[PATCH] Do not sanitize kernel area on 32-bit targets

[Kostya Serebryany]

================
Comment at: lib/sanitizer_common/sanitizer_posix.cc:45
@@ -39,3 +44,3 @@
   // of the address space, so simply checking the stack address is not enough.
-  return (1ULL << 44) - 1;  // 0x00000fffffffffffUL
+  res = (1ULL << 44) - 1;  // 0x00000fffffffffffUL
 # elif defined(__aarch64__)
----------------
please don't change the 64-bit code path

================
Comment at: lib/sanitizer_common/sanitizer_posix.cc:53
@@ +52,3 @@
+  res = (1ULL << 32) - 1;  // 0xffffffff;
+# if SANITIZER_LINUX
+  // Take care of unusable kernel area in top gigabyte
----------------
Move this logic into a separate function. 
Provide a unit test for it, that will work on both 64-bit machine and 32-bit machine (differently, of course). 

================
Comment at: lib/asan/asan_mapping.h:112
@@ -111,2 +111,3 @@
                                        : kZeroBaseShadowStart)
-#define kShadowGapEnd   ((kMidMemBeg ? kMidShadowBeg : kHighShadowBeg) - 1)
+#define kShadowGapEnd   ((kMidMemBeg ? kMidShadowBeg \
+                                     : MEM_TO_SHADOW(kHighShadowEnd + 1)) - 1)
----------------
please extend the comment around "Default Linux/i386 mapping"
(add one more section that reflects the different of mapping on 32- and 64-bit OS)

================
Comment at: lib/asan/asan_rtl.cc:485
@@ -484,1 +484,3 @@
     ProtectGap(kShadowGapBeg, kShadowGapEnd - kShadowGapBeg + 1);
+    // Allow accesses to unprotected part of address space
+    if (kShadowGapEnd != kHighShadowBeg - 1)
----------------
this is weird... This should not be required... hm.. Let me see your comments first. 


http://llvm-reviews.chandlerc.com/D3119