Skip to content

[TSan] Add 2 test cases related to incomplete shadow cleanup in unmap #145472

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jun 25, 2025

Conversation

Camsyn
Copy link
Contributor

@Camsyn Camsyn commented Jun 24, 2025

Once part of PR #144648, follow the reviewer's advice and split into this separate PR.

unmap works at page granularity, but supports an arbitrary non-zero size as an argument, which results in possible shadow undercleaning in the existing TSan implementation when size % kShadowCell != 0.

This change introduces two test cases to verify the shadow cleaning effect in unmap.

  • java_heap_init2.cpp: Imitating java_heap_init cpp, verify the incomplete cleaning of meta
  • munmap_clear_shadow.c: verify the incomplete cleaning of shadow

- java_heap_init2.cpp: Imitating java_heap_init cpp, verify the
  incomplete cleaning of meta
- munmap_clear_shadow.c: verify the incomplete clearning of shadow
@llvmbot
Copy link
Member

llvmbot commented Jun 24, 2025

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

Author: Kunqiu Chen (Camsyn)

Changes

Once part of PR #144648, follow the reviewer's advice and split into this separate PR.

unmap works at page granularity, but supports an arbitrary non-zero size as an argument, which results in possible shadow undercleaning in the existing TSan implementation when size % kShadowCell != 0.

This change introduces two test cases to verify the shadow cleaning effect in unmap.

  • java_heap_init2.cpp: Imitating java_heap_init cpp, verify the incomplete cleaning of meta
  • munmap_clear_shadow.c: verify the incomplete cleaning of shadow

Full diff: https://github.com/llvm/llvm-project/pull/145472.diff

2 Files Affected:

  • (added) compiler-rt/test/tsan/java_heap_init2.cpp (+34)
  • (added) compiler-rt/test/tsan/munmap_clear_shadow.c (+59)
diff --git a/compiler-rt/test/tsan/java_heap_init2.cpp b/compiler-rt/test/tsan/java_heap_init2.cpp
new file mode 100644
index 0000000000000..2e5724d930e8f
--- /dev/null
+++ b/compiler-rt/test/tsan/java_heap_init2.cpp
@@ -0,0 +1,34 @@
+// RUN: %clangxx_tsan -O1 %s -o %t && %run %t 2>&1 | FileCheck %s
+// XFAIL: *
+
+#include "java.h"
+#include <errno.h>
+#include <sys/mman.h>
+
+int main() {
+  // Test a non-regular kHeapSize
+  // Previously __tsan_java_init failed because it encountered non-zero meta
+  // shadow for the destination.
+  size_t const kPageSize = sysconf(_SC_PAGESIZE);
+  int const kSize = kPageSize - 1;
+  jptr jheap2 = (jptr)mmap(0, kSize, PROT_READ | PROT_WRITE,
+                           MAP_ANON | MAP_PRIVATE, -1, 0);
+  if (jheap2 == (jptr)MAP_FAILED)
+    return printf("mmap failed with %d\n", errno);
+  __atomic_store_n((int *)(jheap2 + kSize - 3), 1, __ATOMIC_RELEASE);
+  // Due to the previous incorrect meta-end calculation, the following munmap
+  // did not clear the tail meta shadow.
+  munmap((void *)jheap2, kSize);
+  int const kHeapSize2 = kSize + 1;
+  jheap2 = (jptr)mmap((void *)jheap2, kHeapSize2, PROT_READ | PROT_WRITE,
+                      MAP_ANON | MAP_PRIVATE, -1, 0);
+  if (jheap2 == (jptr)MAP_FAILED)
+    return printf("second mmap failed with %d\n", errno);
+  __tsan_java_init(jheap2, kHeapSize2);
+  __tsan_java_move(jheap2, jheap2 + kHeapSize2 - 8, 8);
+  fprintf(stderr, "DONE\n");
+  return __tsan_java_fini();
+}
+
+// CHECK-NOT: WARNING: ThreadSanitizer: data race
+// CHECK: DONE
diff --git a/compiler-rt/test/tsan/munmap_clear_shadow.c b/compiler-rt/test/tsan/munmap_clear_shadow.c
new file mode 100644
index 0000000000000..8a435a84258f5
--- /dev/null
+++ b/compiler-rt/test/tsan/munmap_clear_shadow.c
@@ -0,0 +1,59 @@
+// RUN: %clang_tsan %s -o %t && %run %t | FileCheck %s
+// XFAIL: *
+
+#include "test.h"
+#include <assert.h>
+#include <pthread.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <sys/mman.h>
+#include <unistd.h>
+
+void __tsan_read1(void *addr);
+
+struct thread_params {
+  char *buf;
+  unsigned int size;
+};
+
+static void *thread_func(void *arg) {
+  struct thread_params *p = (struct thread_params *)arg;
+  // Access 1
+  p->buf[0] = 0x42;
+  p->buf[p->size - 1] = 0x42;
+  barrier_wait(&barrier);
+  return 0;
+}
+
+int main() {
+  const unsigned int kPageSize = sysconf(_SC_PAGESIZE);
+  // The relevant shadow memory size should be exactly multiple of kPageSize,
+  // even if Size = kPageSize - 1.
+  const unsigned int Size = kPageSize - 1;
+
+  barrier_init(&barrier, 2);
+  char *buf = (char *)mmap(NULL, Size, PROT_READ | PROT_WRITE,
+                           MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+  assert(buf != MAP_FAILED);
+  assert(((uintptr_t)buf % kPageSize) == 0);
+
+  pthread_t t;
+  struct thread_params p = {buf, Size};
+  pthread_create(&t, 0, thread_func, &p);
+
+  barrier_wait(&barrier);
+  // Should clear all the shadow memory related to the mmaped memory.
+  munmap(buf, Size);
+
+  // If the shadow memory is cleared completely, the following reads should not
+  // cause races and behave the same. However, previously, __tsan_read1(&buf[0])
+  // would not report a race, while __tsan_read1(&buf[Size - 1]) did.
+  // CHECK-NOT: WARNING: ThreadSanitizer: data race
+  __tsan_read1(&buf[0]);        // Access 2
+  __tsan_read1(&buf[Size - 1]); // Access 2
+  pthread_join(t, 0);
+
+  puts("DONE");
+
+  return 0;
+}

@Camsyn Camsyn merged commit 956bab0 into llvm:main Jun 25, 2025
11 checks passed
anthonyhatran pushed a commit to anthonyhatran/llvm-project that referenced this pull request Jun 26, 2025
…llvm#145472)

Once part of PR llvm#144648, follow the reviewer's advice and split into
this separate PR.

`unmap` works at page granularity, but supports an arbitrary non-zero
size as an argument, which results in possible shadow undercleaning in
the existing TSan implementation when `size % kShadowCell != 0`.

This change introduces two test cases to verify the shadow cleaning
effect in `unmap`.

- java_heap_init2.cpp: Imitating java_heap_init cpp, verify the
incomplete cleaning of meta
- munmap_clear_shadow.c: verify the incomplete cleaning of shadow
@Camsyn Camsyn deleted the tsan-new-test branch June 27, 2025 06:43
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants