maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[libFuzzer] implement crash-resistant merge (https://github.com/google/sanitizers/issues/722). This is a first experimental variant that needs some more testing, thus not yet adding a lit test (but there are unit tests). 

llvm-svn: 289166
This commit is contained in:
Kostya Serebryany 2016-12-09 01:17:24 +00:00
parent 8786754cc3
commit 111e1d69e3
8 changed files with 488 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
llvm/lib/Fuzzer/CMakeLists.txt
View File

@ -1,6 +1,6 @@
set(LIBFUZZER_FLAGS_BASE "${CMAKE_CXX_FLAGS}")
# Disable the coverage and sanitizer instrumentation for the fuzzer itself.
set(CMAKE_CXX_FLAGS "${LIBFUZZER_FLAGS_BASE} -fno-sanitize=all -fno-sanitize-coverage=edge,trace-cmp,indirect-calls,8bit-counters -Werror")
set(CMAKE_CXX_FLAGS "${LIBFUZZER_FLAGS_BASE} -fno-sanitize=all -fno-sanitize-coverage=trace-pc-guard,edge,trace-cmp,indirect-calls,8bit-counters -Werror")
if( LLVM_USE_SANITIZE_COVERAGE )
if(NOT "${LLVM_USE_SANITIZER}" STREQUAL "Address")
message(FATAL_ERROR
@ -18,6 +18,7 @@ if( LLVM_USE_SANITIZE_COVERAGE )
FuzzerIOPosix.cpp
FuzzerIOWindows.cpp
FuzzerLoop.cpp
FuzzerMerge.cpp
FuzzerMutate.cpp
FuzzerSHA1.cpp
FuzzerTracePC.cpp

19
llvm/lib/Fuzzer/FuzzerDriver.cpp
View File

@ -219,8 +219,8 @@ static void WorkerThread(const std::string &Cmd, std::atomic<int> *Counter,
}
}
static std::string CloneArgsWithoutX(const std::vector<std::string> &Args,
const char *X1, const char *X2) {
std::string CloneArgsWithoutX(const std::vector<std::string> &Args,
const char *X1, const char *X2) {
std::string Cmd;
for (auto &S : Args) {
if (FlagValue(S.c_str(), X1) || FlagValue(S.c_str(), X2))
@ -230,11 +230,6 @@ static std::string CloneArgsWithoutX(const std::vector<std::string> &Args,
return Cmd;
}
static std::string CloneArgsWithoutX(const std::vector<std::string> &Args,
const char *X) {
return CloneArgsWithoutX(Args, X, X);
}
static int RunInMultipleProcesses(const std::vector<std::string> &Args,
int NumWorkers, int NumJobs) {
std::atomic<int> Counter(0);
@ -499,6 +494,16 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {
exit(0);
}
if (Flags.experimental_merge) {
if (Options.MaxLen == 0)
F->SetMaxInputLen(kMaxSaneLen);
if (Flags.merge_control_file)
F->CrashResistantMergeInternalStep(Flags.merge_control_file);
else
F->CrashResistantMerge(Args, *Inputs);
exit(0);
}
size_t TemporaryMaxLen = Options.MaxLen ? Options.MaxLen : kMaxSaneLen;
UnitVector InitialCorpus;

3
llvm/lib/Fuzzer/FuzzerFlags.def
View File

@ -37,6 +37,9 @@ FUZZER_FLAG_INT(help, 0, "Print help.")
FUZZER_FLAG_INT(merge, 0, "If 1, the 2-nd, 3-rd, etc corpora will be "
"merged into the 1-st corpus. Only interesting units will be taken. "
"This flag can be used to minimize a corpus.")
FUZZER_FLAG_INT(experimental_merge, 0, "Experimental crash-resistant, "
"may eventually replace -merge.")
FUZZER_FLAG_STRING(merge_control_file, "internal flag")
FUZZER_FLAG_INT(minimize_crash, 0, "If 1, minimizes the provided"
" crash input. Use with -runs=N or -max_total_time=N to limit "
"the number attempts")

