blob: 90a10c9bf52aafb595c2a372216d0ee14f385387 [file] [log] [blame]
/*
*
* honggfuzz - architecture dependent code (LINUX/PTRACE)
* -----------------------------------------
*
* Author: Robert Swiecki <swiecki@google.com>
*
* Copyright 2010-2015 by Google Inc. All Rights Reserved.
*
* Licensed under the Apache License, Version 2.0 (the "License"); you may
* not use this file except in compliance with the License. You may obtain
* a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
* implied. See the License for the specific language governing
* permissions and limitations under the License.
*
*/
#include "linux/trace.h"
#include <ctype.h>
#include <dirent.h>
#include <elf.h>
#include <endian.h>
#include <errno.h>
#include <fcntl.h>
#include <inttypes.h>
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/cdefs.h>
#include <sys/personality.h>
#include <sys/ptrace.h>
#include <sys/resource.h>
#include <sys/stat.h>
#include <sys/syscall.h>
#include <sys/time.h>
#include <sys/types.h>
#include <sys/uio.h>
#include <sys/user.h>
#include <sys/wait.h>
#include <time.h>
#include <unistd.h>
#include "libhfcommon/common.h"
#include "libhfcommon/files.h"
#include "libhfcommon/log.h"
#include "libhfcommon/util.h"
#include "linux/bfd.h"
#include "linux/unwind.h"
#include "sancov.h"
#include "sanitizers.h"
#include "subproc.h"
#include "socketfuzzer.h"
#if defined(__ANDROID__)
#include "capstone.h"
#endif
#if defined(__i386__) || defined(__arm__) || defined(__powerpc__)
#define REG_TYPE uint32_t
#define REG_PM PRIx32
#define REG_PD "0x%08"
#elif defined(__x86_64__) || defined(__aarch64__) || defined(__powerpc64__) || \
defined(__mips__) || defined(__mips64__)
#define REG_TYPE uint64_t
#define REG_PM PRIx64
#define REG_PD "0x%016"
#endif
/*
* Size in characters required to store a string representation of a
* register value (0xdeadbeef style))
*/
#define REGSIZEINCHAR (2 * sizeof(REG_TYPE) + 3)
#if defined(__i386__) || defined(__x86_64__)
#define MAX_INSTR_SZ 16
#elif defined(__arm__) || defined(__powerpc__) || defined(__powerpc64__)
#define MAX_INSTR_SZ 4
#elif defined(__aarch64__)
#define MAX_INSTR_SZ 8
#elif defined(__mips__) || defined(__mips64__)
#define MAX_INSTR_SZ 8
#endif
#if defined(__i386__) || defined(__x86_64__)
struct user_regs_struct_32 {
uint32_t ebx;
uint32_t ecx;
uint32_t edx;
uint32_t esi;
uint32_t edi;
uint32_t ebp;
uint32_t eax;
uint16_t ds, __ds;
uint16_t es, __es;
uint16_t fs, __fs;
uint16_t gs, __gs;
uint32_t orig_eax;
uint32_t eip;
uint16_t cs, __cs;
uint32_t eflags;
uint32_t esp;
uint16_t ss, __ss;
};
struct user_regs_struct_64 {
uint64_t r15;
uint64_t r14;
uint64_t r13;
uint64_t r12;
uint64_t bp;
uint64_t bx;
uint64_t r11;
uint64_t r10;
uint64_t r9;
uint64_t r8;
uint64_t ax;
uint64_t cx;
uint64_t dx;
uint64_t si;
uint64_t di;
uint64_t orig_ax;
uint64_t ip;
uint64_t cs;
uint64_t flags;
uint64_t sp;
uint64_t ss;
uint64_t fs_base;
uint64_t gs_base;
uint64_t ds;
uint64_t es;
uint64_t fs;
uint64_t gs;
};
#define HEADERS_STRUCT struct user_regs_struct_64
#endif /* defined(__i386__) || defined(__x86_64__) */
#if defined(__arm__) || defined(__aarch64__)
#ifndef ARM_pc
#ifdef __ANDROID__ /* Building with NDK headers */
#define ARM_pc uregs[15]
#else /* Building with glibc headers */
#define ARM_pc 15
#endif
#endif /* ARM_pc */
#ifndef ARM_cpsr
#ifdef __ANDROID__ /* Building with NDK headers */
#define ARM_cpsr uregs[16]
#else /* Building with glibc headers */
#define ARM_cpsr 16
#endif
#endif /* ARM_cpsr */
struct user_regs_struct_32 {
uint32_t uregs[18];
};
struct user_regs_struct_64 {
uint64_t regs[31];
uint64_t sp;
uint64_t pc;
uint64_t pstate;
};
#define HEADERS_STRUCT struct user_regs_struct_64
#endif /* defined(__arm__) || defined(__aarch64__) */
#if defined(__powerpc64__) || defined(__powerpc__)
#define HEADERS_STRUCT struct pt_regs
struct user_regs_struct_32 {
uint32_t gpr[32];
uint32_t nip;
uint32_t msr;
uint32_t orig_gpr3;
uint32_t ctr;
uint32_t link;
uint32_t xer;
uint32_t ccr;
uint32_t mq;
uint32_t trap;
uint32_t dar;
uint32_t dsisr;
uint32_t result;
