[ltp] Strange intermittent problem with T23, RH 8.0--best way to sleep?

[Richard Neill]

julian underwood wrote:
>>My investigations so far lead to:
>>
>>1)The problem is that, somehow, the HDD doesn't wake up. On unsuccessful 
>>wakes, you don't hear the start up 'kerchunk', and the system blocks on 
>>HDD i/o. The consequences are:
> 
> 
> Yes, I forgot to mention that the HDD activity light is solid when this
> happens.  Although it doesn't sound like there is any activity.  The
> drive is of course spinning however.

Odd. So, your HDD is responding? (I'm assuming that the HDD light is 
controlled in hardware, not by any software, so can't be lying).

> 
> I've noticed for some reason, if I put my unit to sleep while running
> the latest version of VMWare, it seems to happen _much_ more
> frequently--about 1 out of 3 times.  Which is weird.
> 
> 
>>	*Programs still in RAM appear to work - that's why X seems
>>	 partly normal at first.
>>
>>	*Debugging is impossible - something like
>>          "apm -s >apmerror.txt 2>&1" never works, because if there is an
>>	error, the system can't write to disk.
>>
>>	*You can do "Alt-SysRq-B" to reboot. But none of the other
>>	sysrq things work properly (it may try to sync, but can't)
>>	
>>2)A good test is this script
>>
>>	#!/bin/bash
>>
>>	for i in `seq 1 100` ; do
>>         	echo -e "This is attempt $i"
>>		touch foo	#timestamp of last successful write
>>         	sleep 1
>>         	apm -s
>>	done
> 
> 
> Not sure I fully understand what that does, but I'll

Just sleeps 100 times, printing a log to the screen, and attempting to 
write the HDD. With a fail rate of about 1 in 10 to 1 in 20, I'd expect 
to need 100 successes to be statistically convinced the apm problem is 
fixed.

> 
> 
>>You'll have to keep pressing 'Fn' every 20 secs. But I would be very 
>>surprised if the script gets to the end.
>>
>>
>>3)The only fix I can imagine is to add an SysRq key combination for 
>>"wake the hard disk, yes really". But I don't know how to do that.
>>(I tried to contact the sysrq maintainers, their email is bouncing - I 
>>think that code has been abandoned!)
>>I think it would have to be done in the kernel, since there's no 
>>filesystem to work with if the hdd isn't alive.
>>
>>4)Does anyone know how to do this? Could it actually just be a config 
>>error? Is it a BIOS bug after all? Would the use of ACPI (supposedly 
>>planned for Mandrake 9.1) be better  ?
>>
>>
>>Best wishes
>>
>>Richard
>>
>>
> 
>