Hello Brian,

Sorry for the late reply, but I was on a trip and hardly had usable internet
connectivity.

On 2008.05.06. 18:47:48 Brian Oostenbrink <[EMAIL PROTECTED]> wrote:

> I'm working on adding a new platform to openWRT.  The platform is
> multi-processor MIPS cpu that runs SMP Linux.  I have a few questions:
> 1. We currently have Linux 2.6.20 + busybox + squashfs running and
> stable on our platform (using an old, customized buildroot-based
> system), and have customers that require this kernel version.  Any
> suggestions regarding the best release of kamikaze to work with this
> kernel version?

Personally I would base it on trunk, but that's because I know the current 
status
only, 7.09 was long ago. However I would do that locally and forward port the
needed patches/platform support in the same time, too.

> 2. I'm currently trying to get this working with kamikaze 7.09.  I see
> patch files in target/linux/generic-2.6/patches  (and patches-2.6.22).
> Is it the intent that these patches apply to *all* platforms, and each
> platform can then add additional patches as needed?  Our kernel version
> has many patches, and since it is already working and stable, I'd like
> to apply as few openwrt patches as possible to get a system running
> quickly (and then later work on applying all the patches).  Is this a
> reasonable approach, or are the kernel patches required to support
> openwrt features?

Sounds reasonable. There are some patches that are required, but I think
you can easily figure out those. One example is the rootfs splitting patch.

> 3. As I started building, I quickly came upon this error during the
> "make -j1 -C kernel compile" step:
> cp: cannot stat
> `/home/oostenbr/kamikaze_7.09/build_mips/linux-2.6-pmc71xx/linux-2.6.20.
> 21/net/ipv4/ne
> tfilter/ipt_iprange.ko': No such file or directory
> I used a kernel .config file that already works with our system, which
> statically links all parts of the kernel.  Is it necessary to use
> modules in openwrt?  We compile statically to allow careful control of
> code alignment in memory for optimal cache utilization (=performance),
> so requiring the use of LKMs is a serious drawback in our application.

You can make specific patches depend on !TARGET locally.


Cheers,
Imre
_______________________________________________
openwrt-devel mailing list
[email protected]
http://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel

Reply via email to