boringssl/util/all_tests.go
David Benjamin 17d553d299 Add a CFI tester to CHECK_ABI.
This uses the x86 trap flag and libunwind to test CFI works at each
instruction. For now, it just uses the system one out of pkg-config and
disables unwind tests if unavailable. We'll probably want to stick a
copy into //third_party and perhaps try the LLVM one later.

This tester caught two bugs in P-256 CFI annotations already:
I47b5f9798b3bcee1748e537b21c173d312a14b42 and
I9f576d868850312d6c14d1386f8fbfa85021b347

An earlier design used PTRACE_SINGLESTEP with libunwind's remote
unwinding features. ptrace is a mess around stop signals (see group-stop
discussion in ptrace(2)) and this is 10x faster, so I went with it. The
question of which is more future-proof is complex:

- There are two libunwinds with the same API,
  https://www.nongnu.org/libunwind/ and LLVM's. This currently uses the
  system nongnu.org for convenience. In future, LLVM's should be easier
  to bundle (less complex build) and appears to even support Windows,
  but I haven't tested this.  Moreover, setting the trap flag keeps the
  test single-process, which is less complex on Windows. That suggests
  the trap flag design and switching to LLVM later. However...

- Not all architectures have a trap flag settable by userspace. As far
  as I can tell, ARMv8's PSTATE.SS can only be set from the kernel. If
  we stick with nongnu.org libunwind, we can use PTRACE_SINGLESTEP and
  remote unwinding. Or we implement it for LLVM. Another thought is for
  the ptracer to bounce SIGTRAP back into the process, to share the
  local unwinding code.

- ARMv7 has no trap flag at all and PTRACE_SINGLESTEP fails. Debuggers
  single-step by injecting breakpoints instead. However, ARMv8's trap
  flag seems to work in both AArch32 and AArch64 modes, so we may be
  able to condition it on a 64-bit kernel.

Sadly, neither strategy works with Intel SDE. Adding flags to cpucap
vectors as we do with ARM would help, but it would not emulate CPUs
newer than the host CPU. For now, I've just had SDE tests disable these.

Annoyingly, CMake does not allow object libraries to have dependencies,
so make test_support a proper static library. Rename the target to
test_support_lib to avoid
https://gitlab.kitware.com/cmake/cmake/issues/17785

Update-Note: This adds a new optional test dependency, but it's disabled
by default (define BORINGSSL_HAVE_LIBUNWIND), so consumers do not need
to do anything. We'll probably want to adjust this in the future.

Bug: 181
Change-Id: I817263d7907aff0904a9cee83f8b26747262cc0c
Reviewed-on: https://boringssl-review.googlesource.com/c/33966
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
2019-01-03 22:01:55 +00:00

