On 08.09.2018 05:57, Masami Hiramatsu wrote: > On Fri, 7 Sep 2018 14:50:59 +0200 > Peter Oberparleiter <oberpar@xxxxxxxxxxxxx> wrote: > >> On 06.09.2018 18:42, Masami Hiramatsu wrote: >>> Peter Oberparleiter <oberpar@xxxxxxxxxxxxx> wrote: >>>> I've attached a quick fix that should address both problems. I'd >>>> appreciate if this patch could get some testing before I post proper fix >>>> patches. >>> >>> Hmm, I'm still not able to reproduce it on powerpc cross build even with >>> CONFIG_LD_DEAD_CODE_DATA_ELIMINATION=y ... So, sorry I couldn't test this patch. >> >> Maybe this is related to the compiler/binutils versions used. I'm using >> Fedora 28's gcc-powerpc64 and binutils-powerpc64 packages: >> >> powerpc64-linux-gnu-gcc (GCC) 7.1.1 20170622 (Red Hat Cross 7.1.1-3) >> GNU ld version 2.29.1-4.fc28 >> >> For reference I'm also attaching the config that I used to reproduce the >> problem on kernel v4.18. >> >>>> diff --git a/arch/arm/kernel/vmlinux.lds.h b/arch/arm/kernel/vmlinux.lds.h >>>> index ae5fdff18406..2ca33277a28b 100644 >>>> --- a/arch/arm/kernel/vmlinux.lds.h >>>> +++ b/arch/arm/kernel/vmlinux.lds.h >>>> @@ -48,6 +48,7 @@ >>>> >>>> #define ARM_DISCARD \ >>>> *(.ARM.exidx.exit.text) \ >>>> + *(.ARM.exidx.text.exit) \ >>> >>> BTW, why would we need this? >> >> That's necessary to fix one of the two ARM linker failures reported via >> https://lkml.org/lkml/2018/8/24/345 >> >>>>> `.text.exit' referenced in section `.ARM.exidx.text.exit' of >>>>> kernel/trace/trace_clock.o: defined in discarded section `.text.exit' >>>>> of kernel/trace/trace_clock.o >> >> Section ".ARM.exidx.text.exit" refers to ".text.exit" which was >> discarded. With the change above, the extraneous section >> ".ARM.exidx.text.exit" is also discarded, resolving the linker failure. > > OK, so your patch fixes following issue too? > > https://patchwork.kernel.org/patch/10584685/ Yes it does, by discarding the text.exit section unconditionally. I don't think we need to keep any exit-related section introduced by GCOV profiling. It is my understanding that the code contained in such a section is only relevant for user-space profiling, to process profiling data when a program ends. Also discarding .ARM.exidx.text.exit seems to be more consistent to how .ARM.exidx.exit.text is handled. > In that case, we should drop above patch. Yes, I agree. -- Peter Oberparleiter Linux on Z Development - IBM Germany
