DragonFlyBSD

Same here. I tried replacing DRIVERSLEEP's with DELAY's in ata-all.c
and it made it into single-user mode, so there may be some edge cases
where DRIVERSLEEP is not usable in early boot stage.

Cheers.

:Same here. I tried replacing DRIVERSLEEP's with DELAY's in ata-all.c
:and it made it into single-user mode, so there may be some edge cases
:where DRIVERSLEEP is not usable in early boot stage.
:
:Cheers.

Hmm.    Sascha, try changing the DRIVERSLEEPs to DELAYs in 
    /usr/src/sys/dev/disk/ata/ata-all.c and see if that fixes the problem.

Yonetani, could you put kprintf()'s around the DRIVERSLEEPs and tell
    me if it hangs there or gets past them?

If worse comes to worse I'll back out that change, but I'd like to
    try to track it down.  I thought I was being conservative, I only
    put them in the ata_reset() path (which runs without interrupts enabled).
    It should be able to thread switch in there!

I will try loading HEAD onto my laptops to see if I can reproduce the
    problem.

-Matt
                    Matthew Dillon 
                    <dillon@backplane.com>

Ok, I reproduced the DRIVERSLEEP problem. I'm tracking it down.

-Matt

Ok, its been fixed. Thread0 wasn't being put on the LWKT run queue,
so when DRIVERSLEEP called lwkt_switch(), poof, nothing left but
the idle thread.

Thanks!

-Matt
                    Matthew Dillon 
                    <dillon@backplane.com>