[cfi] Don't pass a uint16_t to memset. Make sure the 16-bit constant is appropriate for us.

[eugenis]

lib/cfi/cfi.cc

138

memset takes an int and converts it to unsigned char, so this masking should be unnecessary.

[filcab]

Thank you,
Filipe

lib/cfi/cfi.cc

138

True, I ended up adding this because one of our static analyzers complained about a suspicious integer being passed to memset. I can fix it locally only if it's preferred, though.

[eugenis]

lib/cfi/cfi.cc

138

On the second thought, this is good for readability. LGTM

lib/cfi/cfi.cc

 void ShadowBuilder::Start() {
   shadow_ = (uptr)MmapNoReserveOrDie(GetShadowSize(), "CFI shadow");
   VReport(1, "CFI: shadow at %zx .. %zx\n", shadow_, shadow_ + GetShadowSize());
 }
 
 void ShadowBuilder::AddUnchecked(uptr begin, uptr end) {
   uint16_t *shadow_begin = MemToShadow(begin, shadow_);
   uint16_t *shadow_end = MemToShadow(end - 1, shadow_) + 1;
-  memset(shadow_begin, kUncheckedShadow,
+  // memset takes a byte, so our unchecked shadow value requires both bytes to
+  // be the same. Make sure we're ok during compilation.
+  static_assert(kUncheckedShadow & 0xff == ((kUncheckedShadow >> 8) & 0xff));
+  memset(shadow_begin, kUncheckedShadow & 0xff,
          (shadow_end - shadow_begin) * sizeof(*shadow_begin));
 }
 
 void ShadowBuilder::Add(uptr begin, uptr end, uptr cfi_check) {
   assert((cfi_check & (kShadowAlign - 1)) == 0);
 
   // Don't fill anything below cfi_check. We can not represent those addresses
   // in the shadow, and must make sure at codegen to place all valid call