Issue 241315
Summary [PGO] clang asserts in ProfileSummary::getDetailedSummaryMD when a summary NumCounts exceeds 32 bits
Labels clang
Assignees
Reporter cactusBalll
    **Generated by Fuzzer.** Compiling with `-fprofile-instr-use=<profdata>` aborts in
`llvm::ProfileSummary::getDetailedSummaryMD()` when any profile-summary cutoff entry has
`NumBlocks > UINT32_MAX`. The reader stores `NumBlocks` as a 64-bit `uint64_t`, but
`getDetailedSummaryMD()` builds an `i32` `ConstantInt` from `Entry.NumCounts`, tripping the
`APInt::isUIntN` assertion in `+assertions` builds.

## Environment

- clang `24.0.0git`, commit `0db002877564df53575e997e985538a949a03327`
- build config `+assertions,+asan,+ubsan`, `x86_64-unknown-linux-gnu`

## Reproduction

The crash does not depend on the compiled source contents; a one-line translation unit is
enough. Save the following as `repro.c`:

```c
int main(void) { return 0; }
```

Save this corrupted indexed profile as `mut.profdata`:

```bash
base64 -d > mut.profdata <<'EOF'
/2xwcm9maYEOAAAAAAAAAQAAAAAAAAAAAAAAAAAAAAAoCwAAAAAAAAAAAAAAAAAAOAwAAAAAAAAA
AAAAAAAAAEAMAAAAAAAABgAAAAAAAAAQAAAAAAAAABYAAAAAAAAASAAAAAAAAAAABAEAAAAAAAAE
AQAAAAAAAEUBAAAAAABTcQEAAAAAABAnAAAAAAAAAAQBAAAAAAABAAAAAAAAAKCGAQAAAAAAAAQB
AAAAAAABAAAAAIkAAEANAwAAAAAAAAQBAAAAAAABAAAAAAAAAOCTBAAAAAAAAAQBAAAAAAABAAAA
AAAAAIAaBgAAAAAAAAQBAAAAAAABAAAAAAAAACChBwAAAAAAAAQBAAAAAAABAAAAAAAAAMAnCQAA
AAAAAAQBAAAAAAABAAAAAAAAAGCuCgAAAAAAAAQBAAAAAAABAAAAAAAAAAA1DAAAAAAAAEEAAAAA
AAACAAAAAAAAAKC7DQAAAAAAkBoAAAAAAAADAAAAAAAAAPB+DgAAAAAAkBoAAAAAAAADAAAAAAAA
ADAbDwAAAAAAIAIAAAAAAAAGAAAAAAAAAFg+DwAAAAAAIAAAAAAAAAANAAAAAAAAANxBDwAAAAAA
AgAAAAAAAAAWAAAAAAAAADZCDwAAAAAAAQAAAAAAAAAYAAAAAAAAAD9CD/EAAAAAAQAAAAAAAAAY
AAAAAAAAAAEAYD5nW1c0/kwGAAAAAAAAAPgAAAAAAAAAZnVuY18xiXQdLjZwoAwYAAAAAAAAAAAE
AQAAAAAAIAAAAAAAAAAAQQAAAAAAAJAaAAAAAAAAoAAAAAAAAAAgAgAAAAAAAAAAAAAAAAAAIAAA
AAAAAAAcAAAAAAAAAFYAAAAAAAAAXgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAgAAAAAAAAAAAAAAAAAAAAIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAACAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAYAAAAAQAAAAAAAAABAAAAAAAAAAAA
AAABAMHGrewKbEj3CQAAAAAAAAA4AAAAAAAAAHBnb190Z3RfN0iFOR0+qgkCAgAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAgAAAAAAAAAAQAjWGbCseW5GgoAAAAAAAAAOAAA
AAAAAABwZ29fdGFpbF8wBYhNKkRa4QoCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAACAAAAAAAAAABAKUKDg740XdfCgAAAAAAAAA4AAAAAAAAAHBnb190YWlsXzIFiE0qRFrh
CgIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAIAAAAAAAAAAEAhnYlqVQo
LxsMAAAAAAAAAEAAAAAAAAAAcGdvX3ZhcmFyZ18w2a+nuWlbcwcDAAAAAAAAAAYAAAAAAAAAAAAA
AAAAAAACAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAIAAAAAAAAAAIAp5hP9Tl8iqoTAAAAAAAAADgA
AAAAAAAAY3NtaXRoX2NvbXB1dGVfaGFzaEiFOR0+qgkCAgAAAAAAAAAIAAAAAAAAAAEAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAgAAAAAAAAAh3DWZeTwz2cJAAAAAAAAADAAAAAAAAAAcGdvX2hvdF8y
////79MKTQoBAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAgAAAAAAAAAAgCrp7DqZIt3
3gkAAAAAAAAAOAAAAAAAAABwZ29fdGd0XzZIhTkdPqoJAgIAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAIAAAAAAAAAMtxf//ODPbgCgAAAAAAAAAwAAAAAAAAAHBnb190YWls
XzP////v0wpNCgEAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACAAAAAAAAAABAKwJtefN
Vm8RCQAAAAAAAAAwAAAAAAAAAHBnb19ob3RfMf///+/TCk0KAQAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAIAAAAAAAAAAIArZMwpxsopYQJAAAAAAAAADgAAAAAAAAAcGdvX3RndF80SIU5
HT6qCQICAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACAAAAAAAAABN98wb
tULWmwkAAAAAAAAAOAAAAAAAAABwZ29fdGd0XzNIhTkdPqoJAgIAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAIAAAAAAAAAAEALtHhzfb93bAJAAAAAAAAADgAAAAAAAAAcGdv
X3RndF8ySIU5HT6qCQICAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACAAA
AAAAAAABALBM7O59pJWYCgAAAAAAAAB4AAAAAAAAAHBnb19jb2xkXzButVddneg2BAoAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACAAAAAAAAAABAPIeMoAmI5vJ
CQAAAAAAAAA4AAAAAAAAAHBnb190Z3RfMEiFOR0+qgkCAgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAgAAAAAAAAAAQC12y4N5kSQfwkAAAAAAAAAOAAAAAAAAABwZ29fdGd0
XzFIhTkdPqoJAgIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAIAAAAAAAA
AAIAtwo9LeJSQjQJAAAAAAAAADAAAAAAAAAAcGdvX2hvdF8w////79MKTQoBAAAAAAAAANIAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAgAAAAAAAAAN/NSOyBlsaoJAAAAAAAAADAAAAAAAAAAcGdvX2hv
dF8z////79MKTQoBAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAgAAAAAAAAAAQD61Y3n
NmSV2wQAAAAAAAAAUAAAAAAAAABtYWluqNfIoasLCAQFAAAAAAAAAAAIAAAAAAAAAAEAAAAAAAAC
AAAAAAAAAAEAAAAAAAAAAAQAAAAAAAAAAAAAAAAAAAAAAAAAAAAACAAAAAAAAAABAPzFy30ecXf6
CgAAAAAAAABIAAAAAAAAAHBnb190YWlsXzEFiE3aEQzzAgIAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAYAAAAAQAAAAAAAAABAAAAAAAAAAAAAAABAN0TvVjqKQ+LCQAAAAAA
AAA4AAAAAAAAAHBnb190Z3RfNUiFOR0+qgkCAgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAgAAAAAAAAAAQDeeXS139Oz5FAAAAAAAAAAMAAAAAAAAAAvZGV2L3NobS9wZ28t
ZnV6ei13b3JrLW1heC9pdGVyXzAwMDY2MV8zMmQ4MzNkOWUyYjk5Yjg0L3Byb2dfMC5jO3RyYW5z
cGFyZW50X2NyY////+/TCgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAuCgAAAAAAAIkKAAAAAAAAAAAAAAAAAAAAAAAAAAAA
AAAAAAAAAAAA
EOF
```

