Re: x86/purgatory: undefined symbol __stack_chk_fail

[Ingo Molnar <mingo@kernel.org>]

* Arvind Sankar <nivedita@alum.mit.edu> wrote:

> On Sat, Sep 28, 2019 at 12:41:29PM +0000, Ingo Molnar wrote:
> > 
> > * Randy Dunlap <rdunlap@infradead.org> wrote:
> > 
> > > On 9/3/19 8:50 AM, Andreas Smas wrote:
> > > > Hi,
> > > > 
> > > > For me, kernels built including this commit
> > > > b059f801a937 (x86/purgatory: Use CFLAGS_REMOVE rather than reset KBUILD_CFLAGS)
> > > > 
> > > > results in kexec() failing to load the kernel:
> > > > 
> > > > kexec: Undefined symbol: __stack_chk_fail
> > > > kexec-bzImage64: Loading purgatory failed
> > > > 
> > > > Can be seen:
> > > > 
> > > > $ readelf -a arch/x86/purgatory/purgatory.ro | grep UND
> > > >      0: 0000000000000000     0 NOTYPE  LOCAL  DEFAULT  UND
> > > >     51: 0000000000000000     0 NOTYPE  GLOBAL DEFAULT  UND __stack_chk_fail
> > > > 
> > > > Using: gcc version 7.4.0 (Ubuntu 7.4.0-1ubuntu1~18.04.1)
> > > > 
> > > > Adding -ffreestanding or -fno-stack-protector to ccflags-y in
> > > > arch/x86/purgatory/Makefile
> > > > fixes the problem. Not sure which would be preferred.
> > > > 
> > > 
> > > Hi,
> > > Do you have a kernel .config file that causes this?
> > > I can't seem to reproduce it.
> > 
> > Does it go away with this fix in x86/urgent:
> > 
> >   ca14c996afe7: ("x86/purgatory: Disable the stackleak GCC plugin for the purgatory")
> > 
> >   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git x86/urgent
> > 
> > ?
> > 
> > Thanks,
> > 
> > 	Ing
> This one was fixed by [1] e16c2983fba0f ("x86/purgatory: Change compiler
> flags from -mcmodel=kernel to -mcmodel=large to fix kexec relocation
> errors") from Steve Wahl, which in addition to changing mcmodel also
> added back -ffreestanding (and -fno-zero-initialized-in-bss). It was
> merged on the 12th. The stackleak one is a different undefined symbol
> error.
> 
> [1] https://marc.info/?l=git-commits-head&m=156829711224800

Great, so all known kexec bugs should be fixed for now in Linus's latest 
kernel, right?

Thanks,

	Ingo