Initial snapshot of the new BPF-enabled seccomp sandbox. This code is

sandbox/linux/seccomp_bpf/sandbox_bpf.cc

Index: sandbox/linux/seccomp_bpf/sandbox_bpf.cc
===================================================================
--- sandbox/linux/seccomp_bpf/sandbox_bpf.cc	(revision 0)
+++ sandbox/linux/seccomp_bpf/sandbox_bpf.cc	(revision 0)
@@ -0,0 +1,296 @@
+// Copyright (c) 2012 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/seccomp_bpf/sandbox_bpf.h"
+
+// The kernel gives us a sandbox, we turn it into a playground :-)
+// This is version 2 of the playground; version 1 was built on top of
+// pre-BPF seccomp mode.
+namespace playground2 {
+
+static Sandbox::ErrorCode probeEvaluator(int signo) {
+  switch (signo) {
+  case __NR_getpid:
+    // Return EPERM so that we can check that the filter actually ran.
+    return (Sandbox::ErrorCode)EPERM;
+  case __NR_exit_group:
+    // Allow exit() with a non-default return code.
+    return Sandbox::SB_ALLOWED;
+  default:
+    // Make everything else fail in an easily recognizable way.
+    return (Sandbox::ErrorCode)EINVAL;
+  }
+}
+
+Sandbox::SandboxStatus Sandbox::supportsSeccompSandbox(int proc_fd) {
+  // It the sandbox is currently active, we clearly must have support for
+  // sandboxing.
+  if (status_ == STATUS_ENABLED) {
+    return status_;
+  }
+
+  // Even if the sandbox was previously available, something might have
+  // changed in our run-time environment. Check one more time.
+  if (status_ == STATUS_AVAILABLE) {
+    if (!isSingleThreaded(proc_fd)) {
+      status_ = STATUS_UNAVAILABLE;
+    }
+    return status_;
+  }
+
+  // If we previously checked and we determined that kernel support was
+  // available but the current state of the run-time environment was
+  // incompatible with the sandbox, we want to perform the full test again.
+  // This deals with more subtle incompatibilities in the run-time
+  // environment.
+  if (status_ == STATUS_UNAVAILABLE &&
+      isSingleThreaded(proc_fd)) {

[Chris Evans, 2012/06/04 19:17:26]

Don't we go straight to STATUS_AVAILABLE at this time? Seems like this case
might be the one we discussed in e-mail: the zygote was multi-threaded at the
time of the warm-up and now we're in a single-threaded renderer and wish to
avoid the main overhead of the check.

[Markus (顧孟勤), 2012/06/04 21:13:40]

OK, I changed the code.

This is a little fragile, if somebody ever changes the state transitions. So, I
put a big comment in there. Hopefully, that's good enough.

+    status_ == STATUS_UNKNOWN;
+  }
+
+  // If we have not previously checked for availability of the sandbox or if
+  // we otherwise don't believe to have a good cached value, we have to
+  // perform a thorough check now.
+  if (status_ == STATUS_UNKNOWN) {

[Chris Evans, 2012/06/04 19:17:26]

This whole function is looking much clearer now, and to finish the look, how
about moving this chunk to a separate function, e.g.
"kernelSupportsSeccompFilters" or some such?

[Markus (顧孟勤), 2012/06/04 21:13:40]

Done.

+    sigset_t oldMask, newMask;
+    if (!sigfillset(&newMask) &&
+        !sigprocmask(SIG_BLOCK, &newMask, &oldMask)) {

[Chris Evans, 2012/06/04 19:17:26]

This is another place where you can avoid the indentation, especially if you've
broken this piece into another function. e.g. if (sig... || sig...) return
STATUS_BLAH;

Actually, it looks like you might want to die(), to be symmetrical with other
cases lower down.

[Markus (顧孟勤), 2012/06/04 21:13:40]

Done.

