[lldb][AArch64] Add testing of save/restore for Linux MTE control register

[DavidSpickett]

Summary

This has always worked but had no coverage. Adding testing now so that
later I can refactor the save/restore code safely.

[DavidSpickett]

Rebase

[omjavaid]

This looks fine but instead of adding a new test cant we add MTE control register tests to already available MTE tests?

[DavidSpickett]

Remove unused includes.

[DavidSpickett]

This looks fine but instead of adding a new test cant we add MTE control register tests to already available MTE tests?

We have these existing tests:
mte_core_file - clearly it can't go here, core files are read only.
mte_memory_region - memory regions have nothing to do with registers.
mte_tag_access - Could go here.
mte_tag_faults - Could go here.

You can call it def test_mte_ctrl_is_restored and that explains it if it fails among a set of somewhat unrelated tests, but it makes it harder to see at a glance if we've tested that register specifically. Doing it in it's own test file is clear what it's doing and it's next to the other register tests and the forthcoming SME tests.

I know I've gone either way on adding new test cases or tacking onto older ones in the past but in this case that's the way I'm leaning.

[DavidSpickett]

Remove HWCAP check from program file. The runner will check cppuinfo
and even if that was fooled, the prctl would fail at runtime if you
didn't have MTE.

lldb/test/API/commands/register/register/aarch64_mte_ctrl_register/Makefile

+C_SOURCES := main.c
+
+include Makefile.rules

lldb/test/API/commands/register/register/aarch64_mte_ctrl_register/TestMTECtrlRegister.py

+"""
+Test that LLDB correctly reads, writes and restores the MTE control register on
+AArch64 Linux.
+"""
+
+import lldb
+from lldbsuite.test.decorators import *
+from lldbsuite.test.lldbtest import *
+from lldbsuite.test import lldbutil
+
+
+class MTECtrlRegisterTestCase(TestBase):
+    @no_debug_info_test
+    @skipIf(archs=no_match(["aarch64"]))
+    @skipIf(oslist=no_match(["linux"]))
+    def test_mte_ctrl_register(self):
+        if not self.isAArch64MTE():
+            self.skipTest("Target must support MTE.")
+
+        self.build()
+        self.line = line_number("main.c", "// Set a break point here.")
+
+        exe = self.getBuildArtifact("a.out")
+        self.runCmd("file " + exe, CURRENT_EXECUTABLE_SET)
+
+        lldbutil.run_break_set_by_file_and_line(
+            self, "main.c", self.line, num_expected_locations=1
+        )
+        self.runCmd("run", RUN_SUCCEEDED)
+
+        self.expect(
+            "process status",
+            STOPPED_DUE_TO_BREAKPOINT,
+            substrs=["stop reason = breakpoint 1."],
+        )
+
+        # Bit 0 = tagged addressing enabled
+        # Bit 1 = synchronous faults
+        # Bit 2 = asynchronous faults
+        # We start enabled with synchronous faults.
+        self.expect("register read mte_ctrl", substrs=["0x0000000000000003"])
+
+        # Change to asynchronous faults.
+        self.runCmd("register write mte_ctrl 5")
+        self.expect("register read mte_ctrl", substrs=["0x0000000000000005"])
+
+        # This would return to synchronous faults if we did not restore the
+        # previous value.
+        self.expect("expression setup_mte()", substrs=["= 0"])
+        self.expect("register read mte_ctrl", substrs=["0x0000000000000005"])
 No newline at end of file

lldb/test/API/commands/register/register/aarch64_mte_ctrl_register/main.c

+#include <sys/prctl.h>
+
+// This is its own function so that lldb can call it.
+int setup_mte() {
+  return prctl(PR_SET_TAGGED_ADDR_CTRL, PR_TAGGED_ADDR_ENABLE | PR_MTE_TCF_SYNC,
+               0, 0, 0);
+}
+
+int main(int argc, char const *argv[]) {
+  if (setup_mte())
+    return 1;
+
+  return 0; // Set a break point here.
+}