Re: [PATCH] usb: hub: try old enumeration scheme first for high speed devices

[Roger Quadros <rogerq@ti.com>]

On 16/08/18 13:59, Zengtao (B) wrote:
> Hi Roger:
> 
>> -----Original Message-----
>> From: Roger Quadros [mailto:rogerq@ti.com]
>> Sent: Thursday, August 16, 2018 3:17 PM
>> To: Zengtao (B) <prime.zeng@hisilicon.com>; Alan Stern
>> <stern@rowland.harvard.edu>
>> Cc: gregkh@linuxfoundation.org; mathias.nyman@linux.intel.com;
>> drinkcat@chromium.org; felipe.balbi@linux.intel.com;
>> drake@endlessm.com; mike.looijmans@topic.nl; joe@perches.com;
>> linux-usb@vger.kernel.org; linux-kernel@vger.kernel.org
>> Subject: Re: [PATCH] usb: hub: try old enumeration scheme first for high
>> speed devices
>>
>> On 16/08/18 09:21, Zengtao (B) wrote:
>>> Hi alan:
>>>
>>>> -----Original Message-----
>>>> From: Alan Stern [mailto:stern@rowland.harvard.edu]
>>>> Sent: Tuesday, August 14, 2018 10:40 PM
>>>> To: Zengtao (B) <prime.zeng@hisilicon.com>
>>>> Cc: Roger Quadros <rogerq@ti.com>; gregkh@linuxfoundation.org;
>>>> mathias.nyman@linux.intel.com; drinkcat@chromium.org;
>>>> felipe.balbi@linux.intel.com; drake@endlessm.com;
>>>> mike.looijmans@topic.nl; joe@perches.com; linux-usb@vger.kernel.org;
>>>> linux-kernel@vger.kernel.org
>>>> Subject: RE: [PATCH] usb: hub: try old enumeration scheme first for
>>>> high speed devices
>>>>
>>>> On Tue, 14 Aug 2018, Zengtao (B) wrote:
>>>>
>>>>> Hi alan:
>>>>>
>>>>>> -----Original Message-----
>>>>>> From: linux-usb-owner@vger.kernel.org
>>>>>> [mailto:linux-usb-owner@vger.kernel.org] On Behalf Of Alan Stern
>>>>>> Sent: Friday, August 10, 2018 10:20 PM
>>>>>> To: Roger Quadros <rogerq@ti.com>
>>>>>> Cc: Zengtao (B) <prime.zeng@hisilicon.com>;
>>>>>> gregkh@linuxfoundation.org; mathias.nyman@linux.intel.com;
>>>>>> drinkcat@chromium.org; felipe.balbi@linux.intel.com;
>>>>>> drake@endlessm.com; mike.looijmans@topic.nl; joe@perches.com;
>>>>>> linux-usb@vger.kernel.org; linux-kernel@vger.kernel.org
>>>>>> Subject: Re: [PATCH] usb: hub: try old enumeration scheme first for
>>>>>> high speed devices
>>>>>>
>>>>>> On Fri, 10 Aug 2018, Roger Quadros wrote:
>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 10/08/18 18:01, Zeng Tao wrote:
>>>>>>>> The new scheme is required just to support legacy low and
>>>>>>>> full-speed devices. For high speed devices, it will slower the
>>>>>>>> enumeration
>>>> speed.
>>>>>>>> So in this patch we try the "old" enumeration scheme first for
>>>>>>>> high speed devices.
>>>>>>>
>>>>>>> How slow does it get? Is it significant?
>>>>>>> Do we risk breaking existing HS devices that work? I don't think
>>>>>>> we can be sure till we run this through testing.
>>>>>>
>>>>>> Indeed.  I am extremely skeptical about a patch like this, unless
>>>>>> somebody can show that Windows uses the "old" scheme for
>> high-speed
>>>> devices.
>>>>>
>>>>> Yes, this is what the windows has done, you can refer to
>>>>>
>> https://blogs.msdn.microsoft.com/usbcoreblog/2013/04/11/usb-2-1-2-0-
>>>>> 1- 1-device-enumeration-changes-in-windows-8/
>>>>
>>>> And that blog post is 5 years old!
>>>>
>>>> Okay, I think we can go ahead and make this change.  However, you
>>>> should update the patch description to mention what Microsoft did in
>>>> Windows 8 and say that the new behavior matches theirs.
>>>>
>>> Okay, I will update it the change log in v2.
>>>
>>>> Also, as Roger mentioned, you should update the documentation to say
>>>> that the old_scheme_first module parameter now applies only to low-
>>>> and full-speed devices, since high- and SuperSpeed devices always use
>> the old scheme first.
>>>>
>>>
>>> Since we should have dedicated enumeration flow for SS, HIGH, low and
>>> full speed devices, So I think the old_scheme_first and use_both_schemes
>> parameters should be removed.
>>> What do you think about it?
>>
>> I think we should retain them as some host controllers can have issues and
>> these parameters give some control to system integrators to workaround if
>> required.
>>
>> I'm aware of one errata [1] that requires the old_scheme_first to be set in
>> certain circumstances.
>>
>> [1] http://www.ti.com/lit/er/sprz429l/sprz429l.pdf
>> Section i897
> 
> I 'd rather to use a quirk to workaround.
> And the main idea is to keep the enumeration flow as simple as possible.

The workaround isn't always required. Only in certain circumstances decided
by the user.
How can a user set the quirk? Now it is as easy as setting it
in the kernel command line.

One option would be to use the old_scheme_first_port via sysfs.
/sys/bus/usb/devices/.../(hub interface)/portX/quirks

But that won't work if devices are hard-wired in the system and get enumerated
before user gets access to sysfs. We need a way to force one type of scheme
before devices are enumerated.

-- 
cheers,
-roger

Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki