morehouse added inline comments.
================
Comment at: clang/include/clang/Basic/Sanitizers.def:55-59
+// Utilize Intel LAM in sanitizers. Currently only used in combination with
+// -fsanitize=hwaddress. This is an experimental flag which may be removed in
+// the future.
+// TODO: Use -mlam instead, if/when it is supported by clang.
+SANITIZER("lam", LAM)
----------------
vitalybuka wrote:
> if it's experimental, why not just "-fsanitize=hwaddress -mllvm
> -havasan-lam=1" ?
Well, `-mllvm` indicates a flag for LLVM, but we need to change the Clang
behavior to link with the LAM-enabled HWASan runtime. It seems to me that we
should use a flag directed to Clang for this.
Maybe it's possible to parse the `-mllvm` flag before the point where we need
to choose a runtime (I'm not sure), but it seems simpler to do it this way.
================
Comment at: compiler-rt/test/hwasan/TestCases/Linux/vfork.c:7
-// Aliasing mode does not support stack tagging.
-// XFAIL: x86_64
----------------
xiangzhangllvm wrote:
> What does here XFAIL mean, do not test in x86_64 ?
It means the test will run in x86_64 and is expected to fail.
Under LAM the test actually passes, since manually tagging the upper bits in
the stack pointers succeeds. So we need to distinguish between x86 LAM mode
and x86 aliasing mode, which we do with the new `pointer-tagging` feature.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D102288/new/
https://reviews.llvm.org/D102288
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
