[PATCH v4 1/2] seccomp_filters: system call filtering using BPF

[PATCH v4 1/2] seccomp_filters: system call filtering using BPF

[Will Drewry]

[This patch depends on luto@mit.edu's no_new_privs patch:
 https://lkml.org/lkml/2012/1/12/446
]

This patch adds support for seccomp mode 2.  This mode enables dynamic
enforcement of system call filtering policy in the kernel as specified
by a userland task.  The policy is expressed in terms of a Berkeley
Packet Filter program, as is used for userland-exposed socket filtering.
Instead of network data, the BPF program is evaluated over struct
user_regs_struct at the time of the system call (as retrieved using
regviews).

A filter program may be installed by a userland task by calling
  prctl(PR_ATTACH_SECCOMP_FILTER, &fprog);
where fprog is of type struct sock_fprog.

If the first filter program allows subsequent prctl(2) calls, then
additional filter programs may be attached.  All attached programs
must be evaluated before a system call will be allowed to proceed.

To avoid CONFIG_COMPAT related landmines, once a filter program is
installed using specific is_compat_task() and current->personality, it
is not allowed to make system calls or attach additional filters which
use a different combination of is_compat_task() and
current->personality.

Filter programs will be inherited across fork/clone and execve.
However, if the task attaching the filter is unprivileged
(!CAP_SYS_ADMIN) the no_new_privs bit will be set on the task.  This
ensures that unprivileged tasks cannot attach filters that affect
privileged tasks (e.g., setuid binary).

There are a number of benefits to this approach. A few of which are
as follows:
- BPF has been exposed to userland for a long time.
- Userland already knows its ABI: expected register layout and system
  call numbers.
- Full register information is provided which may be relevant for
  certain syscalls (fork, rt_sigreturn) or for other userland
  filtering tactics (checking the PC).
- No time-of-check-time-of-use vulnerable data accesses are possible.

This patch includes its own BPF evaluator, but relies on the
net/core/filter.c BPF checking code.  It is possible to share
evaluators, but the performance sensitive nature of the network
filtering path makes it an iterative optimization which (I think :) can
be tackled separately via separate patchsets. (And at some point sharing
BPF JIT code!)

 v4: - adjusted prctl to make room for PR_[SG]ET_NO_NEW_PRIVS
     - now uses current->no_new_privs
         (luto@mit.edu,torvalds@linux-foundation.com)
     - assign names to seccomp modes (rdunlap@xenotime.net)
     - fix style issues (rdunlap@xenotime.net)
     - reworded Kconfig entry (rdunlap@xenotime.net)
 v3: - macros to inline (oleg@redhat.com)
     - init_task behavior fixed (oleg@redhat.com)
     - drop creator entry and extra NULL check (oleg@redhat.com)
     - alloc returns -EINVAL on bad sizing (serge.hallyn@canonical.com)
     - adds tentative use of "always_unprivileged" as per
       torvalds@linux-foundation.org and luto@mit.edu
 v2: - (patch 2 only)

Signed-off-by: Will Drewry <wad@chromium.org>
---
 include/linux/prctl.h   |    3 +
 include/linux/seccomp.h |   70 +++++-
 kernel/Makefile         |    1 +
 kernel/fork.c           |    4 +
 kernel/seccomp.c        |   10 +-
 kernel/seccomp_filter.c |  642 +++++++++++++++++++++++++++++++++++++++++++++++
 kernel/sys.c            |    4 +
 security/Kconfig        |   16 ++
 8 files changed, 746 insertions(+), 4 deletions(-)
 create mode 100644 kernel/seccomp_filter.c

diff --git a/include/linux/prctl.h b/include/linux/prctl.h
index a3baeb2..81c6081 100644
--- a/include/linux/prctl.h
+++ b/include/linux/prctl.h
@@ -64,6 +64,9 @@
 #define PR_GET_SECCOMP	21
 #define PR_SET_SECCOMP	22
 
+/* Set process seccomp filters */
+#define PR_ATTACH_SECCOMP_FILTER	37
+
 /* Get/set the capability bounding set (as per security/commoncap.c) */
 #define PR_CAPBSET_READ 23
 #define PR_CAPBSET_DROP 24
diff --git a/include/linux/seccomp.h b/include/linux/seccomp.h
index cc7a4e9..d73fc35 100644
--- a/include/linux/seccomp.h
+++ b/include/linux/seccomp.h
@@ -5,9 +5,30 @@
 #ifdef CONFIG_SECCOMP
 
 #include <linux/thread_info.h>
+#include <linux/types.h>
 #include <asm/seccomp.h>
 
-typedef struct { int mode; } seccomp_t;
+/* Valid values of seccomp_struct.mode */
+#define SECCOMP_MODE_DISABLED	0 /* seccomp is not in use. */
+#define SECCOMP_MODE_STRICT	1 /* uses hard-coded seccomp.c rules. */
+#define SECCOMP_MODE_FILTER	2 /* system call access determined by filter. */
+
+struct seccomp_filter;
+/**
+ * struct seccomp_struct - the state of a seccomp'ed process
+ *
+ * @mode:  indicates one of the valid values above for controlled
+ *         system calls available to a process.
+ * @filter: Metadata for filter if using CONFIG_SECCOMP_FILTER.
+ *          @filter must only be accessed from the context of current as there
+ *          is no guard.
+ */
+typedef struct seccomp_struct {
+	int mode;
+#ifdef CONFIG_SECCOMP_FILTER
+	struct seccomp_filter *filter;
+#endif
+} seccomp_t;
 
 extern void __secure_computing(int);
 static inline void secure_computing(int this_syscall)
@@ -28,8 +49,7 @@ static inline int seccomp_mode(seccomp_t *s)
 
