I first noticed this problem when I first got a PCMCIA atheros based wireless card to work. The problem begins whenever I start the wireless card and then plug in an external USB 2.0 hard drive, mount it, and then begin browsing the the drive using nautilus, xmms, etc... A quick look at 'top' shows the logging facilities in FC3, klogd and syslogd, taking up almost all of the cpu. I then look at the /var/log/messages log in which i find that the error
serial8250: irq 11 has too much work
is being added to the file at an incredible rate. This of course spurs me to check the IRQs in proc...
cat /proc/interrupts CPU0 0: 1640085 XT-PIC timer 1: 1403 XT-PIC i8042 2: 0 XT-PIC cascade 9: 2 XT-PIC acpi 11: 164675 XT-PIC ohci1394, ALI 5451, ehci_hcd, ohci_hcd, ohci_hcd, yenta, yenta, yenta, yenta, ath0, eth0 12: 685 XT-PIC i8042 14: 33533 XT-PIC ide0 15: 36 XT-PIC ide1 NMI: 0 ERR: 6
I find it completely ridiculous that irq11 is so overloaded while other irqs are either empty or almost completely empty. I then look within the pcmcia.opts configuration file to see whether I can switch yenta to a different interrupt, I instruct the config parser to 'ignore' irq11. Changing this and then either restarting pcmcia or simply restarting the computer has absolutely no effect upon IRQs.
So I guess my question essentially spurs from something that I once read that seemed to suggest to me that pcmcia.opts is not the first file to be parsed, it is overridden by settings somewhere else. Which essentially boils down to this; where can I definitively set the irq used by pcmcia?
You might have to enable the second PCI IRQ in your systems BIOS before trying to divert it through software. I'm wondering why you are getting the error, as IRQ 11 should be your PCI bus interrupt and you should not have an 8250 UART in your system. I have as much as you do in my system, and I do not get this error. However, PCI IRQ movement is more art than science.
-- James McKenzie
