Linux sandbox: start adding syscall wrappers

sandbox/linux/services/syscall_wrappers.h

+// Copyright 2014 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.
+
+#ifndef SANDBOX_LINUX_SERVICES_SYSCALL_WRAPPERS_H_
+#define SANDBOX_LINUX_SERVICES_SYSCALL_WRAPPERS_H_
+
+#include <sys/types.h>
+
+#include "sandbox/sandbox_export.h"
+
+namespace sandbox {
+
+// Provide direct system call wrappers for a few common system calls.
+// These are guaranteed to perform a system call and do not rely on things such
+// as caching the current pid (c.f. getpid()).
+
+SANDBOX_EXPORT pid_t sys_getpid(void);
+
+SANDBOX_EXPORT pid_t sys_gettid(void);
+
+struct pt_regs;
+SANDBOX_EXPORT long sys_clone(unsigned long flags,

[mdempsky, 2014/11/20 00:31:43]

Most of the calls seem to be for sys_clone(flags, 0, 0, 0, 0).  I wonder if it
makes sense to provide an overload for

    sys_clone(unsigned long flags);

that can pass the zero arguments and also sanity check that
CLONE_CHILD_CLEARTID, CLONE_CHILD_SETTID, CLONE_PARENT_SETTID, and CLONE_VM are
all cleared?

Even if not, I think the style guide wants the call sites to use nullptr instead
of 0?

[jln (very slow on Chromium), 2014/11/20 18:04:50]

Yeah, you're right, this needs to be cleaned-up. I started doing that, but I
fear to go down that rabit hole: a bunch of call to sys_clone need to actually
be rewritten properly.

I'll do that in a follow-up CL.

+                              void* child_stack,
+                              void* ptid,
+                              void* ctid,
+                              struct pt_regs* regs);
+
+SANDBOX_EXPORT void sys_exit_group(int status);
+
+}  // namespace sandbox
+
+#endif  // SANDBOX_LINUX_SERVICES_SYSCALL_WRAPPERS_H_