Kees Cook via llvm-dev
2019-Apr-04 20:21 UTC
[llvm-dev] question about --emit-relocs with lld
Hi,
While doing Linux kernel builds linked with lld, I've tracked down a
difference that breaks relocation of the kernel image (e.g. under
KASLR[1]). Some relocations are changed to ABS (weirdly, all are in
.rodata section). Note the difference below in the resulting linked
output.
.L__const._start.instance becomes *ABS* only under lld:
$ cat minimal.c
struct minimal {
void *pointer;
int value;
};
void _start(void)
{
struct minimal instance = {
.value = 1,
};
}
$ llvm-build/x86/bin/clang -c minimal.c
$ /usr/bin/ld.bfd --emit-relocs minimal.o -o minimal.bfd
$ llvm-build/x86/bin/ld.lld --emit-relocs minimal.o -o minimal.lld
$ objdump -Sdr minimal.bfd
...
00000000004000b0 <_start>:
4000b0: 55 push %rbp
4000b1: 48 89 e5 mov %rsp,%rbp
4000b4: 48 8b 04 25 d0 00 40 mov 0x4000d0,%rax
4000bb: 00
4000b8: R_X86_64_32S .rodata
4000bc: 48 89 45 f0 mov %rax,-0x10(%rbp)
4000c0: 48 8b 04 25 d8 00 40 mov 0x4000d8,%rax
4000c7: 00
4000c4: R_X86_64_32S .L__const._start.instance+0x8
4000c8: 48 89 45 f8 mov %rax,-0x8(%rbp)
4000cc: 5d pop %rbp
4000cd: c3 retq
$ objdump -Sdr minimal.lld
...
0000000000201000 <_start>:
201000: 55 push %rbp
201001: 48 89 e5 mov %rsp,%rbp
201004: 48 8b 04 25 20 01 20 mov 0x200120,%rax
20100b: 00
201008: R_X86_64_32S .rodata
20100c: 48 89 45 f0 mov %rax,-0x10(%rbp)
201010: 48 8b 04 25 28 01 20 mov 0x200128,%rax
201017: 00
201014: R_X86_64_32S *ABS*+0x8
201018: 48 89 45 f8 mov %rax,-0x8(%rbp)
20101c: 5d pop %rbp
20101d: c3 retq
I'm not sure where to start looking for solving this...
Thanks!
-Kees
[1] https://github.com/ClangBuiltLinux/linux/issues/404
--
Kees Cook
Tom Stellard via llvm-dev
2019-Apr-04 20:42 UTC
[llvm-dev] question about --emit-relocs with lld
On 04/04/2019 01:21 PM, Kees Cook via llvm-dev wrote:> Hi, > > While doing Linux kernel builds linked with lld, I've tracked down a > difference that breaks relocation of the kernel image (e.g. under > KASLR[1]). Some relocations are changed to ABS (weirdly, all are in > .rodata section). Note the difference below in the resulting linked > output.Can you file a bug for this. -Tom> .L__const._start.instance becomes *ABS* only under lld: > > $ cat minimal.c > struct minimal { > void *pointer; > int value; > }; > > void _start(void) > { > struct minimal instance = { > .value = 1, > }; > } > $ llvm-build/x86/bin/clang -c minimal.c > $ /usr/bin/ld.bfd --emit-relocs minimal.o -o minimal.bfd > $ llvm-build/x86/bin/ld.lld --emit-relocs minimal.o -o minimal.lld > $ objdump -Sdr minimal.bfd > ... > 00000000004000b0 <_start>: > 4000b0: 55 push %rbp > 4000b1: 48 89 e5 mov %rsp,%rbp > 4000b4: 48 8b 04 25 d0 00 40 mov 0x4000d0,%rax > 4000bb: 00 > 4000b8: R_X86_64_32S .rodata > 4000bc: 48 89 45 f0 mov %rax,-0x10(%rbp) > 4000c0: 48 8b 04 25 d8 00 40 mov 0x4000d8,%rax > 4000c7: 00 > 4000c4: R_X86_64_32S .L__const._start.instance+0x8 > 4000c8: 48 89 45 f8 mov %rax,-0x8(%rbp) > 4000cc: 5d pop %rbp > 4000cd: c3 retq > > $ objdump -Sdr minimal.lld > ... > 0000000000201000 <_start>: > 201000: 55 push %rbp > 201001: 48 89 e5 mov %rsp,%rbp > 201004: 48 8b 04 25 20 01 20 mov 0x200120,%rax > 20100b: 00 > 201008: R_X86_64_32S .rodata > 20100c: 48 89 45 f0 mov %rax,-0x10(%rbp) > 201010: 48 8b 04 25 28 01 20 mov 0x200128,%rax > 201017: 00 > 201014: R_X86_64_32S *ABS*+0x8 > 201018: 48 89 45 f8 mov %rax,-0x8(%rbp) > 20101c: 5d pop %rbp > 20101d: c3 retq > > I'm not sure where to start looking for solving this... > > Thanks! > > -Kees > > [1] https://github.com/ClangBuiltLinux/linux/issues/404 >
Kees Cook via llvm-dev
2019-Apr-04 20:48 UTC
[llvm-dev] question about --emit-relocs with lld