 #include <linux/errno.h>
 
-typedef struct { } seccomp_t;
-
+typedef struct seccomp_struct { } seccomp_t;
 #define secure_computing(x) do { } while (0)
 
 static inline long prctl_get_seccomp(void)
@@ -49,4 +69,48 @@ static inline int seccomp_mode(seccomp_t *s)
 
 #endif /* CONFIG_SECCOMP */
 
+#ifdef CONFIG_SECCOMP_FILTER
+
+
+extern long prctl_attach_seccomp_filter(char __user *);
+
+extern struct seccomp_filter *get_seccomp_filter(struct seccomp_filter *);
+extern void put_seccomp_filter(struct seccomp_filter *);
+
+extern int seccomp_test_filters(int);
+extern void seccomp_filter_log_failure(int);
+extern void seccomp_struct_fork(struct seccomp_struct *child,
+				const struct seccomp_struct *parent);
+
+static inline void seccomp_struct_init_task(struct seccomp_struct *seccomp)
+{
+	seccomp->mode = SECCOMP_MODE_DISABLED;
+	seccomp->filter = NULL;
+}
+
+/* No locking is needed here because the task_struct will
+ * have no parallel consumers.
+ */
+static inline void seccomp_struct_free_task(struct seccomp_struct *seccomp)
+{
+	put_seccomp_filter(seccomp->filter);
+	seccomp->filter = NULL;
+}
+
+#else  /* CONFIG_SECCOMP_FILTER */
+
+#include <linux/errno.h>
+
+struct seccomp_filter { };
+/* Macros consume the unused dereference by the caller. */
+#define seccomp_struct_init_task(_seccomp) do { } while (0);
+#define seccomp_struct_fork(_tsk, _orig) do { } while (0);
+#define seccomp_struct_free_task(_seccomp) do { } while (0);
+
+static inline long prctl_attach_seccomp_filter(char __user *a2)
+{
+	return -ENOSYS;
+}
+
+#endif  /* CONFIG_SECCOMP_FILTER */
 #endif /* _LINUX_SECCOMP_H */
diff --git a/kernel/Makefile b/kernel/Makefile
index e898c5b..0584090 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -79,6 +79,7 @@ obj-$(CONFIG_DETECT_HUNG_TASK) += hung_task.o
 obj-$(CONFIG_LOCKUP_DETECTOR) += watchdog.o
 obj-$(CONFIG_GENERIC_HARDIRQS) += irq/
 obj-$(CONFIG_SECCOMP) += seccomp.o
+obj-$(CONFIG_SECCOMP_FILTER) += seccomp_filter.o
 obj-$(CONFIG_RCU_TORTURE_TEST) += rcutorture.o
 obj-$(CONFIG_TREE_RCU) += rcutree.o
 obj-$(CONFIG_TREE_PREEMPT_RCU) += rcutree.o
diff --git a/kernel/fork.c b/kernel/fork.c
index da4a6a1..22f7ec1 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -34,6 +34,7 @@
 #include <linux/cgroup.h>
 #include <linux/security.h>
 #include <linux/hugetlb.h>
+#include <linux/seccomp.h>
 #include <linux/swap.h>
 #include <linux/syscalls.h>
 #include <linux/jiffies.h>
@@ -166,6 +167,7 @@ void free_task(struct task_struct *tsk)
 	free_thread_info(tsk->stack);
 	rt_mutex_debug_task_free(tsk);
 	ftrace_graph_exit_task(tsk);
+	seccomp_struct_free_task(&tsk->seccomp);
 	free_task_struct(tsk);
 }
 EXPORT_SYMBOL(free_task);
@@ -1089,6 +1091,7 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 		goto fork_out;
 
 	ftrace_graph_init_task(p);
+	seccomp_struct_init_task(&p->seccomp);
 
 	rt_mutex_init_task(p);
 
@@ -1375,6 +1378,7 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 	if (clone_flags & CLONE_THREAD)
 		threadgroup_fork_read_unlock(current);
 	perf_event_fork(p);
+	seccomp_struct_fork(&p->seccomp, &current->seccomp);
 	return p;
 
 bad_fork_free_pid:
diff --git a/kernel/seccomp.c b/kernel/seccomp.c
index 57d4b13..b87c1bc 100644
--- a/kernel/seccomp.c
+++ b/kernel/seccomp.c
@@ -36,7 +36,7 @@ void __secure_computing(int this_syscall)
 	int * syscall;
 
