target/target_core_user.c
> +++ b/drivers/target/target_core_user.c
> @@ -1794,11 +1794,12 @@ static int tcmu_netlink_event_send(struct tcmu_dev *udev,
>  
>  	ret = genlmsg_multicast_allns(&tcmu_genl_family, skb, 0,
>  				      TCMU_MCGRP_CONFIG, GFP_KERNEL);
> -	/* We don't care if no one is listening */
> -	if (ret = -ESRCH)
> -		ret = 0;
> -	if (!ret)
> -		ret = tcmu_wait_genl_cmd_reply(udev);
> +
> +	/* Wait during an add as the listener may not be up yet */
> +	if (ret = 0 ||
> +	   (ret = -ESRCH && cmd = TCMU_CMD_ADDED_DEVICE))
> +		return tcmu_wait_genl_cmd_reply(udev);
> +
>  	return ret;
>  }

Looks ok to me.

Acked-by: Mike Christie <mchristi@redhat.com>

Re: [PATCH] tcmu: wait for nl reply only if there are listeners or during an add

[Martin K. Petersen]

Cathy,

> genlmsg_multicast_allns now returns the correct statuses when a
> message is sent to a listener. However in the case of adding a
> device we want to wait for the listener otherwise we may miss the
> the device during startup.

Applied to 5.1/scsi-queue, thanks!

-- 
Martin K. Petersen	Oracle Linux Engineering