Mailinglist Archive: opensuse (5130 mails)

< Previous Next >
Re: [SLE] Changing hardware IRQ's
  • From: Darryl Gregorash <raven@xxxxxxxxxxxxx>
  • Date: Tue, 02 May 2006 05:15:23 -0600
  • Message-id: <44573F4B.70302@xxxxxxxxxxxxx>
On 01/05/06 13:19, S Glasoe wrote:
>On Monday 01 May 2006 11:35 am, Hylton Conacher (ZR1HPC) wrote:
>>S Glasoe wrote:
>>>Moving eth0 to another PCI slot may change its IRQ but the system and
>>>OS may keep it at 5 anyway. Again, don't bother. Unless there is a
>>>known conflict you can actually 'prove' is happening because sound and
>>>eth0 are sharing an IRQ it doesn't matter. It used to matter about 10
>>>years ago...
>>Tnx, Now I feel old as the knowledge I have is about 10-15yrs old.
>>I'll have a look at the BIOS but otherwise I hope it is not going to be
>Know exactly how that feels. How old is this hardware? You mention that it
>matters in Windows which IRQ is set so which version of Windows was that
>and was it running on this same hardware? You may have hardware and BIOS
>that require setting IRQs but that, in general, is going to be more than
>6-8 years old. Back in the day when ISA/EISA was transitioning to PCI, IRQ
>settings mattered because most add-in cards had hardware jumpers for IRQ.
>PCI allowed those hardware settings to be put into firmware on the cards
>and the system/BIOS and discovered by operating systems so physical
>intervention and its risk of static electricity death could be avoided,
>conflicts over sharing of limited IRQs could be resolved via software, less
>labor intensive, more automation via operating system, etc.

Actually, the problem of IRQ conflict is a serious hardware design issue
(he says, thinking fondly back to the days of undergrad electonics, when
popping transistors could be heard throughout the undergrad lab), rather
than a problem with the actual bus design.

In the early days, and through most of the life of the ISA bus,
peripheral cards used a reverse-biased TTL (transistor-to-transistor
logic) to assert/deassert the IRQ line. To deassert an IRQ line, a card
had to open the circuit (ie ground it). If another card on the same IRQ
line then tried to assert the line (ie close the circuit), the first one
would find its transistors overloaded, and they could burn out. The
cheap ones (like the ones they gave us to playwith in undergrad physics
labs) invariably did.

With open-collector logic, which is mostly used now, these
considerations do not arise. The open (grounded) collector on the
transistor is a natural IRQ deassert, and the current levels involved
when some other card tries to assert the line are not sufficient to
cause an overload, because the transistor never has to be reverse-biased
to perform its tasks.

For non-engineers, think of reverse-biasing as sort of like trying to
push on a door that says "Pull" -- you probably aren't going anywhere,
but if you do, the door's going to break first.

< Previous Next >