[clang] [llvm] [Hashing] Use a non-deterministic seed if LLVM_ENABLE_ABI_BREAKING_CHECKS (PR #96282)
Fangrui Song via cfe-commits
cfe-commits at lists.llvm.org
Tue Jul 2 11:22:39 PDT 2024
================
@@ -322,24 +306,20 @@ struct hash_state {
}
};
-
-/// A global, fixed seed-override variable.
-///
-/// This variable can be set using the \see llvm::set_fixed_execution_seed
-/// function. See that function for details. Do not, under any circumstances,
-/// set or read this variable.
-extern uint64_t fixed_seed_override;
-
+/// In LLVM_ENABLE_ABI_BREAKING_CHECKS builds, the seed is non-deterministic
+/// (address of a variable) to prevent having users depend on the particular
+/// hash values. On platforms without ASLR, this is still likely
+/// non-deterministic per build.
inline uint64_t get_execution_seed() {
- // FIXME: This needs to be a per-execution seed. This is just a placeholder
- // implementation. Switching to a per-execution seed is likely to flush out
- // instability bugs and so will happen as its own commit.
- //
- // However, if there is a fixed seed override set the first time this is
- // called, return that instead of the per-execution seed.
- const uint64_t seed_prime = 0xff51afd7ed558ccdULL;
- static uint64_t seed = fixed_seed_override ? fixed_seed_override : seed_prime;
- return seed;
+ // Work around x86-64 negative offset folding for old Clang -fno-pic
+ // https://reviews.llvm.org/D93931
+#if LLVM_ENABLE_ABI_BREAKING_CHECKS && \
+ (!defined(__clang__) || __clang_major__ > 11)
----------------
MaskRay wrote:
LLVM built using GCC are typically release builds and LLVM_ENABLE_ABI_BREAKING_CHECKS is 0. The `else` branch is taken and there is no impact.
If we want to be over-cautious, this condition can be refined to:
`LLVM_ENABLE_ABI_BREAKING_CHECKS && (!defined(__clang__) || __clang_major__ > 11 || defined(__PIC__))`.
We could code the condition to `utils/bazel/llvm_configs/abi-breaking.h.cmake`, but I am hoping that the complexity isn't needed.
https://github.com/llvm/llvm-project/pull/96282
More information about the cfe-commits
mailing list