Date: Sun, 15 Jul 2007 11:17:32 -0700 From: John-Mark Gurney <gurney_j@resnet.uoregon.edu> To: Michiel Boland <michiel@boland.org> Cc: freebsd-current@freebsd.org Subject: Re: sshd broken with UsePrivilegeSeparation=yes on sparc64 Message-ID: <20070715181732.GR1221@funkthat.com> In-Reply-To: <Pine.GSO.4.64.0707151732540.129@neerbosch.nijmegen.internl.net> References: <Pine.GSO.4.64.0707102349300.24847@neerbosch.nijmegen.internl.net> <Pine.GSO.4.64.0707151732540.129@neerbosch.nijmegen.internl.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Michiel Boland wrote this message on Sun, Jul 15, 2007 at 17:43 +0200:
> It looks like gcc mis-compiles /usr/src/crypto/openssh/monitor_fdpass.c on
> sparc64. For some reason it optimizes away the assignment of fd on line
> 132:
>
> fd = (*(int *)CMSG_DATA(cmsg));
>
> So I guess that every call to mm_receive_fd will return an undefined
> value.
>
> If I add -O0 to CFLAGS in /usr/src/secure/lib/libssh/Makefile, ssh with
> UsePrivilegeSeparation=yes works again.
>
> So, obviously a gcc bug. I will try to generate a smaller test-case for
> this.
Could you give us an assembly dump of the two differences? phk thinks
there might be an issue w/ the CMSG_DATA macro, and a quick glance
makes me question it too... It looks scare on platforms that require
aligned accesses...
--
John-Mark Gurney Voice: +1 415 225 5579
"All that I will do, has been done, All that I have, has not."
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070715181732.GR1221>