Re: [tip: objtool/core] jump_label, x86: Allow short NOPs

[Peter Zijlstra]

On Wed, Jun 30, 2021 at 09:07:05AM +0200, Peter Zijlstra wrote:
> On Tue, Jun 29, 2021 at 09:35:37PM +0100, Matthew Wilcox wrote:
> > On Tue, Jun 29, 2021 at 09:01:25PM +0100, Matthew Wilcox wrote:
> > > So this got merged without the corresponding Kbuild update being merged,
> > > and my kernel failed to boot.  Bisect got as far as
> > > 
> > > $ git bisect good
> > > Bisecting: 4 revisions left to test after this (roughly 2 steps)
> > > [ab3257042c26d0cd44793c741e2f89bf38b21fe8] jump_label, x86: Allow short NOPs
> > > 
> > > before my sluggish memory remembered this thread from six weeks ago.
> > > 
> > > So if anybody else hits this, do a make clean.
> > 
> > Actually, this is a different bug with the same symptom.
> > 
> > Applying the patch from Peter, and running it:
> > 
> > $ ./.build_test_kernel-x86_64/tools/objtool/objtool check -abdJsuld .build_test_kernel-x86_64/vmlinux.o
> > nr_sections: 15446
> > section_bits: 13
> > nr_symbols: 116448
> > symbol_bits: 16
> > max_reloc: 8031700
> > tot_reloc: 12477754
> > reloc_bits: 19
> > nr_insns: 2523443
> > .build_test_kernel-x86_64/vmlinux.o: warning: objtool: want_init_on_free()+0x0: jump-label unpatched
> > 
> > This is against a freshly built kernel -- i removed the build directory,
> > copied in a .config file and built a fresh kernel.
> 
> You happen to have said .config for me?

Also GCC version I suppose. The thing I'm wondering about in particular
is what translation unit is responsible for that symbol.

AFAICT the function itself is an inline from linux/mm.h, but I cannot
find any of the files or functions it's used in as being excluded from
objtool coverage :/