2002-04-04 18:37:24

by Brian Litzinger

[permalink] [raw]
Subject: raid,apm,ide, powers down too fast & corrupts raid

When I power off a system running bootable software RAID mirroring the
raid array is unclean when the system boots back up.

Neil Brown pointed out that the problem may be related to APM shutting
the computer off before everybody (md driver and/or ide and/or drives)
are ready.

Compiling the kernel without APM/ACPI support certainly solves the
problem.

I'm running Linux 2.4.18 on this system. Has this problem been
addressed in any later versions?

Or shall I start exploring the APM, ide driver, md driver, ide
drive interactions?

--
Brian Litzinger <[email protected]>

Copyright (c) 2002 By Brian Litzinger, All Rights Reserved


2002-04-04 19:18:05

by Alan

[permalink] [raw]
Subject: Re: raid,apm,ide, powers down too fast & corrupts raid

> I'm running Linux 2.4.18 on this system. Has this problem been
> addressed in any later versions?

If the box is IDE then current 2.4.19pre/2.4.19-ac trees actually send
flush commands to the IDE disks to be sure its cleared everything out. It
might be that.

2002-04-05 09:20:01

by Luigi Genoni

[permalink] [raw]
Subject: Re: raid,apm,ide, powers down too fast & corrupts raid



On Thu, 4 Apr 2002, Alan Cox wrote:

> > I'm running Linux 2.4.18 on this system. Has this problem been
> > addressed in any later versions?
>
> If the box is IDE then current 2.4.19pre/2.4.19-ac trees actually send
> flush commands to the IDE disks to be sure its cleared everything out. It
> might be that.

Does that work well with disk cache?
(incidentally, I do not have actually IDE disks out of production, so I
cannot test)

Luigi



2002-04-05 12:20:21

by Alan

[permalink] [raw]
Subject: Re: raid,apm,ide, powers down too fast & corrupts raid

> > If the box is IDE then current 2.4.19pre/2.4.19-ac trees actually send
> > flush commands to the IDE disks to be sure its cleared everything out. It
> > might be that.
>
> Does that work well with disk cache?
> (incidentally, I do not have actually IDE disks out of production, so I
> cannot test)

Its actually asking the IDE disk to flush its cache yes