[scudo] Mitigate the overhead in cache storing when MTE enabled

[ChiaHungDuan]

mapSecondary() requires two mmap calls and may impact the performance in some cases that use secondary allocator heavily. Only use setMemoryPermission to reduce the time in contention of memory system calls.

[llvmbot]

@llvm/pr-subscribers-compiler-rt-sanitizer

Changes

mapSecondary() requires two mmap calls and may impact the performance in some cases that use secondary allocator heavily. Only use setMemoryPermission to reduce the time in contention of memory system calls.

---

Full diff: https://github.com/llvm/llvm-project/pull/66717.diff

1 Files Affected:

• (modified) compiler-rt/lib/scudo/standalone/secondary.h (+3-11)

diff --git a/compiler-rt/lib/scudo/standalone/secondary.h b/compiler-rt/lib/scudo/standalone/secondary.h
index d0890d1c5e2d3d5..c4192bfd45bba2e 100644
--- a/compiler-rt/lib/scudo/standalone/secondary.h
+++ b/compiler-rt/lib/scudo/standalone/secondary.h
@@ -210,18 +210,10 @@ template <typename Config> class MapAllocatorCache {
     Entry.MemMap = H->MemMap;
     Entry.Time = Time;
     if (useMemoryTagging<Config>(Options)) {
-      if (Interval == 0 && !SCUDO_FUCHSIA) {
-        // Release the memory and make it inaccessible at the same time by
-        // creating a new MAP_NOACCESS mapping on top of the existing mapping.
-        // Fuchsia does not support replacing mappings by creating a new mapping
-        // on top so we just do the two syscalls there.
+      if (Interval == 0)
         Entry.Time = 0;
-        mapSecondary<Config>(Options, Entry.CommitBase, Entry.CommitSize,
-                             Entry.CommitBase, MAP_NOACCESS, Entry.MemMap);
-      } else {
-        Entry.MemMap.setMemoryPermission(Entry.CommitBase, Entry.CommitSize,
-                                         MAP_NOACCESS);
-      }
+      Entry.MemMap.setMemoryPermission(Entry.CommitBase, Entry.CommitSize,
+                                       MAP_NOACCESS);
     } else if (Interval == 0) {
       Entry.MemMap.releasePagesToOS(Entry.CommitBase, Entry.CommitSize);
       Entry.Time = 0;

[ChiaHungDuan]

This is another way to mitigate the problem mentioned in #65942. Now it's only a RFC change.

[andersdellien-arm]

As far as I can tell, this would solve the same problem that #65942 is addressing. So the performance benefit would be the same

[ChiaHungDuan]

As far as I can tell, this would solve the same problem that #65942 is addressing. So the performance benefit would be the same

Right, this is supposed to do the same thing as setting the release interval and has no impact on non-MTE path