 	switch (mode) {
-	case 1:
+	case SECCOMP_MODE_STRICT:
 		syscall = mode1_syscalls;
 #ifdef CONFIG_COMPAT
 		if (is_compat_task())
@@ -47,6 +47,14 @@ void __secure_computing(int this_syscall)
 				return;
 		} while (*++syscall);
 		break;
+#ifdef CONFIG_SECCOMP_FILTER
+	case SECCOMP_MODE_FILTER:
+		if (seccomp_test_filters(this_syscall) == 0)
+			return;
+
+		seccomp_filter_log_failure(this_syscall);
+		break;
+#endif
 	default:
 		BUG();
 	}
diff --git a/kernel/seccomp_filter.c b/kernel/seccomp_filter.c
new file mode 100644
index 0000000..b1c3f60
--- /dev/null
+++ b/kernel/seccomp_filter.c
@@ -0,0 +1,642 @@
+/*
+ * linux/kernel/seccomp_filter.c
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Copyright (C) 2012 The Chromium OS Authors <chromium-os-dev@chromium.org>
+ *
+ * Extends linux/kernel/seccomp.c to allow tasks to install system call
+ * filters using a Berkeley Packet Filter program which is executed over
+ * struct user_regs_struct.
+ */
+
+#include <linux/capability.h>
+#include <linux/compat.h>
+#include <linux/err.h>
+#include <linux/errno.h>
+#include <linux/rculist.h>
+#include <linux/filter.h>
+#include <linux/kallsyms.h>
+#include <linux/kref.h>
+#include <linux/module.h>
+#include <linux/pid.h>
+#include <linux/prctl.h>
+#include <linux/ptrace.h>
+#include <linux/ratelimit.h>
+#include <linux/reciprocal_div.h>
+#include <linux/regset.h>
+#include <linux/seccomp.h>
+#include <linux/security.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/uaccess.h>
+#include <linux/user.h>
+
+
+/**
+ * struct seccomp_filter - container for seccomp BPF programs
+ *
+ * @usage: reference count to manage the object lifetime.
+ *         get/put helpers should be used when accessing an instance
+ *         outside of a lifetime-guarded section.  In general, this
+ *         is only needed for handling filters shared across tasks.
+ * @parent: pointer to the ancestor which this filter will be composed with.
+ * @flags: provide information about filter from creation time.
+ * @personality: personality of the process at filter creation time.
+ * @insns: the BPF program instructions to evaluate
+ * @count: the number of instructions in the program.
+ *
+ * seccomp_filter objects should never be modified after being attached
+ * to a task_struct (other than @usage).
+ */
+struct seccomp_filter {
+	struct kref usage;
+	struct seccomp_filter *parent;
+	struct {
+		uint32_t compat:1,		/* CONFIG_COMPAT */
+			 __reserved:31;
+	} flags;
+	int personality;
+	unsigned short count;  /* Instruction count */
+	struct sock_filter insns[0];
+};
+
+static unsigned int seccomp_run_filter(const u8 *buf,
+				       const size_t buflen,
+				       const struct sock_filter *);
+
+/**
+ * seccomp_filter_alloc - allocates a new filter object
+ * @padding: size of the insns[0] array in bytes
+ *
+ * The @padding should be a multiple of
+ * sizeof(struct sock_filter).
+ *
+ * Returns ERR_PTR on error or an allocated object.
+ */
+static struct seccomp_filter *seccomp_filter_alloc(unsigned long padding)
+{
+	struct seccomp_filter *f;
+	unsigned long bpf_blocks = padding / sizeof(struct sock_filter);
+
+	/* Drop oversized requests. */
+	if (bpf_blocks == 0 || bpf_blocks > BPF_MAXINSNS)
+		return ERR_PTR(-EINVAL);
+
+	/* Padding should always be in sock_filter increments. */
+	if (padding % sizeof(struct sock_filter))
+		return ERR_PTR(-EINVAL);
+
+	f = kzalloc(sizeof(struct seccomp_filter) + padding, GFP_KERNEL);
+	if (!f)
+		return ERR_PTR(-ENOMEM);
+	kref_init(&f->usage);
+	f->count = bpf_blocks;
+	return f;
+}
+
+/**
+ * seccomp_filter_free - frees the allocated filter.
+ * @filter: NULL or live object to be completely destructed.
+ */
+static void seccomp_filter_free(struct seccomp_filter *filter)
+{
+	if (!filter)
+		return;
+	put_seccomp_filter(filter->parent);
+	kfree(filter);
+}
+
+static void __put_seccomp_filter(struct kref *kref)
+{
+	struct seccomp_filter *orig =
+		container_of(kref, struct seccomp_filter, usage);
+	seccomp_filter_free(orig);
+}
+
+void seccomp_filter_log_failure(int syscall)
+{
+	pr_info("%s[%d]: system call %d blocked at 0x%lx\n",
+		current->comm, task_pid_nr(current), syscall,
+		KSTK_EIP(current));
+}
+
+/* put_seccomp_filter - decrements the ref count of @orig and may free. */
+void put_seccomp_filter(struct seccomp_filter *orig)
+{
+	if (!orig)
+		return;
+	kref_put(&orig->usage, __put_seccomp_filter);
+}
+
+/* get_seccomp_filter - increments the reference count of @orig. */
+struct seccomp_filter *get_seccomp_filter(struct seccomp_filter *orig)