/*
* elf.h's ELF_NGREG says it's 48 registers, so kernel fills it in
* with some zeros
*/
uint32_t zero0;
uint32_t zero1;
uint32_t zero2;
uint32_t zero3;
};
struct user_regs_struct_64 {
uint64_t gpr[32];
uint64_t nip;
uint64_t msr;
uint64_t orig_gpr3;
uint64_t ctr;
uint64_t link;
uint64_t xer;
uint64_t ccr;
uint64_t softe;
uint64_t trap;
uint64_t dar;
uint64_t dsisr;
uint64_t result;
/*
* elf.h's ELF_NGREG says it's 48 registers, so kernel fills it in
* with some zeros
*/
uint64_t zero0;
uint64_t zero1;
uint64_t zero2;
uint64_t zero3;
};
#endif /* defined(__powerpc64__) || defined(__powerpc__) */
#if defined(__mips__) || defined(__mips64__)
struct user_regs_struct {
uint64_t regs[32];
uint64_t lo;
uint64_t hi;
uint64_t cp0_epc;
uint64_t cp0_badvaddr;
uint64_t cp0_status;
uint64_t cp0_cause;
};
#define HEADERS_STRUCT struct user_regs_struct
#endif /* defined(__mips__) || defined(__mips64__) */
#if defined(__ANDROID__)
/*
* Some Android ABIs don't implement PTRACE_GETREGS (e.g. aarch64)
*/
#if defined(PTRACE_GETREGS)
#define PTRACE_GETREGS_AVAILABLE 1
#else
#define PTRACE_GETREGS_AVAILABLE 0
#endif /* defined(PTRACE_GETREGS) */
#endif /* defined(__ANDROID__) */
static struct {
const char* descr;
bool important;
} arch_sigs[_NSIG + 1] = {
[0 ...(_NSIG)].important = false,
[0 ...(_NSIG)].descr = "UNKNOWN",
[SIGTRAP].important = false,
[SIGTRAP].descr = "SIGTRAP",
[SIGILL].important = true,
[SIGILL].descr = "SIGILL",
[SIGFPE].important = true,
[SIGFPE].descr = "SIGFPE",
[SIGSEGV].important = true,
[SIGSEGV].descr = "SIGSEGV",
[SIGBUS].important = true,
[SIGBUS].descr = "SIGBUS",
/* Is affected from monitorSIGABRT flag */
[SIGABRT].important = false,
[SIGABRT].descr = "SIGABRT",
/* Is affected from tmoutVTALRM flag */
[SIGVTALRM].important = false,
[SIGVTALRM].descr = "SIGVTALRM-TMOUT",
/* seccomp-bpf kill */
[SIGSYS].important = true,
[SIGSYS].descr = "SIGSYS",
};
#ifndef SI_FROMUSER
#define SI_FROMUSER(siptr) ((siptr)->si_code <= 0)
#endif /* SI_FROMUSER */
extern const char* sys_sigabbrev[];
static __thread char arch_signame[32];
static const char* arch_sigName(int signo) {
if (signo < 0 || signo > _NSIG) {
snprintf(arch_signame, sizeof(arch_signame), "UNKNOWN-%d", signo);
return arch_signame;
}
if (signo > __SIGRTMIN) {
snprintf(arch_signame, sizeof(arch_signame), "SIG%d-RTMIN+%d", signo, signo - __SIGRTMIN);
return arch_signame;
}
#ifdef __ANDROID__
return arch_sigs[signo].descr;
#else
if (sys_sigabbrev[signo] == NULL) {
snprintf(arch_signame, sizeof(arch_signame), "SIG%d", signo);
} else {
snprintf(arch_signame, sizeof(arch_signame), "SIG%s", sys_sigabbrev[signo]);
}
return arch_signame;
#endif /* __ANDROID__ */
}
static size_t arch_getProcMem(pid_t pid, uint8_t* buf, size_t len, REG_TYPE pc) {
/*
* Let's try process_vm_readv first
*/
const struct iovec local_iov = {
.iov_base = buf,
.iov_len = len,
};
const struct iovec remote_iov = {
.iov_base = (void*)(uintptr_t)pc,
.iov_len = len,
};
if (process_vm_readv(pid, &local_iov, 1, &remote_iov, 1, 0) == (ssize_t)len) {
return len;
}
// Debug if failed since it shouldn't happen very often
PLOG_D("process_vm_readv() failed");
/*
* Ok, let's do it via ptrace() then.
* len must be aligned to the sizeof(long)
*/
int cnt = len / sizeof(long);
size_t memsz = 0;
for (int x = 0; x < cnt; x++) {
uint8_t* addr = (uint8_t*)(uintptr_t)pc + (int)(x * sizeof(long));
long ret = ptrace(PTRACE_PEEKDATA, pid, addr, NULL);
if (errno != 0) {
PLOG_W("Couldn't PT_READ_D on pid %d, addr: %p", pid, addr);
break;
}
memsz += sizeof(long);
memcpy(&buf[x * sizeof(long)], &ret, sizeof(long));
}
return memsz;
}
static size_t arch_getPC(pid_t pid, REG_TYPE* pc, REG_TYPE* status_reg HF_ATTR_UNUSED) {
/*
* Some old ARM android kernels are failing with PTRACE_GETREGS to extract
* the correct register values if struct size is bigger than expected. As such the
* 32/64-bit multiplexing trick is not working for them in case PTRACE_GETREGSET
* fails or is not implemented. To cover such cases we explicitly define
* the struct size to 32bit version for arm CPU.
