-CURRENT userland regression

John-Mark Gurney jmg at funkthat.com
Thu Feb 21 16:55:33 UTC 2013

Andriy Gapon wrote this message on Thu, Feb 21, 2013 at 10:19 +0200:
> on 21/02/2013 01:16 Xin Li said the following:
> > I think it's unlikely -- I have r247057 of sys/ which worked fine...
> > 
> > userland 246957 works good by the way.
> Just a very wild guess - are you sure that it is the userland that is to blame?
> It is rather unfortunate that we install boot blocks, including loader which
> gets _really_ installed, as part of installworld (and they are built as part of
> buildworld).  So there is a possibility that it is loader that causes the
> trouble.  I would try to rule that out.

As I posted in a different thread, apparently my gcc AES changes
(r247012) manages to break a clang built kernel...  I'm in the process
of debugging right now...

I did the usual:
make buildworld && make buildkernel && make installkernel && shutdown -r now

and the kernel hung...  I then reverted r247012 and did the above again,
and the machine is running again:
FreeBSD carbon.funkthat.com 10.0-CURRENT FreeBSD 10.0-CURRENT #4 r247075M: Thu Feb 21 00:49:57 PST 2013     jmg at carbon.funkthat.com:/usr/obj/usr/src/sys/GENERIC  amd64

And I haven't installworld yet... so somehow supporting the AES
instructions in gcc causes the kernel to miscompile...

I'm now going to diff the two kernels to find out what's different...

  John-Mark Gurney				Voice: +1 415 225 5579

     "All that I will do, has been done, All that I have, has not."

More information about the freebsd-current mailing list