Re: PCI bridge optimization

Franz Sirl (
Tue, 1 Oct 1996 01:16:51 +0200

At 16:57 Uhr +0200 29.09.1996, Stefan Röhrich wrote:
>On Fri, 27 Sep 1996, Jon Lewis wrote:
>> On a warm reboot, its SCSI disk was not detected. I hit the reset switch
>> again, and again it started to boot and the SDMS info displayed, but no
>> disk was detected. After turning the system off, waiting a few seconds,
>> and turning it back on, all was well.
>> The system is a 1.5 year old AIR dual P90 (only 1 CPU installed) with
>> AIC-7870 (turned off) built into the board and NCR 810 (using NCR53C8XX
>> driver) controlling a Micropolis 3243 and AHA-1540 controlling a Conner
>> DAT.
>I have the same problem with the warm reboot. But I think, it's caused by
>the Adaptec SCSI controller, because this happens even if I turn off PCI
>bridge optimizations.
>I'm using an Adaptec 2940AU controller (I had to patch the kernel to
>recognize the AIC-7861 chip) and if I try a warm reboot, the Adaptec
>doesn't detect any device, but displays a timeout error (so that's a
>different message than if I only disconnect all SCSI devices).
>If I then continue the boot process with a boot floppy (I can't warm boot
>from hard disk, because I've no IDE disks and the SCSI disk isn't
>recognized at the warm reboot), the kernel hangs during initializing the
>Adaptec SCSI controller driver. If I turn the power off before booting (on
>all attached (even external) SCSI devices, too), everything works well,
>except that a hard disk which should be an ultra scsi device according to
>the web site of the manufacturer, is only recognized at 10 MHz
>synchronous, not 20 MHz. Everything works well with Windows NT, but there
>I didn't find any possibility to show, if the disk is treated as an ultra
>There are some more strange things, e.g. at the warm reboot the SCSI
>devices aren't detected and displayed, but if I go into the Adaptec bios
>setup they show up in the hard disk tools (or similar) list.
>Another curious thing is, that even with pci bridge optimization turned
>on, no message like:
>PCI bridge optimization.
> Cache L2: Not supported.
> CPU-PCI posted write: on.
> CPU-Memory posted write: on.
> PCI-Memory posted write: on.
> PCI burst: on.
>(as on my computer at home) is displayed, board is an Asus board with
>Pentium 133, 82371SB Natoma/Triton II PIIX3 and 82437VX Triton II chips
>(from /proc/pci) and an onboad IDE interface which is turned off.
>I have heard from more people having this problem with the AIC 7861 chip,
>maybe it's the same with the AIC-7870. I have this problem with any warm
>reboot, but I've also heard that with some configurations it disappears
>sometimes, but I don't know which configuration this was.
>Does anybody else have this problem and/or a solution?
>Maybe it's a problem with the pci interface of the Adaptec?

Once again:
Soft reboot is _not_ a real option with some BIOSes and chipsets.
With soft reboot and (at least) busmaster PCI-cards you _definitely_ have
to have a BIOS/chipset which supports activating the PCI-Reset-Line on
_any_ reboot.
Otherwise you will run into a _lot_ of really mysterious problems.

I think this info should stated somewhere? Probably with a OK/!OK list of
BIOSes/chipsets/boards? E.g. for ASUS TP4x boards you should use a BIOS
version >= 2.02.

And Linus: if you visit Munich (the Oktoberfest continues) during this
week, I owe you (and a lot of other people ;-) ) at least a beer!


!!! The Oktoberfest just started! More info at

Lauterbach Trace32 emulators/debuggers
phone +49 8104 8943-60 mail Lauterbach Datentechnik GmbH
fax +49 8104 8943-49 c/o Franz Sirl
URLs <> Fichtenstr. 27
<> D-85649 Hofolding
<> Germany