-
Notifications
You must be signed in to change notification settings - Fork 0
/
sanitizers.c
145 lines (129 loc) · 5.8 KB
/
sanitizers.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
#include "sanitizers.h"
#include <ctype.h>
#include <dirent.h>
#include <inttypes.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <sys/types.h>
#include "cmdline.h"
#include "libhfcommon/common.h"
#include "libhfcommon/files.h"
#include "libhfcommon/log.h"
#include "libhfcommon/util.h"
/*
* All clang sanitizers, except ASan, can be activated for target binaries
* with or without the matching runtime library (libcompiler_rt). If runtime
* libraries are included in target fuzzing environment, we can benefit from the
* various Die() callbacks and abort/exit logic manipulation. However, some
* setups (e.g. Android production ARM/ARM64 devices) enable sanitizers, such as
* UBSan, without the runtime libraries. As such, their default ftrap is activated
* which is for most cases a SIGABRT. For these cases end-user needs to enable
* SIGABRT monitoring flag, otherwise these crashes will be missed.
*
* Normally SIGABRT is not a wanted signal to monitor for Android, since it produces
* lots of useless crashes due to way Android process termination hacks work. As
* a result the sanitizer's 'abort_on_error' flag cannot be utilized since it
* invokes abort() internally. In order to not lose crashes a custom exitcode can
* be registered and monitored. Since exitcode is a global flag, it's assumed
* that target is compiled with only one sanitizer type enabled at a time.
*
* For cases where clang runtime library linking is not an option, SIGABRT should
* be monitored even for noisy targets, such as the Android OS, since no viable
* alternative exists.
*
* There might be cases where ASan instrumented targets crash while generating
* reports for detected errors (inside __asan_report_error() proc). Under such
* scenarios target fails to exit or SIGABRT (AsanDie() proc) as defined in
* ASAN_OPTIONS flags, leaving garbage logs. An attempt is made to parse such
* logs for cases where enough data are written to identify potentially missed
* crashes. If ASan internal error results into a SIGSEGV being raised, it
* will get caught from ptrace API, handling the discovered ASan internal crash.
*/
/* 'log_path' output directory for sanitizer reports */
#define kSANLOGDIR "log_path="
/* Raise SIGABRT on error or continue with exitcode logic */
#define kABORT_ENABLED "abort_on_error=1"
#define kABORT_DISABLED "abort_on_error=0"
/*
* Common sanitizer flags
*
* symbolize: Disable symbolication since it changes logs (which are parsed) format
*/
#define kSAN_COMMON "symbolize=0"
/* --{ ASan }-- */
/*
*Sanitizer specific flags (notice that if enabled 'abort_on_error' has priority
* over exitcode')
*/
#define kASAN_COMMON_OPTS \
"allow_user_segv_handler=1:" \
"handle_segv=0:" \
"allocator_may_return_null=1:" kSAN_COMMON ":exitcode=" HF_XSTR(HF_SAN_EXIT_CODE)
/* Platform specific flags */
#if defined(__ANDROID__)
/*
* start_deactivated: Enable on Android to reduce memory usage (useful when not all
* target's DSOs are compiled with sanitizer enabled
*/
#define kASAN_OPTS kASAN_COMMON_OPTS ":start_deactivated=1"
#else
#define kASAN_OPTS kASAN_COMMON_OPTS
#endif
/* --{ UBSan }-- */
#define kUBSAN_OPTS kSAN_COMMON ":exitcode=" STR(HF_SAN_EXIT_CODE)
/* --{ MSan }-- */
#define kMSAN_OPTS \
kSAN_COMMON ":exit_code=" STR(HF_SAN_EXIT_CODE) ":" \
"wrap_signals=0:print_stats=1"
/* If no sanitzer support was requested, simply make it use abort() on errors */
#define kSAN_REGULAR \
"abort_on_error=1:handle_segv=0:handle_sigbus=0:handle_abort=0:" \
"handle_sigill=0:handle_sigfpe=0:allocator_may_return_null=1:" \
"symbolize=1:detect_leaks=0:disable_coredump=0:" \
"detect_odr_violation=0"
/*
* If the program ends with a signal that ASan does not handle (or can not
* handle at all, like SIGKILL), coverage data will be lost. This is a big
* problem on Android, where SIGKILL is a normal way of evicting applications
* from memory. With 'coverage_direct=1' coverage data is written to a
* memory-mapped file as soon as it collected. Non-Android targets can disable
* coverage direct when more coverage data collection methods are implemented.
*/
#define kSAN_COV_OPTS "coverage=1:coverage_direct=1"
static void sanitizers_AddFlag(honggfuzz_t* hfuzz, const char* env, char* buf, size_t buflen) {
const char* abortFlag = hfuzz->cfg.monitorSIGABRT ? kABORT_ENABLED : kABORT_DISABLED;
if (getenv(env)) {
LOG_W("The '%s' envar is already set. Not overriding it!", env);
return;
}
if (!hfuzz->sanitizer.enable) {
snprintf(buf, buflen, "%s=%s", env, kSAN_REGULAR);
} else {
snprintf(buf, buflen, "%s=%s:%s:%s%s/%s", env, kASAN_OPTS, abortFlag, kSANLOGDIR,
hfuzz->io.workDir, kLOGPREFIX);
}
/*
* It will make ASAN to start background thread to check RSS mem use, which
* will prevent the NetDrvier from using unshare(CLONE_NEWNET), which cannot
* be used in multi-threaded contexts
*/
if (!hfuzz->exe.netDriver && hfuzz->exe.rssLimit) {
util_ssnprintf(buf, buflen, ":soft_rss_limit_mb=%" PRId64, hfuzz->exe.rssLimit);
}
cmdlineAddEnv(hfuzz, buf);
LOG_D("%s", env);
}
bool sanitizers_Init(honggfuzz_t* hfuzz) {
static char asanOpts[4096];
sanitizers_AddFlag(hfuzz, "ASAN_OPTIONS", asanOpts, sizeof(asanOpts));
static char ubsanOpts[4096];
sanitizers_AddFlag(hfuzz, "UBSAN_OPTIONS", ubsanOpts, sizeof(ubsanOpts));
static char msanOpts[4096];
sanitizers_AddFlag(hfuzz, "MSAN_OPTIONS", msanOpts, sizeof(msanOpts));
static char lsanOpts[4096];
sanitizers_AddFlag(hfuzz, "LSAN_OPTIONS", lsanOpts, sizeof(lsanOpts));
return true;
}