patch to install unstripped vDSO on disk breaks powerpc kernel build
I'm using gcc 3.4.3 and binutils 2.15.92.0.2. I originally saw the following problem in 2.6.25 but it's present in git head as well:
CALL /home/cfriesen/kernels/2.6.25/linux-2.6.25/scripts/checksyscalls.sh
CHK include/linux/compile.h
CALL/home/cfriesen/kernels/2.6.25/linux-2.6.25/arch/powerpc/kernel/systbl_chk.sh
AS arch/powerpc/kernel/systbl.o
OBJCOPY arch/powerpc/kernel/vdso64/vdso64.so
BFD: arch/powerpc/kernel/vdso64/vdso64.so: The first section in the PT_DYNAMIC segment is not the .dynamic section /home/cfriesen/bin/ppc64-R9a/bin/ppc64-R9a-objcopy: arch/powerpc/kernel/vdso64/vdso64.so: Bad value
BFD: arch/powerpc/kernel/vdso64/vdso64.so: The first section in the PT_DYNAMIC segment is not the .dynamic section /home/cfriesen/bin/ppc64-R9a/bin/ppc64-R9a-objcopy: arch/powerpc/kernel/vdso64/vdso64.so: Bad value make[3]: *** [arch/powerpc/kernel/vdso64/vdso64.so] Error 1
make[2]: *** [arch/powerpc/kernel/vdso64] Error 2
make[1]: *** [arch/powerpc/kernel] Error 2
make: *** [sub-make] Error 2
I installed git and bisected the problem. It turns out that it's due to the following commit:
commit 8150caad02266623b5b9f58088d589f130fccd97
Author: Roland McGrath <[email protected]> Date: Tue Oct 2 13:30:04 2007 -0700
[POWERPC] powerpc vDSO: install unstripped copies on disk
This keeps an unstripped copy of the vDSO images built before they are stripped and embedded in the kernel. The unstripped copies get installed in $(MODLIB)/vdso/ by "make install". These files can be useful when they contain source-level debugging information.
Anyone have any idea what's going on?
Thanks,
Chris
I haven't seen that error before. Can you show the output of {eu-,}readelf -lS on vdso64.so.dbg, and also on the vdso64.so successfully built when you revert the patch?
Thanks,
Roland
Quoted 3 lines · Roland McGrath
I haven't seen that error before. Can you show the output of {eu-,}readelf -lS on vdso64.so.dbg, and also on the vdso64.so successfully built when you revert the patch?
I've attached the raw output from the two commands. The delta between the two is as follows:
-There are 14 section headers, starting at offset 0xd58:
+There are 16 section headers, starting at offset 0xd68:
Section Headers:
[Nr] Name Type Address Offset
@@ -33,7 +33,11 @@
[12] .rela.dyn RELA 0000000000100be8 00000be8
0000000000000000 0000000000000018 A 2 0 8
[13] .shstrtab STRTAB 0000000000000000 00000cd8
- 0000000000000079 0000000000000000 0 0 1
+ 0000000000000089 0000000000000000 0 0 1
+ [14] .symtab SYMTAB 0000000000000000 00001168
+ 0000000000000d80 0000000000000018 15 134 8
+ [15] .strtab STRTAB 0000000000000000 00001ee8
+ 00000000000004a3 0000000000000000 0 0 1 Key to Flags:
W (write), A (alloc), X (execute), M (merge), S (strings) I (info), L (link order), G (group), x (unknown)
Also, I'm going to be away on holidays next week, so I won't be able to test this further until the 19th or so.
