Re: Loading PATA_PDC2027x as a module

["John Treubig" <jtreubig@xxxxxxxxxxx>]

Thanks for the quick response.  Spent most of the day figuring out how to 
build the ramdisk.  Built both drivers (pata_pdc2027x and pdc202xx_old) and 
the ramdisk, loaded the ramdisk at boot and still system is using IDE driver 
for the Promise attached drives.  Have attached the boot log and the verbose 
output of the ramdisk build.  Got any ideas as to what might be the hang up? 
 I am using the M/B controller for system disk and it is not of the Promise 
variety.   The Promise is a PCI card in the I/O card bay.

Best wishes,
John


From: Doug Maxey <dwm@xxxxxxxxxxx>
To: "John Treubig" <jtreubig@xxxxxxxxxxx>
CC: linux-scsi@xxxxxxxxxxxxxxx, linux-ide@xxxxxxxxxxxxxxx,        
dwm@xxxxxxxxxxxxxxxx
Subject: Re: Loading PATA_PDC2027x as a module
Date: Mon, 21 Aug 2006 11:08:52 -0500


On Mon, 21 Aug 2006 10:38:12 CDT, "John Treubig" wrote:
> Can I load pata_pdc2027x as a module or must it be built as part of the
> kernel?  I am using RedHat Fedora with 2.6.17.1 stock kernel.  LibATA has
> been installed in the kernel with ATA_ENABLE_PATA undefined.  The
> pata_pdc2027x driver is not part of the Fedora build, so I have been able 
to
> build pata_pdc2027x as a module with ATA_ENABLE_PATA defined for the 
module.
>   When I try to load the pata_pdc2027x module it will not displace the 
IDE
> native driver.  What's the mechanism that is used to determine if the IDE
> driver grabs the device or LibATA/pata_pdc_2027x?

The base kernel needs the old driver (pdc202xxx...) built as a module
(or not at all) to give the pata_pdc2027x module a chance to load.

Once that is done, you can add the module as required to mkinitrd and
it should be the first loaded, which prevents the other driver from
loading.

>
> You may ask why I don't just live with the IDE driver.  This stems from
> problems we saw in accessing the Promise PDC2027x drives under 2.6.10 
where
> the system lost time.  In discussions with Albert Lee, it was resolved 
that
> the best approach was to convert our access to use LibATA.  After usage 
of
> the LibATA access, the time loses issue has gone away.  I know that there
> have been changes in both the IDE and LibATA structures, so I'm open to
> additional advice.
>
> Best wishes,
> John Treubig
> VT Miltope
> Senior Test Engineer

Attachment: rdbld.log
Description: Binary data

Attachment: startup.log
Description: Binary data