Re: Kernel build with gcc 8 a lot of warnings

[Josh Poimboeuf]

On Tue, May 08, 2018 at 04:25:15PM +0200, Greg KH wrote:
> On Tue, May 08, 2018 at 08:34:47AM -0500, Josh Poimboeuf wrote:
> > On Tue, May 08, 2018 at 07:51:26AM +0200, Greg KH wrote:
> > > On Sun, May 06, 2018 at 11:54:53PM -0500, Josh Poimboeuf wrote:
> > > > On Sat, May 05, 2018 at 09:21:12PM +0200, damian wrote:
> > > > > Helllo together,
> > > > > 
> > > > > Hello everybody,
> > > > > 
> > > > > is something special to note when kernel build with gcc 8? I receive various warnings from the objtool:
> > > > > With GCC 7 works all fine.
> > > > > 
> > > > > kernel/cgroup/cgroup.o: warning: objtool: parse_cgroup_root_flags()+0x40: sibling call from callable instruction with modified stack frame
> > > > > kernel/cgroup/cgroup.o: warning: objtool: cgroup_addrm_files()+0x2a3: sibling call from callable instruction with modified stack frame
> > > > > kernel/cgroup/cgroup.o: warning: objtool: cgroup_apply_control_enable()+0x20b: sibling call from callable instruction with modified stack frame
> > > > > kernel/cgroup/cgroup.o: warning: objtool: rebind_subsystems()+0x296: sibling call from callable instruction with modified stack frame
> > > > > kernel/cgroup/cgroup.o: warning: objtool: cgroup_sk_alloc_disable()+0x10: sibling call from callable instruction with modified stack frame
> > > > > kernel/cgroup/cgroup.o: warning: objtool: parse_cgroup_root_flags.cold.45()+0xa: call without frame pointer save/setup
> > > > > kernel/cgroup/cgroup.o: warning: objtool: cgroup_addrm_files.cold.46()+0x17: call without frame pointer save/setup
> > > > > kernel/cgroup/cgroup.o: warning: objtool: cgroup_apply_control_enable.cold.47()+0x24: call without frame pointer save/setup
> > > > > kernel/cgroup/cgroup.o: warning: objtool: rebind_subsystems.cold.48()+0x10: call without frame pointer save/setup
> > > > > kernel/cgroup/cgroup.o: warning: objtool: cgroup_sk_alloc_disable.cold.49()+0x7: call without frame pointer save/setup
> > > > >   CC      kernel/cgroup/stat.o
> > > > >   CC [M]  arch/x86/kvm/../../../virt/kvm/kvm_main.o
> > > > > arch/x86/kernel/acpi/boot.o: warning: objtool: acpi_register_lapic()+0x10: sibling call from callable instruction with modified stack frame
> > > > > arch/x86/kernel/acpi/boot.o: warning: objtool: acpi_map_cpu()+0x2c: sibling call from callable instruction with modified stack frame
> > > > > arch/x86/kernel/acpi/boot.o: warning: objtool: acpi_register_lapic.cold.6()+0x7: call without frame pointer save/setup
> > > > > arch/x86/kernel/acpi/boot.o: warning: objtool: acpi_map_cpu.cold.7()+0x7: call without frame pointer save/setup
> > > > 
> > > > Thanks, this is a known issue with GCC 8 that I haven't gotten a chance
> > > > to fix yet.  They're related to the fact that GCC 8 is splitting out
> > > > 'unlikely' code into 'cold' subfunctions in .text.unlikely.
> > > > 
> > > > I've got some old patches that fix it.  Now that GCC 8 is out, I'll need
> > > > get the patches dusted off and cleaned up for merging.
> > > 
> > > Any pointers to these patches?  Archi linux just changed their default
> > > compiler to gcc 8 and now I'm seeing this in my local builds :(
> > 
> > I'm still dusting the code off, but this fixes most of the warnings.
> > The rest of the code should hopefully be ready soon.
> > 
> > From: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
> > Subject: [PATCH] objtool: Detect GCC 8 cold subfunctions
> > 
> > GCC 8 moves a lot of unlikely code out of line to "cold" subfunctions in
> > .text.unlikely.  Properly detect the new subfunctions and treat them as
> > extensions of the original functions.
> > 
> > This fixes a bunch of warnings like:
> > 
> >   kernel/cgroup/cgroup.o: warning: objtool: parse_cgroup_root_flags()+0x33: sibling call from callable instruction with modified stack frame
> >   kernel/cgroup/cgroup.o: warning: objtool: cgroup_addrm_files()+0x290: sibling call from callable instruction with modified stack frame
> >   kernel/cgroup/cgroup.o: warning: objtool: cgroup_apply_control_enable()+0x25b: sibling call from callable instruction with modified stack frame
> >   kernel/cgroup/cgroup.o: warning: objtool: rebind_subsystems()+0x325: sibling call from callable instruction with modified stack frame
> > 
> > Reported-by: Arnd Bergmann <arnd@xxxxxxxx>
> > Signed-off-by: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
> > ---
> >  tools/objtool/check.c | 7 ++++---
> >  tools/objtool/elf.c   | 3 +++
> >  tools/objtool/elf.h   | 1 +
> >  3 files changed, 8 insertions(+), 3 deletions(-)
> 
> <snip>
> 
> Much nicer, thanks for the patch, seems to build things "quieter" now.
> For the other warnings, they are "safe" to ignore, right?  It's just
> objtool checking to see if all is ok, but the result should be fine no
> matter what, right?

Yeah, from what I can tell, there's nothing catastrophic in the rest of
the warnings.  Worst case, the ORC unwinder might get confused in a few
places.

-- 
Josh