+      pid_t pid = fork();
+      if (pid >= 0) {
+        if (!pid) {
+          // Test a very simple sandbox policy to verify that we can
+          // successfully turn on sandboxing.
+          suppressLogging_ = true;
+          setSandboxPolicy(probeEvaluator, NULL);
+          setProcFd(proc_fd);
+          startSandbox();
+          if (syscall(__NR_getpid) < 0 && errno == EPERM) {
+            syscall(__NR_exit_group, (intptr_t)100);
+          }
+          die(NULL);

[Chris Evans, 2012/06/04 19:17:26]

I missed what we planned to do about the die(NULL)?

[Markus (顧孟勤), 2012/06/04 21:13:40]

die(NULL) is explicitly safe and support. For all practical purposes, think of
it as our version of _exit().

+        }
+        if (sigprocmask(SIG_SETMASK, &oldMask, NULL)) {
+          die("sigprocmask() failed");
+        }
+        int status;
+        if (HANDLE_EINTR(waitpid(pid, &status, 0)) != pid) {
+          status_ = STATUS_UNAVAILABLE;

[Chris Evans, 2012/06/04 19:17:26]

Maybe use die() here? We know we forked just a single child so this we can
expect it to work a bit more forcefully.

[Markus (顧孟勤), 2012/06/04 21:13:40]

Done.

+          return status_;
+        }
+        status_ = WIFEXITED(status) && WEXITSTATUS(status) == 100
+                ? STATUS_AVAILABLE : STATUS_UNSUPPORTED;
+      } else {
+        // fork() failed. Consider this a temporary failure returning
+        // STATUS_UNSUPPORTED.

[Markus (顧孟勤), 2012/06/04 21:13:40]

In light of all the other changes, I felt if safer to now die() whenever fork()
fails. Going back and forth with Julien on this decision, we feel that it
probably is saner and easier to review that way. I added a new comment to
explain our reasoning.

+        if (sigprocmask(SIG_SETMASK, &oldMask, NULL)) {
+          // This should never happen. Really no good way to recover.
+          die("sigprocmask() failed");
+        }
+      }
+    }
+    // As we are performing our tests from a child process, the run-time
+    // environment that is visible to the sandbox is always guaranteed to be
+    // single-threaded. Let's check here whether the caller is single-
+    // threaded. Otherwise, we mark the sandbox as temporarily unavailable.
+    if (status_ == STATUS_AVAILABLE && !isSingleThreaded(proc_fd)) {
+      status_ = STATUS_UNAVAILABLE;
+    }
+  }
+  return status_;
+}
+
+void Sandbox::setProcFd(int proc_fd) {
+  proc_fd_ = proc_fd;
+}
+
+void Sandbox::startSandbox() {
+  if (status_ == STATUS_UNSUPPORTED || status_ == STATUS_UNAVAILABLE) {
+    die("Trying to start sandbox, even though it is known to be unavailable");
+  } else if (status_ == STATUS_ENABLED) {
+    die("Cannot start sandbox recursively. Use multiple calls to "
+        "setSandboxPolicy() to stack policies instead");
+  }
+  if (proc_fd_ < 0) {
+    proc_fd_ = open("/proc", O_RDONLY|O_DIRECTORY);
+  }
+  if (proc_fd_ < 0) {
+    // For now, continue in degraded mode, if we can't access /proc.
+    // In the future, we might want to tighten this requirement.
+  }
+  if (!isSingleThreaded(proc_fd_)) {
+    die("Cannot start sandbox, if process is already multi-threaded");
+  }
+
+  // We no longer need access to any files in /proc. We want to do this
+  // before installing the filters, just in case that our policy denies
+  // close().
+  if (proc_fd_ >= 0) {
+    if (HANDLE_EINTR(close(proc_fd_))) {
+      die("Failed to close file descriptor for /proc");
+    }
+    proc_fd_ = -1;
+  }
+
+  // Install the filters.
+  installFilter();
+
+  // We are now inside the sandbox.
+  status_ = STATUS_ENABLED;
+}
+
+bool Sandbox::isSingleThreaded(int proc_fd) {
+  if (proc_fd < 0) {
+    // Cannot determine whether program is single-threaded. Hope for
+    // the best...
+    return true;
+  }
+
+  struct stat sb;
+  int task = -1;
+  if (proc_fd < 0 ||

[Chris Evans, 2012/06/04 19:17:26]

Don't need to check proc_fd, you did it just above.

