[PATCH 2/3] elf: Replace a --defsym trick with an object file to be compatible with lld

Fri Jan 29 18:04:57 GMT 2021

On 29/01/2021 12:29, H.J. Lu wrote:
> On Fri, Jan 29, 2021 at 7:26 AM Adhemerval Zanella via Libc-alpha
> <libc-alpha@sourceware.org> wrote:
>>
>>
>>
>> On 27/01/2021 22:03, Fāng-ruì Sòng via Libc-alpha wrote:
>>> On Wed, Jan 20, 2021 at 10:51 AM Fāng-ruì Sòng <maskray@google.com> wrote:
>>>>
>>>> On Tue, Jan 19, 2021 at 4:31 AM H.J. Lu <hjl.tools@gmail.com> wrote:
>>>>>
>>>>> On Tue, Jan 19, 2021 at 12:50 AM Fāng-ruì Sòng <maskray@google.com> wrote:
>>>>>>
>>>>>> On Mon, Jan 18, 2021 at 4:04 PM H.J. Lu <hjl.tools@gmail.com> wrote:
>>>>>>>
>>>>>>> On Mon, Dec 28, 2020 at 11:50 AM Fangrui Song via Libc-alpha
>>>>>>> <libc-alpha@sourceware.org> wrote:
>>>>>>>>
>>>>>>>> The existing code specifies -Wl,--defsym=malloc=0 and other malloc.os
>>>>>>>> definitions before libc_pic.a so that libc_pic.a(malloc.os) is not
>>>>>>>> fetched. This trick is used to avoid multiple definition errors which
>>>>>>>> would happen as a chain result:
>>>>>>>>
>>>>>>>>   dl-allobjs.os has an undefined __libc_scratch_buffer_set_array_size
>>>>>>>>   __libc_scratch_buffer_set_array_size fetches libc_pic.a(scratch_buffer_set_array_size.os)
>>>>>>>>   libc_pic.a(scratch_buffer_set_array_size.os) has an undefined free
>>>>>>>>   free fetches libc_pic.a(malloc.os)
>>>>>>>>   libc_pic.a(malloc.os) has an undefined __libc_message
>>>>>>>>   __libc_message fetches libc_pic.a(libc_fatal.os)
>>>>>>>>
>>>>>>>>   libc_fatal.os will cause a multiple definition error (__GI___libc_fatal)
>>>>>>>>   >>> defined at dl-fxstatat64.c
>>>>>>>>   >>>            /tmp/p/glibc/Release/elf/dl-allobjs.os:(__GI___libc_fatal)
>>>>>>>>   >>> defined at libc_fatal.c
>>>>>>>>   >>>            libc_fatal.os:(.text+0x240) in archive /tmp/p/glibc/Release/libc_pic.a
>>>>>>>>
>>>>>>>> lld processes --defsym after all input files, so this trick does not
>>>>>>>> suppress multiple definition errors with lld. Split the step into two
>>>>>>>> and use an object file to make the intention more obvious and make lld
>>>>>>>> work.
>>>>>>>>
>>>>>>>> This is conceptually more appropriate because --defsym defines a SHN_ABS
>>>>>>>> symbol while a normal definition is relative to the image base.
>>>>>>>>
>>>>>>>
>>>>>>> This is irrelevant since  --defsym was used to create a temporary file which was
>>>>>>> removed immediately after it was created:
>>>>>>>
>>>>>>> /usr/bin/gcc   -nostdlib -nostartfiles -r -o
>>>>>>> /export/users/hjl/build/gnu/tools-build/glibc-gitlab/build-x86_64-linux/elf/librtld.map.o
>>>>>>> -Wl,--defsym=calloc=0 -Wl,--defsym=free=0 -Wl,--defsym=malloc=0
>>>>>>> -Wl,--defsym=realloc=0 -Wl,--defsym=__stack_chk_fail=0
>>>>>>> -Wl,--defsym=__stack_chk_fail_local=0 \
>>>>>>>         '-Wl,-('
>>>>>>> /export/users/hjl/build/gnu/tools-build/glibc-gitlab/build-x86_64-linux/elf/dl-allobjs.os
>>>>>>> /export/users/hjl/build/gnu/tools-build/glibc-gitlab/build-x86_64-linux/libc_pic.a
>>>>>>> -lgcc '-Wl,-)' -Wl,-Map,/export/users/hjl/build/gnu/tools-build/glibc-gitlab/build-x86_64-linux/elf/librtld.mapT
>>>>>>> rm -f /export/users/hjl/build/gnu/tools-build/glibc-gitlab/build-x86_64-linux/elf/librtld.map.o
>>>>>>> mv -f /export/users/hjl/build/gnu/tools-build/glibc-gitlab/build-x86_64-linux/elf/librtld.mapT
>>>>>>> /export/users/hjl/build/gnu/tools-build/glibc-gitlab/build-x86_64-linux/elf/librtld.map
>>>>>>>
>>>>>>>> ---
>>>>>>>>  elf/Makefile | 11 ++++-------
>>>>>>>>  1 file changed, 4 insertions(+), 7 deletions(-)
>>>>>>>>
>>>>>>>> diff --git a/elf/Makefile b/elf/Makefile
>>>>>>>> index 0b4d78c874..299bf24b49 100644
>>>>>>>> --- a/elf/Makefile
>>>>>>>> +++ b/elf/Makefile
>>>>>>>> @@ -524,10 +524,6 @@ rtld-stubbed-symbols = \
>>>>>>>>    malloc \
>>>>>>>>    realloc \
>>>>>>>>
>>>>>>>> -# The GCC arguments that implement $(rtld-stubbed-symbols).
>>>>>>>> -rtld-stubbed-symbols-args = \
>>>>>>>> -  $(patsubst %,-Wl$(comma)--defsym=%=0, $(rtld-stubbed-symbols))
>>>>>>>> -
>>>>>>>>  ifeq ($(have-ssp),yes)
>>>>>>>>  # rtld is not built with the stack protector, so these references will
>>>>>>>>  # go away in the rebuilds.
>>>>>>>> @@ -536,9 +532,10 @@ endif
>>>>>>>>
>>>>>>>>  $(objpfx)librtld.map: $(objpfx)dl-allobjs.os $(common-objpfx)libc_pic.a
>>>>>>>>         @-rm -f $@T
>>>>>>>> -       $(reloc-link) -o $@.o $(rtld-stubbed-symbols-args) \
>>>>>>>> -               '-Wl,-(' $^ -lgcc '-Wl,-)' -Wl,-Map,$@T
>>>>>>>> -       rm -f $@.o
>>>>>>>> +       echo '$(patsubst %,.globl %;,$(rtld-stubbed-symbols)) $(patsubst %,%:,$(rtld-stubbed-symbols))' | \
>>>>>>>> +               $(CC) -o $@T.o $(ASFLAGS) -c -x assembler -
>>>>>>>
>>>>>>> Please generate
>>>>>>>
>>>>>>> .globl symbol;
>>>>>>> symbol = 0;
>>>>>>>
>>>>>>> to make it closer to -Wl,--defsym=symbol=0.
>>>>>>
>>>>>> If symbol: and symbol = 0; work, isn't symbol: slightly better because
>>>>>> it looks more normal?
>>>>>> Or is the intention here using SHN_ABS to make it clear these symbols
>>>>>> are special?
>>>>>
>>>>> Generate machine independent assembly codes is tricky.   We should
>>>>> make before and after the change as close as possible to avoid ANY
>>>>> surprises.   BTW, you need to check HAVE_ASM_SET_DIRECTIVE
>>>>> to decide using ".set" vs "=".
>>>>>
>>>>> --
>>>>> H.J.
>>>>
>>>> Thanks for the tip. In this case I think the patch as is (sticking
>>>> with `foo:`, don't use `foo = 0` or set) is better.
>>>> foo: works everywhere.
>>>
>>> Ping. I think the patch as-is is better than the alternative approach
>>> using either `=` or `.set`.
>>
>> The patch breaks both arc-linux-gnuhf and hppa-linux-gnu with recent
>> ld.bfd 2.26, as below.
>>
>> We will need to sort these out.