+{
+	if (!orig)
+		return NULL;
+	kref_get(&orig->usage);
+	return orig;
+}
+
+static int seccomp_check_personality(struct seccomp_filter *filter)
+{
+	if (filter->personality != current->personality)
+		return -EACCES;
+#ifdef CONFIG_COMPAT
+	if (filter->flags.compat != (!!(is_compat_task())))
+		return -EACCES;
+#endif
+	return 0;
+}
+
+static const struct user_regset *
+find_prstatus(const struct user_regset_view *view)
+{
+	const struct user_regset *regset;
+	int n;
+
+	/* Skip 0. */
+	for (n = 1; n < view->n; ++n) {
+		regset = view->regsets + n;
+		if (regset->core_note_type == NT_PRSTATUS)
+			return regset;
+	}
+
+	return NULL;
+}
+
+/**
+ * seccomp_get_regs - returns a pointer to struct user_regs_struct
+ * @scratch: preallocated storage of size @available
+ * @available: pointer to the size of scratch.
+ *
+ * Returns NULL if the registers cannot be acquired or copied.
+ * Returns a populated pointer to @scratch by default.
+ * Otherwise, returns a pointer to a a u8 array containing the struct
+ * user_regs_struct appropriate for the task personality.  The pointer
+ * may be to the beginning of @scratch or to an externally managed data
+ * structure.  On success, @available should be updated with the
+ * valid region size of the returned pointer.
+ *
+ * If the architecture overrides the linkage, then the pointer may pointer to
+ * another location.
+ */
+__weak u8 *seccomp_get_regs(u8 *scratch, size_t *available)
+{
+	/*
+	 *regset is usually returned based on task personality, not current
+	 * system call convention.  This behavior makes it unsafe to execute
+	 * BPF programs over regviews if is_compat_task or the personality
+	 * have changed since the program was installed.
+	 */
+	const struct user_regset_view *view = task_user_regset_view(current);
+	const struct user_regset *regset = &view->regsets[0];
+	size_t scratch_size = *available;
+	if (regset->core_note_type != NT_PRSTATUS) {
+		/* The architecture should override this method for speed. */
+		regset = find_prstatus(view);
+		if (!regset)
+			return NULL;
+	}
+	*available = regset->n * regset->size;
+	/* Make sure the scratch space isn't exceeded. */
+	if (*available > scratch_size)
+		*available = scratch_size;
+	if (regset->get(current, regset, 0, *available, scratch, NULL))
+		return NULL;
+	return scratch;
+}
+
+/**
+ * seccomp_test_filters - tests 'current' against the given syscall
+ * @syscall: number of the system call to test
+ *
+ * Returns 0 on ok and non-zero on error/failure.
+ */
+int seccomp_test_filters(int syscall)
+{
+	struct seccomp_filter *filter;
+	u8 regs_tmp[sizeof(struct user_regs_struct)], *regs;
+	size_t regs_size = sizeof(struct user_regs_struct);
+	int ret = -EACCES;
+
+	filter = current->seccomp.filter; /* uses task ref */
+	if (!filter)
+		goto out;
+
+	/*
+	 * All filters in the list are required to share the same system call
+	 * convention so only the first filter is ever checked.
+	 */
+	if (seccomp_check_personality(filter))
+		goto out;
+
+	/*
+	 * Grab the user_regs_struct.  Normally, regs == &regs_tmp, but
+	 * that is not mandatory.  E.g., it may return a point to
+	 * task_pt_regs(current).  NULL checking is mandatory.
+	 */
+	regs = seccomp_get_regs(regs_tmp, &regs_size);
+	if (!regs)
+		goto out;
+
+	/* Only allow a system call if it is allowed in all ancestors. */
+	ret = 0;
+	for ( ; filter != NULL; filter = filter->parent) {
+		/* Allowed if return value is the size of the data supplied. */
+		if (seccomp_run_filter(regs, regs_size, filter->insns) !=
+		    regs_size)
+			ret = -EACCES;
+	}
+out:
+	return ret;
+}
+
+/**
+ * seccomp_attach_filter: Attaches a seccomp filter to current.
+ * @fprog: BPF program to install
+ *
+ * Context: User context only. This function may sleep on allocation and
+ *          operates on current. current must be attempting a system call
+ *          when this is called (usually prctl).
+ *
+ * This function may be called repeatedly to install additional filters.
+ * Every filter successfully installed will be evaluated (in reverse order)
+ * for each system call the thread makes.
+ *
+ * Returns 0 on success or an errno on failure.
+ */
+long seccomp_attach_filter(struct sock_fprog *fprog)
+{
+	struct seccomp_filter *filter = NULL;
+	/* Note, len is a short so overflow should be impossible. */
+	unsigned long fp_size = fprog->len * sizeof(struct sock_filter);
+	long ret = -EPERM;
+
+	/* Allocate a new seccomp_filter */
+	filter = seccomp_filter_alloc(fp_size);
