[scudo] Do a M_PURGE call before changing release interval on Android

[ChiaHungDuan]

No description provided.

[llvmbot]

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

Author: None (ChiaHungDuan)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/87110.diff

1 Files Affected:

• (modified) compiler-rt/lib/scudo/standalone/wrappers_c.inc (+5)

diff --git a/compiler-rt/lib/scudo/standalone/wrappers_c.inc b/compiler-rt/lib/scudo/standalone/wrappers_c.inc
index 56d8ef20156e28..21d5b7add51275 100644
--- a/compiler-rt/lib/scudo/standalone/wrappers_c.inc
+++ b/compiler-rt/lib/scudo/standalone/wrappers_c.inc
@@ -247,6 +247,11 @@ void SCUDO_PREFIX(malloc_postinit)() {
 INTERFACE WEAK int SCUDO_PREFIX(mallopt)(int param, int value) {
   if (param == M_DECAY_TIME) {
     if (SCUDO_ANDROID) {
+      // Before changing the interval, reset the memory usage status by doing a
+      // M_PURGE call so that we can minimize the impact of any unreleased pages
+      // introduced by interval transition.
+      SCUDO_ALLOCATOR.releaseToOS(scudo::ReleaseToOS::Force);
+
       if (value == 0) {
         // Will set the release values to their minimum values.
         value = INT32_MIN;

[cferris1000]

Should this be ForceAll? I presume no, but I wanted to ask.

[ChiaHungDuan]

Based on the use case, They show similar memory footprint reduction but have noticeable latency difference. As a result, I think Force is good enough so far.

[cferris1000]

LGTM.