3
llvm/lib/Fuzzer/FuzzerInternal.h
View File

@ -89,6 +89,9 @@ public:
// Merge Corpora[1:] into Corpora[0].
void Merge(const std::vector<std::string> &Corpora);
void CrashResistantMerge(const std::vector<std::string> &Args,
const std::vector<std::string> &Corpora);
void CrashResistantMergeInternalStep(const std::string &ControlFilePath);
// Returns a subset of 'Extra' that adds coverage to 'Initial'.
UnitVector FindExtraUnits(const UnitVector &Initial, const UnitVector &Extra);
MutationDispatcher &GetMD() { return MD; }

255
llvm/lib/Fuzzer/FuzzerMerge.cpp Normal file
View File

@ -0,0 +1,255 @@
//===- FuzzerMerge.cpp - merging corpora ----------------------------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
// Merging corpora.
//===----------------------------------------------------------------------===//
#include "FuzzerInternal.h"
#include "FuzzerIO.h"
#include "FuzzerMerge.h"
#include "FuzzerTracePC.h"
#include "FuzzerUtil.h"
#include <fstream>
#include <sstream>
namespace fuzzer {
bool Merger::Parse(const std::string &Str, bool ParseCoverage) {
std::istringstream SS(Str);
return Parse(SS, ParseCoverage);
}
void Merger::ParseOrExit(std::istream &IS, bool ParseCoverage) {
if (!Parse(IS, ParseCoverage)) {
Printf("MERGE: failed to parse the control file (unexpected error)\n");
exit(1);
}
}
// The control file example:
//
// 3 # The number of inputs
// 1 # The number of inputs in the first corpus, <= the previous number
// file0
// file1
// file2 # One file name per line.
// STARTED 0 123 # FileID, file size
// DONE 0 1 4 6 8 # FileID COV1 COV2 ...
// STARTED 1 456 # If DONE is missing, the input crashed while processing.
// STARTED 2 567
// DONE 2 8 9
bool Merger::Parse(std::istream &IS, bool ParseCoverage) {
LastFailure.clear();
std::string Line;
// Parse NumFiles.
if (!std::getline(IS, Line, '\n')) return false;
std::istringstream L1(Line);
size_t NumFiles = 0;
L1 >> NumFiles;
if (NumFiles == 0 || NumFiles > 10000000) return false;
// Parse NumFilesInFirstCorpus.
if (!std::getline(IS, Line, '\n')) return false;
std::istringstream L2(Line);
NumFilesInFirstCorpus = NumFiles + 1;
L2 >> NumFilesInFirstCorpus;
if (NumFilesInFirstCorpus > NumFiles) return false;
// Parse file names.
Files.resize(NumFiles);
for (size_t i = 0; i < NumFiles; i++)
if (!std::getline(IS, Files[i].Name, '\n'))
return false;
// Parse STARTED and DONE lines.
size_t ExpectedStartMarker = 0;
const size_t kInvalidStartMarker = -1;
size_t LastSeenStartMarker = kInvalidStartMarker;
while (std::getline(IS, Line, '\n')) {
std::istringstream ISS1(Line);
std::string Marker;
size_t N;
ISS1 >> Marker;
ISS1 >> N;
if (Marker == "STARTED") {
// STARTED FILE_ID FILE_SIZE
if (ExpectedStartMarker != N)
return false;
ISS1 >> Files[ExpectedStartMarker].Size;
LastSeenStartMarker = ExpectedStartMarker;
assert(ExpectedStartMarker < Files.size());
ExpectedStartMarker++;
} else if (Marker == "DONE") {
// DONE FILE_SIZE COV1 COV2 COV3 ...
size_t CurrentFileIdx = N;
if (CurrentFileIdx != LastSeenStartMarker)
return false;
LastSeenStartMarker = kInvalidStartMarker;
if (ParseCoverage) {
while (!ISS1.rdstate()) {
ISS1 >> std::hex >> N;
Files[CurrentFileIdx].Features.insert(N);
}
}
} else {
return false;
}
}
if (LastSeenStartMarker != kInvalidStartMarker)
LastFailure = Files[LastSeenStartMarker].Name;
FirstNotProcessedFile = ExpectedStartMarker;
return true;
}
// Decides which files need to be merged (add thost to NewFiles).
// Returns the number of new features added.
size_t Merger::Merge(std::vector<std::string> *NewFiles) {
NewFiles->clear();
assert(NumFilesInFirstCorpus <= Files.size());
std::set<size_t> AllFeatures;
// What features are in the initial corpus?
for (size_t i = 0; i < NumFilesInFirstCorpus; i++) {
auto &Cur = Files[i].Features;
AllFeatures.insert(Cur.begin(), Cur.end());
}
size_t InitialNumFeatures = AllFeatures.size();
// Remove all features that we already know from all other inputs.
for (size_t i = NumFilesInFirstCorpus; i < Files.size(); i++) {
auto &Cur = Files[i].Features;
std::set<size_t> Tmp;
std::set_difference(Cur.begin(), Cur.end(), AllFeatures.begin(),