*/
#if defined(__arm__)
struct user_regs_struct_32 regs;
#else
HEADERS_STRUCT regs;
#endif
const struct iovec pt_iov = {
.iov_base = &regs,
.iov_len = sizeof(regs),
};
if (ptrace(PTRACE_GETREGSET, pid, NT_PRSTATUS, &pt_iov) == -1L) {
PLOG_D("ptrace(PTRACE_GETREGSET) failed");
// If PTRACE_GETREGSET fails, try PTRACE_GETREGS if available
#if PTRACE_GETREGS_AVAILABLE
if (ptrace(PTRACE_GETREGS, pid, 0, &regs)) {
PLOG_D("ptrace(PTRACE_GETREGS) failed");
LOG_W("ptrace PTRACE_GETREGSET & PTRACE_GETREGS failed to extract target registers");
return 0;
}
#else
return 0;
#endif
}
#if defined(__i386__) || defined(__x86_64__)
/*
* 32-bit
*/
if (pt_iov.iov_len == sizeof(struct user_regs_struct_32)) {
struct user_regs_struct_32* r32 = (struct user_regs_struct_32*)&regs;
*pc = r32->eip;
*status_reg = r32->eflags;
return pt_iov.iov_len;
}
/*
* 64-bit
*/
if (pt_iov.iov_len == sizeof(struct user_regs_struct_64)) {
struct user_regs_struct_64* r64 = (struct user_regs_struct_64*)&regs;
*pc = r64->ip;
*status_reg = r64->flags;
return pt_iov.iov_len;
}
LOG_W("Unknown registers structure size: '%zd'", pt_iov.iov_len);
return 0;
#endif /* defined(__i386__) || defined(__x86_64__) */
#if defined(__arm__) || defined(__aarch64__)
/*
* 32-bit
*/
if (pt_iov.iov_len == sizeof(struct user_regs_struct_32)) {
struct user_regs_struct_32* r32 = (struct user_regs_struct_32*)&regs;
#ifdef __ANDROID__
*pc = r32->ARM_pc;
*status_reg = r32->ARM_cpsr;
#else
*pc = r32->uregs[ARM_pc];
*status_reg = r32->uregs[ARM_cpsr];
#endif
return pt_iov.iov_len;
}
/*
* 64-bit
*/
if (pt_iov.iov_len == sizeof(struct user_regs_struct_64)) {
struct user_regs_struct_64* r64 = (struct user_regs_struct_64*)&regs;
*pc = r64->pc;
*status_reg = r64->pstate;
return pt_iov.iov_len;
}
LOG_W("Unknown registers structure size: '%zd'", pt_iov.iov_len);
return 0;
#endif /* defined(__arm__) || defined(__aarch64__) */
#if defined(__powerpc64__) || defined(__powerpc__)
/*
* 32-bit
*/
if (pt_iov.iov_len == sizeof(struct user_regs_struct_32)) {
struct user_regs_struct_32* r32 = (struct user_regs_struct_32*)&regs;
*pc = r32->nip;
return pt_iov.iov_len;
}
/*
* 64-bit
*/
if (pt_iov.iov_len == sizeof(struct user_regs_struct_64)) {
struct user_regs_struct_64* r64 = (struct user_regs_struct_64*)&regs;
*pc = r64->nip;
return pt_iov.iov_len;
}
LOG_W("Unknown registers structure size: '%zd'", pt_iov.iov_len);
return 0;
#endif /* defined(__powerpc64__) || defined(__powerpc__) */
#if defined(__mips__) || defined(__mips64__)
*pc = regs.cp0_epc;
return pt_iov.iov_len;
#endif /* defined(__mips__) || defined(__mips64__) */
LOG_D("Unknown/unsupported CPU architecture");
return 0;
}
static void arch_getInstrStr(pid_t pid, REG_TYPE* pc, char* instr) {
/*
* We need a value aligned to 8
* which is sizeof(long) on 64bit CPU archs (on most of them, I hope;)
*/
uint8_t buf[MAX_INSTR_SZ];
size_t memsz;
REG_TYPE status_reg = 0;
snprintf(instr, _HF_INSTR_SZ, "%s", "[UNKNOWN]");
size_t pcRegSz = arch_getPC(pid, pc, &status_reg);
if (!pcRegSz) {
LOG_W("Current architecture not supported for disassembly");
return;
}
if ((memsz = arch_getProcMem(pid, buf, sizeof(buf), *pc)) == 0) {
snprintf(instr, _HF_INSTR_SZ, "%s", "[NOT_MMAPED]");
return;
}
#if !defined(__ANDROID__)
arch_bfdDisasm(pid, buf, memsz, instr);
#else
cs_arch arch;
cs_mode mode;
#if defined(__arm__) || defined(__aarch64__)
arch = (pcRegSz == sizeof(struct user_regs_struct_64)) ? CS_ARCH_ARM64 : CS_ARCH_ARM;
if (arch == CS_ARCH_ARM) {
mode = (status_reg & 0x20) ? CS_MODE_THUMB : CS_MODE_ARM;
} else {
mode = CS_MODE_ARM;
}
#elif defined(__i386__) || defined(__x86_64__)
arch = CS_ARCH_X86;
mode = (pcRegSz == sizeof(struct user_regs_struct_64)) ? CS_MODE_64 : CS_MODE_32;
#else
LOG_E("Unknown/Unsupported Android CPU architecture");
#endif
csh handle;
cs_err err = cs_open(arch, mode, &handle);
if (err != CS_ERR_OK) {
LOG_W("Capstone initialization failed: '%s'", cs_strerror(err));
return;
}
cs_insn* insn;
size_t count = cs_disasm(handle, buf, sizeof(buf), *pc, 0, &insn);
if (count < 1) {
LOG_W("Couldn't disassemble the assembler instructions' stream: '%s'",
cs_strerror(cs_errno(handle)));
cs_close(&handle);
return;
}
snprintf(instr, _HF_INSTR_SZ, "%s %s", insn[0].mnemonic, insn[0].op_str);
cs_free(insn, count);
cs_close(&handle);
#endif /* defined(__ANDROID__) */
for (int x = 0; instr[x] && x < _HF_INSTR_SZ; x++) {
if (instr[x] == '/' || instr[x] == '\\' || isspace(instr[x]) || !isprint(instr[x])) {
instr[x] = '_';
}
}
return;
}
static void arch_hashCallstack(run_t* run, funcs_t* funcs, size_t funcCnt, bool enableMasking) {
uint64_t hash = 0;
for (size_t i = 0; i < funcCnt && i < run->global->linux.numMajorFrames; i++) {
/*
* Convert PC to char array to be compatible with hash function
*/
char pcStr[REGSIZEINCHAR] = {0};
snprintf(pcStr, REGSIZEINCHAR, REG_PD REG_PM, (REG_TYPE)(long)funcs[i].pc);
/*
* Hash the last three nibbles
*/
hash ^= util_hash(&pcStr[strlen(pcStr) - 3], 3);
}
/*
* If only one frame, hash is not safe to be used for uniqueness. We mask it
* here with a constant prefix, so analyzers can pick it up and create filenames
* accordingly. 'enableMasking' is controlling masking for cases where it should
* not be enabled (e.g. fuzzer worker is from verifier).
