getting new driver into kernel, boot.iso

David Kewley kewley at cns.caltech.edu
Wed Aug 13 04:16:39 UTC 2003


I got the updated sk98lin (version 6.14) driver disk working, and 
successfully used it along with a modified boot.iso CD to do a network 
kickstart using the on-board ethernet chip.  I'll be writing up what I did, 
and posting the writeup and the driver disk files to the web soon.  Thanks 
again, Michael, for the pointer to Doug Ledford's driver disk kit.

I haven't contacted SysKonnect yet, but I'll be doing that, to urge them to 
submit their driver for inclusion in the kernel, and to offer them the 
driver disks I made, and the method for making them in the future.

I wanted to reply to a few of Dave Jones' comments:

Dave Jones wrote on Friday 08 August 2003 09:01:
> (click the disk icon 8-)

http://www.syskonnect.com/syskonnect/support/driver/zip/linux/sk98lin_2.4.21_patch.gz

Yeah, I know. :)  The problem is, the previous version of the patch (for 
kernel 2.4.20) didn't apply cleanly to the RH kernel, and I didn't feel 
like struggling through the minutae of making it fit.

> From a quick skim of the code, it seems largely made up of gratuitous
> whitespace changes, which makes it hard to see the real changes.
> 
> Lots of changing comments from..
> 
> 	/* foo */
> 
> to
> 	/*
> 	** foo
> 	*/
> 
> Seems someone got a new text editor with macros for xmas..

Thanks for taking a look at the patch, but I beg to differ. :)  There are a 
fair number of whitespace changes, but there are (by my eye) lots more 
logic changes, plus many added comments and a fair number of non-whitespace 
reformats.  I actually saw very few of the comment reformats of the type 
you show above.  Of course, this all matters little. :)  I guess my point 
is that it seems to me the patch is not trivial (which is what your comment 
suggested to me).  It weighs in at nearly 18000 lines and 1/2 MB, and I'd 
guess around 1/4 of that is substantive.

> Lots of hard coded values changed to #defines - a good thing.
> Some new cards supported - also a good thing.

I didn't notice the hard-coded values changing, but yes, there are lots of 
#defines added or changed.  And the key thing to me, I think, is the new 
card support.
 
> That aside ISTR someone mentioning that this update backs out
> some stuff that has been fixed in mainline since they last did a push,
> so that needs fixing,

Huh, that could be, I don't know the mainline history of changes to this 
driver (and didn't attempt to look them up).  I see that there's one RH 
patch in 2.4.21 (rawhide) that affects sk98lin slightly, in 
linux-2.4.1-compilefailure.patch.

It'd be useful for someone who knows the history of sk98lin to give SK 
feedback about what needs patching...

> as do the memory leaks they introduce in their
> ioctls. They may also be other problems, but thats all I picked up
> from a quick 2 minute skim of the diff.

OK, I'll take your word for it.  I'm way too inexperienced to judge that, 
even in my 30-minute skim of the diff. :)

> Fishing out the good bits of the patch is a worthwhile thing for
> someone to do, but its not a 10 minute job.

I imagine so.  What would be the most useful way for this to happen?  Who 
would best do this?  The SK developers?  (Then I guess they'd need to know 
what was fixed in the mainline kernel version of sk98lin.)  Or Alan or a 
kernel net driver hacker?

David





More information about the devel mailing list