[Markus (顧孟勤), 2012/06/04 21:13:40]

Done.

+      (task = openat(proc_fd, "self/task", O_RDONLY|O_DIRECTORY)) < 0 ||
+      fstat(task, &sb) != 0 ||
+      sb.st_nlink != 3 ||
+      HANDLE_EINTR(close(task))) {
+    if (task >= 0) {
+      HANDLE_EINTR(close(task));
+    }
+    return false;
+  }
+  return true;
+}
+
+void Sandbox::setSandboxPolicy(EvaluateSyscall syscallEvaluator,
+                               EvaluateArguments argumentEvaluator) {
+  evaluators_.push_back(std::make_pair(syscallEvaluator, argumentEvaluator));
+}
+
+void Sandbox::installFilter() {
+  // Verify that the user pushed a policy.
+  if (evaluators_.empty()) {
+  filter_failed:
+    die("Failed to configure system call filters");
+  }
+
+  // Set new SIGSYS handler
+  struct sigaction sa;
+  memset(&sa, 0, sizeof(sa));
+  sa.sa_sigaction = &sigSys;
+  sa.sa_flags = SA_SIGINFO;
+  if (sigaction(SIGSYS, &sa, NULL) < 0) {
+    goto filter_failed;
+  }
+
+  // Unmask SIGSYS
+  sigset_t mask;
+  sigemptyset(&mask);
+  sigaddset(&mask, SIGSYS);
+  if (sigprocmask(SIG_UNBLOCK, &mask, NULL)) {
+    goto filter_failed;
+  }
+
+  // We can't handle stacked evaluators, yet. We'll get there eventually
+  // though. Hang tight.
+  if (evaluators_.size() != 1) {
+    die("Not implemented");
+  }
+
+  // If the architecture doesn't match SECCOMP_ARCH, disallow the
+  // system call.
+  std::vector<struct sock_filter> program;
+  program.push_back((struct sock_filter)
+                    BPF_STMT(BPF_LD+BPF_W+BPF_ABS,
+                             offsetof(struct arch_seccomp_data, arch)));
+  program.push_back((struct sock_filter)
+    BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, SECCOMP_ARCH, 1, 0));
+  program.push_back((struct sock_filter)
+    BPF_STMT(BPF_RET+BPF_K, SECCOMP_RET_ERRNO + SECCOMP_DENY_ERRNO));
+
+  // Grab the system call number, so that we can implement jump tables.
+  program.push_back((struct sock_filter)
+    BPF_STMT(BPF_LD+BPF_W+BPF_ABS, offsetof(struct arch_seccomp_data, nr)));
+
+  // Evaluate all possible system calls and depending on their
+  // exit codes generate a BPF filter.
+  // This is very inefficient right now. We need to be much smarter
+  // eventually.
+  EvaluateSyscall evaluateSyscall = evaluators_.begin()->first;
+  for (int sysnum = MIN_SYSCALL; sysnum <= MAX_SYSCALL; ++sysnum) {
+    ErrorCode err = evaluateSyscall(sysnum);
+    int ret;
+    switch (err) {
+    case SB_INSPECT_ARG_1...SB_INSPECT_ARG_6:
+      die("Not implemented");
+    case SB_TRAP:
+      ret = SECCOMP_RET_TRAP;
+      break;
+    case SB_ALLOWED:
+      ret = SECCOMP_RET_ALLOW;
+      break;
+    default:
+      if (err >= static_cast<ErrorCode>(1) &&
+          err <= static_cast<ErrorCode>(4096)) {
+        // We limit errno values to a reasonable range. In fact, the Linux ABI
+        // doesn't support errno values outside of this range.
+        ret = SECCOMP_RET_ERRNO + err;
+      } else {
+        die("Invalid ErrorCode reported by sandbox system call evaluator");
+      }
+      break;
+    }
+    program.push_back((struct sock_filter)
+      BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, sysnum, 0, 1));
+    program.push_back((struct sock_filter)
+      BPF_STMT(BPF_RET+BPF_K, ret));
+  }
+
+  // Everything that isn't allowed is forbidden. Eventually, we would
+  // like to have a way to log forbidden calls, when in debug mode.
+  program.push_back((struct sock_filter)
+    BPF_STMT(BPF_RET+BPF_K, SECCOMP_RET_ERRNO + SECCOMP_DENY_ERRNO));
+
+  // Install BPF filter program
+  const struct sock_fprog prog = { program.size(), &program[0] };
+  if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) ||
+      prctl(PR_SET_SECCOMP, SECCOMP_MODE_FILTER, &prog)) {
+    goto filter_failed;
+  }
+
+  return;
+}
+
+void Sandbox::sigSys(int nr, siginfo_t *info, void *void_context) {
+  if (nr != SIGSYS || info->si_code != SYS_SECCOMP || !void_context) {
+    die("Unexpected SIGSYS received");

