Re: [07/18] thunderbolt: Handle rejected Thunderbolt devices

From: Jeremy McNicoll <jmcnicol@redhat.com>
To: Mika Westerberg <mika.westerberg@linux.intel.com>
Cc: linux-kernel@vger.kernel.org,
	Andreas Noever <andreas.noever@gmail.com>,
	Michael Jamet <michael.jamet@intel.com>,
	Yehezkel Bernat <yehezkel.bernat@intel.com>,
	Bjorn Helgaas <bhelgaas@google.com>,
	Mario.Limonciello@dell.com,
	Radion Mirchevsky <radion.mirchevsky@intel.com>
Subject: Re: [07/18] thunderbolt: Handle rejected Thunderbolt devices
Date: Tue, 27 Feb 2018 14:27:26 -0800	[thread overview]
Message-ID: <9a0ba004-1420-daa8-3ad1-299faf5ed5a8@redhat.com> (raw)
In-Reply-To: <20180227092627.GD2424@lahna.fi.intel.com>

On 2018-02-27 1:26 AM, Mika Westerberg wrote:
> On Mon, Feb 26, 2018 at 12:15:28PM -0800, Jeremy McNicoll wrote:
>> On 2018-02-26 11:46 AM, Mika Westerberg wrote:
>>> On Mon, Feb 26, 2018 at 11:28:16AM -0800, Jeremy McNicoll wrote:
>>>> On 2018-02-26 5:38 AM, Mika Westerberg wrote:
>>>>> On Mon, Feb 26, 2018 at 12:20:29PM +0200, Mika Westerberg wrote:
>>>>>> On Thu, Feb 22, 2018 at 03:17:38PM -0800, Jeremy McNicoll wrote:
>>>>>>>> +	if (pkg->link_info & ICM_LINK_INFO_REJECTED) {
>>>>>>>> +		tb_info(tb, "switch at %u.%u was rejected by ICM firmware\n",
>>>>>>>> +			link, depth);
>>>>>>>
>>>>>>> This kind of condition sounds more like an error instead of info.
>>>>>>> Please bump this up to tb_WARN/tb_warn ideally tb_err().
>>>>>>
>>>>>> No, this is not an error.
>>>>>
>>>>> To be more clear, it is totally fine to have the firmware to reject some
>>>>> devices. For example in case of the new usbonly security level the
>>>>> firmware rejects other devices but the first.
>>>>>
>>>>
>>>> Ok. Is that kind of information available to the kernel?  What security
>>>> mode we are in?
>>>>
>>>> ie) if (LINK_REJECTED && !USB_SECURITY)
>>>>          print "Error switch %u was rejected since its not usbonly"
>>>>       endif
>>>>
>>>> I am sure something like that simplified pseudo code above would
>>>> be somewhat useful to users when debugging.
>>>
>>> That's why it is on info level so it goes to dmesg but does not scare
>>> the user :-)
>>>
>>
>> The point I am trying to make is that it would be nice to be able to
>> know WHY the link was rejected and not just that it was rejected.
> 
> Fair enough. In practice (since we ask the firmware to accept any
> device) the only reason for rejection is that the topology limit is
> exceeded (too many devices in the chain).
> 
> I'm thinking to change the message to something like:
> 
> 	tb_info(tb, "switch at %u.%u was rejected by ICM firmware because topology limit exceeded\n",
> 		link, depth);
> 
> And do the same for Titan Ridge in patch [18/18].
> 
> Security level can be read directly from "security" sysfs attribute of
> the domain so that information does not need to be duplicated IMHO.
> 
> Does that work for you?
> 

Sounds good to me.

-jeremy