Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Wed, 12 May 2021 03:00:49 +0200
From:   Krzysztof =?utf-8?Q?Wilczy=C5=84ski?= <kw@linux.com>
To:     Rajat Jain <rajatja@google.com>
Cc:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        "Rafael J. Wysocki" <rafael@kernel.org>,
        Bjorn Helgaas <bhelgaas@google.com>,
        Alan Stern <stern@rowland.harvard.edu>,
        linux-kernel@vger.kernel.org, linux-pci@vger.kernel.org,
        linux-usb@vger.kernel.org, helgaas@kernel.org,
        rajatxjain@gmail.com, jsbarnes@google.com, dtor@google.com
Subject: Re: [PATCH v2 1/2] driver core: Move the "removable" attribute from
 USB to core
Message-ID: <20210512010049.GA89346@rocinante.localdomain>
References: <20210424021631.1972022-1-rajatja@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <20210424021631.1972022-1-rajatja@google.com>
Precedence: bulk

Hi Rajat,

I have few questions below, but to add in advance, I might be confusing
the role that "type->supports_removable" and "dev->removable" plays
here, and if so then I apologise.

[...]
> @@ -2504,8 +2523,16 @@ static int device_add_attrs(struct device *dev)
>  			goto err_remove_dev_online;
>  	}
>  
> +	if (type && type->supports_removable) {
> +		error = device_create_file(dev, &dev_attr_removable);
> +		if (error)
> +			goto err_remove_dev_waiting_for_supplier;
> +	}
> +
>  	return 0;

Would a check for "dev->removable == DEVICE_REMOVABLE" here be more
appropriate?

Unless you wanted to add sysfs objects when the device hints that it has
a notion of being removable even though it might be set to "unknown" or
"fixed" (if that state is at all possible then), and in which case using
the dev_is_removable() helper would also not be an option since it does
a more complex check internally.

Technically, you could always add this sysfs object (similarly to what
USB core did) as it would then show the correct state depending on
"dev->removable".

Also, I suppose, it's not possible for a device to have
"supports_removable" set to true, but "removable" would be different
than "DEVICE_REMOVABLE", correct?

[...]
> +enum device_removable {
> +	DEVICE_REMOVABLE_UNKNOWN = 0,
> +	DEVICE_REMOVABLE,
> +	DEVICE_FIXED,
> +};

I know this was moved from the USB core, but I personally find it
a little bit awkward to read, would something like that be acceptable?

enum device_removable {
	DEVICE_STATE_UNKNOWN = 0,
	DEVICE_STATE_REMOVABLE,
	DEVICE_STATE_FIXED,
};

The addition of state to the name follows the removable_show() function
where the local variable is called "state", and I think it makes sense
to call this as such.  What do you think?

> +static inline bool dev_is_removable(struct device *dev)
> +{
> +	return dev && dev->type && dev->type->supports_removable
> +	    && dev->removable == DEVICE_REMOVABLE;
> +}

Similarly to my question about - would a simple check to see if
"dev->removable" is set to "DEVICE_REMOVABLE" here be enough?

Krzysztof