[BOLT][Instrumentation] Fix indirect call profile in PIE

Because indirect call tables use static addresses for call sites, but pc
values recorded by runtime may be subject to ASLR in PIE, we couldn't
find indirect call descriptions by their runtime address in PIE. It
resulted in [unknown] entries in profile for all indirect calls. We need
to substract base address of .text from runtime addresses to get the
corresponding static addresses. Here we create a getter for base address
of .text and substract it's return value from recorded PC values. It
converts them to static addresses, which then may be used to find the
corresponding indirect call descriptions.

Reviewed By: rafauler

Differential Revision: https://reviews.llvm.org/D154121

Author: treapster

bolt/lib/Rewrite/RewriteInstance.cpp

@@ -1845,8 +1845,9 @@ void RewriteInstance::adjustCommandLineOptions() {
     exit(1);
   }
 
-  if (opts::ReorderFunctions != ReorderFunctions::RT_NONE &&
-      !opts::HotText.getNumOccurrences()) {
+  if (opts::Instrument ||
+      (opts::ReorderFunctions != ReorderFunctions::RT_NONE &&
+       !opts::HotText.getNumOccurrences())) {
     opts::HotText = true;
   } else if (opts::HotText && !BC->HasRelocations) {
     errs() << "BOLT-WARNING: hot text is disabled in non-relocation mode\n";

bolt/runtime/common.h

@@ -165,6 +165,20 @@ int memcmp(const void *s1, const void *s2, size_t n) {
 // Anonymous namespace covering everything but our library entry point
 namespace {
 
+// Get the difference between runtime addrress of .text section and
+// static address in section header table. Can be extracted from arbitrary
+// pc value recorded at runtime to get the corresponding static address, which
+// in turn can be used to search for indirect call description. Needed because
+// indirect call descriptions are read-only non-relocatable data.
+uint64_t getTextBaseAddress() {
+  uint64_t DynAddr;
+  uint64_t StaticAddr;
+  __asm__ volatile("leaq __hot_end(%%rip), %0\n\t"
+                   "movabsq $__hot_end, %1\n\t"
+                   : "=r"(DynAddr), "=r"(StaticAddr));
+  return DynAddr - StaticAddr;
+}
+
 constexpr uint32_t BufSize = 10240;
 
 #define _STRINGIFY(x) #x

bolt/runtime/instr.cpp

@@ -215,6 +215,12 @@ class BumpPtrAllocator {
 /// __bolt_instr_setup, our initialization routine.
 BumpPtrAllocator *GlobalAlloc;
 
+// Base address which we substract from recorded PC values when searching for
+// indirect call description entries. Needed because indCall descriptions are
+// mapped read-only and contain static addresses. Initialized in
+// __bolt_instr_setup.
+uint64_t TextBaseAddress = 0;
+
 // Storage for GlobalAlloc which can be shared if not using
 // instrumentation-file-append-pid.
 void *GlobalMetadataStorage;
@@ -1389,7 +1395,7 @@ void visitIndCallCounter(IndirectCallHashTable::MapEntry &Entry,
   const IndCallDescription *CallsiteDesc =
       &Ctx->IndCallDescriptions[CallsiteID];
   const IndCallTargetDescription *TargetDesc =
-      Ctx->lookupIndCallTarget(Entry.Key);
+      Ctx->lookupIndCallTarget(Entry.Key - TextBaseAddress);
   if (!TargetDesc) {
     DEBUG(report("Failed to lookup indirect call target\n"));
     char LineBuf[BufSize];
@@ -1609,6 +1615,7 @@ extern "C" void __bolt_instr_indirect_tailcall();
 extern "C" void __attribute((force_align_arg_pointer)) __bolt_instr_setup() {
   __bolt_ind_call_counter_func_pointer = __bolt_instr_indirect_call;
   __bolt_ind_tailcall_counter_func_pointer = __bolt_instr_indirect_tailcall;
+  TextBaseAddress = getTextBaseAddress();
 
   const uint64_t CountersStart =
       reinterpret_cast<uint64_t>(&__bolt_instr_locations[0]);