ocf-linux package

ocf-linux package

[Radu Moisan]

Do we need ocf-linux for something other than headers? The current 
recipe looks like all it's doing is to install the headers. However, the 
README file from the package talks about some kernel modules it provides 
and patching the kernel for embedded systems. I don't have any 
experience with this package, so, if somebody does, some additional info 
about how we use it (or planning to use it) would be greatly appreciated.

Thanks,
Radu

Re: ocf-linux package

[Bruce Ashfield]

On Tue, Aug 7, 2012 at 3:03 AM, Radu Moisan <radu.moisan@intel.com> wrote:
> Do we need ocf-linux for something other than headers? The current recipe
> looks like all it's doing is to install the headers. However, the README
> file from the package talks about some kernel modules it provides and
> patching the kernel for embedded systems. I don't have any experience with
> this package, so, if somebody does, some additional info about how we use it
> (or planning to use it) would be greatly appreciated.

There are a series of kernel patches, I've been carrying them in various trees
since 2.6.27. I do have a set of OCF patches that have been tested against
the 3.4 kernel, so if there's interest, it's easy enough for me to make them
available.

The test steps for OCF can be found within the kernel patches, which
is basically
exercising /dev/cryptodev via a test application, and using it with
openssl/ssh for
acceleration. That being said, it is coupled to the BSP such that you
need appropriate
hardware, and kernel options enabled for any acceleration to happen.

I've also seen everything from SMP to module loading kernel traps in
the past, so it
is definitely something that needs to be done in coordination with a
use case, and
a particular board.

If we don't have a use case, or a board in mind, I'd suggest that it
stay latent for
now, or I could enable the base plumbing in the kernel and it remain
"buyer beware".

Cheers,

Bruce

>
> Thanks,
> Radu
>
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core



-- 
"Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end"

Re: ocf-linux package

[Radu Moisan]

On 08/07/2012 03:50 PM, Bruce Ashfield wrote:
> On Tue, Aug 7, 2012 at 3:03 AM, Radu Moisan <radu.moisan@intel.com> wrote:
>> Do we need ocf-linux for something other than headers? The current recipe
>> looks like all it's doing is to install the headers. However, the README
>> file from the package talks about some kernel modules it provides and
>> patching the kernel for embedded systems. I don't have any experience with
>> this package, so, if somebody does, some additional info about how we use it
>> (or planning to use it) would be greatly appreciated.
> There are a series of kernel patches, I've been carrying them in various trees
> since 2.6.27. I do have a set of OCF patches that have been tested against
> the 3.4 kernel, so if there's interest, it's easy enough for me to make them
> available.
>
> The test steps for OCF can be found within the kernel patches, which
> is basically
> exercising /dev/cryptodev via a test application, and using it with
> openssl/ssh for
> acceleration. That being said, it is coupled to the BSP such that you
> need appropriate
> hardware, and kernel options enabled for any acceleration to happen.
>
> I've also seen everything from SMP to module loading kernel traps in
> the past, so it
> is definitely something that needs to be done in coordination with a
> use case, and
> a particular board.
>
> If we don't have a use case, or a board in mind, I'd suggest that it
> stay latent for
> now, or I could enable the base plumbing in the kernel and it remain
> "buyer beware".

Thanks Bruce for this info. I agree, if there is not immediate necessity 
we should leave it aside for now.

radu