+	if (IS_ERR(filter)) {
+		ret = PTR_ERR(filter);
+		goto out;
+	}
+
+	/* Lock the process personality and calling convention. */
+#ifdef CONFIG_COMPAT
+	if (is_compat_task())
+		filter->flags.compat = 1;
+#endif
+	filter->personality = current->personality;
+
+	/*
+	 * If a process lacks CAP_SYS_ADMIN in its namespace, force
+	 * this process and all descendents to run with no_new_privs.
+	 * A privileged process will need to set this bit independently,
+	 * if desired.
+	 */
+	if (security_real_capable_noaudit(current, current_user_ns(),
+					  CAP_SYS_ADMIN) != 0)
+		current->no_new_privs = 1;
+
+	/* Copy the instructions from fprog. */
+	ret = -EFAULT;
+	if (copy_from_user(filter->insns, fprog->filter, fp_size))
+		goto out;
+
+	/* Check the fprog */
+	ret = sk_chk_filter(filter->insns, filter->count);
+	if (ret)
+		goto out;
+
+	/*
+	 * If there is an existing filter, make it the parent
+	 * and reuse the existing task-based ref.
+	 */
+	filter->parent = current->seccomp.filter;
+
+	/* Force all filters to use one system call convention. */
+	ret = -EINVAL;
+	if (filter->parent) {
+		if (filter->parent->flags.compat != filter->flags.compat)
+			goto out;
+		if (filter->parent->personality != filter->personality)
+			goto out;
+	}
+
+	/*
+	 * Double claim the new filter so we can release it below simplifying
+	 * the error paths earlier.
+	 */
+	ret = 0;
+	get_seccomp_filter(filter);
+	current->seccomp.filter = filter;
+	/* Engage seccomp if it wasn't. This doesn't use PR_SET_SECCOMP. */
+	if (current->seccomp.mode == SECCOMP_MODE_DISABLED) {
+		current->seccomp.mode = SECCOMP_MODE_FILTER;
+		set_thread_flag(TIF_SECCOMP);
+	}
+
+out:
+	put_seccomp_filter(filter);  /* for get or task, on err */
+	return ret;
+}
+
+long prctl_attach_seccomp_filter(char __user *user_filter)
+{
+	struct sock_fprog fprog;
+	long ret = -EINVAL;
+
+	ret = -EFAULT;
+	if (!user_filter)
+		goto out;
+
+	if (copy_from_user(&fprog, user_filter, sizeof(fprog)))
+		goto out;
+
+	ret = seccomp_attach_filter(&fprog);
+out:
+	return ret;
+}
+
+/**
+ * seccomp_struct_fork: manages inheritance on fork
+ * @child: forkee's seccomp_struct
+ * @parent: forker's seccomp_struct
+ *
+ * Ensures that @child inherits seccomp mode and state iff
+ * seccomp filtering is in use.
+ */
+void seccomp_struct_fork(struct seccomp_struct *child,
+			 const struct seccomp_struct *parent)
+{
+	child->mode = parent->mode;
+	if (parent->mode != SECCOMP_MODE_FILTER)
+		return;
+	child->filter = get_seccomp_filter(parent->filter);
+}
+
+/**
+ * load_pointer: checks and returns a pointer to the requested offset
+ * @buf: u8 array to index into
+ * @buflen: length of the @buf array
+ * @offset: offset to return data from
+ * @size: size of the data to retrieve at offset
+ * @unused: placeholder which net/core/filter.c uses for for temporary
+ *          storage.  Ideally, the two code paths can be merged.
+ *
+ * Returns a pointer to the BPF evaluator after checking the offset and size
+ * boundaries.
+ */
+static const void *load_pointer(const u8 *buf, size_t buflen,
+				int offset, size_t size,
+				void *unused)
+{
+	if (offset >= buflen)
+		goto fail;
+	if (offset < 0)
+		goto fail;
+	if (size > buflen - offset)
+		goto fail;
+	return buf + offset;
+fail:
+	return NULL;
+}
+
+/**
+ * seccomp_run_filter - evaluate BPF (over user_regs_struct)
+ *	@buf: buffer to execute the filter over
+ *	@buflen: length of the buffer
+ *	@fentry: filter to apply
+ *
+ * Decode and apply filter instructions to the buffer.
+ * Return length to keep, 0 for none. @buf is a regset we are
+ * filtering, @filter is the array of filter instructions.
+ * Because all jumps are guaranteed to be before last instruction,
+ * and last instruction guaranteed to be a RET, we dont need to check
+ * flen.
+ *
+ * See core/net/filter.c as this is nearly an exact copy.
+ * At some point, it would be nice to merge them to take advantage of
+ * optimizations (like JIT).
+ *
+ * A successful filter must return the full length of the data. Anything less
+ * will currently result in a seccomp failure.  In the future, it may be
+ * possible to use that for hard filtering registers on the fly so it is
+ * ideal for consumers to return 0 on intended failure.
+ */
+static unsigned int seccomp_run_filter(const u8 *buf,
+				       const size_t buflen,
+				       const struct sock_filter *fentry)
+{
+	const void *ptr;
+	u32 A = 0;			/* Accumulator */
+	u32 X = 0;			/* Index Register */
+	u32 mem[BPF_MEMWORDS];		/* Scratch Memory Store */
+	u32 tmp;
