Re: BUG, 2.6.20-rc3 raid autodetection

[Kasper Sandberg <lkml@metanurb.dk>]

On Thu, 2007-01-04 at 23:05 +0100, Bartlomiej Zolnierkiewicz wrote:
> On 1/4/07, Kasper Sandberg <lkml@metanurb.dk> wrote:
> > On Thu, 2007-01-04 at 22:06 +0100, Bartlomiej Zolnierkiewicz wrote:
> > > On 1/4/07, Kasper Sandberg <lkml@metanurb.dk> wrote:
> > > > On Thu, 2007-01-04 at 21:07 +0100, Bartlomiej Zolnierkiewicz wrote:
> > > > > On 1/4/07, Kasper Sandberg <lkml@metanurb.dk> wrote:
> > > > > > On Thu, 2007-01-04 at 20:07 +0100, Bartlomiej Zolnierkiewicz wrote:
> > > > > > > On 1/4/07, Kasper Sandberg <lkml@metanurb.dk> wrote:
> > > > > > > > Hello.
> > > > > > > >
> > > > > > > > i just attempted to test .20-rc3-git4 on a box, which has 6 drives in
> > > > > > > > raid5. it uses raid autodetection, and 2 ide controllers (via and
> > > > > > > > promise 20269).
> > > > > > > >
> > > > > > > > there are two problems.
> > > > > > > >
> > > > > > > > first, and most importantly, it doesent autodetect, i attempted with
> > > > > > > > both the old ide drivers, and the new pata on libata drivers, the drives
> > > > > > > > appears to be found, but the raid autoassembling just doesent happen.
> > > > > > > >
> > > > > > > > this is .17, which works:
> > > > > > > > http://sh.nu/p/8001
> > > > > > > >
> > > > > > > > this is .20-rc3-git4 which doesent work, in pata-on-libata mode:
> > > > > > > > http://sh.nu/p/8000
> > > > > > > >
> > > > > > > > this is .20-rc3-git4 which doesent work, in old ide mode:
> > > > > > > > http://sh.nu/p/8002
> > > > > > >
> > > > > > > For some reason IDE disk driver is not claiming IDE devices.
> > > > > > >
> > > > > > > Could you please double check that IDE disk driver is built-in
> > > > > > > (CONFIG_BLK_DEV_IDEDISK=y in the kernel configuration)
> > > > > > > and not compiled as module?
> > > > > > i need not check even once, i do not have module support enabled, so
> > > > >
> > > > > OK
> > > > >
> > > > > > everything 1000000% surely is built in. this is the case for .17 too
> > > > > > (and earlier, this box was started with .15 i think.)
> > > > >
> > > > > Could you send me your config?
> > > > > I'll try to reproduce this locally.
> > > > sure thing.
> > > >
> > > > http://sh.nu/p/8004 <-- .17 working
> > >
> > > CONFIG_BLK_DEV_IDEDISK=y
> > >
> > > > http://sh.nu/p/8005 <-- .20-rc3-git4 nonworking, idemode, the one with
> > > > libata i dont have anymore, but the only difference is that i use the
> > > > libata drivers, but as its same result, shouldnt matter
> > >
> > > # CONFIG_BLK_DEV_IDEDISK is not set
> > >
> > > so not everything is "1000000% surely" built-in ;)
> > i see your point, im afraid i may have misinterpreted you, since you
> > said "and not compiled as module", so i thought you meant i had to make
> > sure it wasnt moduled only.
> 
> You were right - I forgot about "CONFIG_BLK_DEV_IDEDISK=n" case.
> 
> > i will try with this now, what about pataonlibata, do i need this for
> 
> Please report if this fixes the issue.
yeah it did.
> 
> > that too?
> 
> for libata you need SCSI disk driver: CONFIG_BLK_DEV_SD=y
> 
> [ libata is not independent subsystem like IDE but depends on SCSI,
>   and I really don't know why this is not mentioned in config help ]
i knew that, i just thought it may select what it required, though i can
see that it isnt strictly required, just what one wishes often. :)

thanks for your help, will try with libata now.
>