*/
if (enableMasking && funcCnt == 1) {
hash |= _HF_SINGLE_FRAME_MASK;
}
run->backtrace = hash;
}
static void arch_traceGenerateReport(
pid_t pid, run_t* run, funcs_t* funcs, size_t funcCnt, siginfo_t* si, const char* instr) {
run->report[0] = '\0';
util_ssnprintf(run->report, sizeof(run->report), "ORIG_FNAME: %s\n", run->origFileName);
util_ssnprintf(run->report, sizeof(run->report), "FUZZ_FNAME: %s\n", run->crashFileName);
util_ssnprintf(run->report, sizeof(run->report), "PID: %d\n", pid);
util_ssnprintf(run->report, sizeof(run->report), "SIGNAL: %s (%d)\n",
arch_sigName(si->si_signo), si->si_signo);
util_ssnprintf(run->report, sizeof(run->report), "FAULT ADDRESS: %p\n",
SI_FROMUSER(si) ? NULL : si->si_addr);
util_ssnprintf(run->report, sizeof(run->report), "INSTRUCTION: %s\n", instr);
util_ssnprintf(run->report, sizeof(run->report), "STACK HASH: %016llx\n", run->backtrace);
util_ssnprintf(run->report, sizeof(run->report), "STACK:\n");
for (size_t i = 0; i < funcCnt; i++) {
#ifdef __HF_USE_CAPSTONE__
util_ssnprintf(
run->report, sizeof(run->report), " <" REG_PD REG_PM "> ", (REG_TYPE)(long)funcs[i].pc);
if (funcs[i].func[0] != '\0')
util_ssnprintf(run->report, sizeof(run->report), "[%s() + 0x%x at %s]\n", funcs[i].func,
funcs[i].line, funcs[i].mapName);
else
util_ssnprintf(run->report, sizeof(run->report), "[]\n");
#else
util_ssnprintf(run->report, sizeof(run->report), " <" REG_PD REG_PM "> [%s():%u at %s]\n",
(REG_TYPE)(long)funcs[i].pc, funcs[i].func, funcs[i].line, funcs[i].mapName);
#endif
}
// libunwind is not working for 32bit targets in 64bit systems
#if defined(__aarch64__)
if (funcCnt == 0) {
util_ssnprintf(run->report, sizeof(run->report),
" !ERROR: If 32bit fuzz target"
" in aarch64 system, try ARM 32bit build\n");
}
#endif
return;
}
static void arch_traceAnalyzeData(run_t* run, pid_t pid) {
REG_TYPE pc = 0, status_reg = 0;
size_t pcRegSz = arch_getPC(pid, &pc, &status_reg);
if (!pcRegSz) {
LOG_W("ptrace arch_getPC failed");
return;
}
/*
* Unwind and resolve symbols
*/
funcs_t* funcs = util_Malloc(_HF_MAX_FUNCS * sizeof(funcs_t));
defer {
free(funcs);
};
memset(funcs, 0, _HF_MAX_FUNCS * sizeof(funcs_t));
#if !defined(__ANDROID__)
size_t funcCnt = arch_unwindStack(pid, funcs);
arch_bfdResolveSyms(pid, funcs, funcCnt);
#else
size_t funcCnt = arch_unwindStack(pid, funcs);
#endif
/*
* If unwinder failed (zero frames), use PC from ptrace GETREGS if not zero.
* If PC reg zero return and callers should handle zero hash case.
*/
if (funcCnt == 0) {
if (pc) {
/* Manually update major frame PC & frames counter */
funcs[0].pc = (void*)(uintptr_t)pc;
funcCnt = 1;
} else {
return;
}
}
/*
* Calculate backtrace callstack hash signature
*/
arch_hashCallstack(run, funcs, funcCnt, false);
}
static void arch_traceSaveData(run_t* run, pid_t pid) {
REG_TYPE pc = 0;
/* Local copy since flag is overridden for some crashes */
bool saveUnique = run->global->io.saveUnique;
char instr[_HF_INSTR_SZ] = "\x00";
siginfo_t si;
bzero(&si, sizeof(si));
if (ptrace(PTRACE_GETSIGINFO, pid, 0, &si) == -1) {
PLOG_W("Couldn't get siginfo for pid %d", pid);
}
arch_getInstrStr(pid, &pc, instr);
LOG_D("Pid: %d, signo: %d, errno: %d, code: %d, addr: %p, pc: %" REG_PM ", instr: '%s'", pid,
si.si_signo, si.si_errno, si.si_code, si.si_addr, pc, instr);
if (!SI_FROMUSER(&si) && pc && si.si_addr < run->global->linux.ignoreAddr) {
LOG_I("Input is interesting (%s), but the si.si_addr is %p (below %p), skipping",
arch_sigName(si.si_signo), si.si_addr, run->global->linux.ignoreAddr);
return;
}
/*
* Unwind and resolve symbols
*/
funcs_t* funcs = util_Malloc(_HF_MAX_FUNCS * sizeof(funcs_t));
defer {
free(funcs);
};
memset(funcs, 0, _HF_MAX_FUNCS * sizeof(funcs_t));
#if !defined(__ANDROID__)
size_t funcCnt = arch_unwindStack(pid, funcs);
arch_bfdResolveSyms(pid, funcs, funcCnt);
#else
size_t funcCnt = arch_unwindStack(pid, funcs);
#endif
/*
* If unwinder failed (zero frames), use PC from ptrace GETREGS if not zero.
* If PC reg zero, temporarily disable uniqueness flag since callstack
* hash will be also zero, thus not safe for unique decisions.
*/
if (funcCnt == 0) {
if (pc) {
/* Manually update major frame PC & frames counter */
funcs[0].pc = (void*)(uintptr_t)pc;
funcCnt = 1;
} else {
saveUnique = false;
}
}
/*
* Temp local copy of previous backtrace value in case worker hit crashes into multiple
* tids for same target master thread. Will be 0 for first crash against target.
*/
uint64_t oldBacktrace = run->backtrace;
/*
* Calculate backtrace callstack hash signature
*/
arch_hashCallstack(run, funcs, funcCnt, saveUnique);
/*
* If fuzzing with sanitizer coverage feedback increase crashes counter used
* as metric for dynFile evolution
*/
if (run->global->useSanCov) {
run->sanCovCnts.crashesCnt++;
}
/*
* If unique flag is set and single frame crash, disable uniqueness for this crash
* to always save (timestamp will be added to the filename)
*/
if (saveUnique && (funcCnt == 1)) {
saveUnique = false;
}
/*
* If worker crashFileName member is set, it means that a tid has already crashed
* from target master thread.
