// Copyright 2016 The Bazel Authors. 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. package com.google.devtools.build.lib.shell; import com.google.devtools.build.lib.shell.SubprocessBuilder.StreamAction; import java.io.File; import java.io.IOException; import java.io.InputStream; import java.io.OutputStream; import java.lang.ProcessBuilder.Redirect; import java.util.concurrent.TimeUnit; import java.util.concurrent.atomic.AtomicBoolean; /** * A subprocess factory that uses {@link java.lang.ProcessBuilder}. */ public class JavaSubprocessFactory implements SubprocessFactory { /** * A subprocess backed by a {@link java.lang.Process}. */ private static class JavaSubprocess implements Subprocess { private final Process process; private final long deadlineMillis; private final AtomicBoolean deadlineExceeded = new AtomicBoolean(); private JavaSubprocess(Process process, long deadlineMillis) { this.process = process; this.deadlineMillis = deadlineMillis; } @Override public boolean destroy() { process.destroy(); return true; } @Override public int exitValue() { return process.exitValue(); } @Override public boolean finished() { if (deadlineMillis > 0 && System.currentTimeMillis() > deadlineMillis && deadlineExceeded.compareAndSet(false, true)) { // We use compareAndSet here to avoid calling destroy multiple times. Note that destroy // returns immediately, and we don't want to wait in this method. process.destroy(); } // this seems to be the only non-blocking call for checking liveness return !process.isAlive(); } @Override public boolean timedout() { return deadlineExceeded.get(); } @Override public void waitFor() throws InterruptedException { if (deadlineMillis > 0) { // Careful: I originally used Long.MAX_VALUE if there's no timeout. This is safe with // Process, but not for the UNIXProcess subclass, which has an integer overflow for very // large timeouts. As of this writing, it converts the passed in value to nanos (which // saturates at Long.MAX_VALUE), then adds 999999 to round up (which overflows), converts // back to millis, and then calls Object.wait with a negative timeout, which throws. long waitTimeMillis = deadlineMillis - System.currentTimeMillis(); boolean exitedInTime = process.waitFor(waitTimeMillis, TimeUnit.MILLISECONDS); if (!exitedInTime && deadlineExceeded.compareAndSet(false, true)) { process.destroy(); // The destroy call returns immediately, so we still need to wait for the actual exit. The // sole caller assumes that waitFor only exits when the process is gone (or throws). process.waitFor(); } } else { process.waitFor(); } } @Override public OutputStream getOutputStream() { return process.getOutputStream(); } @Override public InputStream getErrorStream() { return process.getErrorStream(); } @Override public InputStream getInputStream() { return process.getInputStream(); } @Override public void close() { // java.lang.Process doesn't give us a way to clean things up other than #destroy(), which was // already called by this point. } } public static final JavaSubprocessFactory INSTANCE = new JavaSubprocessFactory(); private JavaSubprocessFactory() { // We are a singleton } // since we are a singleton, we represent an ideal global lock for // process invocations, which is required due to the following race condition: // Linux does not provide a safe API for a multi-threaded program to fork a subprocess. // Consider the case where two threads both write an executable file and then try to execute // it. It can happen that the first thread writes its executable file, with the file // descriptor still being open when the second thread forks, with the fork inheriting a copy // of the file descriptor. Then the first thread closes the original file descriptor, and // proceeds to execute the file. At that point Linux sees an open file descriptor to the file // and returns ETXTBSY (Text file busy) as an error. This race is inherent in the fork / exec // duality, with fork always inheriting a copy of the file descriptor table; if there was a // way to fork without copying the entire file descriptor table (e.g., only copy specific // entries), we could avoid this race. // // I was able to reproduce this problem reliably by running significantly more threads than // there are CPU cores on my workstation - the more threads the more likely it happens. // // As a workaround, we put a synchronized block around the fork. private synchronized Process start(ProcessBuilder builder) throws IOException { return builder.start(); } @Override public Subprocess create(SubprocessBuilder params) throws IOException { ProcessBuilder builder = new ProcessBuilder(); builder.command(params.getArgv()); if (params.getEnv() != null) { builder.environment().clear(); builder.environment().putAll(params.getEnv()); } builder.redirectOutput(getRedirect(params.getStdout(), params.getStdoutFile())); builder.redirectError(getRedirect(params.getStderr(), params.getStderrFile())); builder.redirectErrorStream(params.redirectErrorStream()); builder.directory(params.getWorkingDirectory()); // Deadline is now + given timeout. long deadlineMillis = params.getTimeoutMillis() > 0 ? Math.addExact(System.currentTimeMillis(), params.getTimeoutMillis()) : 0; return new JavaSubprocess(start(builder), deadlineMillis); } /** * Returns a {@link java.lang.ProcessBuilder.Redirect} appropriate for the parameters. If a file * redirected to exists, deletes the file before redirecting to it. */ private Redirect getRedirect(StreamAction action, File file) { switch (action) { case DISCARD: return Redirect.to(new File("/dev/null")); case REDIRECT: // We need to use Redirect.appendTo() here, because on older Linux kernels writes are // otherwise not atomic and might result in lost log messages: // https://lkml.org/lkml/2014/3/3/308 if (file.exists()) { file.delete(); } return Redirect.appendTo(file); case STREAM: return Redirect.PIPE; default: throw new IllegalStateException(); } } }