+	int k;
+
+	/*
+	 * Process array of filter instructions.
+	 */
+	for (;; fentry++) {
+#if defined(CONFIG_X86_32)
+#define	K (fentry->k)
+#else
+		const u32 K = fentry->k;
+#endif
+
+		switch (fentry->code) {
+		case BPF_S_ALU_ADD_X:
+			A += X;
+			continue;
+		case BPF_S_ALU_ADD_K:
+			A += K;
+			continue;
+		case BPF_S_ALU_SUB_X:
+			A -= X;
+			continue;
+		case BPF_S_ALU_SUB_K:
+			A -= K;
+			continue;
+		case BPF_S_ALU_MUL_X:
+			A *= X;
+			continue;
+		case BPF_S_ALU_MUL_K:
+			A *= K;
+			continue;
+		case BPF_S_ALU_DIV_X:
+			if (X == 0)
+				return 0;
+			A /= X;
+			continue;
+		case BPF_S_ALU_DIV_K:
+			A = reciprocal_divide(A, K);
+			continue;
+		case BPF_S_ALU_AND_X:
+			A &= X;
+			continue;
+		case BPF_S_ALU_AND_K:
+			A &= K;
+			continue;
+		case BPF_S_ALU_OR_X:
+			A |= X;
+			continue;
+		case BPF_S_ALU_OR_K:
+			A |= K;
+			continue;
+		case BPF_S_ALU_LSH_X:
+			A <<= X;
+			continue;
+		case BPF_S_ALU_LSH_K:
+			A <<= K;
+			continue;
+		case BPF_S_ALU_RSH_X:
+			A >>= X;
+			continue;
+		case BPF_S_ALU_RSH_K:
+			A >>= K;
+			continue;
+		case BPF_S_ALU_NEG:
+			A = -A;
+			continue;
+		case BPF_S_JMP_JA:
+			fentry += K;
+			continue;
+		case BPF_S_JMP_JGT_K:
+			fentry += (A > K) ? fentry->jt : fentry->jf;
+			continue;
+		case BPF_S_JMP_JGE_K:
+			fentry += (A >= K) ? fentry->jt : fentry->jf;
+			continue;
+		case BPF_S_JMP_JEQ_K:
+			fentry += (A == K) ? fentry->jt : fentry->jf;
+			continue;
+		case BPF_S_JMP_JSET_K:
+			fentry += (A & K) ? fentry->jt : fentry->jf;
+			continue;
+		case BPF_S_JMP_JGT_X:
+			fentry += (A > X) ? fentry->jt : fentry->jf;
+			continue;
+		case BPF_S_JMP_JGE_X:
+			fentry += (A >= X) ? fentry->jt : fentry->jf;
+			continue;
+		case BPF_S_JMP_JEQ_X:
+			fentry += (A == X) ? fentry->jt : fentry->jf;
+			continue;
+		case BPF_S_JMP_JSET_X:
+			fentry += (A & X) ? fentry->jt : fentry->jf;
+			continue;
+		case BPF_S_LD_W_ABS:
+			k = K;
+load_w:
+			ptr = load_pointer(buf, buflen, k, 4, &tmp);
+			if (ptr != NULL) {
+				/*
+				 * Note, unlike on network data, values are not
+				 * byte swapped.
+				 */
+				A = *(const u32 *)ptr;
+				continue;
+			}
+			return 0;
+		case BPF_S_LD_H_ABS:
+			k = K;
+load_h:
+			ptr = load_pointer(buf, buflen, k, 2, &tmp);
+			if (ptr != NULL) {
+				A = *(const u16 *)ptr;
+				continue;
+			}
+			return 0;
+		case BPF_S_LD_B_ABS:
+			k = K;
+load_b:
+			ptr = load_pointer(buf, buflen, k, 1, &tmp);
+			if (ptr != NULL) {
+				A = *(const u8 *)ptr;
+				continue;
+			}
+			return 0;
+		case BPF_S_LD_W_LEN:
+			A = buflen;
+			continue;
+		case BPF_S_LDX_W_LEN:
+			X = buflen;
+			continue;
+		case BPF_S_LD_W_IND:
+			k = X + K;
+			goto load_w;
+		case BPF_S_LD_H_IND:
+			k = X + K;
+			goto load_h;
+		case BPF_S_LD_B_IND:
+			k = X + K;
+			goto load_b;
+		case BPF_S_LDX_B_MSH:
+			ptr = load_pointer(buf, buflen, K, 1, &tmp);
+			if (ptr != NULL) {
+				X = (*(u8 *)ptr & 0xf) << 2;
+				continue;
+			}
+			return 0;
+		case BPF_S_LD_IMM:
+			A = K;
+			continue;
+		case BPF_S_LDX_IMM:
+			X = K;
+			continue;
+		case BPF_S_LD_MEM:
+			A = mem[K];
+			continue;
+		case BPF_S_LDX_MEM:
+			X = mem[K];
+			continue;
+		case BPF_S_MISC_TAX:
+			X = A;
+			continue;
+		case BPF_S_MISC_TXA:
+			A = X;
+			continue;
+		case BPF_S_RET_K:
+			return K;
+		case BPF_S_RET_A:
+			return A;
+		case BPF_S_ST:
+			mem[K] = A;
+			continue;
+		case BPF_S_STX:
+			mem[K] = X;
+			continue;
+		case BPF_S_ANC_PROTOCOL:
+		case BPF_S_ANC_PKTTYPE:
+		case BPF_S_ANC_IFINDEX:
+		case BPF_S_ANC_MARK:
+		case BPF_S_ANC_QUEUE:
+		case BPF_S_ANC_HATYPE:
+		case BPF_S_ANC_RXHASH:
+		case BPF_S_ANC_CPU:
+		case BPF_S_ANC_NLATTR:
+		case BPF_S_ANC_NLATTR_NEST:
+			/* ignored */
+			continue;
+		default:
+			WARN_RATELIMIT(1, "Unknown code:%u jt:%u tf:%u k:%u\n",
+				       fentry->code, fentry->jt,
+				       fentry->jf, fentry->k);
+			return 0;
+		}
+	}
+
+	return 0;
+}
diff --git a/kernel/sys.c b/kernel/sys.c
index 481611f..77f2eda 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -1783,6 +1783,10 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
 		case PR_SET_SECCOMP:
 			error = prctl_set_seccomp(arg2);
 			break;
+		case PR_ATTACH_SECCOMP_FILTER:
+			error = prctl_attach_seccomp_filter((char __user *)
+								arg2);
+			break;
 		case PR_GET_TSC:
 			error = GET_TSC_CTL(arg2);
 			break;
diff --git a/security/Kconfig b/security/Kconfig
index 51bd5a0..6491bb1 100644
--- a/security/Kconfig
+++ b/security/Kconfig
@@ -84,6 +84,22 @@ config SECURITY_DMESG_RESTRICT
 
