[scudo] Always zero on linux even if the memory cannot be released. #167788
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
If a caller has locked memory, then the madvise call will fail. In that case, zero the memory so that we don't return non-zeroed memory for calloc calls since we thought the memory had been released.
llvmbot added compiler-rt compiler-rt:scudo 
Member
| @llvm/pr-subscribers-compiler-rt-sanitizer Author: Christopher Ferris (cferris1000) ChangesIf a caller has locked memory, then the madvise call will fail. In that case, zero the memory so that we don't return non-zeroed memory for calloc calls since we thought the memory had been released. Full diff: https://github.com/llvm/llvm-project/pull/167788.diff 2 Files Affected:
diff --git a/compiler-rt/lib/scudo/standalone/mem_map_linux.cpp b/compiler-rt/lib/scudo/standalone/mem_map_linux.cpp index 783c4f0d9ab0f..df3e54cab6695 100644 --- a/compiler-rt/lib/scudo/standalone/mem_map_linux.cpp +++ b/compiler-rt/lib/scudo/standalone/mem_map_linux.cpp @@ -122,7 +122,12 @@ void MemMapLinux::setMemoryPermissionImpl(uptr Addr, uptr Size, uptr Flags) { void MemMapLinux::releaseAndZeroPagesToOSImpl(uptr From, uptr Size) { void *Addr = reinterpret_cast<void *>(From); - while (madvise(Addr, Size, MADV_DONTNEED) == -1 && errno == EAGAIN) { + int rc; + while ((rc = madvise(Addr, Size, MADV_DONTNEED)) == -1 && errno == EAGAIN) { + } + if (rc == -1) { + // If we can't madvies the memory, then we still need to zero it. + memset(Addr, 0, Size); } } diff --git a/compiler-rt/lib/scudo/standalone/tests/map_test.cpp b/compiler-rt/lib/scudo/standalone/tests/map_test.cpp index cc7d3ee4dc6c2..fadc0e00c7844 100644 --- a/compiler-rt/lib/scudo/standalone/tests/map_test.cpp +++ b/compiler-rt/lib/scudo/standalone/tests/map_test.cpp @@ -14,6 +14,10 @@ #include <string.h> #include <unistd.h> +#if SCUDO_LINUX +#include <sys/mman.h> +#endif + static const char *MappingName = "scudo:test"; TEST(ScudoMapTest, PageSize) { @@ -89,3 +93,40 @@ TEST(ScudoMapTest, MapGrowUnmap) { memset(reinterpret_cast<void *>(Q), 0xbb, PageSize); MemMap.unmap(); } + +// Verify that zeroing works properly. +TEST(ScudoMapTest, Zeroing) { + scudo::ReservedMemoryT ReservedMemory; + const scudo::uptr PageSize = scudo::getPageSizeCached(); + const scudo::uptr Size = 3 * PageSize; + ReservedMemory.create(/*Addr=*/0U, Size, MappingName); + ASSERT_TRUE(ReservedMemory.isCreated()); + + scudo::MemMapT MemMap = ReservedMemory.dispatch(ReservedMemory.getBase(), ReservedMemory.getCapacity()); + EXPECT_TRUE(MemMap.remap(MemMap.getBase(), MemMap.getCapacity(), MappingName)); + unsigned char *Data = reinterpret_cast<unsigned char*>(MemMap.getBase()); + memset(Data, 1U, MemMap.getCapacity()); + // Spot check some values. + EXPECT_EQ(1U, Data[0]); + EXPECT_EQ(1U, Data[PageSize]); + EXPECT_EQ(1U, Data[PageSize * 2]); + MemMap.releaseAndZeroPagesToOS(MemMap.getBase(), MemMap.getCapacity()); + EXPECT_EQ(0U, Data[0]); + EXPECT_EQ(0U, Data[PageSize]); + EXPECT_EQ(0U, Data[PageSize * 2]); + +#if SCUDO_LINUX + // Now verify that if madvise fails, the data is still zeroed. + memset(Data, 1U, MemMap.getCapacity()); + EXPECT_NE(-1, mlock(Data, MemMap.getCapacity())); + EXPECT_EQ(1U, Data[0]); + EXPECT_EQ(1U, Data[PageSize]); + EXPECT_EQ(1U, Data[PageSize * 2]); + MemMap.releaseAndZeroPagesToOS(MemMap.getBase(), MemMap.getCapacity()); + EXPECT_EQ(0U, Data[0]); + EXPECT_EQ(0U, Data[PageSize]); + EXPECT_EQ(0U, Data[PageSize * 2]); +#endif + + MemMap.unmap(); +} |
| ✅ With the latest revision this PR passed the C/C++ code formatter. |
ChiaHungDuan approved these changes Nov 13, 2025
Comment on lines +128 to +130
| if (rc == -1) { | ||
| // If we can't madvies the memory, then we still need to zero it. | ||
| memset(Addr, 0, Size); |
Contributor
There was a problem hiding this comment.
I was thinking if we want to check something like "errno == EPERM". It seems not making any difference to the problem but maybe it helps with some other cases. I guess not...
Contributor Author
There was a problem hiding this comment.
Yeah, when I checked this the errno was EINVAL, which is not what I was expecting. It seems safer to just memset when it fails for whatever reason.
Member
| The new test is failing for me while testing on Gentoo/arm64: This is on top of d9dfe75. Neoverse-N1. glibc 2.41. |
Contributor
| The new test also fails on ppc64le starting with https://lab.llvm.org/buildbot/#/builders/145/builds/10835 |
Contributor
| @cferris1000 Do you mind having a look at the ppc64le failures, possibly reverting until you can investigate further? |
Contributor Author
| This looks like mlock/munlock are not supported everywhere, so I'll fix the test right now. |
Contributor Author
| I believe this should pull request should fix this: |
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit. This suggestion is invalid because no changes were made to the code. Suggestions cannot be applied while the pull request is closed. Suggestions cannot be applied while viewing a subset of changes. Only one suggestion per line can be applied in a batch. Add this suggestion to a batch that can be applied as a single commit. Applying suggestions on deleted lines is not supported. You must change the existing code in this line in order to create a valid suggestion. Outdated suggestions cannot be applied. This suggestion has been applied or marked resolved. Suggestions cannot be applied from pending reviews. Suggestions cannot be applied on multi-line comments. Suggestions cannot be applied while the pull request is queued to merge. Suggestion cannot be applied right now. Please check back later.
If a caller has locked memory, then the madvise call will fail. In that case, zero the memory so that we don't return non-zeroed memory for calloc calls since we thought the memory had been released.