[PATCH] D111342: [scudo] Reduce the scope of AllocAfterFork

Kostya Kortchinsky via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Thu Oct 7 12:30:58 PDT 2021


cryptoad created this revision.
cryptoad added reviewers: leonardchan, pcc, vitalybuka, hctim, cferris.
cryptoad requested review of this revision.
Herald added a project: Sanitizers.
Herald added a subscriber: Sanitizers.

`ScudoWrappersCppTest.AllocAfterFork` was failing obscurely sometimes.
Someone pointed us to Linux's `vm.max_map_count` that can be
significantly lower on some machines than others. It turned out that
on a machine with that setting set to 65530, some `ENOMEM` errors
would occur with `mmap` & `mprotect` during that specific test.

Reducing the number of times we fork, and the maximum size allocated
during that test makes it pass on those machines.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D111342

Files:
  compiler-rt/lib/scudo/standalone/tests/wrappers_cpp_test.cpp


Index: compiler-rt/lib/scudo/standalone/tests/wrappers_cpp_test.cpp
===================================================================
--- compiler-rt/lib/scudo/standalone/tests/wrappers_cpp_test.cpp
+++ compiler-rt/lib/scudo/standalone/tests/wrappers_cpp_test.cpp
@@ -130,10 +130,6 @@
 }
 
 #if !SCUDO_FUCHSIA
-// TODO(kostyak): for me, this test fails in a specific configuration when ran
-//                by itself with some Scudo or GWP-ASan violation. Other people
-//                can't seem to reproduce the failure. Consider skipping this in
-//                the event it fails on the upstream bots.
 TEST(ScudoWrappersCppTest, AllocAfterFork) {
   std::atomic_bool Stop;
 
@@ -142,7 +138,7 @@
   for (size_t N = 0; N < 5; N++) {
     std::thread *T = new std::thread([&Stop] {
       while (!Stop) {
-        for (size_t SizeLog = 3; SizeLog <= 21; SizeLog++) {
+        for (size_t SizeLog = 3; SizeLog <= 20; SizeLog++) {
           char *P = new char[1UL << SizeLog];
           EXPECT_NE(P, nullptr);
           // Make sure this value is not optimized away.
@@ -155,10 +151,10 @@
   }
 
   // Create a thread to fork and allocate.
-  for (size_t N = 0; N < 100; N++) {
+  for (size_t N = 0; N < 50; N++) {
     pid_t Pid;
     if ((Pid = fork()) == 0) {
-      for (size_t SizeLog = 3; SizeLog <= 21; SizeLog++) {
+      for (size_t SizeLog = 3; SizeLog <= 20; SizeLog++) {
         char *P = new char[1UL << SizeLog];
         EXPECT_NE(P, nullptr);
         // Make sure this value is not optimized away.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D111342.377963.patch
Type: text/x-patch
Size: 1528 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20211007/3d8e9d23/attachment.bin>


More information about the llvm-commits mailing list