Chris
good:
[cfriesen@localhost linux-2.6]$ /home/cfriesen/bin/ppc64-R9a/bin/ppc64-R9a-readelf -lS ../g5/arch/powerpc/kernel/vdso64/vdso64.so There are 14 section headers, starting at offset 0xd58:
Section Headers:
[Nr] Name Type Address Offset
Size EntSize Flags Link Info Align
[ 0] NULL 0000000000000000 00000000
0000000000000000 0000000000000000 0 0 0
[ 1] .hash HASH 0000000000100120 00000120
000000000000004c 0000000000000004 A 2 0 8
[ 2] .dynsym DYNSYM 0000000000100170 00000170
0000000000000150 0000000000000018 A 3 4 8
[ 3] .dynstr STRTAB 00000000001002c0 000002c0
00000000000000ef 0000000000000000 A 0 0 1
[ 4] .gnu.version VERSYM 00000000001003b0 000003b0
000000000000001c 0000000000000002 A 2 0 2
[ 5] .gnu.version_d VERDEF 00000000001003d0 000003d0
0000000000000038 0000000000000000 A 3 2 8
[ 6] .note NOTE 0000000000100408 00000408
0000000000000018 0000000000000000 A 0 0 4
[ 7] .text PROGBITS 0000000000100420 00000420
00000000000002ec 0000000000000000 AX 0 0 8
[ 8] .eh_frame_hdr PROGBITS 0000000000100710 00000710
0000000000000008 0000000000000000 A 0 0 4
[ 9] .eh_frame PROGBITS 0000000000100718 00000718
00000000000004c4 0000000000000000 A 0 0 8
[10] .got PROGBITS 0000000000100be0 00000be0
0000000000000008 0000000000000008 WA 0 0 8
[11] .dynamic DYNAMIC 0000000000100be8 00000be8
00000000000000f0 0000000000000010 WA 3 0 8
[12] .rela.dyn RELA 0000000000100be8 00000be8
0000000000000000 0000000000000018 A 2 0 8
[13] .shstrtab STRTAB 0000000000000000 00000cd8
0000000000000079 0000000000000000 0 0 1Key to Flags:
W (write), A (alloc), X (execute), M (merge), S (strings)
I (info), L (link order), G (group), x (unknown)
O (extra OS processing required) o (OS specific), p (processor specific)
Elf file type is DYN (Shared object file) Entry point 0x0
There are 4 program headers, starting at offset 64
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
LOAD 0x0000000000000000 0x0000000000100000 0x0000000000100000
0x0000000000000cd8 0x0000000000000cd8 R E 10000
NOTE 0x0000000000000408 0x0000000000100408 0x0000000000100408
0x0000000000000018 0x0000000000000018 R 4
DYNAMIC 0x0000000000000be8 0x0000000000100be8 0x0000000000100be8
0x00000000000000f0 0x00000000000000f0 R 8
GNU_EH_FRAME 0x0000000000000710 0x0000000000100710 0x0000000000100710
0x0000000000000008 0x0000000000000008 R 4 Section to Segment mapping:
Segment Sections...
00 .hash .dynsym .dynstr .gnu.version .gnu.version_d .note .text .eh_frame_hdr .eh_frame .got .dynamic 01 .note 02 .dynamic 03 .eh_frame_hdr
bad:
[cfriesen@localhost linux-2.6]$ /home/cfriesen/bin/ppc64-R9a/bin/ppc64-R9a-readelf -lS ../g5/arch/powerpc/kernel/vdso64/vdso64.so.dbg There are 16 section headers, starting at offset 0xd68:
Section Headers:
[Nr] Name Type Address Offset
Size EntSize Flags Link Info Align
[ 0] NULL 0000000000000000 00000000
0000000000000000 0000000000000000 0 0 0
[ 1] .hash HASH 0000000000100120 00000120
000000000000004c 0000000000000004 A 2 0 8
[ 2] .dynsym DYNSYM 0000000000100170 00000170
0000000000000150 0000000000000018 A 3 4 8
[ 3] .dynstr STRTAB 00000000001002c0 000002c0
00000000000000ef 0000000000000000 A 0 0 1
[ 4] .gnu.version VERSYM 00000000001003b0 000003b0
000000000000001c 0000000000000002 A 2 0 2
[ 5] .gnu.version_d VERDEF 00000000001003d0 000003d0
0000000000000038 0000000000000000 A 3 2 8
[ 6] .note NOTE 0000000000100408 00000408
0000000000000018 0000000000000000 A 0 0 4
[ 7] .text PROGBITS 0000000000100420 00000420
00000000000002ec 0000000000000000 AX 0 0 8
