cvs commit: src Makefile.inc1 src/gnu/lib Makefile src/gnu/lib/csu Makefile src/gnu/lib/libssp Makefile src/lib/csu Makefile.inc src/lib/libc Makefile src/lib/libstand Makefile src/lib/libthr Makefile src/libexec/rtld-elf Makefile src/release Makefile ...

[Bernd Walter]

On Mon, Jul 21, 2008 at 11:39:51PM +0200, Jeremie Le Hen wrote:
> Hi Bernd,
> 
> Again, sorry for the late reply, I'm abroad this week once again for
> work, so I have very little time for my personal stuff.
> 
> On Fri, Jul 18, 2008 at 01:40:59PM +0200, Bernd Walter wrote:
> > No change - with SSP all dynamic linked binaries fail the same as the
> > 7-stable did.
> > Without SSP only some binaries fail, but those are the same.
> > And of course since my 7-stable binaries were failing as well this
> > sounds like a kernel thing and not a userland one.
> > 
> > What comes to mind after reading the original commit when reading the
> > compiler options used:
> > cc -mlittle-endian -c -O2 -pipe -fno-strict-aliasing -mcpu=arm9 -std=c99  -Wall -Wredundant-decls -Wnested-externs -Wstrict-prototypes  -Wmissing-prototypes -Wpointer-arith -Winline -Wcast-qual  -Wundef -Wno-pointer-sign -fformat-extensions -nostdinc  -I. -I../../.. -I../../../contrib/altq -D_KERNEL -DHAVE_KERNEL_OPTION_HEADERS -include opt_global.h -fno-common -finline-limit=8000 --param inline-unit-growth=100 --param large-function-growth=1000 -mcpu=arm9 -ffreestanding -fstack-protector -Werror  ../../../arm/at91/at91_twi.c
> > The original message says that -fstack-protector shouldn't be used for
> > the kernel.
> > So why it is there?
> > I never explicitly enabled it anywhere.
> 
> Actually, there was one patch to enable SSP for userland and another one
> for kernel, but they have been committed together and their commitlog
> has been concatenated.  If you read it further, you will see the
> kernel part commitlog.

Yes - I saw later that those were different compiler options.

> Concerning the automatic activation, this is intended so as to give SSP
> more exposure.  We'll need to reach a consensus for the release about
> whether it should be opt-in or opt-out.  For now, you can disable it for
> both userland and kernel using the WITHOUT_SPP knob in src.conf(5).  If
> you only want to disable it for kernel build, simply use "make
> -DWITHOUT_SSP buildkernel".
> 
> I'm keen to hear about any clue you will gather about the problem.

Well Olivier Houchard went in and disabled SSP support on arm.
He mentioned in his commit log a compiler bug, but you should ask him
about the details.
Unfortunately he missed one file, but after disabling this as well
and recompiling the kernel, my problems were gone.

-- 
B.Walter <bernd at bwct.de> http://www.bwct.de
Modbus/TCP Ethernet I/O Baugruppen, ARM basierte FreeBSD Rechner uvm.