Re: binary kernel drivers re. hpt370 and redhat

Re: binary kernel drivers re. hpt370 and redhat

[Stephen Hemminger]

On Wed, 27 Aug 2003 18:40:30 -0400
joe briggs <jbriggs@briggsmedia.com> wrote:

> I have a client who has a raid controller currently supported under windows, 
> and now wants to support linux as a bootable device.  Currently, some of 
> their trade secrets are contained in the driver as opposed to the controller 
> firmware, etc., so for now they wish to release a binary-only driver to 
> certain beta customers.  (i.e., 1st stage of porting is similar functionality 
> as windows). Am I correct that in order to boot off of this device that the 
> driver would have to be statically linked in vs. a module which could be 
> distributed as a binary-only driver keyed to the kernel.revision of the 
> distribution's kernel?  I would like to avoid any flames and ask that all 
> recognize that some hardware providers are having to ease into the pond a toe 
> at a time.  Any constructive thoughts, suggestions, references, tips, etc. 
> highly appreciated.

The driver could be a module and live in initramfs.  If you can
get the initial Linux image and initramfs loaded, you would be okay.

The problem is more in the bootloader (LILO or GRUB) would not no how
to do raid. The /boot partition would have to be on a non-raid partition.
Same problem if driver is statically linked in the kernel.

Re: binary kernel drivers re. hpt370 and redhat

[Alan Cox]

On Mer, 2003-08-27 at 22:57, Stephen Hemminger wrote:
> On Wed, 27 Aug 2003 18:40:30 -0400
> joe briggs <jbriggs@briggsmedia.com> wrote:
> 
> > I have a client who has a raid controller currently supported under windows, 
> > and now wants to support linux as a bootable device.  Currently, some of 
> > their trade secrets are contained in the driver as opposed to the controller

Standard practice - its not IMHO so much trade secrets as "improving the
barrier to vendor change" 8). Pretty much all of the older PATA
controllers don't actually do hardware raid but bios/driver raid - ie
its the equivalent (or roughly so) of the md layer but locks you into
the vendor. The notable exception here is the 3ware card (there are a 
couple of others too - Promise Supertrak100, SX6000)

We know some of these formats (eg see the hptraid driver in 2.4.2x)

> The problem is more in the bootloader (LILO or GRUB) would not no how
> to do raid. The /boot partition would have to be on a non-raid partition.
> Same problem if driver is statically linked in the kernel.

Plus little issues like the GPL 8)

binary kernel drivers re. hpt370 and redhat

[joe briggs]

I have a client who has a raid controller currently supported under windows, 
and now wants to support linux as a bootable device.  Currently, some of 
their trade secrets are contained in the driver as opposed to the controller 
firmware, etc., so for now they wish to release a binary-only driver to 
certain beta customers.  (i.e., 1st stage of porting is similar functionality 
as windows). Am I correct that in order to boot off of this device that the 
driver would have to be statically linked in vs. a module which could be 
distributed as a binary-only driver keyed to the kernel.revision of the 
distribution's kernel?  I would like to avoid any flames and ask that all 
recognize that some hardware providers are having to ease into the pond a toe 
at a time.  Any constructive thoughts, suggestions, references, tips, etc. 
highly appreciated.


-- 
Joe Briggs
Briggs Media Systems
105 Burnsen Ave.
Manchester NH 01304 USA
TEL 603-232-3115 FAX 603-625-5809 MOBILE 603-493-2386
www.briggsmedia.com

Re: binary kernel drivers re. hpt370 and redhat

[Samuel Flory]

Stephen Hemminger wrote:

>On Wed, 27 Aug 2003 18:40:30 -0400
>joe briggs <jbriggs@briggsmedia.com> wrote:
>
>  
>
>>I have a client who has a raid controller currently supported under windows, 
>>and now wants to support linux as a bootable device.  Currently, some of 
>>their trade secrets are contained in the driver as opposed to the controller 
>>firmware, etc., so for now they wish to release a binary-only driver to 
>>certain beta customers.  (i.e., 1st stage of porting is similar functionality 
>>as windows). Am I correct that in order to boot off of this device that the 
>>driver would have to be statically linked in vs. a module which could be 
>>distributed as a binary-only driver keyed to the kernel.revision of the 
>>distribution's kernel?  I would like to avoid any flames and ask that all 
>>recognize that some hardware providers are having to ease into the pond a toe 
>>at a time.  Any constructive thoughts, suggestions, references, tips, etc. 
>>highly appreciated.
>>    
>>
>
>The driver could be a module and live in initramfs.  If you can
>get the initial Linux image and initramfs loaded, you would be okay.
>

   Rather an initrd under Linux.  Note that there is a partial source 
driver, and RH driver's disks here: (look under raid IC with the right 
chipset for partial source.)
http://www.highpoint-tech.com/usaindex.htm

>
>The problem is more in the bootloader (LILO or GRUB) would not no how
>to do raid. The /boot partition would have to be on a non-raid partition.
>Same problem if driver is statically linked in the kernel.
>  
>
   If you are doing raid 1.  Lilo should work.  It doesn't really matter 
if lilo isn't aware of of the data on the other drive.  Each has a full 
copy of everything.