AllFeatures.end(), std::inserter(Tmp, Tmp.begin()));
Cur.swap(Tmp);
}
// Sort. Give preference to
// * smaller files
// * files with more features.
std::sort(Files.begin() + NumFilesInFirstCorpus, Files.end(),
[&](const MergeFileInfo &a, const MergeFileInfo &b) -> bool {
if (a.Size != b.Size)
return a.Size < b.Size;
return a.Features.size() > b.Features.size();
});
// One greedy pass: add the file's features to AllFeatures.
// If new features were added, add this file to NewFiles.
for (size_t i = NumFilesInFirstCorpus; i < Files.size(); i++) {
auto &Cur = Files[i].Features;
// Printf("%s -> sz %zd ft %zd\n", Files[i].Name.c_str(),
// Files[i].Size, Cur.size());
size_t OldSize = AllFeatures.size();
AllFeatures.insert(Cur.begin(), Cur.end());
if (AllFeatures.size() > OldSize)
NewFiles->push_back(Files[i].Name);
}
return AllFeatures.size() - InitialNumFeatures;
}
// Inner process. May crash if the target crashes.
void Fuzzer::CrashResistantMergeInternalStep(const std::string &CFPath) {
Printf("MERGE-INNER: using the control file '%s'\n", CFPath.c_str());
Merger M;
std::ifstream IF(CFPath);
M.ParseOrExit(IF, false);
IF.close();
if (!M.LastFailure.empty())
Printf("MERGE-INNER: '%s' caused a failure at the previous merge step\n",
M.LastFailure.c_str());
Printf("MERGE-INNER: %zd total files;"
" %zd processed earlier; will process %zd files now\n",
M.Files.size(), M.FirstNotProcessedFile,
M.Files.size() - M.FirstNotProcessedFile);
std::ofstream OF(CFPath, std::ofstream::out | std::ofstream::app);
for (size_t i = M.FirstNotProcessedFile; i < M.Files.size(); i++) {
auto U = FileToVector(M.Files[i].Name);
std::ostringstream StartedLine;
// Write the pre-run marker.
OF << "STARTED " << std::dec << i << " " << U.size() << "\n";
OF.flush(); // Flush is important since ExecuteCommand may crash.
// Run.
TPC.ResetMaps();
ExecuteCallback(U.data(), U.size());
// Collect coverage.
std::set<size_t> Features;
TPC.CollectFeatures([&](size_t Feature) -> bool {
Features.insert(Feature);
return true;
});
// Show stats.
TotalNumberOfRuns++;
if (!(TotalNumberOfRuns & (TotalNumberOfRuns - 1)))
PrintStats("pulse ");
// Write the post-run marker and the coverage.
OF << "DONE " << i;
for (size_t F : Features)
OF << " " << std::hex << F;
OF << "\n";
}
}
// Outer process. Does not call the target code and thus sohuld not fail.
void Fuzzer::CrashResistantMerge(const std::vector<std::string> &Args,
const std::vector<std::string> &Corpora) {
if (Corpora.size() <= 1) {
Printf("Merge requires two or more corpus dirs\n");
return;
}
std::vector<std::string> AllFiles;
ListFilesInDirRecursive(Corpora[0], nullptr, &AllFiles, /*TopDir*/true);
size_t NumFilesInFirstCorpus = AllFiles.size();
for (size_t i = 1; i < Corpora.size(); i++)
ListFilesInDirRecursive(Corpora[i], nullptr, &AllFiles, /*TopDir*/true);
Printf("MERGE-OUTER: %zd files, %zd in the initial corpus\n",
AllFiles.size(), NumFilesInFirstCorpus);
std::string CFPath =
"libFuzzerTemp." + std::to_string(GetPid()) + ".txt";
// Write the control file.
DeleteFile(CFPath);
std::ofstream ControlFile(CFPath);
ControlFile << AllFiles.size() << "\n";
ControlFile << NumFilesInFirstCorpus << "\n";
for (auto &Path: AllFiles)
ControlFile << Path << "\n";
ControlFile.close();
// Execute the inner process untill it passes.
// Every inner process should execute at least one input.
std::string BaseCmd = CloneArgsWithoutX(Args, "keep-all-flags");
for (size_t i = 1; i <= AllFiles.size(); i++) {
Printf("MERGE-OUTER: attempt %zd\n", i);
auto ExitCode =
ExecuteCommand(BaseCmd + " -merge_control_file=" + CFPath);
if (!ExitCode) {
Printf("MERGE-OUTER: succesfull in %zd attempt(s)\n", i);
break;
}
}
// Read the control file and do the merge.
Merger M;
std::ifstream IF(CFPath);
M.ParseOrExit(IF, true);
IF.close();
std::vector<std::string> NewFiles;
size_t NumNewFeatures = M.Merge(&NewFiles);
Printf("MERGE-OUTER: %zd new files with %zd new features added\n",
NewFiles.size(), NumNewFeatures);
for (auto &F: NewFiles)
WriteToOutputCorpus(FileToVector(F));
// We are done, delete the control file.
DeleteFile(CFPath);
}
} // namespace fuzzer

