[Fwd: cvs commit: src/sys/dev/acpica acpi_pci_link.c]

[Nate Lawson]

Michael Reifenberger wrote:
> On Wed, 29 Sep 2004, Nate Lawson wrote:
> ...
> 
>>> If this change isn't MFC'd jet than it could fix a system freeze I had
>>> during boot on RELENG_5 (see my mail to -current a few days ago)
>>> which where fixed when booting -current.
>>
>>
>> Maybe.  But that also was probably the floppy hang problem that Warner 
>> just MFCd.  To test, cvsup to a BETA5 kernel and then copy in this 
>> file to src/sys/dev/acpica/acpi_pci_link.c:
>>
> 
> No. It hang after:
> ...
> ioapic0: routing intpin 23 (PCI IRQ 23) to cluster 0
> ...
> This where after fdc probe. Or is fdc relevant for mountroot too?
> 
> Unfortunately I can't downgrade this machine ATM for the next few days
> to get shure.

If it's the floppy problem, that is where it's likely to hang.  There 
are now 2 probes, 1`for floppy hardware done where the fdc0 messages are 
and 1 done for floppy media by GEOM, right at the end of boot (while 
discovering disks).  If the first one gets the config wrong, the second 
triggers the hang.  This was previously hidden by the fact that in the 
past, we didn't bother with media in the drive during boot, only when 
someone made a mount or read attempt from /dev/fd0 at runtime.

The fix is to get the config right.  At the moment, I believe both 
RELENG_5 and -current take into account all known settings.  It's quite 
possible there are other config issues, possibly triggered by 
nonstandard hardware (i.e., media line change), that we'll have to 
account for.

-- 
Nate