*/
if (run->crashFileName[0] != '\0') {
LOG_D("Multiple crashes detected from worker against attached tids group");
/*
* If stackhashes match, don't re-analyze. This will avoid duplicates
* and prevent verifier from running multiple passes. Depth of check is
* always 1 (last backtrace saved only per target iteration).
*/
if (oldBacktrace == run->backtrace) {
return;
}
}
/* Increase global crashes counter */
ATOMIC_POST_INC(run->global->cnts.crashesCnt);
/*
* Check if backtrace contains whitelisted symbol. Whitelist overrides
* both stackhash and symbol blacklist. Crash is always kept regardless
* of the status of uniqueness flag.
*/
if (run->global->linux.symsWl) {
char* wlSymbol = arch_btContainsSymbol(
run->global->linux.symsWlCnt, run->global->linux.symsWl, funcCnt, funcs);
if (wlSymbol != NULL) {
saveUnique = false;
LOG_D("Whitelisted symbol '%s' found, skipping blacklist checks", wlSymbol);
}
} else {
/*
* Check if stackhash is blacklisted
*/
if (run->global->blacklist && (fastArray64Search(run->global->blacklist,
run->global->blacklistCnt, run->backtrace) != -1)) {
LOG_I("Blacklisted stack hash '%" PRIx64 "', skipping", run->backtrace);
ATOMIC_POST_INC(run->global->cnts.blCrashesCnt);
return;
}
/*
* Check if backtrace contains blacklisted symbol
*/
char* blSymbol = arch_btContainsSymbol(
run->global->linux.symsBlCnt, run->global->linux.symsBl, funcCnt, funcs);
if (blSymbol != NULL) {
LOG_I("Blacklisted symbol '%s' found, skipping", blSymbol);
ATOMIC_POST_INC(run->global->cnts.blCrashesCnt);
return;
}
}
/* If non-blacklisted crash detected, zero set two MSB */
ATOMIC_POST_ADD(run->global->dynFileIterExpire, _HF_DYNFILE_SUB_MASK);
void* sig_addr = si.si_addr;
if (run->global->linux.disableRandomization == false) {
pc = 0UL;
sig_addr = NULL;
}
/* User-induced signals don't set si.si_addr */
if (SI_FROMUSER(&si)) {
sig_addr = NULL;
}
/* If dry run mode, copy file with same name into workspace */
if (run->global->mutationsPerRun == 0U && run->global->useVerifier) {
snprintf(run->crashFileName, sizeof(run->crashFileName), "%s/%s", run->global->io.crashDir,
run->origFileName);
} else if (saveUnique) {
snprintf(run->crashFileName, sizeof(run->crashFileName),
"%s/%s.PC.%" REG_PM ".STACK.%" PRIx64 ".CODE.%d.ADDR.%p.INSTR.%s.%s",
run->global->io.crashDir, arch_sigName(si.si_signo), pc, run->backtrace, si.si_code,
sig_addr, instr, run->global->io.fileExtn);
} else {
char localtmstr[PATH_MAX];
util_getLocalTime("%F.%H:%M:%S", localtmstr, sizeof(localtmstr), time(NULL));
snprintf(run->crashFileName, sizeof(run->crashFileName),
"%s/%s.PC.%" REG_PM ".STACK.%" PRIx64 ".CODE.%d.ADDR.%p.INSTR.%s.%s.%d.%s",
run->global->io.crashDir, arch_sigName(si.si_signo), pc, run->backtrace, si.si_code,
sig_addr, instr, localtmstr, pid, run->global->io.fileExtn);
}
/* Target crashed (no duplicate detection yet) */
if (run->global->socketFuzzer) {
LOG_D("SocketFuzzer: trace: Crash Identified");
run->hasCrashed = true;
}
if (files_exists(run->crashFileName)) {
LOG_I("Crash (dup): '%s' already exists, skipping", run->crashFileName);
// Clear filename so that verifier can understand we hit a duplicate
memset(run->crashFileName, 0, sizeof(run->crashFileName));
return;
}
if (files_writeBufToFile(run->crashFileName, run->dynamicFile, run->dynamicFileSz,
O_CREAT | O_EXCL | O_WRONLY | O_CLOEXEC) == false) {
LOG_E("Couldn't write to '%s'", run->crashFileName);
return;
}
/* Unique new crash, notify fuzzer */
if (run->global->socketFuzzer) {
LOG_D("SocketFuzzer: trace: New Uniqu Crash");
fuzz_notifySocketFuzzerCrash(run);
}
LOG_I("Crash: saved as '%s'", run->crashFileName);
ATOMIC_POST_INC(run->global->cnts.uniqueCrashesCnt);
/* If unique crash found, reset dynFile counter */
ATOMIC_CLEAR(run->global->dynFileIterExpire);
arch_traceGenerateReport(pid, run, funcs, funcCnt, &si, instr);
}
/* TODO: Add report parsing support for other sanitizers too */
static int arch_parseAsanReport(
run_t* run, pid_t pid, funcs_t* funcs, void** crashAddr, char** op) {
char crashReport[PATH_MAX] = {0};
const char* const crashReportCpy = crashReport;
snprintf(
crashReport, sizeof(crashReport), "%s/%s.%d", run->global->io.workDir, kLOGPREFIX, pid);
FILE* fReport = fopen(crashReport, "rb");
if (fReport == NULL) {
PLOG_D("Couldn't open '%s' - R/O mode", crashReport);
return -1;
}
defer {
fclose(fReport);
};
defer {
unlink(crashReportCpy);
};
char header[35] = {0};
snprintf(header, sizeof(header), "==%d==ERROR: AddressSanitizer:", pid);
size_t headerSz = strlen(header);
bool headerFound = false;
uint8_t frameIdx = 0;
char framePrefix[5] = {0};
snprintf(framePrefix, sizeof(framePrefix), "#%" PRIu8, frameIdx);
char *lineptr = NULL, *cAddr = NULL;
size_t n = 0;
defer {
free(lineptr);
};
for (;;) {
if (getline(&lineptr, &n, fReport) == -1) {
break;
}
/* First step is to identify header */
if (headerFound == false) {