 	  If you are unsure how to answer this question, answer N.
 
+config SECCOMP_FILTER
+	bool "Enable seccomp-based system call filtering"
+	select SECCOMP
+	help
+	  This option provide support for limiting the accessibility of
+	  systems calls at a task-level using a dynamically defined policy.
+	
+	  System call filtering policy is expressed by the user using
+	  a Berkeley Packet Filter program.  The program is attached using
+	  prctl(2).  For every system call the task makes, its register
+	  state will be evaluated by the attached filter program.  The
+	  result determines if the system call may proceed or if the
+	  task should be terminated.
+
+	  See Documentation/prctl/seccomp_filter.txt for more detail.
+
 config SECURITY
 	bool "Enable different security models"
 	depends on SYSFS
-- 
1.7.5.4

[PATCH v4 2/2] Documentation: prctl/seccomp_filter

[Will Drewry]

Documents how system call filtering using Berkeley Packet
Filter programs works and how it may be used.
Includes an example for x86 (32-bit).

v4: - update for no_new_privs use
    - minor tweaks
v3: - call out BPF <-> Berkeley Packet Filter (rdunlap@xenotime.net)
    - document use of tentative always-unprivileged
    - guard sample compilation for i386 and x86_64
v2: - move code to samples (corbet@lwn.net)

Signed-off-by: Will Drewry <wad@chromium.org>
---
 Documentation/prctl/seccomp_filter.txt |   94 ++++++++++++++++++++++++++++++++
 samples/Makefile                       |    2 +-
 samples/seccomp/Makefile               |   18 ++++++
 samples/seccomp/bpf-example.c          |   74 +++++++++++++++++++++++++
 4 files changed, 187 insertions(+), 1 deletions(-)
 create mode 100644 Documentation/prctl/seccomp_filter.txt
 create mode 100644 samples/seccomp/Makefile
 create mode 100644 samples/seccomp/bpf-example.c

diff --git a/Documentation/prctl/seccomp_filter.txt b/Documentation/prctl/seccomp_filter.txt
new file mode 100644
index 0000000..0cf8120
--- /dev/null
+++ b/Documentation/prctl/seccomp_filter.txt
@@ -0,0 +1,94 @@
+		Seccomp filtering
+		=================
+
+Introduction
+------------
+
+A large number of system calls are exposed to every userland process
+with many of them going unused for the entire lifetime of the process.
+As system calls change and mature, bugs are found and eradicated.  A
+certain subset of userland applications benefit by having a reduced set
+of available system calls.  The resulting set reduces the total kernel
+surface exposed to the application.  System call filtering is meant for
+use with those applications.
+
+Seccomp filtering provides a means for a process to specify a filter
+for incoming system calls.  The filter is expressed as a Berkeley Packet
+Filter (BPF) program, as with socket filters, except that the data
+operated on is the current user_regs_struct.  This allows for expressive
+filtering of system calls using the pre-existing system call ABI and
+using a filter program language with a long history of being exposed to
+userland.  Additionally, BPF makes it impossible for users of seccomp to
+fall prey to time-of-check-time-of-use (TOCTOU) attacks that are common
+in system call interposition frameworks because the evaluated data is
+solely register state just after system call entry.
+
+What it isn't
+-------------
+
+System call filtering isn't a sandbox.  It provides a clearly defined
+mechanism for minimizing the exposed kernel surface.  Beyond that,
+policy for logical behavior and information flow should be managed with
+a combinations of other system hardening techniques and, potentially, a
+LSM of your choosing.  Expressive, dynamic filters provide further options down
+this path (avoiding pathological sizes or selecting which of the multiplexed
+system calls in socketcall() is allowed, for instance) which could be
+construed, incorrectly, as a more complete sandboxing solution.
+
+Usage
+-----
+
+An additional seccomp mode is added, but they are not directly set by the
+consuming process.  The new mode, '2', is only available if
+CONFIG_SECCOMP_FILTER is set and enabled using prctl with the
+PR_ATTACH_SECCOMP_FILTER argument.
+
+Interacting with seccomp filters is done using one prctl(2) call.
+
+PR_ATTACH_SECCOMP_FILTER:
+	Allows the specification of a new filter using a BPF program.
+	The BPF program will be executed over a user_regs_struct data
+	reflecting system call time except with the system call number
+	resident in orig_[register].  To allow a system call, the size
+	of the data must be returned.  At present, all other return values
+	result in the system call being blocked, but it is recommended to
+	return 0 in those cases.  This will allow for future custom return
+	values to be introduced, if ever desired.
+
+	Usage:
+		prctl(PR_ATTACH_SECCOMP_FILTER, prog);
+
+	The 'prog' argument is a pointer to a struct sock_fprog which will
+	contain the filter program.  If the program is invalid, the call
+	will return -1 and set errno to -EINVAL.
+
+	The struct user_regs_struct the @prog will see is based on the
+	personality of the task at the time of this prctl call.  Additionally,
+	is_compat_task is also tracked for the @prog.  This means that once set
+	the calling task will have all of its system calls blocked if it
+	switches its system call ABI (via personality or other means).
+
+	If fork/clone and execve are allowed by @prog, any child processes will
+	be constrained to the same filters and syscal call ABI as the parent.
+
+	When called from an unprivileged process (lacking CAP_SYS_ADMIN), the
+	"no_new_privs" bit is enabled for the process.
+
+	Additionally, if prctl(2) is allowed by the attached filter,
+	additional filters may be layered on which will increase evaluation
+	time, but allow for further decreasing the attack surface during
+	execution of a process.
+
+The above call returns 0 on success and non-zero on error.
+
+Example
+-------
+
+samples/seccomp-bpf-example.c shows an example process that allows read from stdin,
+write to stdout/err, exit and signal returns for 32-bit x86.
+
+Adding architecture support
+-----------------------
+
+Any platform with seccomp support will support seccomp filters
+as long as CONFIG_SECCOMP_FILTER is enabled.
diff --git a/samples/Makefile b/samples/Makefile
index 6280817..f29b19c 100644
--- a/samples/Makefile
+++ b/samples/Makefile
@@ -1,4 +1,4 @@
 # Makefile for Linux samples code
 
