Re: [BUG REPORT] Sony VAIO, 2.4.7: CardBus failures with Tulip & 3c575

Linus Torvalds (torvalds@transmeta.com)
Sun, 22 Jul 2001 13:59:02 -0700


In article <3B5B1F77.D8B45FFA@candelatech.com> you write:
>
>This report contains information about my failure to get my
>CardBus NICs working correctly. Hardware involved is:
>
>Sony VAIO PCG-FX210 laptop (800Mhz Duron...)
>DFE-650 16-bit PCMCIA NIC x2
>3Com Megahertz 32-bit 3CCFE575BT NIC x2
>AmbiCom 32-bit 8100 NIC (tulip) x2

This looks suspiciously like your slot #1 gets the PCI interrupt routing
wrong.

Note especially the kernel reports:

Linux Kernel Card Services 3.1.22
options: [pci] [cardbus] [pm]
PCI: Assigned IRQ 9 for device 00:0a.0
PCI: Assigned IRQ 10 for device 00:0a.1
IRQ routing conflict for 00:07.5, have irq 5, want irq 10
IRQ routing conflict for 00:07.6, have irq 5, want irq 10
PCI: Sharing IRQ 10 with 00:10.0

it really looks like your slot 1 controller (00:0a.1) really wants irq5,
based on the fact that other devices are reported to have irq5.

However, if they _really_ have irq5 already routed, I'm surprised that
the PCI irq router "r->get()" function didn't pick up on that fact, and
that the "set" function apparently didn't work correctly.

So I'd guess that when you insert a card in slot #1, you get a constant
stream of interrupts on irq5, which is not where the kernel is expecting
them, so your machine locks up.

Can you do the following:
- run dump_pirq on your machine (attached)
- run "lspci -vvxxx" as root

send me and Jeff the output. Jeff also suggested enabling debugging in
yenta.c and that might be useful too.

Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/