[safestack] Support USP in TCB for musl libc

[mlemay-intel]

Summary

This patch stores the Unsafe Stack Pointer (USP) at offset 9*sizeof(uintptr_t)
in the Thread Control Block (TCB) for musl environments.

[mlemay-intel]

Revised this to be specific to musl libc on X86.

[mlemay-intel]

Rebase.

[mlemay-intel]

Revised patch so that it no longer modifies X86 subtarget.

[mlemay-intel]

Updated so that USP is stored in TCB for all musl libc environments on X86,
rather than just those on Linux with the separate stack segment feature enabled.

My original musl libc patches just supported SafeStack with the separate stack
segment feature enabled, but I recently divided them into two series. The first
series provides architecture-independent support for SafeStack including the
support for storing USP in the TCB. The second series adds support for the
separate stack segment feature.

By the way, I would welcome feedback on my musl libc patches from the LLVM
community, especially from those of you that have experience developing runtime
support for SafeStack such as @pcc and @eugenis. My recent posts are listed at
http://www.openwall.com/lists/musl/2016/10/28/ .

lib/Target/X86/X86ISelLowering.cpp

if (!V) {
       if (V->getValueType() != StackPtrTy)
         report_fatal_error(Twine(Name) + " must have void* type");
       if (V->isThreadLocal())
         report_fatal_error(Twine(Name) + " must not be thread-local");
     }
     return V;
   }
 
-  if (!Subtarget.isTargetAndroid())
+  bool MuslSeparateStack =
+      Subtarget.isTargetMusl() &&
+      Subtarget.isTargetLinux() &&
+      Subtarget.useSeparateStackSeg();
+
+  if (!(MuslSeparateStack || Subtarget.isTargetAndroid()))
     return TargetLowering::getSafeStackPointerLocation(IRB);
 
   // Android provides a fixed TLS slot for the SafeStack pointer. See the
   // definition of TLS_SLOT_SAFESTACK in
   // https://android.googlesource.com/platform/bionic/+/master/libc/private/bionic_tls.h
   unsigned AddressSpace, Offset;
 
   // %fs:0x48, unless we're using a Kernel code model, in which case it's %gs:

lib/Target/X86/X86Subtarget.h

public:
   bool isTargetKFreeBSD() const { return TargetTriple.isOSKFreeBSD(); }
   bool isTargetGlibc() const { return TargetTriple.isOSGlibc(); }
   bool isTargetAndroid() const { return TargetTriple.isAndroid(); }
   bool isTargetNaCl() const { return TargetTriple.isOSNaCl(); }
   bool isTargetNaCl32() const { return isTargetNaCl() && !is64Bit(); }
   bool isTargetNaCl64() const { return isTargetNaCl() && is64Bit(); }
   bool isTargetMCU() const { return TargetTriple.isOSIAMCU(); }
   bool isTargetContiki() const { return TargetTriple.isOSContiki(); }
+  bool isTargetMusl() const { return TargetTriple.isMusl(); }
 
   bool isTargetWindowsMSVC() const {
     return TargetTriple.isWindowsMSVCEnvironment();
   }
 
   bool isTargetKnownWindowsMSVC() const {
     return TargetTriple.isKnownWindowsMSVCEnvironment();
   }

test/Transforms/SafeStack/addr-taken.ll

 ; RUN: opt -safe-stack -S -mtriple=i386-pc-linux-gnu < %s -o - | FileCheck %s
 ; RUN: opt -safe-stack -S -mtriple=x86_64-pc-linux-gnu < %s -o - | FileCheck %s
+; RUN: opt -safe-stack -S -mtriple=i386-pc-linux-musl -mattr=+separate-stack-seg < %s -o - | FileCheck --check-prefix TCB %s
 
 @.str = private unnamed_addr constant [4 x i8] c"%s\0A\00", align 1
 
 ; Address-of local taken (j = &a)
 ; Requires protector.
 
 define void @foo() nounwind uwtable safestack {
 entry:
   ; CHECK: __safestack_unsafe_stack_ptr
+  ; TCB: %unsafe_stack_ptr = load i8*, i8* addrspace(256)* inttoptr (i32 36 to i8* addrspace(256)*)
+  ; TCB-NEXT: %unsafe_stack_static_top = getelementptr i8, i8* %unsafe_stack_ptr, i32 -16
+  ; TCB-NEXT: store i8* %unsafe_stack_static_top, i8* addrspace(256)* inttoptr (i32 36 to i8* addrspace(256)*)
   %retval = alloca i32, align 4
   %a = alloca i32, align 4
   %j = alloca i32*, align 8
   store i32 0, i32* %retval
   %0 = load i32, i32* %a, align 4
   %add = add nsw i32 %0, 1
   store i32 %add, i32* %a, align 4
   store i32* %a, i32** %j, align 8
+  ; TCB: store i8* %unsafe_stack_ptr, i8* addrspace(256)* inttoptr (i32 36 to i8* addrspace(256)*)
   ret void
 }