[HWASAN] Run sanitizer_common tests

[vitalybuka]

Summary

ninja check-sanitizer will run tests with hwasan as well

[vitalybuka]

remove xfails

[dyung]

Hi @vitalybuka , you may already be aware of it (I see you are making fixes in this area in follow-up commits), but this change caused a test failure on at least one bot that still seems to be failing. Please take a look and fix it if you can or XFAIL or revert if not.

https://lab.llvm.org/buildbot/#/builders/247/builds/3924

[vitalybuka]

In D149234#4300589, @dyung wrote:

Hi @vitalybuka , you may already be aware of it (I see you are making fixes in this area in follow-up commits), but this change caused a test failure on at least one bot that still seems to be failing. Please take a look and fix it if you can or XFAIL or revert if not.

https://lab.llvm.org/buildbot/#/builders/247/builds/3924

Thanks for pointer, I'll fix or revert soon.

[DavidSpickett]

These tests appear to fail on bots that are not building lld: https://lab.llvm.org/buildbot/#/builders/179/builds/6024

It is required that they use lld, should there be a dependency here?

[awarzynski]

In D149234#4301524, @DavidSpickett wrote:

These tests appear to fail on bots that are not building lld: https://lab.llvm.org/buildbot/#/builders/179/builds/6024

It is required that they use lld, should there be a dependency here?

@vitalybuka Could you take a look, please? I am very unfamiliar with HWSAN, but all the failing tests are "hwsan" related and started appearing after 3 HWSAN related changes landed. This one and:

• https://reviews.llvm.org/D149271
• https://reviews.llvm.org/D149219

All authored by you :) If you have no access to an AArch64 machine, I am happy to help with testing.

[vitalybuka]

In D149234#4312384, @awarzynski wrote:

In D149234#4301524, @DavidSpickett wrote:

These tests appear to fail on bots that are not building lld: https://lab.llvm.org/buildbot/#/builders/179/builds/6024

It is required that they use lld, should there be a dependency here?

@vitalybuka Could you take a look, please? I am very unfamiliar with HWSAN, but all the failing tests are "hwsan" related and started appearing after 3 HWSAN related changes landed. This one and:

• https://reviews.llvm.org/D149271
• https://reviews.llvm.org/D149219

All authored by you :) If you have no access to an AArch64 machine, I am happy to help with testing.

Sorry, I just notices. I will fix today.
We have arm but we use lld there :) There is a bug in ld which manifests with hwasan.
So I guess we need to disabled hwasan without lld.
What is confuses me, that we landed similar stuff for lsan+hwasan few month ago, and I don't see failures on that bot. And seems check-all does not run lsan tests.

[awarzynski]

In D149234#4312801, @vitalybuka wrote:

In D149234#4312384, @awarzynski wrote:

In D149234#4301524, @DavidSpickett wrote:

These tests appear to fail on bots that are not building lld: https://lab.llvm.org/buildbot/#/builders/179/builds/6024

It is required that they use lld, should there be a dependency here?

@vitalybuka Could you take a look, please? I am very unfamiliar with HWSAN, but all the failing tests are "hwsan" related and started appearing after 3 HWSAN related changes landed. This one and:

• https://reviews.llvm.org/D149271
• https://reviews.llvm.org/D149219

All authored by you :) If you have no access to an AArch64 machine, I am happy to help with testing.

Sorry, I just notices. I will fix today.

That would be much appreciated. This is affecting our SVE bot, and these are rare :)

We have arm but we use lld there :) There is a bug in ld which manifests with hwasan.
So I guess we need to disabled hwasan without lld.

+1

Once this is resolved, we could discuss installing lld on that bot. For that, however, we'd need to wait for @DavidSpickett to be back (he'll be away most of May).

[DavidSpickett]

This is correct but I’m not the only person who can deal with bots. Write
to the mailing list
https://lists.linaro.org/archives/list/[e-mail]/
and one of my Linaro colleagues can help you figure it out.

[vitalybuka]

LLD issue should be fixed now

compiler-rt/test/sanitizer_common/CMakeLists.txt

 set(SANITIZER_COMMON_LIT_SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR})
 
 set(SANITIZER_COMMON_TEST_DEPS ${SANITIZER_COMMON_LIT_TEST_DEPS})
 
 set(SANITIZER_COMMON_TESTSUITES)
 
 # FIXME(dliew): We should switch to COMPILER_RT_SANITIZERS_TO_BUILD instead of
 # the hard coded `SUPPORTED_TOOLS_INIT` list once we know that the other
 # sanitizers work.
-set(SUPPORTED_TOOLS_INIT asan lsan msan tsan ubsan)
+set(SUPPORTED_TOOLS_INIT asan lsan hwasan msan tsan ubsan)
 set(SUPPORTED_TOOLS)
   foreach(SANITIZER_TOOL ${SUPPORTED_TOOLS_INIT})
     string(TOUPPER ${SANITIZER_TOOL} SANITIZER_TOOL_UPPER)
     if (COMPILER_RT_HAS_${SANITIZER_TOOL_UPPER})
       list(APPEND SUPPORTED_TOOLS ${SANITIZER_TOOL})
     endif()
   endforeach()
 

compiler-rt/test/sanitizer_common/lit.common.cfg.py

 # -*- Python -*-
 
 # Setup source root.
 config.test_source_root = os.path.join(os.path.dirname(__file__), "TestCases")
 
 config.name = "SanitizerCommon-" + config.name_suffix
 
 default_tool_options = []
 collect_stack_traces = ""
 if config.tool_name == "asan":
   tool_cflags = ["-fsanitize=address"]
   tool_options = "ASAN_OPTIONS"
+elif config.tool_name == "hwasan":
+  tool_cflags = ["-fsanitize=hwaddress", "-fuse-ld=lld"]
+  if config.target_arch == "x86_64":
+    tool_cflags += ["-fsanitize-hwaddress-experimental-aliasing"]
+  tool_options = "HWASAN_OPTIONS"
 elif config.tool_name == "tsan":
   tool_cflags = ["-fsanitize=thread"]
   tool_options = "TSAN_OPTIONS"
 elif config.tool_name == "msan":
   tool_cflags = ["-fsanitize=memory"]
   tool_options = "MSAN_OPTIONS"
   collect_stack_traces = "-fsanitize-memory-track-origins"
 elif config.tool_name == "lsan":