On 2005-01-04 matt@bodgit-n-scarper.com wrote:
> On Mon, 2005-01-03 at 23:56, Matt Dainty wrote:
> >
> > pcmcia at pcic0 controller 1 socket 0 not configured
> > pcmcia at pcic0 controller 1 socket 1 not configured
>
> Just rechecked the dmesg from the RAMDISK kernel, bit more context might
> make more sense:
>
> pcic0 at isa0 port 0x3e0-0x3e1 iomem 0x70000-0x73fff irq 589827
> pcic0: controller 0 (Intel 82365SL Revision 1) has sockets A and B
> pcic0: controller 1 (Intel 82365SL Revision 1) has sockets A and B
> pcmcia0 at pcic0 controller 0 socket 0
> pcmcia1 at pcic0 controller 0 socket 1
> pcmcia at pcic0 controller 1 socket 0 not configured
> pcmcia at pcic0 controller 1 socket 1 not configured
>
> >From the GENERIC kernel, I just get:
>
> pcic0 at isa0 port 0x3e0-0x3e1 iomem 0x70000-0x73fff irq 589827
> pcic0: controller 1 (Intel 82365SL Revision 1) has sockets A and B
> pcmcia at pcic0 controller 1 socket 0 not configured
> pcmcia at pcic0 controller 1 socket 1 not configured
>
> Both same sources.

  This is strange.  Does this only happen with the patch? If so, does
contoller 1 appear at all without the patch? Does it happen with 2.0 also?
Did you verify the md5 of the actual booted image with the compiled
version (in case of bad disk, seems unlikely but just in case...)?

  AFAICT, this means bad bus reads or bad memory (a bad disk could do
arbitrarily wierd stuff too).  I have seen two issues that seem related to
bus reads.  One is the bad reads for the CF USB card I am seeing on my
Clio 1000 that I have posted about before.  With that card I get
occasional bad bus reads, but the size of the kernel seems to affect where
the bad reads occur.  However, even doing splhigh around every bus access
did not solve the problem (and it was present before my patch as well).
I have not seen any problems with my flash cards (either of the two that
work on that Clio, however others I have do not work even though they work
on my i386 machine).  I bought a Clio 1050 to see if I could get the card
to work there and noticed that under 1.6.X pcic was detected but under 2.0
it was not (this was with unpatched kernels).  The problem there was a bad
isaportoffset, but under 1.6 the bus read apparently succeeded anyway (as
it turns out, the CIS read of the CF USB card fails on the 1050,
apparently even under CE, so I don't know if the bad bus reads would
happen or not).  I never figured out what changed between 1.6 and 2.0 that
caused the difference on the 1050 (or if it was just a matter of different
kernels).

Matthew Orgass
darkstar@city-net.com