[PATCH] nvme: make ANA support independent on native nvme multipath

[PATCH] nvme: make ANA support independent on native nvme multipath

[Hannes Reinecke]

NVMe native multipathing is an implementation detail on the host,
and ANA support is actually independent on that.
So we shouldn't check for native NVMe multipathing when trying to
evaluate whether ANA is supported; not doing so results in the
ANA sysfs attributes not to be present when the 'multipath=false'
module option is set.

Signed-off-by: Hannes Reinecke <hare at suse.com>
---
 drivers/nvme/host/multipath.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c
index 4944ffdf6831..f7efcaeb2d98 100644
--- a/drivers/nvme/host/multipath.c
+++ b/drivers/nvme/host/multipath.c
@@ -22,7 +22,7 @@ MODULE_PARM_DESC(multipath,
 
 inline bool nvme_ctrl_use_ana(struct nvme_ctrl *ctrl)
 {
-	return multipath && ctrl->subsys && (ctrl->subsys->cmic & (1 << 3));
+	return ctrl->subsys && (ctrl->subsys->cmic & (1 << 3));
 }
 
 /*
-- 
2.16.4

[PATCH] nvme: make ANA support independent on native nvme multipath

[Sagi Grimberg]

Reviewed-by: Sagi Grimberg <sagi at grimberg.me>

[PATCH] nvme: make ANA support independent on native nvme multipath

[Christoph Hellwig]

On Mon, Nov 05, 2018@04:41:36PM +0100, Hannes Reinecke wrote:
> NVMe native multipathing is an implementation detail on the host,
> and ANA support is actually independent on that.
> So we shouldn't check for native NVMe multipathing when trying to
> evaluate whether ANA is supported; not doing so results in the
> ANA sysfs attributes not to be present when the 'multipath=false'
> module option is set.

No, we should not enable ANA support without multipathing, as we
don't want people to rely on any ANA related side effects if multipathing
is not enabled.

nvme: make ANA support independent on native nvme multipath

[Mike Snitzer]

On Mon, Nov 05 2018 at 10:41am -0500,
Hannes Reinecke <hare@suse.de> wrote:

> NVMe native multipathing is an implementation detail on the host,
> and ANA support is actually independent on that.
> So we shouldn't check for native NVMe multipathing when trying to
> evaluate whether ANA is supported; not doing so results in the
> ANA sysfs attributes not to be present when the 'multipath=false'
> module option is set.
> 
> Signed-off-by: Hannes Reinecke <hare at suse.com>
> ---
>  drivers/nvme/host/multipath.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c
> index 4944ffdf6831..f7efcaeb2d98 100644
> --- a/drivers/nvme/host/multipath.c
> +++ b/drivers/nvme/host/multipath.c
> @@ -22,7 +22,7 @@ MODULE_PARM_DESC(multipath,
>  
>  inline bool nvme_ctrl_use_ana(struct nvme_ctrl *ctrl)
>  {
> -	return multipath && ctrl->subsys && (ctrl->subsys->cmic & (1 << 3));
> +	return ctrl->subsys && (ctrl->subsys->cmic & (1 << 3));
>  }
>  
>  /*
> -- 
> 2.16.4
> 

So you're going for ANA to work with CONFIG_NVME_MULTIPATH=Y but
nvme_core.multipath=N

Yes please.

Acked-by: Mike Snitzer <snitzer at redhat.com>

nvme: make ANA support independent on native nvme multipath

[Mike Snitzer]

On Thu, Nov 08 2018 at  3:48am -0500,
Christoph Hellwig <hch@lst.de> wrote:

> On Mon, Nov 05, 2018@04:41:36PM +0100, Hannes Reinecke wrote:
> > NVMe native multipathing is an implementation detail on the host,
> > and ANA support is actually independent on that.
> > So we shouldn't check for native NVMe multipathing when trying to
> > evaluate whether ANA is supported; not doing so results in the
> > ANA sysfs attributes not to be present when the 'multipath=false'
> > module option is set.
> 
> No, we should not enable ANA support without multipathing, as we
> don't want people to rely on any ANA related side effects if multipathing
> is not enabled.

Could you be more specific?

Do you have an example side-effect that you hold to be problematic?

As you can imagine I've long hoped for ANA to be decoupled from
multipathing.  Would welcome your throughts on why this is "bad".

Mike