[llvm] r303637 - [RuntimeDyld, PowerPC] Fix check for external symbols when detecting reloction overflow
Ulrich Weigand via llvm-commits
llvm-commits at lists.llvm.org
Tue May 23 07:51:18 PDT 2017
Author: uweigand
Date: Tue May 23 09:51:18 2017
New Revision: 303637
URL: http://llvm.org/viewvc/llvm-project?rev=303637&view=rev
Log:
[RuntimeDyld, PowerPC] Fix check for external symbols when detecting reloction overflow
The PowerPC part of processRelocationRef currently assumes that external
symbols can be identified by checking for SymType == SymbolRef::ST_Unknown.
This is actually incorrect in some cases, causing relocation overflows to
be mis-detected. The correct check is to test whether Value.SymbolName
is null.
Includes test case. Note that it is a bit tricky to replicate the exact
condition that triggers the bug in a test case. The one included here
seems to fail reliably (before the fix) across different operating
system versions on Power, but it still makes a few assumptions (called
out in the test case comments).
Also add ppc64le platform name to the supported list in the lit.local.cfg
files for the MCJIT and OrcMCJIT directories, since those tests were
currently not run at all.
Fixes PR32650.
Reviewer: hfinkel
Differential Revision: https://reviews.llvm.org/D33402
Added:
llvm/trunk/test/ExecutionEngine/OrcMCJIT/pr32650.ll
Modified:
llvm/trunk/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp
llvm/trunk/test/ExecutionEngine/MCJIT/lit.local.cfg
llvm/trunk/test/ExecutionEngine/OrcMCJIT/lit.local.cfg
Modified: llvm/trunk/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp?rev=303637&r1=303636&r2=303637&view=diff
==============================================================================
--- llvm/trunk/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp (original)
+++ llvm/trunk/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp Tue May 23 09:51:18 2017
@@ -1324,12 +1324,12 @@ RuntimeDyldELF::processRelocationRef(
Obj.getPlatformFlags(AbiVariant);
AbiVariant &= ELF::EF_PPC64_ABI;
// A PPC branch relocation will need a stub function if the target is
- // an external symbol (Symbol::ST_Unknown) or if the target address
+ // an external symbol (Value.SymbolName set) or if the target address
// is not within the signed 24-bits branch address.
SectionEntry &Section = Sections[SectionID];
uint8_t *Target = Section.getAddressWithOffset(Offset);
bool RangeOverflow = false;
- if (SymType != SymbolRef::ST_Unknown) {
+ if (!Value.SymbolName) {
if (AbiVariant != 2) {
// In the ELFv1 ABI, a function call may point to the .opd entry,
// so the final symbol value is calculated based on the relocation
@@ -1348,15 +1348,12 @@ RuntimeDyldELF::processRelocationRef(
// If it is within 26-bits branch range, just set the branch target
if (SignExtend64<26>(delta) == delta) {
RelocationEntry RE(SectionID, Offset, RelType, Value.Addend);
- if (Value.SymbolName)
- addRelocationForSymbol(RE, Value.SymbolName);
- else
- addRelocationForSection(RE, Value.SectionID);
+ addRelocationForSection(RE, Value.SectionID);
} else {
RangeOverflow = true;
}
}
- if (SymType == SymbolRef::ST_Unknown || RangeOverflow) {
+ if (Value.SymbolName || RangeOverflow) {
// It is an external symbol (SymbolRef::ST_Unknown) or within a range
// larger than 24-bits.
StubMap::const_iterator i = Stubs.find(Value);
@@ -1412,7 +1409,7 @@ RuntimeDyldELF::processRelocationRef(
RelType, 0);
Section.advanceStubOffset(getMaxStubSize());
}
- if (SymType == SymbolRef::ST_Unknown) {
+ if (Value.SymbolName) {
// Restore the TOC for external calls
if (AbiVariant == 2)
writeInt32BE(Target + 4, 0xE8410018); // ld r2,28(r1)
Modified: llvm/trunk/test/ExecutionEngine/MCJIT/lit.local.cfg
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/ExecutionEngine/MCJIT/lit.local.cfg?rev=303637&r1=303636&r2=303637&view=diff
==============================================================================
--- llvm/trunk/test/ExecutionEngine/MCJIT/lit.local.cfg (original)
+++ llvm/trunk/test/ExecutionEngine/MCJIT/lit.local.cfg Tue May 23 09:51:18 2017
@@ -9,7 +9,8 @@ else:
# FIXME: autoconf and cmake produce different arch names. We should normalize
# them before getting here.
if root.host_arch not in ['i386', 'x86', 'x86_64', 'AMD64',
- 'AArch64', 'ARM', 'Mips', 'PowerPC', 'ppc64', 'SystemZ']:
+ 'AArch64', 'ARM', 'Mips',
+ 'PowerPC', 'ppc64', 'ppc64le', 'SystemZ']:
config.unsupported = True
if 'armv7' in root.host_arch:
Modified: llvm/trunk/test/ExecutionEngine/OrcMCJIT/lit.local.cfg
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/ExecutionEngine/OrcMCJIT/lit.local.cfg?rev=303637&r1=303636&r2=303637&view=diff
==============================================================================
--- llvm/trunk/test/ExecutionEngine/OrcMCJIT/lit.local.cfg (original)
+++ llvm/trunk/test/ExecutionEngine/OrcMCJIT/lit.local.cfg Tue May 23 09:51:18 2017
@@ -9,7 +9,8 @@ else:
# FIXME: autoconf and cmake produce different arch names. We should normalize
# them before getting here.
if root.host_arch not in ['i386', 'x86', 'x86_64', 'AMD64',
- 'AArch64', 'ARM', 'Mips', 'PowerPC', 'ppc64', 'SystemZ']:
+ 'AArch64', 'ARM', 'Mips',
+ 'PowerPC', 'ppc64', 'ppc64le', 'SystemZ']:
config.unsupported = True
if 'armv7' in root.host_arch:
Added: llvm/trunk/test/ExecutionEngine/OrcMCJIT/pr32650.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/ExecutionEngine/OrcMCJIT/pr32650.ll?rev=303637&view=auto
==============================================================================
--- llvm/trunk/test/ExecutionEngine/OrcMCJIT/pr32650.ll (added)
+++ llvm/trunk/test/ExecutionEngine/OrcMCJIT/pr32650.ll Tue May 23 09:51:18 2017
@@ -0,0 +1,28 @@
+; RUN: %lli -jit-kind=orc-mcjit %s
+
+; This test is intended to verify that a function weakly defined in
+; JITted code, and strongly defined in the main executable, can be
+; correctly resolved when called from elsewhere in JITted code.
+
+; This test makes the assumption that the lli executable in compiled
+; to export symbols (e.g. --export-dynamic), and that is actually does
+; contain the symbol LLVMInitializeCodeGen. (Note that this function
+; is not actually called by the test. The test simply verifes that
+; the reference can be resolved without relocation errors.)
+
+define linkonce_odr void @LLVMInitializeCodeGen() {
+entry:
+ ret void
+}
+
+define void @test() {
+entry:
+ call void @LLVMInitializeCodeGen()
+ ret void
+}
+
+define i32 @main() {
+entry:
+ ret i32 0
+}
+
More information about the llvm-commits
mailing list