if ((strlen(lineptr) > headerSz) && (strncmp(header, lineptr, headerSz) == 0)) {
headerFound = true;
/* Parse crash address */
cAddr = strstr(lineptr, "address 0x");
if (cAddr) {
cAddr = cAddr + strlen("address ");
char* endOff = strchr(cAddr, ' ');
cAddr[endOff - cAddr] = '\0';
*crashAddr = (void*)((size_t)strtoull(cAddr, NULL, 16));
} else {
*crashAddr = 0x0;
}
}
continue;
} else {
char* pLineLC = lineptr;
/* Trim leading spaces */
while (*pLineLC != '\0' && isspace(*pLineLC)) {
++pLineLC;
}
/* End separator for crash thread stack trace is an empty line */
if ((*pLineLC == '\0') && (frameIdx != 0)) {
break;
}
/* Basic length checks */
if (strlen(pLineLC) < 10) {
continue;
}
/* If available parse the type of error (READ/WRITE) */
if (cAddr && strstr(pLineLC, cAddr)) {
if (strncmp(pLineLC, "READ", 4) == 0) {
*op = "READ";
} else if (strncmp(pLineLC, "WRITE", 5) == 0) {
*op = "WRITE";
}
cAddr = NULL;
}
/* Check for crash thread frames */
if (strncmp(pLineLC, framePrefix, strlen(framePrefix)) == 0) {
/* Abort if max depth */
if (frameIdx >= _HF_MAX_FUNCS) {
break;
}
/*
* Frames have following format:
#0 0xaa860177 (/system/lib/libc.so+0x196177)
*/
char* savePtr = NULL;
strtok_r(pLineLC, " ", &savePtr);
funcs[frameIdx].pc =
(void*)((size_t)strtoull(strtok_r(NULL, " ", &savePtr), NULL, 16));
/* DSO & code offset parsing */
char* targetStr = strtok_r(NULL, " ", &savePtr);
char* startOff = strchr(targetStr, '(') + 1;
char* plusOff = strchr(targetStr, '+');
char* endOff = strrchr(targetStr, ')');
targetStr[endOff - startOff] = '\0';
if ((startOff == NULL) || (endOff == NULL) || (plusOff == NULL)) {
LOG_D("Invalid ASan report entry (%s)", lineptr);
} else {
size_t dsoSz =
MIN(sizeof(funcs[frameIdx].mapName), (size_t)(plusOff - startOff));
memcpy(funcs[frameIdx].mapName, startOff, dsoSz);
char* codeOff = targetStr + (plusOff - startOff) + 1;
funcs[frameIdx].line = strtoull(codeOff, NULL, 16);
}
frameIdx++;
snprintf(framePrefix, sizeof(framePrefix), "#%" PRIu8, frameIdx);
}
}
}
return frameIdx;
}
/*
* Special book keeping for cases where crashes are detected based on exitcode and not
* a raised signal. Such case is the ASan fuzzing for Android. Crash file name maintains
* the same format for compatibility with post campaign tools.
*/
static void arch_traceExitSaveData(run_t* run, pid_t pid) {
REG_TYPE pc = 0;
void* crashAddr = 0;
char* op = "UNKNOWN";
pid_t targetPid = (run->global->linux.pid > 0) ? run->global->linux.pid : run->pid;
/* Save only the first hit for each worker */
if (run->crashFileName[0] != '\0') {
return;
}
/* Increase global crashes counter */
ATOMIC_POST_INC(run->global->cnts.crashesCnt);
ATOMIC_POST_AND(run->global->dynFileIterExpire, _HF_DYNFILE_SUB_MASK);
/*
* If fuzzing with sanitizer coverage feedback increase crashes counter used
* as metric for dynFile evolution
*/
if (run->global->useSanCov) {
run->sanCovCnts.crashesCnt++;
}
/* If sanitizer produces reports with stack traces (e.g. ASan), they're parsed manually */
int funcCnt = 0;
funcs_t* funcs = util_Malloc(_HF_MAX_FUNCS * sizeof(funcs_t));
defer {
free(funcs);
};
memset(funcs, 0, _HF_MAX_FUNCS * sizeof(funcs_t));
/* Sanitizers save reports against parent PID */
if (targetPid != pid) {
return;
}
funcCnt = arch_parseAsanReport(run, pid, funcs, &crashAddr, &op);
/*
* -1 error indicates a file not found for report. This is expected to happen often since
* ASan report is generated once for crashing TID. Ptrace arch is not guaranteed to parse
* that TID first. Not setting the 'crashFileName' variable will ensure that this branch
* is executed again for all TIDs until the matching report is found
*/
if (funcCnt == -1) {
return;
}
/* Since crash address is available, apply ignoreAddr filters */
if (crashAddr < run->global->linux.ignoreAddr) {
LOG_I("Input is interesting, but the crash addr is %p (below %p), skipping", crashAddr,
run->global->linux.ignoreAddr);
return;
}
/* If frames successfully recovered, calculate stack hash & populate crash PC */
arch_hashCallstack(run, funcs, funcCnt, false);
pc = (uintptr_t)funcs[0].pc;
/*
* Check if stackhash is blacklisted
*/
if (run->global->blacklist && (fastArray64Search(run->global->blacklist,
run->global->blacklistCnt, run->backtrace) != -1)) {
LOG_I("Blacklisted stack hash '%" PRIx64 "', skipping", run->backtrace);
ATOMIC_POST_INC(run->global->cnts.blCrashesCnt);
return;
}
/* If dry run mode, copy file with same name into workspace */
if (run->global->mutationsPerRun == 0U && run->global->useVerifier) {
snprintf(run->crashFileName, sizeof(run->crashFileName), "%s/%s", run->global->io.crashDir,
run->origFileName);
} else {
/* Keep the crashes file name format identical */
if (run->backtrace != 0ULL && run->global->io.saveUnique) {
snprintf(run->crashFileName, sizeof(run->crashFileName),
"%s/%s.PC.%" REG_PM ".STACK.%" PRIx64 ".CODE.%s.ADDR.%p.INSTR.%s.%s",
