FDC/ACPI/GEOM problems (was: RE: 5.3-BETA5-i386-disc1 bootstrap failures)

On Mon, 2004-09-27 at 22:42, David Boyd wrote:
> Yes, I am tracking RELENG_5.
> 
> The kernel build with the files you suggested resulted in the same hang.
> 
> I filtered the commits by hour and then by 20 minutes intervals until I
> experienced the hang.
> 
> The hang occurs after building the kernel with the follow commits:
> 
>  Edit src/sys/dev/fdc/fdc.c
>   Add delta 1.283.2.1 2004.09.18.04.57.55 imp
>  Edit src/sys/dev/fdc/fdc_acpi.c
>   Add delta 1.2.2.2 2004.09.18.04.57.55 imp
>  Edit src/sys/dev/fdc/fdc_isa.c
>   Add delta 1.9.2.2 2004.09.18.04.57.55 imp
>  Edit src/sys/dev/fdc/fdc_pccard.c
>   Add delta 1.8.2.1 2004.09.18.04.57.55 imp
>  Delete src/sys/dev/fdc/fdcreg.h
>  Edit src/sys/dev/fdc/fdcvar.h
>   Add delta 1.3.2.1 2004.09.18.04.57.55 imp
>  Edit src/sys/sys/bio.h
>   Add delta 1.139.2.1 2004.09.18.04.56.06 imp
>  Edit src/sys/sys/fdcio.h
>   Add delta 1.17.8.1 2004.09.18.04.57.55 imp
> 
> I have some "free" time to pursue this now!
> I'll try almost anything you suggest.
> 
> ------------------------------------------------
> 
> P.S.  The system isn't actually dead...touching the power switch results in
> 
> 		"ACPI: suspend request ignored (not ready yet)"
> 
> 	as often as the power switch is depressed briefly.

I have spent a little time today looking at this problem in more detail.
The hangs on BETA5 that several people have reported are indeed related
to the above file revisions. Apparently systems only hang with ACPI
enabled, and seem to hang in g_waitidle(), presumably waiting for the
floppy drive to become idle.

db> tr 1
sched_switch(c1501320,0,1) at sched_switch+0x14f
mi_switch(1,0) at mi_switch+0x264
sleepq_switch(c08b7980,1,cbc81c90,c060ee72,c08b7980) at
sleepq_switch+0xe0
sleepq_timedwait(c08b7980,0,0,0,0) at sleepq_timedwait+0x34
msleep(c08b7980,0,68,c07f4bb1,14) at msleep+0x2ca
g_waitidle(c1500e00,c05e7600,0,cbc81d1c,c05e7652) at g_waitidle+0x48
vfs_mountroot(0,c1500e00,c1501320,0,0) at vfs_mountroot+0x10
start_init(0,cbc81d48,0,c05e7600,0) at start_init+0x52
fork_exit(c05e7600,0,cbc81d48) at fork_exit+0xa4
fork_trampoline() at fork_trampoline+0x8
--- trap 0x1, eip = 0, esp = 0xcbc81d7c, ebp = 0 ---

On my system, I can't prove that the issue is related to ACPI but others
have reported this issue goiung away with ACPI disabled. On my system,
the floppy drive LED comes on at the point of the hang. Having a floppy 
disk in the drive makes no difference.

Gavin