[vortex] Re: [Neomagic] neomagic 256AV/3com 3c575_cb problem

Egbert Eich eich@xfree86.org
Tue Jun 25 03:13:00 2002


Armin Obersteiner writes:
 > hi!
 > 
 > this goes to both neomagic an vortex mailing list, because without starting X
 > i don't have problems, but the problems are with then network card ...
 > (and it worked, some releases back ...)
 > 
 > hardware: sony vaio F350, neomagic 256AV 2.5MB, 3com 575 pcmcia card
 > software: linux 2.4.19-pre10/linux 2.4.18, pcmcia-cs-3.1.33/pcmcia-cs-3.1.34,
 >           xfree86-4.2.0
 > distibution: gentoo 1.2 (compiled *everything*)/suse 8.0
 > (items seperated with / mean i tried it with BOTH versions)
 > 
 > problem: when i start x (X or X -probeonly) i get PCI errors with my network card (pcmcia)
 >          it seems to be the fault of the pci probing, its possibly an irq conflict
 > 
 > question: can it be really the pci probing? can i disable the probing?
 >           how to force the networkcard/gfx-chip to a certain irq?

I don't think PCI probing is at fault. You cannot disable probing
as this is a central piece of the resource management of the Xserver.
X doesn't deal with interrupts at all at the moment. Therefore I
doubt that this causes the trouble.

 > 
 > info: 
 >  * worked the last time with suse 7.3 (xfree86-4.0.2 or 4.1.0, pcmcia-3.1.29, linux 2.4.10)
 >  * starting sound does not interfere in any way (neomagic 256 AV has sound too)
 >  * it worked once, but i could not reproduce it (i beleive for once the networkcard
 >    was not on irq9 (shared), which ist the irq of the neomagic)
 >  * current start order: pcmcia, sound, network, X
 >  
 > attached: xfree86 config, /var/log/messges, XF86 start log, /proc/interrupts, dmesg
 > 

The X log file showes the same device on bus 5 appearing multiple times
under different function numbers. Ths device doesn't show up at all
in the lspci output. 
To me this device looks broken - however we may be missing something
in the Xserver code, too. Probing of PCI devices should not affect
these, though.
Please try to run 'scanpci -v' and check if this causes the same
problems.

Regards,
	Egbert.