[ Adding PeterZ and Jiri to the participants. ]
Looks like 5.18.13 added that commit 9bb2ec608a20 ("objtool: Update
Retpoline validation") but I don't see 3131ef39fb03 ("x86/asm/32: Fix
ANNOTATE_UNRET_SAFE use on 32-bit") in that list.
That said, 3131ef39fb03 should have fixed a completely different issue
on 32-bit, not the "naked ret" thing.
PeterZ, Jiri, any ideas? Limited quoting below, see thread at
https://lore.kernel.org/all/CA+G9fYsJBBbEXowA-3kxDNqcfbtcqmxBrEnJSkCnLUsMzNfJZw@xxxxxxxxxxxxxx/
for more details.
Linus
On Wed, Jul 20, 2022 at 9:37 AM Justin Forbes <jforbes@xxxxxxxxxxxxxxxxx> wrote:
>
> On Tue, Jul 19, 2022 at 12:32:48PM -0700, Linus Torvalds wrote:
> > On Tue, Jul 19, 2022 at 10:57 AM Naresh Kamboju
> > <naresh.kamboju@xxxxxxxxxx> wrote:
> > >
> > >
> > > 2. Large number of build warnings on x86 with gcc-11,
> > > I do not see these build warnings on mainline,
> > ..
> > > 'naked' return found in RETPOLINE build
> >
> > Hmm. Does your cross-compiler support '-mfunction-return=thunk-extern'?
> >
> > Your build does magic things with 'scripts/kconfig/merge_config.sh',
> > and I'm wondering if you perhaps end up enabling CONFIG_RETHUNK with a
> > compiler that doesn't actually support it, or something like that?
>
> I am seeing these 'naked' return found in RETPOLINE build on the
> standard fedora 36 toolchain as well. No cross compiling, nothing fancy.
> These were not seen with mainline, or with the 5.18.12-rc1 retbleed
> patches.
>
> Justin
