[JITLink][AArch32] Implement R_ARM_PREL31 and process .ARM.exidx sections #79044
Conversation
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
| @@ -397,6 +397,11 @@ template <typename ELFT> Error ELFLinkGraphBuilder<ELFT>::graphifySections() { | |||
| orc::ExecutorAddr(Sec.sh_addr), | |||
| Sec.sh_addralign, 0); | |||
|
|
|||
| if (Sec.sh_type == ELF::SHT_ARM_EXIDX) { | |||
| // Add live symbol to avoid dead-stripping for .ARM.exidx sections | |||
| G->addAnonymousSymbol(*B, orc::ExecutorAddrDiff(), orc::ExecutorAddrDiff(), false, true); | |||
There was a problem hiding this comment.
The compiler emits a R_ARM_NONE relocation from an exception-handling table section to the required personality routine to indicate the dependence, but the edge alone doesn't keep the section alive in JITLink. We add the anonymous live symbol here to preserve it from dead-stripping.
There was a problem hiding this comment.
That's surprising -- any valid edge should keep the routine alive? Do we actually build an edge, or do we discard R_ARM_NONE relocations?
There was a problem hiding this comment.
IIUC the issue is that here https://github.com/llvm/llvm-project/blob/release/17.x/llvm/lib/ExecutionEngine/JITLink/JITLinkGeneric.cpp#L271 we start from a set of live symbols and then follow edges and mark their target blocks live. Blocks without any symbols and only outgoing edges will remain dead.
I implemented R_ARM_NONE as no-op here 565470e. Not sure we could use it to mark the block live.
R_ARM_PREL31is a 31-bits relative data relocation where the most-significant bit is preserved. It's used primarily in.ARM.exidxsections, which we skipped processing until now, because we didn't support the relocation type. This was implemented in RuntimeDyld with https://reviews.llvm.org/D25069 and I implemented it in a similar way in JITLink in order to reach feature parity.