> I suspected IRQ problems, and discovered that both USB and the LTmodem
> were sharing IRQ5.
> In the bios
...
> is a facility to "reserve" irq's.
As far as I can make out (writers of mobo manuals should be shot,
drawn, quartered, and shot again), this takes the specified IRQs out of
the pool available for auto-allocation. The only thing this is ever
useful for, as far as I can make out, is if you have an ISA card which
is fixed-wired to that particular IRQ, and can't share that IRQ with
other devices. I.e., limited functionality these days of this "reserve
IRQ".
> There is a separate facilty to
> allocate an irq to usb, although not a specific IRQ number, just
> allocate or not allocate.
Yes, effectively this is nothing more than "disable USB".
> Can anyone point me to the right resolution here?
Good question, I asked this a few days (2 weeks?) ago about USB 2.0
cards[1]. The real issue is "how do I control the IRQ allocation to
cards/PCIslots/devices in auto mode". I don't know, and there were no
replies with an answer to this.
> I suspect I should go
> into the bios and set it to "reserve" IRQ5 so that it does not get
> automatically allocated to USB, and is available for LTmodem to grab.
This could work, *if* ltmodem can be told to grab a specified IRQ.
> However I have never played with this stuff in the bios much and I am
> not 100% aware of the implications.
You can't beak anything. Things just stop working temporarily. Ooops
well, if the harddisk interrupt gets meddled and causes data
corruption... (very unlikely in my experience).
> well its an idea, but I am not sure why suse would do any differently.
Whether IRQ allocation and things like handling of USB devices produces
a functional system afterwards is influenced by:
* kernel + driver (module) versions
* kernel compilation options selected
* vendor kernel patches applied
* hotplug scripting
* ACPI / APM handling by bios and kernel ("interrupt routing")
* precise(!!) hardware used (both mobo and cards)
All but the last of these factors can easily differ between distros. The
last point is crucial for success/failure stories, but as most of these
reports leave out essential info they're nothing more than indications.
Volker
[1] It works fine now with SuSE 9.1 kernel 2.5.4/2.6.5, but I have only
done quick stress testing (read 256MB from flash card). Incidentally,
the time required for that read is far short (15%) of the read access
time specified for the flash chip.
--
Volker Kuhlmann is possibly list0570 with the domain in header
http://volker.dnsalias.net/ Please do not CC list postings to me.