70
llvm/lib/Fuzzer/FuzzerMerge.h Normal file
View File

@ -0,0 +1,70 @@
//===- FuzzerMerge.h - merging corpa ----------------------------*- C++ -* ===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
// Merging Corpora.
//
// The task:
// Take the existing corpus (possibly empty) and merge new inputs into
// it so that only inputs with new coverage ('features') are added.
// The process should tolerate the crashes, OOMs, leaks, etc.
//
// Algorithm:
// The outter process collects the set of files and writes their names
// into a temporary "control" file, then repeatedly launches the inner
// process until all inputs are processed.
// The outer process does not actually execute the target code.
//
// The inner process reads the control file and sees a) list of all the inputs
// and b) the last processed input. Then it starts processing the inputs one
// by one. Before processing every input it writes one line to control file:
// STARTED INPUT_ID INPUT_SIZE
// After processing an input it write another line:
// DONE INPUT_ID Feature1 Feature2 Feature3 ...
// If a crash happens while processing an input the last line in the control
// file will be "STARTED INPUT_ID" and so the next process will know
// where to resume.
//
// Once all inputs are processed by the innner process(es) the outer process
// reads the control files and does the merge based entirely on the contents
// of control file.
// It uses a single pass greedy algorithm choosing first the smallest inputs
// within the same size the inputs that have more new features.
//
//===----------------------------------------------------------------------===//
#ifndef LLVM_FUZZER_MERGE_H
#define LLVM_FUZZER_MERGE_H
#include "FuzzerDefs.h"
#include <istream>
#include <set>
namespace fuzzer {
struct MergeFileInfo {
std::string Name;
size_t Size = 0;
std::set<size_t> Features;
};
struct Merger {
std::vector<MergeFileInfo> Files;
size_t NumFilesInFirstCorpus = 0;
size_t FirstNotProcessedFile = 0;
std::string LastFailure;
bool Parse(std::istream &IS, bool ParseCoverage);
bool Parse(const std::string &Str, bool ParseCoverage);
void ParseOrExit(std::istream &IS, bool ParseCoverage);
size_t Merge(std::vector<std::string> *NewFiles);
};
} // namespace fuzzer
#endif // LLVM_FUZZER_MERGE_H