PS-  Newer linux kernels should be able to support the "raid" controller 
as a normal ide controller.  You could then just configure linux 
software raid.

-- 
Once you have their hardware. Never give it back.
(The First Rule of Hardware Acquisition)
Sam Flory  <sflory@rackable.com>

Re: binary kernel drivers re. hpt370 and redhat

[joe briggs]

I believe that companies will eventually see it costs less and the benefits 
higher to open source their drivers.  But that is a conclusion and paradigm 
that they will have to evolve to themselves, and shoving it down their 
throats will only lengthen the process.  The first step is to support their 
hardware under the platform (linux), and that is what I am focusing on.  

Thanks for the technical analysis and suggestions.

Joe

On Wednesday 27 August 2003 06:17 pm, you wrote:
> On Mer, 2003-08-27 at 22:57, Stephen Hemminger wrote:
> > On Wed, 27 Aug 2003 18:40:30 -0400
> >
> > joe briggs <jbriggs@briggsmedia.com> wrote:
> > > I have a client who has a raid controller currently supported under
> > > windows, and now wants to support linux as a bootable device. 
> > > Currently, some of their trade secrets are contained in the driver as
> > > opposed to the controller
>
> Standard practice - its not IMHO so much trade secrets as "improving the
> barrier to vendor change" 8). Pretty much all of the older PATA
> controllers don't actually do hardware raid but bios/driver raid - ie
> its the equivalent (or roughly so) of the md layer but locks you into
> the vendor. The notable exception here is the 3ware card (there are a
> couple of others too - Promise Supertrak100, SX6000)
>
> We know some of these formats (eg see the hptraid driver in 2.4.2x)
>
> > The problem is more in the bootloader (LILO or GRUB) would not no how
> > to do raid. The /boot partition would have to be on a non-raid partition.
> > Same problem if driver is statically linked in the kernel.
>
> Plus little issues like the GPL 8)

-- 
Joe Briggs
Briggs Media Systems
105 Burnsen Ave.
Manchester NH 01304 USA
TEL 603-232-3115 FAX 603-625-5809 MOBILE 603-493-2386
www.briggsmedia.com

Re: binary kernel drivers re. hpt370 and redhat

[Samuel Flory]

Samuel Flory wrote:

> Stephen Hemminger wrote:
>
>> On Wed, 27 Aug 2003 18:40:30 -0400
>> joe briggs <jbriggs@briggsmedia.com> wrote:
>>
>>  
>>
>>> I have a client who has a raid controller currently supported under 
>>> windows, and now wants to support linux as a bootable device.  
>>> Currently, some of their trade secrets are contained in the driver 
>>> as opposed to the controller firmware, etc., so for now they wish to 
>>> release a binary-only driver to certain beta customers.  (i.e., 1st 
>>> stage of porting is similar functionality as windows). Am I correct 
>>> that in order to boot off of this device that the driver would have 
>>> to be statically linked in vs. a module which could be distributed 
>>> as a binary-only driver keyed to the kernel.revision of the 
>>> distribution's kernel?  I would like to avoid any flames and ask 
>>> that all recognize that some hardware providers are having to ease 
>>> into the pond a toe at a time.  Any constructive thoughts, 
>>> suggestions, references, tips, etc. highly appreciated.
>>>   
>>
>>
>> The driver could be a module and live in initramfs.  If you can
>> get the initial Linux image and initramfs loaded, you would be okay.
>>
>
>   Rather an initrd under Linux.

  Sorry I meant Linux 2.4. ;-)  Initramfs is a 2.6 thing.

>   Note that there is a partial source driver, and RH driver's disks 
> here: (look under raid IC with the right chipset for partial source.)
> http://www.highpoint-tech.com/usaindex.htm
>
>>
>> The problem is more in the bootloader (LILO or GRUB) would not no how
>> to do raid. The /boot partition would have to be on a non-raid 
>> partition.
>> Same problem if driver is statically linked in the kernel.
>>  
>>
>   If you are doing raid 1.  Lilo should work.  It doesn't really 
> matter if lilo isn't aware of of the data on the other drive.  Each 
> has a full copy of everything.
>
> PS-  Newer linux kernels should be able to support the "raid" 
> controller as a normal ide controller.  You could then just configure 
> linux software raid.
>


-- 
Once you have their hardware. Never give it back.
(The First Rule of Hardware Acquisition)
Sam Flory  <sflory@rackable.com>