[Chris Evans, 2012/06/04 19:17:26]

I still don't think LOG(FATAL) (one particular expansion of die()) is
async-signal-safe.

At a minimum, I think it deserves an apologetic comment :) However, these really
are should-not-ever-happen conditions so I'd be supportive of simply _exit(1) if
you wanted.

[Markus (顧孟勤), 2012/06/04 21:13:40]

I hear you -- you are really preaching to the choir, and I am usually the one
who keeps yelling at people for not writing async-safe code.

No, I am not very happy with LOG(FATAL).

But Julien very correctly pointed out that it is the best we have right now, if
we want proper logging and reporting. And I am stuck between getting the LGTM
from you or getting the OK from Julien...

In the long run, we probably need to come up with something a little better. But
that'll introduce some crazy code complexity that I don't want to touch at this
point (e.g. we could have a ring-buffer that is periodically checked by another
thread; of course, when we are about to die, we have no guarantee that the other
thread still runs; so, we'd have to think of some best-effort approach that is
likely to succeed most of the time).

+  }
+  ucontext_t *ctx = reinterpret_cast<ucontext_t *>(void_context);
+  int old_errno   = errno;
+
+  // In case of error, set the REG_RESULT CPU register to the default
+  // errno value (i.e. EPERM).
+  // We need to be very careful when doing this, as some of our target
+  // platforms have pointer types and CPU registers that are wider than
+  // ints. Furthermore, the kernel ABI requires us to return a negative
+  // value, but errno values are usually positive. And in fact, it would
+  // be perfectly reasonable for somebody to have defined them as unsigned
+  // properties. This makes the correct incantation of type casts rather
+  // subtle. Sometimes, C++ is just too smart for its own good.
+  void *rc        = (void *)(intptr_t)-(int)SECCOMP_DENY_ERRNO;
+
+  // This is where we can add extra code to handle complex system calls.
+  // ...
+
+  ctx->uc_mcontext.gregs[REG_RESULT] = reinterpret_cast<greg_t>(rc);
+  errno                              = old_errno;
+  return;
+}
+
+
+bool Sandbox::suppressLogging_          = false;
+Sandbox::SandboxStatus Sandbox::status_ = STATUS_UNKNOWN;
+int    Sandbox::proc_fd_                = -1;
+std::vector<std::pair<Sandbox::EvaluateSyscall,
+                      Sandbox::EvaluateArguments> > Sandbox::evaluators_;
+
+} // namespace
Property changes on: sandbox/linux/seccomp_bpf/sandbox_bpf.cc
___________________________________________________________________
Added: svn:eol-style
   + LF