482 lines
13 KiB
Go

/* Copyright (c) 2015, Google Inc.
*
* Permission to use, copy, modify, and/or distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
* SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
package main
import (
"bufio"
"bytes"
"encoding/json"
"errors"
"flag"
"fmt"
"math/rand"
"os"
"os/exec"
"path"
"runtime"
"strconv"
"strings"
"sync"
"syscall"
"boringssl.googlesource.com/boringssl/util/testresult"
)
// TODO(davidben): Link tests with the malloc shim and port -malloc-test to this runner.
var (
useValgrind = flag.Bool("valgrind", false, "If true, run code under valgrind")
useCallgrind = flag.Bool("callgrind", false, "If true, run code under valgrind to generate callgrind traces.")
useGDB = flag.Bool("gdb", false, "If true, run BoringSSL code under gdb")
useSDE = flag.Bool("sde", false, "If true, run BoringSSL code under Intel's SDE for each supported chip")
sdePath = flag.String("sde-path", "sde", "The path to find the sde binary.")
buildDir = flag.String("build-dir", "build", "The build directory to run the tests from.")
numWorkers = flag.Int("num-workers", runtime.NumCPU(), "Runs the given number of workers when testing.")
jsonOutput = flag.String("json-output", "", "The file to output JSON results to.")
mallocTest = flag.Int64("malloc-test", -1, "If non-negative, run each test with each malloc in turn failing from the given number onwards.")
mallocTestDebug = flag.Bool("malloc-test-debug", false, "If true, ask each test to abort rather than fail a malloc. This can be used with a specific value for --malloc-test to identity the malloc failing that is causing problems.")
simulateARMCPUs = flag.Bool("simulate-arm-cpus", simulateARMCPUsDefault(), "If true, runs tests simulating different ARM CPUs.")
)
func simulateARMCPUsDefault() bool {
return runtime.GOOS == "linux" && (runtime.GOARCH == "arm" || runtime.GOARCH == "arm64")
}
type test struct {
args []string
shard, numShards int
// cpu, if not empty, contains a code to simulate. For SDE, run `sde64
// -help` to get a list of these codes. For ARM, see gtest_main.cc for
// the supported values.
cpu string
}
type result struct {
Test test
Passed bool
Error error
}
// sdeCPUs contains a list of CPU code that we run all tests under when *useSDE
// is true.
var sdeCPUs = []string{
"p4p", // Pentium4 Prescott
"mrm", // Merom
"pnr", // Penryn
"nhm", // Nehalem
"wsm", // Westmere
"snb", // Sandy Bridge
"ivb", // Ivy Bridge
"hsw", // Haswell
"bdw", // Broadwell
"skx", // Skylake Server
"skl", // Skylake Client
"cnl", // Cannonlake
"knl", // Knights Landing
"slt", // Saltwell
"slm", // Silvermont
"glm", // Goldmont
"knm", // Knights Mill
}
var armCPUs = []string{
"none", // No support for any ARM extensions.
"neon", // Support for NEON.
"crypto", // Support for NEON and crypto extensions.
}
func valgrindOf(dbAttach bool, path string, args ...string) *exec.Cmd {
valgrindArgs := []string{"--error-exitcode=99", "--track-origins=yes", "--leak-check=full", "--quiet"}
if dbAttach {
valgrindArgs = append(valgrindArgs, "--db-attach=yes", "--db-command=xterm -e gdb -nw %f %p")
}
valgrindArgs = append(valgrindArgs, path)
valgrindArgs = append(valgrindArgs, args...)
return exec.Command("valgrind", valgrindArgs...)
}
func callgrindOf(path string, args ...string) *exec.Cmd {
valgrindArgs := []string{"-q", "--tool=callgrind", "--dump-instr=yes", "--collect-jumps=yes", "--callgrind-out-file=" + *buildDir + "/callgrind/callgrind.out.%p"}
valgrindArgs = append(valgrindArgs, path)
valgrindArgs = append(valgrindArgs, args...)
return exec.Command("valgrind", valgrindArgs...)
}
func gdbOf(path string, args ...string) *exec.Cmd {
xtermArgs := []string{"-e", "gdb", "--args"}
xtermArgs = append(xtermArgs, path)
xtermArgs = append(xtermArgs, args...)
return exec.Command("xterm", xtermArgs...)
}
func sdeOf(cpu, path string, args ...string) *exec.Cmd {
sdeArgs := []string{"-" + cpu}
// The kernel's vdso code for gettimeofday sometimes uses the RDTSCP
// instruction. Although SDE has a -chip_check_vsyscall flag that
// excludes such code by default, it does not seem to work. Instead,