Then run:

```bash
clang -O2 -fprofile-instr-use=mut.profdata -c repro.c
```

`mut.profdata` is accepted by `llvm-profdata show` (the fuzzer metadata records
`profile_valid = true`); the crash happens later in the `pgo-instr-use` pass.

## Output

```text
clang-24: llvm/include/llvm/ADT/APInt.h:128:
llvm::APInt::APInt(unsigned int, uint64_t, bool, bool):
Assertion `llvm::isUIntN(BitWidth, val) &&
"Value is not an N-bit unsigned value"' failed.
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/
Stack dump:
...
#18 llvm::ConstantInt::get(llvm::IntegerType *, unsigned long, bool, bool)
 llvm/lib/IR/Constants.cpp:1012
#21 llvm::ProfileSummary::getDetailedSummaryMD(llvm::LLVMContext &)
 llvm/lib/IR/ProfileSummary.cpp:63
#22 llvm::ProfileSummary::getMD(llvm::LLVMContext &, bool, bool)
 llvm/lib/IR/ProfileSummary.cpp:99
#23 annotateAllFunctions(...)
 llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp
#24 llvm::PGOInstrumentationUse::run(...)
 llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp:2283
```

## Root cause

`llvm/lib/IR/ProfileSummary.cpp`:

```cpp
Metadata *EntryMD[3] = {
    ConstantAsMetadata::get(ConstantInt::get(Int32Ty, Entry.Cutoff)),
    ConstantAsMetadata::get(ConstantInt::get(Int64Ty, Entry.MinCount)),
    ConstantAsMetadata::get(ConstantInt::get(Int32Ty, Entry.NumCounts))};
```

`ProfileSummaryEntry::NumCounts` is `uint64_t`
(`llvm/include/llvm/IR/ProfileSummary.h`), and `IndexedInstrProfReader::readSummary()`
passes the on-disk `NumBlocks` through without a range check. A corrupted profile can therefore
provide a value that does not fit in 32 bits.

## Relation to existing issues

- #230018 / PR #231727 cover structural summary bounds (`NumSummaryFields`,
 `NumCutoffEntries`, remaining buffer size), not the semantic range of `NumBlocks`.
- #238440: a text-profile internal counter overflow aborts
 `InstrProfSummaryBuilder::addInternalCount()` in `llvm-profdata merge`; the input format,
  tool, processing stage, and failure site are all different from the `i32` metadata
  construction here.
- #65416 (closed): requested that `llvm-profdata` report unreasonable block counts; that is a
 counter-range diagnostic request, not an `APInt::isUIntN` crash in
 `ProfileSummary::getDetailedSummaryMD()`.
- #225435 covers on-disk hash-table record decoding.


_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs

Reply via email to