Re: [PATCH V6 3/7] libxl: add pvusb API

[Juergen Gross <jgross@suse.com>]

On 09/14/2015 12:36 PM, George Dunlap wrote:
> On Mon, Sep 14, 2015 at 4:48 AM, Juergen Gross <jgross@suse.com> wrote:
>> On 09/11/2015 04:41 PM, Ian Campbell wrote:
>>>
>>> On Fri, 2015-09-11 at 16:18 +0200, Juergen Gross wrote:
>>>>
>>>> On 09/11/2015 04:09 PM, Ian Campbell wrote:
>>>>>
>>>>> On Fri, 2015-09-11 at 15:55 +0200, Juergen Gross wrote:
>>>>>>
>>>>>> On 09/11/2015 03:26 PM, Ian Campbell wrote:
>>>>>>>
>>>>>>> On Thu, 2015-09-10 at 23:42 -0600, Chun Yan Liu wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>>> Do these fields have any particular size requirements arising
>>>>>>>>> from
>>>>>>>>> e.g. the
>>>>>>>>> USB spec or from possible dom0 implementations?
>>>>>>>>>
>>>>>>>>> If they have a well defined fixed size from a USB spec then
>>>>>>>>> maybe
>>>>>>>>> we
>>>>>>>>> could
>>>>>>>>> use the appropriate fixed size types?
>>>>>>>>
>>>>>>>>
>>>>>>>> Di> dn't see the size limitation. In Linux kernel code, busnum
>>>>>>>> and
>>>>>>>> devnum (here
>>>>>>>> 'hostbus, hostaddr') are both 'int' type.
>>>>>>>
>>>>>>>
>>>>>>> Is that a Linux-specific implementation detail or a fundamental
>>>>>>> property of
>>>>>>> USB? We should be designing the interface around Linux
>>>>>>> implementation
>>>>>>> details. It seems like something in the USB spec ought to define
>>>>>>> precisely
>>>>>>> the number of bits in both a bus number and a device address within
>>>>>>> that
>>>>>>> bus.
>>>>>>
>>>>>>
>>>>>> The USB spec is only about _the_ bus. How many buses a host can
>>>>>> operate and how they are numbered is outside the USB spec.
>>>>>>
>>>>>> Devices are addressed via their ports in the USB protocol. devnum
>>>>>> is a unique index for a device on the bus, the USB protocol
>>>>>> equivalent
>>>>>> is a list of ports of:
>>>>>> - 1 member in case of direct attached devices
>>>>>> - multiple members in case of hubs between bus and device
>>>>>
>>>>>
>>>>> Thanks for the info. So an "address" in the USB protocol is actually a
>>>>> "path" and "hostbus" is an implementation dependent shorthand for all
>>>>> but
>>>>> the last link in that path.
>>>>
>>>>
>>>> I'm not sure in which direction you are looking. "address" is a path.
>>>> A path is normally a list of ports starting at the host and walking
>>>> through all hubs until you reach the device. The "bus" is the root
>>>> of that path. So the number of buses the host knows of is the number
>>>> of USB host adapters without any hub.
>>>
>>>
>>> OK, I thought I understood but the above suggests not.
>>>
>>> In USB speak, the address is a list of port numbers, which you follow from
>>> the host bus which is the root.
>>>
>>> In Linux speak a "bus" is actually each hub along that path.
>>
>>
>> No. Each hub is just a port which happens to have more ports behind it.
>>
>>> Let me try a worked example and see if I've got it right. Lets take this
>>> topology:
>>>
>>> ROOT0
>>>    |-PORT0 ----+--HUB1
>>>    |-PORT1-,   |-PORT0 -- DEVICE A
>>>            |   `-PORT1 -- DEVICE B
>>>            |
>>>            `--HUB2
>>>                |-PORT0 -- DEVICE C
>>>                `-PORT1 -- HUB3
>>>                            |-PORT0 -- DEVICE D
>>>                            `-PORT1 -x
>>>
>>> ROOT1 -- ... other stuff
>>>
>>> In the USB protocol there are two buses corresponding to ROOT0 and ROOT1.
>>>
>>> So in the protocol the address of DEVICE D on the bus associated with
>>> ROOT0
>>> is [1,1,0], that is PORT1 on ROOT0 => PORT1 on HUB2 => PORT0 on HUB3.
>>>
>>> DEVICE A is [0,0] on the bus associated with ROOT0, similarly.
>>>
>>> In the Linux numbering scheme each ROOTn or HUBn is given a bus number,
>>> somewhat arbitrarily (although I'm sure there is a scheme by which they
>>> allocated). So perhaps:
>>>
>>> ROOT0==BUS1
>>
>>
>> Correct.
>>
>>> HUB1==BUS2
>>
>>
>> No, Just Bus1-Port0 or Bus1:Devnum1
>>
>>> HUB2==BUS2
>>
>>
>> Bus1-Port1 or Bus1:Devnum2
>>
>>> HUB3==BUS4
>>
>>
>> Bus1-Port1.Port1 or Bus1:Devnum3
>>
>>> ROOT1==BUS42
>>
>>
>> Bus2
>>
>>> And in this scheme the address is hostbus+hostaddr, so DEVICE D is [3,0],
>>> that is hostbus==3==HUB3, and port 0. And DEVICE A is [2,0]
>>
>>
>> Device D: Bus1-Port1.Port1.Port0 or e.g. Bus1:Devnum4
>> Device A: Bus1-Port0.Port0 or e.g. Bus1:Devnum5
>>
>>> Is that right?
>>>
>>>> One bus can have up to 31 ports.
>>>
>>>
>>> So the answer is that hostaddr can be 5 bits?
>>
>>
>> 5*8 (7 hubs and a device at the last level) == 40, that's the 1 trillion
>> I suggested before. Things are a little bit more complicated. A devnum
>> in a bus is never assigned twice. So when you plug in a device, it might
>> get devnum 6. Unplug it and replug it again will lead to devnum 7.
>>
>>>>    In theory I think up to 7 cascaded
>>>> hubs are possible, but I don't think the resulting theoretical maximum
>>>> of about 1 trillion devices on a bus is to be considered. :-)
>>>
>>>
>>> And this suggests that in principal a Linux hostbus could be 5*7 bits ==
>>> 35
>>> bits, maybe. Or at least that any USB address can be encoded in that many
>>> bits.
>>
>>
>> Busnum can grow to arbitrary values. A new USB bus detected will get a
>> new bus number. Removing it and adding it again will again use a new
>> number.
>
> FWIW libusb seems to define these as uint8:
>
> http://libusb.org/static/api-1.0/group__dev.html#gaf2718609d50c8ded2704e4051b3d2925
>
> (I *think* that "bus number" and "device address" correspond to busnum
> and devnum...)
>
> Anyone want to look into the Linux source code to find out how big it
> will allow busnum / devnum to grow?

drivers/usb/core/hcd.c is using a bitmap to find the next bus number
currently not in use. It's size is USB_MAXBUS which in turn has the
value 64.

choose_devnum() in drivers/usb/core/hub.c is doing a similar job for
device numbers. Here the highest number supported is 127.


Juergen