> > Have just spent a couple of days fighting to try to get a Speedtouch USB
> > modem running on an Ultra 5 (sparc64).  The translation between 32 bit
> > ioctls and 64 bit handlers isn't quite finished and usbdevfs is one of
> > the areas that it isn't finished yet (up to 2.6.5).  Given a lot more of
> > this is now common to all 64 bit platforms I suspect this means it won't
> > work on any 64 bit platform as of yet.  If anyone could confirm or deny
> > this / provide more information I'd be interested.  Given time I will be
> > attempting to remedy this state of affairs in a couple of months.
> This isn't a very informative post.
Sorry - was justintended as a note to warn anyone else rather than a bug
report or a complaint.

The system was a fairly standard Sun Ultra 5 w/ and extra PCI USB card
(using OHCI) running Debian stable (3.0).  I tried with kernels 2.4.18,
2.4.19, 2.4.22, 2.6.3.

> The only thing i can say is that i have no access to a 64 bit machine at
> all (neither Alpha,  nor Ultra Sparc, nor AMD64), so i  can't say at all
> if this driver has a chance to work correctly.
> 
> The driver's code is probably  safe (there is no abuse of pointers<->int
> conversions),
pppoa2, pppoa3 and modem_run all compile and work as intended (as far as
I can tell), both speedtouch 1.1 and speedtouch 1.2.

>  but as you  say, maybe  the usb  interface is  not mature
> enough in 64bit... dunno. Maybe duncan knows a bit more.
The problem is with any ioctls that submit URBs IE
pusb_endpoint_rw_no_timeout in pusb-linux.c  These seem fundamental to
the way to program works so I didn't try to work around them in
userspace.  These functions are called in both modem_run and pppoa3 (via
pusb_endpoint_read) and do not work correctly as kernel support is
missing.  strace suggest that -1 is being returned and the speedtouch
code is registering this an an error and continuing.

The sections that handle these ioctls from 32 to 64 bit mode (originally
in arch/sparc64/kernel/ioctl32.c moved to fs/compat_ioctl.c in the 2.6
series and shared across the 64 bit architectures), see
http://lxr.linux.no/source/fs/compat_ioctl.c?v=2.6.1;a=sparc64#L2621

This is a kernel features problem not the speedtouch driver itself. 
When I have time I will look into the problem and try to fix it myself. 
If anyone else wants a go I'll happily provide any more info that would
be useful (and I know).  I just wanted to warn anyone else who looked to
do this in the future.

HTH

Sweet Dreams,
 - Martin
        
-- 
Martin
[EMAIL PROTECTED]
"Seasons change, things come to pass"


Liste de diffusion modem ALCATEL SpeedTouch USB
Pour se d�sinscrire : mailto:[EMAIL PROTECTED]

        

Reply via email to