"use register arguments" option enabled.

Mike Honeyfield mike.honeyfield at gmail.com
Tue May 17 08:23:07 UTC 2005


On 5/17/05, Arjan van de Ven <arjan at infradead.org> wrote:
> > So you disagree that this is actually "experimental" as per the kernel config?
> 
> yes I do.
> EXPERIMENTAL in the kernel config in practice is meaningless; it hardly
> ever gets removed even after years of high production use of stuff. I
> wonder if the EXPERIMENTAL got removed from EXT3 yet .. ;)

I see. 

> > Yes, RH's choice in kernel configure option is theirs and yes, it is
> > true, binary only vendor can "fix" (not sure waht the quotes imply),
> > we have "fixed" our drivers. Worth noting, our driver was written for
> > 2.6 and in RHEL it was broken, but not other platforms we support,
> > like debian and suse.
> 
> I assume that since your driver was written with linux in mind, that it
> is gpl, right? So then you wouldn't have this problem....
> (if it's not, make sure your lawyers have read the COPYING.modules file
> from the kernel rpms)

Yes, we are fine there.

> > However, I was curious because it seemed ironic to possibly break ABI
> > for binary only vendors in the enterprise product.
> 
> The linux kernel has no ABI. So it's also not possible to break it. SLES
> uses this option too afaik, and it has been there for a long time in
> 2.6, any driver not capable of running with it is of course very buggy
> or at least not adjusted to 2.6 properly.

Sorry, I confused myself, re: ABI. The kernel config states its uses a
different ABI when this option is set, not break.

IIRC, SLES 9 ran fine. Debian has been the platform of choice by the
developers, largely, but customers also use RHEL and SLES.

As previously stated, the issue was corrected. The question was raised
to find out why.

As I understand it, it offers some performance improvements, not sure how much. 

Thanks for the clarification all. 

Mike




More information about the fedora-devel-list mailing list