On Thu, Apr 4, 2019 at 1:42 PM Tom Stellard <tstellar at redhat.com> wrote:> > On 04/04/2019 01:21 PM, Kees Cook via llvm-dev wrote: > > Hi, > > > > While doing Linux kernel builds linked with lld, I've tracked down a > > difference that breaks relocation of the kernel image (e.g. under > > KASLR[1]). Some relocations are changed to ABS (weirdly, all are in > > .rodata section). Note the difference below in the resulting linked > > output. > > Can you file a bug for this.Sure! Done: https://bugs.llvm.org/show_bug.cgi?id=41385 Thanks! -- Kees Cook
Rui Ueyama via llvm-dev
2019-Apr-05 01:22 UTC
[llvm-dev] question about --emit-relocs with lld
This seems like a bug in lld. Here is the cause of the bug. Symbol whose name begins with ".L" are local symbols. Usually such symbols are discarded by the assembler because relocations relative to local symbols can be replaced by ones that are relative to beginning of sections. But there's one case in which the assembler cannot do that. Local symbols in mergeable sections cannot be replaced, because mergeable sections are split by the linker and reassembled, so their offsets from beginning of sections are not computable at assemble-time. The linker recognizes remaining ".L" symbols and discards them. lld implements that behavior, but it discards the symbols even when --emit-relocs is given. This is the cause of the bug. We should keep ".L" symbols when --emit-relocs is given. Looks like that's what GNU linkers do as well. On Fri, Apr 5, 2019 at 5:21 AM Kees Cook via llvm-dev < llvm-dev at lists.llvm.org> wrote:> Hi, > > While doing Linux kernel builds linked with lld, I've tracked down a > difference that breaks relocation of the kernel image (e.g. under > KASLR[1]). Some relocations are changed to ABS (weirdly, all are in > .rodata section). Note the difference below in the resulting linked > output. > .L__const._start.instance becomes *ABS* only under lld: > > $ cat minimal.c > struct minimal { > void *pointer; > int value; > }; > > void _start(void) > { > struct minimal instance = { > .value = 1, > }; > } > $ llvm-build/x86/bin/clang -c minimal.c > $ /usr/bin/ld.bfd --emit-relocs minimal.o -o minimal.bfd > $ llvm-build/x86/bin/ld.lld --emit-relocs minimal.o -o minimal.lld > $ objdump -Sdr minimal.bfd > ... > 00000000004000b0 <_start>: > 4000b0: 55 push %rbp > 4000b1: 48 89 e5 mov %rsp,%rbp > 4000b4: 48 8b 04 25 d0 00 40 mov 0x4000d0,%rax > 4000bb: 00 > 4000b8: R_X86_64_32S .rodata > 4000bc: 48 89 45 f0 mov %rax,-0x10(%rbp) > 4000c0: 48 8b 04 25 d8 00 40 mov 0x4000d8,%rax > 4000c7: 00 > 4000c4: R_X86_64_32S > .L__const._start.instance+0x8 > 4000c8: 48 89 45 f8 mov %rax,-0x8(%rbp) > 4000cc: 5d pop %rbp > 4000cd: c3 retq > > $ objdump -Sdr minimal.lld > ... > 0000000000201000 <_start>: > 201000: 55 push %rbp > 201001: 48 89 e5 mov %rsp,%rbp > 201004: 48 8b 04 25 20 01 20 mov 0x200120,%rax > 20100b: 00 > 201008: R_X86_64_32S .rodata > 20100c: 48 89 45 f0 mov %rax,-0x10(%rbp) > 201010: 48 8b 04 25 28 01 20 mov 0x200128,%rax > 201017: 00 > 201014: R_X86_64_32S *ABS*+0x8 > 201018: 48 89 45 f8 mov %rax,-0x8(%rbp) > 20101c: 5d pop %rbp > 20101d: c3 retq > > I'm not sure where to start looking for solving this... > > Thanks! > > -Kees > > [1] https://github.com/ClangBuiltLinux/linux/issues/404 > > -- > Kees Cook > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org > https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20190405/118b895c/attachment.html>
Seemingly Similar Threads
- Creating yyyymm regexp strings on the fly for aggregation.
- [lld][ELF] Addends adjustment for relocatable object
- [llvm-mc] FreeBSD kernel module performance impact when upgrading clang
- lld mishandling R_X86_64_PC32 relocations
- Compiling R with --enable-R-shlib for rpy2 error