USB low-speed bulk transfers

USB low-speed bulk transfers

[Steinar H. Gunderson]

[Please Cc on reply]

Hi,

I recently bought an USB MIDI interface from ESI (called “ESI MIDI Mate”). It
claims to work with Linux, but doesn't -- I've already asked the manufacturer
for an explanation, but as I was impatient, I hacked a bit on the drivers to
actually make it work...

The /proc/bus/usb/devices entry looks like this:

  T:  Bus=04 Lev=01 Prnt=01 Port=00 Cnt=01 Dev#=  4 Spd=1.5 MxCh= 0
  D:  Ver= 2.00 Cls=00(>ifc ) Sub=00 Prot=00 MxPS= 8 #Cfgs=  1
  P:  Vendor=0a92 ProdID=1001 Rev= 1.04
  S:  Manufacturer=ESI
  S:  Product=ESI MIDI Mate
  C:* #Ifs= 2 Cfg#= 1 Atr=80 MxPwr= 20mA
  I:* If#= 0 Alt= 0 #EPs= 0 Cls=01(audio) Sub=01 Prot=00 Driver=snd-usb-audio
  I:* If#= 1 Alt= 0 #EPs= 2 Cls=01(audio) Sub=03 Prot=00 Driver=snd-usb-audio
  E:  Ad=81(I) Atr=02(Bulk) MxPS=   4 Ivl=0ms
  E:  Ad=01(O) Atr=02(Bulk) MxPS=   4 Ivl=0ms

There are two points worth noting here:

 - The device is USB low speed. snd-usb-audio simply checks for full/high
   speed, and refuses any device that isn't. I can hack around this,
   inverting a few checks etc., and it seems to work reasonably well
   (probably since the device has no PCM parts).
 - Both endpoint descriptors are bulk. The HCD driver plain refuses bulk
   transfers for low-speed; it looks like they are disallowed in the USB
   standard somehow. If I comment out the check, the driver works
   (perfectly!), but I guess this isn't acceptable for upstream?

Could the check for low-speed bulk transfers be replaced by a kernel warning
somehow? I can't see any big harm by allowing them, and obviously, Windows XP
and Mac OS X does so.

I can supply a patch for the snd-usb-audio specific parts if desired, but I
can't guarantee it's the correct fix for cards that support PCM. Not that I
know of any PCM-capable low-speed USB sound cards out there...

/* Steinar */
-- 
Homepage: http://www.sesse.net/

Re: USB low-speed bulk transfers

[Andrew Morton]

On Wed, 6 Jun 2007 23:20:46 +0200 "Steinar H. Gunderson" <sgunderson@bigfoot.com> wrote:

> [Please Cc on reply]
> 
> Hi,
> 
> I recently bought an USB MIDI interface from ESI (called “ESI MIDI Mate”). It
> claims to work with Linux, but doesn't -- I've already asked the manufacturer
> for an explanation, but as I was impatient, I hacked a bit on the drivers to
> actually make it work...
> 
> The /proc/bus/usb/devices entry looks like this:
> 
>   T:  Bus=04 Lev=01 Prnt=01 Port=00 Cnt=01 Dev#=  4 Spd=1.5 MxCh= 0
>   D:  Ver= 2.00 Cls=00(>ifc ) Sub=00 Prot=00 MxPS= 8 #Cfgs=  1
>   P:  Vendor=0a92 ProdID=1001 Rev= 1.04
>   S:  Manufacturer=ESI
>   S:  Product=ESI MIDI Mate
>   C:* #Ifs= 2 Cfg#= 1 Atr=80 MxPwr= 20mA
>   I:* If#= 0 Alt= 0 #EPs= 0 Cls=01(audio) Sub=01 Prot=00 Driver=snd-usb-audio
>   I:* If#= 1 Alt= 0 #EPs= 2 Cls=01(audio) Sub=03 Prot=00 Driver=snd-usb-audio
>   E:  Ad=81(I) Atr=02(Bulk) MxPS=   4 Ivl=0ms
>   E:  Ad=01(O) Atr=02(Bulk) MxPS=   4 Ivl=0ms
> 
> There are two points worth noting here:
> 
>  - The device is USB low speed. snd-usb-audio simply checks for full/high
>    speed, and refuses any device that isn't. I can hack around this,
>    inverting a few checks etc., and it seems to work reasonably well
>    (probably since the device has no PCM parts).
>  - Both endpoint descriptors are bulk. The HCD driver plain refuses bulk
>    transfers for low-speed; it looks like they are disallowed in the USB
>    standard somehow. If I comment out the check, the driver works
>    (perfectly!), but I guess this isn't acceptable for upstream?
> 
> Could the check for low-speed bulk transfers be replaced by a kernel warning
> somehow? I can't see any big harm by allowing them, and obviously, Windows XP
> and Mac OS X does so.
> 
> I can supply a patch for the snd-usb-audio specific parts if desired, but I
> can't guarantee it's the correct fix for cards that support PCM. Not that I
> know of any PCM-capable low-speed USB sound cards out there...
> 

(added linux-usb-devel to cc)