SSD on the Netwinder

Ralph Siemsen ralphs at
Wed Apr 17 15:41:29 PDT 2019

On Wed, Apr 17, 2019 at 05:38:42PM -0400, D. Hugh Redelmeier wrote:
>| From: Ralph Siemsen <ralphs at>
>| Decided to give a go at building a new kernel. Kind of an ironic version jump,
>| from 2.4.19 back in 2003 (!!!), to v4.19 some 16 years later... Now to be
>| fair, I did also do a 3.18 build back in 2015.
>How has the size changed?

Looks roughly as follows:
2.2.18: 1.3MB
2.4.18: 1.4MB
3.18:  2.0MB
4.19: 2.0MB
Keep in mind the configurations here are not necessarily identical.. and 
no effort has been made to trim the 3.x and 4.x, they may contain a lot 
of features that do not make sense.

>| Anyhow, it worked "out of the box" using netwinder_defconfig from the 4.19
>| mainline kernel.
>Amazing!  I would have thought that some drivers would have been
>abandoned over the decades.

All our "old friends" are still there, like nwbutton and such. Amazing 

>| In contrast with my attempt in 2015, I did not have to fight with gcc to build
>| for armv4 (no need for --fix-v4bx anymore). So either gcc7 got smarter, or the
>| buildroot folks may have patched it... I didn't look except to note it works.
>| Also I was able to build glibc that seems to work.
>That too is amazing.  I thought that armv4 had been let go.

I guess there must still be some bigger users...

>| Ignoring unrecognised tag 0x54410004
>| ATAG_INITRD is deprecated; please update your bootloader.
>What does that mean?  Does it matter?

We're using a long-ago deprecated way of passing arguments from 
bootloader to kernel. And we're passing some tag the kernel doesn't 
understand. It is probably harmless for now.

But really we should move to device tree like everybody else ;-)

>| atkbd serio1: keyboard reset failed on isa0060/serio1

I was using serial console, and did not have a PS/2 keyboard plugged in, 
hence this warning.

>| # uname -a
>| Linux buildroot 4.19.25 #10 Wed Apr 17 15:49:55 EDT 2019 armv4l GNU/Linux
>| # free
>|             total       used       free     shared    buffers     cached
>| Mem:        125024       6488     118536       1540          0       1540
>So the kernel + initial processes are taking only 6.5M out of 128?
>That seems surprisingly good.

It's a really minimal image, just the kernel, and userspace consists of 
nothing more than busybox.


More information about the Netwinder mailing list