run->global->io.crashDir, "SAN", pc, run->backtrace, op, crashAddr, "[UNKNOWN]",
run->global->io.fileExtn);
} else {
/* If no stack hash available, all crashes treated as unique */
char localtmstr[PATH_MAX];
util_getLocalTime("%F.%H:%M:%S", localtmstr, sizeof(localtmstr), time(NULL));
snprintf(run->crashFileName, sizeof(run->crashFileName),
"%s/%s.PC.%" REG_PM ".STACK.%" PRIx64 ".CODE.%s.ADDR.%p.INSTR.%s.%s.%s",
run->global->io.crashDir, "SAN", pc, run->backtrace, op, crashAddr, "[UNKNOWN]",
localtmstr, run->global->io.fileExtn);
}
}
int fd = open(run->crashFileName, O_WRONLY | O_EXCL | O_CREAT, 0600);
if (fd == -1 && errno == EEXIST) {
LOG_I("It seems that '%s' already exists, skipping", run->crashFileName);
return;
} else if (fd == -1) {
PLOG_E("Cannot create output file '%s'", run->crashFileName);
return;
} else {
defer {
close(fd);
};
if (files_writeToFd(fd, run->dynamicFile, run->dynamicFileSz)) {
LOG_I("Ok, that's interesting, saved new crash as '%s'", run->crashFileName);
/* Clear stack hash so that verifier can understand we hit a duplicate */
run->backtrace = 0ULL;
/* Increase unique crashes counters */
ATOMIC_POST_INC(run->global->cnts.uniqueCrashesCnt);
ATOMIC_CLEAR(run->global->dynFileIterExpire);
} else {
LOG_E("Couldn't save crash to '%s'", run->crashFileName);
/* In case of write error, clear crashFileName to so that other monitored TIDs can retry
*/
memset(run->crashFileName, 0, sizeof(run->crashFileName));
return;
}
}
/* Generate report */
run->report[0] = '\0';
util_ssnprintf(run->report, sizeof(run->report), "EXIT_CODE: %s\n", HF_SAN_EXIT_CODE);
util_ssnprintf(run->report, sizeof(run->report), "ORIG_FNAME: %s\n", run->origFileName);
util_ssnprintf(run->report, sizeof(run->report), "FUZZ_FNAME: %s\n", run->crashFileName);
util_ssnprintf(run->report, sizeof(run->report), "PID: %d\n", pid);
util_ssnprintf(run->report, sizeof(run->report), "OPERATION: %s\n", op);
util_ssnprintf(run->report, sizeof(run->report), "FAULT ADDRESS: %p\n", crashAddr);
if (funcCnt > 0) {
util_ssnprintf(run->report, sizeof(run->report), "STACK HASH: %016llx\n", run->backtrace);
util_ssnprintf(run->report, sizeof(run->report), "STACK:\n");
for (int i = 0; i < funcCnt; i++) {
util_ssnprintf(run->report, sizeof(run->report), " <" REG_PD REG_PM "> ",
(REG_TYPE)(long)funcs[i].pc);
if (funcs[i].mapName[0] != '\0') {
util_ssnprintf(run->report, sizeof(run->report), "[%s + 0x%x]\n", funcs[i].mapName,
funcs[i].line);
} else {
util_ssnprintf(run->report, sizeof(run->report), "[]\n");
}
}
}
}
static void arch_traceExitAnalyzeData(run_t* run, pid_t pid) {
void* crashAddr = 0;
char* op = "UNKNOWN";
int funcCnt = 0;
funcs_t* funcs = util_Malloc(_HF_MAX_FUNCS * sizeof(funcs_t));
defer {
free(funcs);
};
memset(funcs, 0, _HF_MAX_FUNCS * sizeof(funcs_t));
funcCnt = arch_parseAsanReport(run, pid, funcs, &crashAddr, &op);
/*
* -1 error indicates a file not found for report. This is expected to happen often since
* ASan report is generated once for crashing TID. Ptrace arch is not guaranteed to parse
* that TID first. Not setting the 'crashFileName' variable will ensure that this branch
* is executed again for all TIDs until the matching report is found
*/
if (funcCnt == -1) {
return;
}
/* If frames successfully recovered, calculate stack hash & populate crash PC */
arch_hashCallstack(run, funcs, funcCnt, false);
}
void arch_traceExitAnalyze(run_t* run, pid_t pid) {
if (run->mainWorker) {
/* Main fuzzing threads */
arch_traceExitSaveData(run, pid);
} else {
/* Post crash analysis (e.g. crashes verifier) */
arch_traceExitAnalyzeData(run, pid);
}
}
#define __WEVENT(status) ((status & 0xFF0000) >> 16)
static void arch_traceEvent(run_t* run, int status, pid_t pid) {
LOG_D("PID: %d, Ptrace event: %d", pid, __WEVENT(status));
switch (__WEVENT(status)) {
case PTRACE_EVENT_EXIT: {
unsigned long event_msg;
if (ptrace(PTRACE_GETEVENTMSG, pid, NULL, &event_msg) == -1) {
PLOG_E("ptrace(PTRACE_GETEVENTMSG,%d) failed", pid);
return;
}
if (WIFEXITED(event_msg)) {
LOG_D("PID: %d exited with exit_code: %lu", pid,
(unsigned long)WEXITSTATUS(event_msg));
if (WEXITSTATUS(event_msg) == (unsigned long)HF_SAN_EXIT_CODE) {
arch_traceExitAnalyze(run, pid);
}
} else if (WIFSIGNALED(event_msg)) {
LOG_D(
"PID: %d terminated with signal: %lu", pid, (unsigned long)WTERMSIG(event_msg));
} else {
LOG_D("PID: %d exited with unknown status: %lu", pid, event_msg);
}
} break;
default:
break;
}
ptrace(PTRACE_CONT, pid, 0, 0);
}
void arch_traceAnalyze(run_t* run, int status, pid_t pid) {
/*
* It's a ptrace event, deal with it elsewhere
*/
if (WIFSTOPPED(status) && __WEVENT(status)) {
return arch_traceEvent(run, status, pid);
}
if (WIFSTOPPED(status)) {
/*
* If it's an interesting signal, save the testcase
*/
if (arch_sigs[WSTOPSIG(status)].important) {
/*
* If fuzzer worker is from core fuzzing process run full
* analysis. Otherwise just unwind and get stack hash signature.