[ 8] .eh_frame_hdr PROGBITS 0000000000100710 00000710
0000000000000008 0000000000000000 A 0 0 4
[ 9] .eh_frame PROGBITS 0000000000100718 00000718
00000000000004c4 0000000000000000 A 0 0 8
[10] .got PROGBITS 0000000000100be0 00000be0
0000000000000008 0000000000000008 WA 0 0 8
[11] .dynamic DYNAMIC 0000000000100be8 00000be8
00000000000000f0 0000000000000010 WA 3 0 8
[12] .rela.dyn RELA 0000000000100be8 00000be8
0000000000000000 0000000000000018 A 2 0 8
[13] .shstrtab STRTAB 0000000000000000 00000cd8
0000000000000089 0000000000000000 0 0 1
[14] .symtab SYMTAB 0000000000000000 00001168
0000000000000d80 0000000000000018 15 134 8
[15] .strtab STRTAB 0000000000000000 00001ee8
00000000000004a3 0000000000000000 0 0 1Key to Flags:
W (write), A (alloc), X (execute), M (merge), S (strings)
I (info), L (link order), G (group), x (unknown)
O (extra OS processing required) o (OS specific), p (processor specific)
Elf file type is DYN (Shared object file) Entry point 0x0
There are 4 program headers, starting at offset 64
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
LOAD 0x0000000000000000 0x0000000000100000 0x0000000000100000
0x0000000000000cd8 0x0000000000000cd8 R E 10000
NOTE 0x0000000000000408 0x0000000000100408 0x0000000000100408
0x0000000000000018 0x0000000000000018 R 4
DYNAMIC 0x0000000000000be8 0x0000000000100be8 0x0000000000100be8
0x00000000000000f0 0x00000000000000f0 R 8
GNU_EH_FRAME 0x0000000000000710 0x0000000000100710 0x0000000000100710
0x0000000000000008 0x0000000000000008 R 4 Section to Segment mapping:
Segment Sections...
00 .hash .dynsym .dynstr .gnu.version .gnu.version_d .note .text .eh_frame_hdr .eh_frame .got .dynamic 01 .note 02 .dynamic 03 .eh_frame_hdr
Ok. That looks like a bug in the older binutils (objcopy) you are using. It is confused by the empty .rela.dyn section right next to the .dynamic section. Current binutils don't seem to have a problem with this.
I haven't tried to get an old binutils running to reproduce this. I suspect the following change will work around the problem. It should not cause any problem with newer or older binutils.
Thanks,
Roland
---
[PATCH] powerpc vdso64 linker script tweak
This should work around bugs in older binutils' objcopy. The placement of these sections does not really matter, but it confused the buggy old BFD libraries.
--- arch/powerpc/kernel/vdso64/vdso64.lds.S | 6 +++--- 1 files changed, 3 insertions(+), 3 deletions(-)
arch/powerpc/kernel/vdso64/vdso64.lds.S +3 -3
@@ -43,15 +43,15 @@ SECTIONS .rodata : { *(.rodata .rodata.* .gnu.linkonce.r.*) } .rodata1 : { *(.rodata1) } + .dynamic : { *(.dynamic) } :text :dynamic + .eh_frame_hdr : { *(.eh_frame_hdr) } :text :eh_frame_hdr .eh_frame : { KEEP (*(.eh_frame)) } :text .gcc_except_table : { *(.gcc_except_table) } + .rela.dyn ALIGN(8) : { *(.rela.dyn) } .opd ALIGN(8) : { KEEP (*(.opd)) } .got ALIGN(8) : { *(.got .toc) } - .rela.dyn ALIGN(8) : { *(.rela.dyn) } - - .dynamic : { *(.dynamic) } :text :dynamic _end = .; PROVIDE(end = .);
Quoted 7 lines · Roland McGrath
Ok. That looks like a bug in the older binutils (objcopy) you are using. It is confused by the empty .rela.dyn section right next to the .dynamic section. Current binutils don't seem to have a problem with this.
I haven't tried to get an old binutils running to reproduce this. I suspect the following change will work around the problem. It should not cause any problem with newer or older binutils.
Thanks, this seems to have fixed the problem.
Chris