| Issue |
172777
|
| Summary |
[lld/ELF] eh_frame_hdr generation only supports 32-bit values
|
| Labels |
lld:ELF
|
| Assignees |
|
| Reporter |
aeubanks
|
trying the large code model on some binaries, we're running into lld complaining about eh_frame_hdr values not fitting in 32 bits:
```
$ cat /tmp/a.c
int main() {}
$ cat /tmp/bigtext.lds
SECTIONS
{
.ltext :
{
*(.ltext)
*(.ltext.*)
. = ALIGN(4);
. = MAX(., 0x100000000);
}
}
$ clang -fuse-ld=lld -mcmodel=large -T /tmp/bigtext.lds /tmp/a.c -o /tmp/a
ld.lld: error: /tmp/a-bdc8cb.o:(.eh_frame): PC offset is too large: 0xffffffff000002ac
```
according to https://refspecs.linuxfoundation.org/LSB_1.3.0/gLSB/gLSB/ehframehdr.html, 64 bit values are allowed by specifying `DW_EH_PE_sdata8`
lld code that assumes 32 bit eh_frame_hdr values:
https://github.com/llvm/llvm-project/blob/796fafeff92fe5d2d20594859e92607116e30a16/lld/ELF/SyntheticSections.cpp#L616
https://github.com/llvm/llvm-project/blob/796fafeff92fe5d2d20594859e92607116e30a16/lld/ELF/SyntheticSections.cpp#L638
would lld be ok with going to 64 bit values in eh_frame_hdr when any value doesn't fit in 32 bits?
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
