[PATCH] D157365: [lld][PPC64] Fix location of the least significant byte when fixing up R_PPC64_ADDR14 for little-endian target

[Kai Luo via Phabricator via llvm-commits]

lkail added inline comments.


================
Comment at: lld/ELF/Arch/PPC64.cpp:1224
+    uint8_t aalk = loc[config->isLE ? 0 : 3];
+    write16(config->isLE ? loc : loc + 2, (aalk & 3) | (val & 0xfffc));
     break;
----------------
lkail wrote:
> nemanjai wrote:
> > Is this backwards? Shouldn't it be offset by 2 bytes for LE?
> We should agree with `aalk` is at `loc[0]` for LE. Then we should also agree with `loc[1]` contains the most significant 8-bits of the intermediate number. `write16(loc, ...)` writes `loc[0]` and `loc[1]` while `write(loc + 2, ...)` writes `loc[2]` and `loc[3]`. IIUC, when it's LE, where we want to write at should be `loc[0]` and `loc[1]`.
Here's an example
```
t_be.o:	file format elf64-powerpc

Disassembly of section .text:

0000000000000000 <bca_target>:
       0: 4e 80 00 20  	blr

0000000000000004 <check_reloc_addr14>:
       4: 42 9f 00 02  	bca 20, 31, 0
```

```
t_le.o:	file format elf64-powerpcle

Disassembly of section .text:

0000000000000000 <bca_target>:
       0: 20 00 80 4e  	blr

0000000000000004 <check_reloc_addr14>:
       4: 02 00 9f 42  	bca 20, 31, 0
```


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D157365/new/

https://reviews.llvm.org/D157365