8
llvm/lib/Fuzzer/FuzzerUtil.h
View File

@ -66,5 +66,13 @@ FILE *OpenProcessPipe(const char *Command, const char *Mode);
const void *SearchMemory(const void *haystack, size_t haystacklen,
const void *needle, size_t needlelen);
std::string CloneArgsWithoutX(const std::vector<std::string> &Args,
const char *X1, const char *X2);
inline std::string CloneArgsWithoutX(const std::vector<std::string> &Args,
const char *X) {
return CloneArgsWithoutX(Args, X, X);
}
} // namespace fuzzer
#endif // LLVM_FUZZER_UTIL_H

135
llvm/lib/Fuzzer/test/FuzzerUnittest.cpp
View File

@ -8,6 +8,7 @@
#include "FuzzerCorpus.h"
#include "FuzzerInternal.h"
#include "FuzzerDictionary.h"
#include "FuzzerMerge.h"
#include "FuzzerMutate.h"
#include "FuzzerRandom.h"
#include "gtest/gtest.h"
@ -598,3 +599,137 @@ TEST(Corpus, Distribution) {
EXPECT_GT(Hist[i], TriesPerUnit / N / 3);
}
}
TEST(Merge, Bad) {
const char *kInvalidInputs[] = {
"",
"x",
"3\nx",
"2\n3",
"2\n2",
"2\n2\nA\n",
"2\n2\nA\nB\nC\n",
"0\n0\n",
"1\n1\nA\nDONE 0",
"1\n1\nA\nSTARTED 1",
};
Merger M;
for (auto S : kInvalidInputs) {
// fprintf(stderr, "TESTING:\n%s\n", S);
EXPECT_FALSE(M.Parse(S, false));
}
}
void EQ(const std::set<size_t> &A, const std::set<size_t> &B) {
EXPECT_EQ(A, B);
}
void EQ(const std::vector<std::string> &A, const std::vector<std::string> &B) {
std::set<std::string> a(A.begin(), A.end());
std::set<std::string> b(B.begin(), B.end());
EXPECT_EQ(a, b);
}
static void Merge(const std::string &Input,
const std::vector<std::string> Result,
size_t NumNewFeatures) {
Merger M;
std::vector<std::string> NewFiles;
EXPECT_TRUE(M.Parse(Input, true));
EXPECT_EQ(NumNewFeatures, M.Merge(&NewFiles));
EQ(NewFiles, Result);
}
TEST(Merge, Good) {
Merger M;
EXPECT_TRUE(M.Parse("1\n0\nAA\n", false));
EXPECT_EQ(M.Files.size(), 1U);
EXPECT_EQ(M.NumFilesInFirstCorpus, 0U);
EXPECT_EQ(M.Files[0].Name, "AA");
EXPECT_TRUE(M.LastFailure.empty());
EXPECT_EQ(M.FirstNotProcessedFile, 0U);
EXPECT_TRUE(M.Parse("2\n1\nAA\nBB\nSTARTED 0 42\n", false));
EXPECT_EQ(M.Files.size(), 2U);
EXPECT_EQ(M.NumFilesInFirstCorpus, 1U);