// pass the -chip_check_exe_only flag which retains test coverage when
// statically linked and excludes the vdso.
if cpu == "p4p" || cpu == "pnr" || cpu == "mrm" || cpu == "slt" {
sdeArgs = append(sdeArgs, "-chip_check_exe_only")
}
sdeArgs = append(sdeArgs, "--", path)
sdeArgs = append(sdeArgs, args...)
return exec.Command(*sdePath, sdeArgs...)
}
var (
errMoreMallocs = errors.New("child process did not exhaust all allocation calls")
errTestSkipped = errors.New("test was skipped")
)
func runTestOnce(test test, mallocNumToFail int64) (passed bool, err error) {
prog := path.Join(*buildDir, test.args[0])
args := append([]string{}, test.args[1:]...)
if *simulateARMCPUs && test.cpu != "" {
args = append(args, "--cpu=" + test.cpu)
}
if *useSDE {
// SDE is neither compatible with the unwind tester nor automatically
// detected.
args = append(args, "--no_unwind_tests")
}
var cmd *exec.Cmd
if *useValgrind {
cmd = valgrindOf(false, prog, args...)
} else if *useCallgrind {
cmd = callgrindOf(prog, args...)
} else if *useGDB {
cmd = gdbOf(prog, args...)
} else if *useSDE {
cmd = sdeOf(test.cpu, prog, args...)
} else {
cmd = exec.Command(prog, args...)
}
var outBuf bytes.Buffer
cmd.Stdout = &outBuf
cmd.Stderr = &outBuf
if mallocNumToFail >= 0 {
cmd.Env = os.Environ()
cmd.Env = append(cmd.Env, "MALLOC_NUMBER_TO_FAIL="+strconv.FormatInt(mallocNumToFail, 10))
if *mallocTestDebug {
cmd.Env = append(cmd.Env, "MALLOC_ABORT_ON_FAIL=1")
}
cmd.Env = append(cmd.Env, "_MALLOC_CHECK=1")
}
if err := cmd.Start(); err != nil {
return false, err
}
if err := cmd.Wait(); err != nil {
if exitError, ok := err.(*exec.ExitError); ok {
switch exitError.Sys().(syscall.WaitStatus).ExitStatus() {
case 88:
return false, errMoreMallocs
case 89:
fmt.Print(string(outBuf.Bytes()))
return false, errTestSkipped
}
}
fmt.Print(string(outBuf.Bytes()))
return false, err
}
// Account for Windows line-endings.
stdout := bytes.Replace(outBuf.Bytes(), []byte("\r\n"), []byte("\n"), -1)
if bytes.HasSuffix(stdout, []byte("PASS\n")) &&
(len(stdout) == 5 || stdout[len(stdout)-6] == '\n') {
return true, nil
}
// Also accept a googletest-style pass line. This is left here in
// transition until the tests are all converted and this script made
// unnecessary.
if bytes.Contains(stdout, []byte("\n[ PASSED ]")) {
return true, nil
}
fmt.Print(string(outBuf.Bytes()))
return false, nil
}
func runTest(test test) (bool, error) {
if *mallocTest < 0 {
return runTestOnce(test, -1)
}
for mallocNumToFail := int64(*mallocTest); ; mallocNumToFail++ {
if passed, err := runTestOnce(test, mallocNumToFail); err != errMoreMallocs {
if err != nil {
err = fmt.Errorf("at malloc %d: %s", mallocNumToFail, err)
}
return passed, err
}
}
}
// setWorkingDirectory walks up directories as needed until the current working
// directory is the top of a BoringSSL checkout.
func setWorkingDirectory() {
for i := 0; i < 64; i++ {
if _, err := os.Stat("BUILDING.md"); err == nil {
return
}
os.Chdir("..")
}
panic("Couldn't find BUILDING.md in a parent directory!")
}
func parseTestConfig(filename string) ([]test, error) {
in, err := os.Open(filename)
if err != nil {
return nil, err
}
defer in.Close()
decoder := json.NewDecoder(in)
var testArgs [][]string
if err := decoder.Decode(&testArgs); err != nil {
return nil, err
}
var result []test
for _, args := range testArgs {
result = append(result, test{args: args})
}
return result, nil
}
func worker(tests <-chan test, results chan<- result, done *sync.WaitGroup) {
defer done.Done()
for test := range tests {
passed, err := runTest(test)
results <- result{test, passed, err}
}
}
func (t test) shortName() string {
return t.args[0] + t.shardMsg() + t.cpuMsg()
}
func (t test) longName() string {
return strings.Join(t.args, " ") + t.cpuMsg()
}
func (t test) shardMsg() string {
if t.numShards == 0 {
return ""
}
return fmt.Sprintf(" [shard %d/%d]", t.shard+1, t.numShards)
}
func (t test) cpuMsg() string {
if len(t.cpu) == 0 {
return ""
}
return fmt.Sprintf(" (for CPU %q)", t.cpu)
}
func (t test) getGTestShards() ([]test, error) {
