On Wed, 7 Jun 2000, Petko Manolov wrote:
> > Unfortunately I have new information. After about 2 hours of remote X
> > activity, using Cyrille's patch, the ethernet dropped out and repeated the
> > "TX status -100" message over and over again. Looks like it just
> > delayed the problem from occuring.
>
> Thanx. That is very valuable info. It appears that resetting MAC don't
> fix the problem.
Got new problems recently. It seems that on some occasions, the OHCI
driver disables itself (when the UE bit (as Unrecoverable Error) is set).
Doing "ifdown eth0 ; rmmod pegasus ; rmmod usb-ohci ; modprobe usb-ohci ;
modprobe pegasus ; sleep 1 ; ifup eth0" restarts the interface, which
usually works for a few hours then.
The OHCI docs says UE should almost never happen ; most cases would be
related to some contention on the PCI bus (the related linux source line
says roughly the same).
Following my usual "kick it in the balls till it works" way of patching, I
tried to make the OHCI driver reset itself when it falls on an UE. For the
moment, that tactic fails miserably, but I for now do that in
hc_interrupt(). I'll try to see whether a slightly delayed tasklet would
allow me to reset the OHCI bus and unlock the UE (or look deeper in the
OHCI spec if there's a way out of UE without rebuilding the whole software
model of the bus topology ; which simply sucks).
Maybe OHCI wizards could have an idea on this ?
-- Cyrille
------------------------------------------------------------------------------
Grumpf.
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