 obj-$(CONFIG_SAMPLES)	+= kobject/ kprobes/ tracepoints/ trace_events/ \
-			   hw_breakpoint/ kfifo/ kdb/ hidraw/
+			   hw_breakpoint/ kfifo/ kdb/ hidraw/ seccomp/
diff --git a/samples/seccomp/Makefile b/samples/seccomp/Makefile
new file mode 100644
index 0000000..cdf0282
--- /dev/null
+++ b/samples/seccomp/Makefile
@@ -0,0 +1,18 @@
+# This sample is x86-only.
+ifeq ($(filter-out x86_64 i386,$(KBUILD_BUILDHOST)),)
+# kbuild trick to avoid linker error. Can be omitted if a module is built.
+obj- := dummy.o
+
+# List of programs to build
+hostprogs-y := bpf-example
+bpf-example-objs := bpf-example.o
+
+# Tell kbuild to always build the programs
+always := $(hostprogs-y)
+
+HOSTCFLAGS_bpf-example.o += -I$(objtree)/usr/include
+ifeq ($(KBUILD_BUILDHOST),x86_64)
+HOSTCFLAGS_bpf-example.o += -m32
+HOSTLOADLIBES_bpf-example += -m32
+endif
+endif  # host arch is x86
diff --git a/samples/seccomp/bpf-example.c b/samples/seccomp/bpf-example.c
new file mode 100644
index 0000000..f98b70a
--- /dev/null
+++ b/samples/seccomp/bpf-example.c
@@ -0,0 +1,74 @@
+/*
+ * Seccomp BPF example
+ *
+ * Copyright (c) 2012 The Chromium OS Authors <chromium-os-dev@chromium.org>
+ * Author: Will Drewry <wad@chromium.org>
+ *
+ * The code may be used by anyone for any purpose,
+ * and can serve as a starting point for developing
+ * applications using prctl(PR_ATTACH_SECCOMP_FILTER).
+ */
+
+#include <asm/unistd.h>
+#include <linux/filter.h>
+#include <stdio.h>
+#include <stddef.h>
+#include <sys/prctl.h>
+#include <sys/user.h>
+#include <unistd.h>
+
+#ifndef PR_ATTACH_SECCOMP_FILTER
+#	define PR_ATTACH_SECCOMP_FILTER 36
+#endif
+
+#define regoffset(_reg) (offsetof(struct user_regs_struct, _reg))
+static int install_filter(void)
+{
+	struct sock_filter filter[] = {
+		/* Grab the system call number */
+		BPF_STMT(BPF_LD+BPF_W+BPF_IND, regoffset(orig_eax)),
+		/* Jump table for the allowed syscalls */
+		BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, __NR_rt_sigreturn, 10, 0),
+		BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, __NR_sigreturn, 9, 0),
+		BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, __NR_exit_group, 8, 0),
+		BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, __NR_exit, 7, 0),
+		BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, __NR_read, 1, 0),
+		BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, __NR_write, 2, 6),
+
+		/* Check that read is only using stdin. */
+		BPF_STMT(BPF_LD+BPF_W+BPF_IND, regoffset(ebx)),
+		BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, STDIN_FILENO, 3, 4),
+
+		/* Check that write is only using stdout/stderr */
+		BPF_STMT(BPF_LD+BPF_W+BPF_IND, regoffset(ebx)),
+		BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, STDOUT_FILENO, 1, 0),
+		BPF_JUMP(BPF_JMP+BPF_JEQ+BPF_K, STDERR_FILENO, 0, 1),
+
+		/* Put the "accept" value in A */
+		BPF_STMT(BPF_LD+BPF_W+BPF_LEN, 0),
+
+		BPF_STMT(BPF_RET+BPF_A,0),
+	};
+	struct sock_fprog prog = {
+		.len = (unsigned short)(sizeof(filter)/sizeof(filter[0])),
+		.filter = filter,
+	};
+	if (prctl(PR_ATTACH_SECCOMP_FILTER, &prog)) {
+		perror("prctl");
+		return 1;
+	}
+	return 0;
+}
+
+#define payload(_c) _c, sizeof(_c)
+int main(int argc, char **argv) {
+	char buf[4096];
+	ssize_t bytes = 0;
+	if (install_filter())
+		return 1;
+	syscall(__NR_write, STDOUT_FILENO, payload("OHAI! WHAT IS YOUR NAME? "));
+	bytes = syscall(__NR_read, STDIN_FILENO, buf, sizeof(buf));
+	syscall(__NR_write, STDOUT_FILENO, payload("HELLO, "));
+	syscall(__NR_write, STDOUT_FILENO, buf, bytes);
+	return 0;
+}
-- 
1.7.5.4