EXPECT_EQ(M.Files[0].Name, "AA");
EXPECT_EQ(M.Files[1].Name, "BB");
EXPECT_EQ(M.LastFailure, "AA");
EXPECT_EQ(M.FirstNotProcessedFile, 1U);
EXPECT_TRUE(M.Parse("3\n1\nAA\nBB\nC\n"
"STARTED 0 1000\n"
"DONE 0 1 2 3\n"
"STARTED 1 1001\n"
"DONE 1 4 5 6 \n"
"STARTED 2 1002\n"
"", true));
EXPECT_EQ(M.Files.size(), 3U);
EXPECT_EQ(M.NumFilesInFirstCorpus, 1U);
EXPECT_EQ(M.Files[0].Name, "AA");
EXPECT_EQ(M.Files[0].Size, 1000U);
EXPECT_EQ(M.Files[1].Name, "BB");
EXPECT_EQ(M.Files[1].Size, 1001U);
EXPECT_EQ(M.Files[2].Name, "C");
EXPECT_EQ(M.Files[2].Size, 1002U);
EXPECT_EQ(M.LastFailure, "C");
EXPECT_EQ(M.FirstNotProcessedFile, 3U);
EQ(M.Files[0].Features, {1, 2, 3});
EQ(M.Files[1].Features, {4, 5, 6});
std::vector<std::string> NewFiles;
EXPECT_TRUE(M.Parse("3\n2\nAA\nBB\nC\n"
"STARTED 0 1000\nDONE 0 1 2 3\n"
"STARTED 1 1001\nDONE 1 4 5 6 \n"
"STARTED 2 1002\nDONE 2 6 1 3 \n"
"", true));
EXPECT_EQ(M.Files.size(), 3U);
EXPECT_EQ(M.NumFilesInFirstCorpus, 2U);
EXPECT_TRUE(M.LastFailure.empty());
EXPECT_EQ(M.FirstNotProcessedFile, 3U);
EQ(M.Files[0].Features, {1, 2, 3});
EQ(M.Files[1].Features, {4, 5, 6});
EQ(M.Files[2].Features, {1, 3, 6});
EXPECT_EQ(0U, M.Merge(&NewFiles));
EQ(NewFiles, {});
EXPECT_TRUE(M.Parse("3\n1\nA\nB\nC\n"
"STARTED 0 1000\nDONE 0 1 2 3\n"
"STARTED 1 1001\nDONE 1 4 5 6 \n"
"STARTED 2 1002\nDONE 2 6 1 3 \n"
"", true));
EXPECT_EQ(3U, M.Merge(&NewFiles));
EQ(NewFiles, {"B"});
}
TEST(Merge, Merge) {
Merge("3\n1\nA\nB\nC\n"
"STARTED 0 1000\nDONE 0 1 2 3\n"
"STARTED 1 1001\nDONE 1 4 5 6 \n"
"STARTED 2 1002\nDONE 2 6 1 3 \n",
{"B"}, 3);
Merge("3\n0\nA\nB\nC\n"
"STARTED 0 2000\nDONE 0 1 2 3\n"
"STARTED 1 1001\nDONE 1 4 5 6 \n"
"STARTED 2 1002\nDONE 2 6 1 3 \n",
{"A", "B", "C"}, 6);
Merge("4\n0\nA\nB\nC\nD\n"
"STARTED 0 2000\nDONE 0 1 2 3\n"
"STARTED 1 1101\nDONE 1 4 5 6 \n"
"STARTED 2 1102\nDONE 2 6 1 3 100 \n"
"STARTED 3 1000\nDONE 3 1 \n",
{"A", "B", "C", "D"}, 7);
Merge("4\n1\nA\nB\nC\nD\n"
"STARTED 0 2000\nDONE 0 4 5 6 7 8\n"
"STARTED 1 1100\nDONE 1 1 2 3 \n"
"STARTED 2 1100\nDONE 2 2 3 \n"
"STARTED 3 1000\nDONE 3 1 \n",
{"B", "D"}, 3);
}