Re: linux-2.1.44 on i586: Immediate crash on boot

Linus Torvalds (torvalds@transmeta.com)
8 Jul 1997 04:09:36 GMT


In article <199707080258.WAA02273@sleipnir.valparaiso.cl>,
Horst von Brand <vonbrand@sleipnir.valparaiso.cl> wrote:
>Linus was right today: This _is_ a "featureful" kernel. Sure is a pretty
>sight when all the colors flash by ;-)

Hey, that's a new one. I have some features on my machine, but no
pretty colours.

>The kernel crashes after mounting all disks, briefly flashes a console full
>of weirdly colored characters and reboots. No messages make it to
>/var/log/messages, nothing can be seen when it goes down either..
>
>Besides, I've got this too when building:
>
> inode.c: In function `fat_read_super':
> inode.c:332: structure has no member named `s_mounted'

Yup, 2.1.44 only supports ext2, /proc, NFS and autofs. You can draw your
own conclusions as to which filesystems I tend to use personally ;)

>How about stopping this nonsense where "testing" kernels do mostly work
>(2.1.44-pre[23]), "beta" ones (2.1.4[34]) just don't?

Well, the problem was/is that both 43 and 44 had _lots_ of very new code
in them. Some of the pre-44's (but by no means all of them) had fixes
for the worst breakage, but then the later 44's ("for-davem") generally
took a radically new approach to the dcache and didn't boot very well at
all.

The real 2.1.44 boots for me and is usable enough that I'm actually
doing all my work on it - it's great performance-wise. However, I
certainly agree that it is featureful.

The reason I released it is (a) I wanted people to find all the
features, and hoped that some people out there would actually start to
work on them and (b) I had half a meg worth of patches (mainly different
ports) that I needed to get out before some of the people out there lose
track of my source tree..

So anybody out there with a hankering for kernel hacking, go right
ahead. It shouldn't blow up quite as spectacularly as it seems to do
for you, but I notice that you have quota enabled and seem to be using a
few filesystem modules and that might trigger behaviour that I haven't
seen.

Never fear, we'll get this sorted out in a few patch-levels, but it's
going to be a lot easier for me if there are people who get their hands
dirty..

Linus