s390/boot: workaround current 'llvm-objdump -t -j ...' behavior

When building with OBJDUMP=llvm-objdump, there are a series of warnings
from the section comparisons that arch/s390/boot/Makefile performs
between vmlinux and arch/s390/boot/vmlinux:

  llvm-objdump: warning: section '.boot.preserved.data' mentioned in a -j/--section option, but not found in any input file
  llvm-objdump: warning: section '.boot.data' mentioned in a -j/--section option, but not found in any input file
  llvm-objdump: warning: section '.boot.preserved.data' mentioned in a -j/--section option, but not found in any input file
  llvm-objdump: warning: section '.boot.data' mentioned in a -j/--section option, but not found in any input file

The warning is a little misleading, as these sections do exist in the
input files. It is really pointing out that llvm-objdump does not match
GNU objdump's behavior of respecting '-j' / '--section' in combination
with '-t' / '--syms':

  $ s390x-linux-gnu-objdump -t -j .boot.data vmlinux.full

  vmlinux.full:     file format elf64-s390

  SYMBOL TABLE:
  0000000001951000 l     O .boot.data     0000000000003000 sclp_info_sccb
  00000000019550e0 l     O .boot.data     0000000000000001 sclp_info_sccb_valid
  00000000019550e2 g     O .boot.data     0000000000001000 early_command_line
  ...

  $ llvm-objdump -t -j .boot.data vmlinux.full

  vmlinux.full:   file format elf64-s390

  SYMBOL TABLE:
  0000000000100040 l     O .text  0000000000000010 dw_psw
  0000000000000000 l    df *ABS*  0000000000000000 main.c
  00000000001001b0 l     F .text  00000000000000c6 trace_event_raw_event_initcall_level
  0000000000100280 l     F .text  0000000000000100 perf_trace_initcall_level
  ...

It may be possible to change llvm-objdump's behavior to match GNU
objdump's behavior but the difficulty of that task has not yet been
explored. The combination of '$(OBJDUMP) -t -j' is not common in the
kernel tree on a whole, so workaround this tool difference by grepping
for the sections in the full symbol table output in a similar manner to
the sed invocation. This results in no visible change for GNU objdump
users while fixing the warnings for OBJDUMP=llvm-objdump, further
enabling use of LLVM=1 for ARCH=s390 with versions of LLVM that have
support for s390 in ld.lld and llvm-objcopy.

Reported-by: Heiko Carstens <hca@linux.ibm.com>
Closes: https://lore.kernel.org/20240219113248.16287-C-hca@linux.ibm.com/
Link: https://github.com/ClangBuiltLinux/linux/issues/859
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Link: https://lore.kernel.org/r/20240220-s390-work-around-llvm-objdump-t-j-v1-1-47bb0366a831@kernel.org
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>

diff --git a/arch/s390/boot/Makefile b/arch/s390/boot/Makefile
index aecafabc20543e64f30b335a174dc1be7c2cffc5..294f08a8811a2211442bbf659cbaddc93cd59dd6 100644 (file)
--- a/arch/s390/boot/Makefile
+++ b/arch/s390/boot/Makefile
@@ -60,9 +60,9 @@ clean-files += vmlinux.map
 
 quiet_cmd_section_cmp = SECTCMP $*
 define cmd_section_cmp
-       s1=`$(OBJDUMP) -t -j "$*" "$<" | sort | \
+       s1=`$(OBJDUMP) -t "$<" | grep "\s$*\s\+" | sort | \
                sed -n "/0000000000000000/! s/.*\s$*\s\+//p" | sha256sum`; \
-       s2=`$(OBJDUMP) -t -j "$*" "$(word 2,$^)" | sort | \
+       s2=`$(OBJDUMP) -t "$(word 2,$^)" | grep "\s$*\s\+" | sort | \
                sed -n "/0000000000000000/! s/.*\s$*\s\+//p" | sha256sum`; \
        if [ "$$s1" != "$$s2" ]; then \
                echo "error: section $* differs between $< and $(word 2,$^)" >&2; \