if *numWorkers == 1 || len(t.args) != 1 {
return []test{t}, nil
}
// Only shard the three GTest-based tests.
if t.args[0] != "crypto/crypto_test" && t.args[0] != "ssl/ssl_test" && t.args[0] != "decrepit/decrepit_test" {
return []test{t}, nil
}
prog := path.Join(*buildDir, t.args[0])
cmd := exec.Command(prog, "--gtest_list_tests")
var stdout bytes.Buffer
cmd.Stdout = &stdout
if err := cmd.Start(); err != nil {
return nil, err
}
if err := cmd.Wait(); err != nil {
return nil, err
}
var group string
var tests []string
scanner := bufio.NewScanner(&stdout)
for scanner.Scan() {
line := scanner.Text()
// Remove the parameter comment and trailing space.
if idx := strings.Index(line, "#"); idx >= 0 {
line = line[:idx]
}
line = strings.TrimSpace(line)
if len(line) == 0 {
continue
}
if line[len(line)-1] == '.' {
group = line
continue
}
if len(group) == 0 {
return nil, fmt.Errorf("found test case %q without group", line)
}
tests = append(tests, group+line)
}
const testsPerShard = 20
if len(tests) <= testsPerShard {
return []test{t}, nil
}
// Slow tests which process large test vector files tend to be grouped
// together, so shuffle the order.
shuffled := make([]string, len(tests))
perm := rand.Perm(len(tests))
for i, j := range perm {
shuffled[i] = tests[j]
}
var shards []test
for i := 0; i < len(shuffled); i += testsPerShard {
n := len(shuffled) - i
if n > testsPerShard {
n = testsPerShard
}
shard := t
shard.args = []string{shard.args[0], "--gtest_filter=" + strings.Join(shuffled[i:i+n], ":")}
shard.shard = len(shards)
shards = append(shards, shard)
}
for i := range shards {
shards[i].numShards = len(shards)
}
return shards, nil
}
func main() {
flag.Parse()
setWorkingDirectory()
testCases, err := parseTestConfig("util/all_tests.json")
if err != nil {
fmt.Printf("Failed to parse input: %s\n", err)
os.Exit(1)
}
var wg sync.WaitGroup
tests := make(chan test, *numWorkers)
results := make(chan result, *numWorkers)
for i := 0; i < *numWorkers; i++ {
wg.Add(1)
go worker(tests, results, &wg)
}
go func() {
for _, test := range testCases {
if *useSDE {
// SDE generates plenty of tasks and gets slower
// with additional sharding.
for _, cpu := range sdeCPUs {
testForCPU := test
testForCPU.cpu = cpu
tests <- testForCPU
}
} else if *simulateARMCPUs {
// This mode is run instead of the default path,
// so also include the native flow.
tests <- test
for _, cpu := range armCPUs {
testForCPU := test
testForCPU.cpu = cpu
tests <- testForCPU
}
} else {
shards, err := test.getGTestShards()
if err != nil {
fmt.Printf("Error listing tests: %s\n", err)
os.Exit(1)
}
for _, shard := range shards {
tests <- shard
}
}
}
close(tests)
wg.Wait()
close(results)
}()
testOutput := testresult.NewResults()
var failed, skipped []test
for testResult := range results {
test := testResult.Test
args := test.args
if testResult.Error == errTestSkipped {
fmt.Printf("%s\n", test.longName())
fmt.Printf("%s was skipped\n", args[0])
skipped = append(skipped, test)
testOutput.AddSkip(test.longName())
} else if testResult.Error != nil {
fmt.Printf("%s\n", test.longName())
fmt.Printf("%s failed to complete: %s\n", args[0], testResult.Error)
failed = append(failed, test)
testOutput.AddResult(test.longName(), "CRASH")
} else if !testResult.Passed {
fmt.Printf("%s\n", test.longName())
fmt.Printf("%s failed to print PASS on the last line.\n", args[0])
failed = append(failed, test)
testOutput.AddResult(test.longName(), "FAIL")
} else {
fmt.Printf("%s\n", test.shortName())
testOutput.AddResult(test.longName(), "PASS")
}
}
if *jsonOutput != "" {
if err := testOutput.WriteToFile(*jsonOutput); err != nil {
fmt.Fprintf(os.Stderr, "Error: %s\n", err)
}
}
if len(skipped) > 0 {
fmt.Printf("\n%d of %d tests were skipped:\n", len(skipped), len(testCases))
for _, test := range skipped {
fmt.Printf("\t%s%s\n", strings.Join(test.args, " "), test.cpuMsg())
}
}
if len(failed) > 0 {
fmt.Printf("\n%d of %d tests failed:\n", len(failed), len(testCases))
for _, test := range failed {
fmt.Printf("\t%s%s\n", strings.Join(test.args, " "), test.cpuMsg())
}
os.Exit(1)
}
fmt.Printf("\nAll tests passed!\n")
}