Re: drivers/scsi/.tmp_mc_st.s:3: Error: invalid operands (*UND* and *UND* sections) for `^'

From: Greentime Hu
Date: Fri Dec 06 2019 - 05:15:50 EST


Greentime Hu <green.hu@xxxxxxxxx> æ 2019å12æ6æ éä äå4:38åéï
>
> Greentime Hu <green.hu@xxxxxxxxx> æ 2019å12æ6æ éä äå4:19åéï
> >
> > Arnd Bergmann <arnd@xxxxxxxx> æ 2019å12æ6æ éä äå4:00åéï
> > >
> > > On Fri, Dec 6, 2019 at 2:05 AM kbuild test robot <lkp@xxxxxxxxx> wrote:
> > > >
> > > > tree: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
> > > > head: 2f13437b8917627119d163d62f73e7a78a92303a
> > > > commit: 1207045da5a7c94344e0ea9a9e7495985eef499a compat_ioctl: move tape handling into drivers
> > > > date: 6 weeks ago
> > > > config: nds32-allyesconfig (attached as .config)
> > > > compiler: nds32le-linux-gcc (GCC) 9.2.0
> > > > reproduce:
> > > > wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
> > > > chmod +x ~/bin/make.cross
> > > > git checkout 1207045da5a7c94344e0ea9a9e7495985eef499a
> > > > # save the attached .config to linux build tree
> > > > GCC_VERSION=9.2.0 make.cross ARCH=nds32
> > > >
> > > > If you fix the issue, kindly add following tag
> > > > Reported-by: kbuild test robot <lkp@xxxxxxxxx>
> > > >
> > > > All errors (new ones prefixed by >>):
> > > >
> > > > drivers/scsi/.tmp_mc_st.s: Assembler messages:
> > > > >> drivers/scsi/.tmp_mc_st.s:3: Error: invalid operands (*UND* and *UND* sections) for `^'
> > > > drivers/scsi/.tmp_mc_st.s:4: Error: invalid operands (*UND* and *UND* sections) for `^'
> > > > drivers/scsi/.tmp_mc_st.s:5: Error: invalid operands (*UND* and *UND* sections) for `^'
> > > > drivers/scsi/.tmp_mc_st.s:6: Error: invalid operands (*UND* and *UND* sections) for `^'
> > > > drivers/scsi/.tmp_mc_st.s:7: Error: invalid operands (*UND* and *UND* sections) for `^'
> > > > drivers/scsi/.tmp_mc_st.s:8: Error: invalid operands (*UND* and *UND* sections) for `^'
> > > > drivers/scsi/.tmp_mc_st.s:9: Error: invalid operands (*UND* and *UND* sections) for `^'
> > >
> > > Adding nds32 maintainers to Cc:
> > >
> > > It looks like a regression caused by my patch, but I don't think it's something
> > > I did wrong, but rather a toolchain bug being uncovered by the modified sources.
> > >
> > > Are you able to reproduce this?
> >
> > Hi Arnd,
> >
> > I am trying to reproduce this problem, but it happened to me.
> >
> > greentimeh@gamma07:/scratch/greentimeh/nds32/linux <(1207045da5a7...)>
> > $ GCC_VERSION=9.2.0 make.cross ARCH=nds32
> > cd: received redirection to `https://download.01.org/0day-ci/cross-package/'
> > Cannot find nds32-linux under
> > https://download.01.org/0day-ci/cross-package check
> > /tmp/crosstool-files
> >
> > Can you reproduce it?
>
> I can reproduce it now by adding these 2 lines to make.cross.
> 177 nds32)
> 178 gcc_arch=nds32le-linux
>
> It will be built failed with gcc-9.2.0-nolibc toolchain, but it can be
> built pass with gcc-8.1.0-nolibc toolchain.

It seems nds32le-linux-objdump -hdr xxx.o will generate different strings.
For example:
00000050 <.L2^B1>:
This string is different when generated by 9.2 toolchain and 8.1 toolchain.
"^B" is treated as 2 character in 9.2 toolchain
"^B" is treated as 1 character in 8.1 toolchain

This causes ./scripts/recordmcount.pl to call `$cc -o $mcount_o -c
$mcount_s`; failed to cause this issue.
I have a quick patch to fix this build error, but I am not sure if it
is a good solution.
Maybe Nick could have a look at it?

diff --git a/scripts/recordmcount.pl b/scripts/recordmcount.pl
index 3f77a5d695c1..807b61fc5f5d 100755
--- a/scripts/recordmcount.pl
+++ b/scripts/recordmcount.pl
@@ -551,7 +551,8 @@ while (<IN>) {
# if this is either a local function or a weak function
# keep looking for functions that are global that
# we can use safely.
- if (!defined($locals{$text}) && !defined($weak{$text})) {
+ if (!defined($locals{$text}) && !defined($weak{$text}) &&
+ $text !~ /^\.L/) {
$ref_func = $text;
$read_function = 0;
$offset = hex $1;