Add namespace sandbox class.

sandbox/linux/services/namespace_sandbox.cc

+// Copyright 2015 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "sandbox/linux/services/namespace_sandbox.h"
+
+#include <sched.h>
+#include <stdlib.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include <string>
+#include <utility>
+
+#include "base/bind.h"
+#include "base/command_line.h"
+#include "base/environment.h"
+#include "base/files/scoped_file.h"
+#include "base/logging.h"
+#include "base/posix/eintr_wrapper.h"
+#include "base/process/launch.h"
+#include "base/process/process.h"
+#include "base/strings/stringprintf.h"
+#include "sandbox/linux/services/namespace_utils.h"
+
+namespace sandbox {
+
+namespace {
+const char kPipeValue = '\xcc';
+
+class BlockOnPipeDelegate : public base::LaunchOptions::PreExecDelegate {
+ public:
+  explicit BlockOnPipeDelegate(int fd) : fd_(fd) {}
+
+  ~BlockOnPipeDelegate() override {}
+
+  void RunAsyncSafe() override {
+    char c;
+    RAW_CHECK(HANDLE_EINTR(read(fd_, &c, 1)) == 1);
+    RAW_CHECK(IGNORE_EINTR(close(fd_)) == 0);
+    RAW_CHECK(c == kPipeValue);
+  }
+
+ private:
+  int fd_;
+  DISALLOW_COPY_AND_ASSIGN(BlockOnPipeDelegate);
+};
+
+void SetEnvironForNamespaceType(base::EnvironmentMap* environ,
+                                base::NativeEnvironmentString env_var,
+                                bool value) {
+  // An empty string causes the env var to be unset in the child process.
+  (*environ)[env_var] = value ? "1" : "";
+}
+
+const char kSandboxUSERNSEnvironmentVarName[] = "SBX_USER_NS";
+const char kSandboxPIDNSEnvironmentVarName[] = "SBX_PID_NS";
+const char kSandboxNETNSEnvironmentVarName[] = "SBX_NET_NS";
+
+}  // namespace
+
+NamespaceSandbox::NamespaceSandbox() : launch_called_(false) {
+}
+
+NamespaceSandbox::~NamespaceSandbox() {
+}
+
+void NamespaceSandbox::AddFdMapping(int from_fd, int to_fd) {
+  fds_to_remap_.push_back(std::make_pair(from_fd, to_fd));
+}
+
+base::Process NamespaceSandbox::Launch(const base::CommandLine& cmdline) {
+  CHECK(!launch_called_) << "NamespaceSandbox may only be used once.";
+  launch_called_ = true;
+
+  int clone_flags = 0;
+  int ns_types[] = {CLONE_NEWUSER, CLONE_NEWPID, CLONE_NEWNET};
+  for (const int ns_type : ns_types) {
+    if (NamespaceUtils::KernelSupportsUnprivilegedNamespace(ns_type)) {
+      clone_flags |= ns_type;
+    }
+  }
+  CHECK(clone_flags & CLONE_NEWUSER);
+
+  int fds[2];
+  PCHECK(pipe(fds) == 0);
+  read_fd_.reset(fds[0]);
+  write_fd_.reset(fds[1]);
+  fds_to_remap_.push_back(std::make_pair(read_fd_.get(), read_fd_.get()));
+
+  BlockOnPipeDelegate block_on_pipe_delegate(read_fd_.get());
+  launch_options_.pre_exec_delegate = &block_on_pipe_delegate;
+  launch_options_.fds_to_remap = &fds_to_remap_;
+  launch_options_.clone_flags = clone_flags;
+
+  const std::pair<int, const char*> clone_flag_environ[] = {
+      std::make_pair(CLONE_NEWUSER, kSandboxUSERNSEnvironmentVarName),
+      std::make_pair(CLONE_NEWPID, kSandboxPIDNSEnvironmentVarName),
+      std::make_pair(CLONE_NEWNET, kSandboxNETNSEnvironmentVarName),
+  };
+
+  base::EnvironmentMap* environ = &launch_options_.environ;
+  for (const auto& entry : clone_flag_environ) {
+    const int flag = entry.first;
+    const char* environ_name = entry.second;
+    SetEnvironForNamespaceType(environ, environ_name, clone_flags & flag);
+  }
+
+  base::Process process = base::LaunchProcess(cmdline, launch_options_);
+  read_fd_.reset();
+
+  if (!process.IsValid()) {
+    write_fd_.reset();

[jln (very slow on Chromium), 2015/01/30 20:12:28]

Having to "remember" to do that is problematic.

read_fd_ and write_fd_ are mostly used in this function.

I would not make them class members and instead just have the ScopedFD in this
method.

Then you can have WriteUidGidMapAndUnblockProcess() take a ScopedFD (that you
can Pass()) and I believe that everything will look right, with "automatic"
closing of the ScopedFD.

[rickyz (no longer on Chrome), 2015/01/31 02:33:53]

Ah, that's much nicer, done

+    return process.Pass();
+  }
+
+  WriteUidGidMapAndUnblockProcess(process.Pid());
+  return process.Pass();
+}
+
+void NamespaceSandbox::WriteUidGidMapAndUnblockProcess(pid_t pid) {
+  const std::string uid_map_path = base::StringPrintf("/proc/%d/uid_map", pid);
+  const std::string gid_map_path = base::StringPrintf("/proc/%d/gid_map", pid);
+  NamespaceUtils::WriteToIdMapFile(uid_map_path.c_str(), getuid());
+  NamespaceUtils::WriteToIdMapFile(gid_map_path.c_str(), getgid());
+
+  PCHECK(HANDLE_EINTR(write(write_fd_.get(), &kPipeValue, 1)) == 1);

[jln (very slow on Chromium), 2015/01/30 20:12:28]

If the child process gets killed or disappears for any reason, you'll get
SIGPIP-ed.

You could use send with MSG_NOSIGNAL. Or you could simply close the pipe to
unblock the child, but then the child won't immediately know if the parent died
or signaled it (which is probably no big deal).

[rickyz (no longer on Chrome), 2015/01/31 02:33:53]

Done.

+  write_fd_.reset();
+}
+
+bool NamespaceSandbox::InNewUserNamespace() {
+  return getenv(kSandboxUSERNSEnvironmentVarName) != nullptr;
+}
+
+bool NamespaceSandbox::InNewPidNamespace() {
+  return getenv(kSandboxPIDNSEnvironmentVarName) != nullptr;
+}
+
+bool NamespaceSandbox::InNewNetNamespace() {
+  return getenv(kSandboxNETNSEnvironmentVarName) != nullptr;
+}
+
+}  // namespace sandbox