*/
if (run->mainWorker) {
arch_traceSaveData(run, pid);
} else {
arch_traceAnalyzeData(run, pid);
}
}
/* Do not deliver SIGSTOP, as we don't support PTRACE_LISTEN anyway */
int sig = (WSTOPSIG(status) != SIGSTOP) ? WSTOPSIG(status) : 0;
ptrace(PTRACE_CONT, pid, 0, sig);
return;
}
/*
* Resumed by delivery of SIGCONT
*/
if (WIFCONTINUED(status)) {
return;
}
/*
* Process exited
*/
if (WIFEXITED(status)) {
/*
* Target exited with sanitizer defined exitcode (used when SIGABRT is not monitored)
*/
if (WEXITSTATUS(status) == (unsigned long)HF_SAN_EXIT_CODE) {
arch_traceExitAnalyze(run, pid);
}
return;
}
if (WIFSIGNALED(status)) {
return;
}
abort(); /* NOTREACHED */
}
static bool arch_listThreads(int tasks[], size_t thrSz, int pid) {
char path[512];
snprintf(path, sizeof(path), "/proc/%d/task", pid);
/* An optimization, the number of threads is st.st_nlink - 2 (. and ..) */
struct stat st;
if (stat(path, &st) != -1) {
if (st.st_nlink == 3) {
tasks[0] = pid;
tasks[1] = 0;
return true;
}
}
size_t count = 0;
DIR* dir = opendir(path);
if (!dir) {
PLOG_E("Couldn't open dir '%s'", path);
return false;
}
defer {
closedir(dir);
};
for (;;) {
errno = 0;
const struct dirent* res = readdir(dir);
if (res == NULL && errno != 0) {
PLOG_E("Couldn't read contents of '%s'", path);
return false;
}
if (res == NULL) {
break;
}
pid_t pid = (pid_t)strtol(res->d_name, (char**)NULL, 10);
if (pid == 0) {
LOG_D("The following dir entry couldn't be converted to pid_t '%s'", res->d_name);
continue;
}
tasks[count++] = pid;
LOG_D("Added pid '%d' from '%s/%s'", pid, path, res->d_name);
if (count >= thrSz) {
break;
}
}
PLOG_D("Total number of threads in pid '%d': '%zd'", pid, count);
tasks[count + 1] = 0;
if (count < 1) {
return false;
}
return true;
}
bool arch_traceWaitForPidStop(pid_t pid) {
for (;;) {
int status;
pid_t ret = wait4(pid, &status, __WALL | WUNTRACED, NULL);
if (ret == -1 && errno == EINTR) {
continue;
}
if (ret == -1) {
PLOG_W("wait4(pid=%d) failed", pid);
return false;
}
if (!WIFSTOPPED(status)) {
LOG_W("PID %d not in a stopped state - status:%d", pid, status);
return false;
}
return true;
}
}
#define MAX_THREAD_IN_TASK 4096
bool arch_traceAttach(run_t* run, pid_t pid) {
/*
* It should be present since, at least, Linux kernel 3.8, but
* not always defined in kernel-headers
*/
#if !defined(PTRACE_O_EXITKILL)
#define PTRACE_O_EXITKILL (1 << 20)
#endif /* !defined(PTRACE_O_EXITKILL) */
long seize_options = PTRACE_O_TRACECLONE | PTRACE_O_TRACEFORK | PTRACE_O_TRACEVFORK;
if (run->global->linux.pid == 0) {
seize_options |= PTRACE_O_EXITKILL;
}
/* The event is only used with sanitizers */
if (run->global->enableSanitizers) {
seize_options |= PTRACE_O_TRACEEXIT;
}
if (run->global->linux.pid == 0 && !arch_traceWaitForPidStop(pid)) {
return false;
}
if (ptrace(PTRACE_SEIZE, pid, NULL, seize_options) == -1) {
PLOG_W("Couldn't ptrace(PTRACE_SEIZE) to pid: %d", pid);
return false;
}
LOG_D("Attached to PID: %d", pid);
/* It only makes sense to attach to threads with -p */
if (run->global->linux.pid == 0) {
return true;
}
int tasks[MAX_THREAD_IN_TASK + 1] = {0};
if (!arch_listThreads(tasks, MAX_THREAD_IN_TASK, pid)) {
LOG_E("Couldn't read thread list for pid '%d'", pid);
return false;
}
for (int i = 0; i < MAX_THREAD_IN_TASK && tasks[i]; i++) {
if (tasks[i] == pid) {
continue;
}
if (ptrace(PTRACE_SEIZE, tasks[i], NULL, seize_options) == -1) {
PLOG_W("Couldn't ptrace(PTRACE_SEIZE) to pid: %d", tasks[i]);
continue;
}
LOG_D("Attached to PID: %d (thread_group:%d)", tasks[i], pid);
}
return true;
}
void arch_traceDetach(pid_t pid) {
if (syscall(__NR_kill, pid, 0) == -1 && errno == ESRCH) {
LOG_D("PID: %d no longer exists", pid);
return;
}
int tasks[MAX_THREAD_IN_TASK + 1] = {0};
if (!arch_listThreads(tasks, MAX_THREAD_IN_TASK, pid)) {
LOG_E("Couldn't read thread list for pid '%d'", pid);
return;
}
for (int i = 0; i < MAX_THREAD_IN_TASK && tasks[i]; i++) {
ptrace(PTRACE_INTERRUPT, tasks[i], NULL, NULL);
arch_traceWaitForPidStop(tasks[i]);
ptrace(PTRACE_DETACH, tasks[i], NULL, NULL);
}
}
void arch_traceSignalsInit(honggfuzz_t* hfuzz) {
/* Default is true for all platforms except Android */
arch_sigs[SIGABRT].important = hfuzz->monitorSIGABRT;
/* Default is false */
arch_sigs[SIGVTALRM].important = hfuzz->timing.tmoutVTALRM;
}