The issue was in fact that hppa and arc uses no default asm line separators
(represented by ASM_LINE_SEP within code): hppa uses '!' while arc '`'. 
By replacing echo with printf and issuing a new line explicit I see that
all targets build:

---

diff --git a/elf/Makefile b/elf/Makefile
index 5d666b1b0c..e5c04f682c 100644
--- a/elf/Makefile
+++ b/elf/Makefile
@@ -526,10 +526,6 @@ rtld-stubbed-symbols = \
   malloc \
   realloc \

-# The GCC arguments that implement $(rtld-stubbed-symbols).
-rtld-stubbed-symbols-args = \
-  $(patsubst %,-Wl$(comma)--defsym=%=0, $(rtld-stubbed-symbols))
-
 ifeq ($(have-ssp),yes)
 # rtld is not built with the stack protector, so these references will
 # go away in the rebuilds.
@@ -538,9 +534,10 @@ endif

 $(objpfx)librtld.map: $(objpfx)dl-allobjs.os $(common-objpfx)libc_pic.a
        @-rm -f $@T
-       $(reloc-link) -o $@.o $(rtld-stubbed-symbols-args) \
-               '-Wl,-(' $^ -lgcc '-Wl,-)' -Wl,-Map,$@T
-       rm -f $@.o
+       printf '$(patsubst %,.globl %;\n,$(rtld-stubbed-symbols)) $(patsubst %,% = 0;\n,$(rtld-stubbed-symbols))' | \
+               $(CC) -o $@T.o $(ASFLAGS) -c -x assembler -
+       $(reloc-link) -o $@.o $@T.o '-Wl,-(' $^ -lgcc '-Wl,-)' -Wl,-Map,$@T
+       rm -f %@T.o $@.o
        mv -f $@T $@

 # For lld, skip preceding addresses and values before matching the archive and the member.
