From mboxrd@z Thu Jan  1 00:00:00 1970
From: hch@lst.de (Christoph Hellwig)
Date: Tue, 17 Apr 2018 17:28:15 +0200
Subject: [PATCH v1 3/3] nvmet: Add port transport state flag
In-Reply-To: <5f7a6f5e-cddc-0612-ab29-0b2a34a0dd76@grimberg.me>
References: <20180412080656.1691-1-sagi@grimberg.me>
 <20180412080656.1691-4-sagi@grimberg.me> <20180413171421.GA23432@lst.de>
 <5f7a6f5e-cddc-0612-ab29-0b2a34a0dd76@grimberg.me>
Message-ID: <20180417152815.GE25823@lst.de>

On Sun, Apr 15, 2018@11:54:59AM +0300, Sagi Grimberg wrote:
>
>
>>> +{
>>> +	struct nvmet_port *port = to_nvmet_port(item);
>>> +
>>> +	return sprintf(page, "%s\n",
>>> +		nvmet_is_port_active(port) ? "up" : "down");
>>> +}
>>> +
>>> +CONFIGFS_ATTR_RO(nvmet_, trstate);
>>> +
>>>   /*
>>>    * Namespace structures & file operation functions below
>>>    */
>>> @@ -870,6 +880,7 @@ static struct configfs_attribute *nvmet_port_attrs[] = {
>>>   	&nvmet_attr_addr_traddr,
>>>   	&nvmet_attr_addr_trsvcid,
>>>   	&nvmet_attr_addr_trtype,
>>> +	&nvmet_attr_trstate,
>>
>> Please don't create attributes without a group ever.  Just create a
>> different group which isn't serialized, e.g. state as the group name
>> and then transport as the attribute name inside it, or something similar.
>
> We already have ns enable which is not within a group. How is that
> different?

enable is a magic trigger that we can't handle automatically when
restoring the state in nvmetcli.  trstate is a read-only field that
make sense to show with generic code.