Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752016AbaANU7b (ORCPT ); Tue, 14 Jan 2014 15:59:31 -0500 Received: from mail-pd0-f173.google.com ([209.85.192.173]:50516 "EHLO mail-pd0-f173.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751274AbaANU71 (ORCPT ); Tue, 14 Jan 2014 15:59:27 -0500 MIME-Version: 1.0 In-Reply-To: References: <1389645028-17157-1-git-send-email-wad@chromium.org> <1389645028-17157-2-git-send-email-wad@chromium.org> <52D47890.2060901@mit.edu> Date: Tue, 14 Jan 2014 14:59:26 -0600 Message-ID: Subject: Re: [PATCH 2/2] sys, seccomp: add PR_SECCOMP_EXT and SECCOMP_EXT_ACT_TSYNC From: Will Drewry To: Andy Lutomirski Cc: LKML , Nicolas Schichan , Kees Cook , James Morris , Andrew Morton , Oleg Nesterov , Alexander Viro , Julien Tinnes Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Jan 14, 2014 at 2:24 PM, Andy Lutomirski wrote: > On Tue, Jan 14, 2014 at 10:59 AM, Will Drewry wrote: >> On Mon, Jan 13, 2014 at 5:36 PM, Andy Lutomirski wrote: >>> On 01/13/2014 12:30 PM, Will Drewry wrote: >>>> Applying restrictive seccomp filter programs to large or diverse >>>> codebases often requires handling threads which may be started early in >>>> the process lifetime (e.g., by code that is linked in). While it is >>>> possible to apply permissive programs prior to process start up, it is >>>> difficult to further restrict the kernel ABI to those threads after that >>>> point. >>>> >>>> This change adds a new seccomp "extension" for synchronizing thread >>>> group seccomp filters and a prctl() for accessing that functionality. >>>> The need for the added prctl() is due to the lack of reserved arguments >>>> in PR_SET_SECCOMP. >>>> >>>> When prctl(PR_SECCOMP_EXT, SECCOMP_EXT_ACT_TSYNC, 0, 0) is called, it >>>> will attempt to synchronize all threads in current's threadgroup to its >>>> seccomp filter program. This is possible iff all threads are using a >>>> filter that is an ancestor to the filter current is attempting to >>>> synchronize to. NULL filters (where the task is running as >>>> SECCOMP_MODE_NONE) are also treated as ancestors allowing threads to be >>>> transitioned into SECCOMP_MODE_FILTER. On success, 0 is returned. On >>>> failure, the pid of one of the failing threads will be returned. >>>> >>>> Suggested-by: Julien Tinnes >>>> Signed-off-by: Will Drewry >>>> --- >>>> include/linux/seccomp.h | 7 +++ >>>> include/uapi/linux/prctl.h | 6 ++ >>>> include/uapi/linux/seccomp.h | 6 ++ >>>> kernel/seccomp.c | 128 ++++++++++++++++++++++++++++++++++++++++++ >>>> kernel/sys.c | 3 + >>>> 5 files changed, 150 insertions(+) >>>> >>>> diff --git a/include/linux/seccomp.h b/include/linux/seccomp.h >>>> index 85c0895..3163db6 100644 >>>> --- a/include/linux/seccomp.h >>>> +++ b/include/linux/seccomp.h >>>> @@ -77,6 +77,8 @@ static inline int seccomp_mode(struct seccomp *s) >>>> extern void put_seccomp_filter(struct task_struct *tsk); >>>> extern void get_seccomp_filter(struct task_struct *tsk); >>>> extern u32 seccomp_bpf_load(int off); >>>> +extern long prctl_seccomp_ext(unsigned long, unsigned long, >>>> + unsigned long, unsigned long); >>>> #else /* CONFIG_SECCOMP_FILTER */ >>>> static inline void put_seccomp_filter(struct task_struct *tsk) >>>> { >>>> @@ -86,5 +88,10 @@ static inline void get_seccomp_filter(struct task_struct *tsk) >>>> { >>>> return; >>>> } >>>> +static inline long prctl_seccomp_ext(unsigned long arg2, unsigned long arg3, >>>> + unsigned long arg4, unsigned long arg5) >>>> +{ >>>> + return -EINVAL; >>>> +} >>>> #endif /* CONFIG_SECCOMP_FILTER */ >>>> #endif /* _LINUX_SECCOMP_H */ >>>> diff --git a/include/uapi/linux/prctl.h b/include/uapi/linux/prctl.h >>>> index 289760f..5dcd5d3 100644 >>>> --- a/include/uapi/linux/prctl.h >>>> +++ b/include/uapi/linux/prctl.h >>>> @@ -149,4 +149,10 @@ >>>> >>>> #define PR_GET_TID_ADDRESS 40 >>>> >>>> +/* >>>> + * Access seccomp extensions >>>> + * See Documentation/prctl/seccomp_filter.txt for more details. >>>> + */ >>>> +#define PR_SECCOMP_EXT 41 >>>> + >>>> #endif /* _LINUX_PRCTL_H */ >>>> diff --git a/include/uapi/linux/seccomp.h b/include/uapi/linux/seccomp.h >>>> index ac2dc9f..49b5279 100644 >>>> --- a/include/uapi/linux/seccomp.h >>>> +++ b/include/uapi/linux/seccomp.h >>>> @@ -10,6 +10,12 @@ >>>> #define SECCOMP_MODE_STRICT 1 /* uses hard-coded filter. */ >>>> #define SECCOMP_MODE_FILTER 2 /* uses user-supplied filter. */ >>>> >>>> +/* Valid extension types as arg2 for prctl(PR_SECCOMP_EXT) */ >>>> +#define SECCOMP_EXT_ACT 1 >>>> + >>>> +/* Valid extension actions as arg3 to prctl(PR_SECCOMP_EXT, SECCOMP_EXT_ACT) */ >>>> +#define SECCOMP_EXT_ACT_TSYNC 1 /* attempt to synchronize thread filters */ >>>> + >>>> /* >>>> * All BPF programs must return a 32-bit value. >>>> * The bottom 16-bits are for optional return data. >>>> diff --git a/kernel/seccomp.c b/kernel/seccomp.c >>>> index 71512e4..8a0de7b 100644 >>>> --- a/kernel/seccomp.c >>>> +++ b/kernel/seccomp.c >>>> @@ -24,6 +24,7 @@ >>>> #ifdef CONFIG_SECCOMP_FILTER >>>> #include >>>> #include >>>> +#include >>>> #include >>>> #include >>>> #include >>>> @@ -220,6 +221,108 @@ static u32 seccomp_run_filters(int syscall) >>>> return ret; >>>> } >>>> >>>> +/* Returns 1 if the candidate is an ancestor. */ >>>> +static int is_ancestor(struct seccomp_filter *candidate, >>>> + struct seccomp_filter *child) >>>> +{ >>>> + /* NULL is the root ancestor. */ >>>> + if (candidate == NULL) >>>> + return 1; >>>> + for (; child; child = child->prev) >>>> + if (child == candidate) >>>> + return 1; >>>> + return 0; >>>> +} >>>> + >>>> +/** >>>> + * seccomp_sync_threads: sets all threads to use current's filter >>>> + * >>>> + * Returns 0 on success or the pid of a thread which was either not >>>> + * in the correct seccomp mode or it did not have an ancestral >>>> + * seccomp filter. current must be in seccomp.mode=2 already. >>>> + */ >>>> +static pid_t seccomp_sync_threads(void) >>>> +{ >>>> + struct task_struct *thread, *caller; >>>> + pid_t failed = 0; >>>> + thread = caller = current; >>>> + >>>> + read_lock(&tasklist_lock); >>>> + if (thread_group_empty(caller)) >>>> + goto done; >>>> + while_each_thread(caller, thread) { >>>> + task_lock(thread); >>>> + /* >>>> + * All threads must not be in SECCOMP_MODE_STRICT to >>>> + * be eligible for synchronization. >>>> + */ >>>> + if ((thread->seccomp.mode == SECCOMP_MODE_DISABLED || >>>> + thread->seccomp.mode == SECCOMP_MODE_FILTER) && >>>> + is_ancestor(thread->seccomp.filter, >>>> + caller->seccomp.filter)) { >>>> + /* Get a task reference for the new leaf node. */ >>>> + get_seccomp_filter(caller); >>>> + /* >>>> + * Drop the task reference to the shared ancestor since >>>> + * current's path will hold a reference. (This also >>>> + * allows a put before the assignment.) >>>> + */ >>>> + put_seccomp_filter(thread); >>>> + thread->seccomp.filter = caller->seccomp.filter; >>>> + /* Opt the other thread into seccomp if needed. >>>> + * As threads are considered to be trust-realm >>>> + * equivalent (see ptrace_may_access), it is safe to >>>> + * allow one thread to transition the other. >>>> + */ >>>> + if (thread->seccomp.mode == SECCOMP_MODE_DISABLED) { >>>> + thread->seccomp.mode = SECCOMP_MODE_FILTER; >>>> + /* >>>> + * Don't let an unprivileged task work around >>>> + * the no_new_privs restriction by creating >>>> + * a thread that sets it up, enters seccomp, >>>> + * then dies. >>>> + */ >>>> + if (caller->no_new_privs) >>>> + thread->no_new_privs = 1; >>>> + set_tsk_thread_flag(thread, TIF_SECCOMP); >>> >>> no_new_privs is a bitfield, and some of the other bits in there look >>> like things that might not want to be read and written back from another >>> thread. >> >> Ah :/ Good catch! >> >>> Would it be too annoying to require that the other threads already have >>> no_new_privs set? >> >> Hrm, it's pretty painful in the edge cases where you don't control the >> process initialization which might setup threads you need to ensnare. >> >> Would it be crazy to do something like below in sched.h? >> - unsigned no_new_privs:1; >> + unsigned no_new_privs; > > set_bit, etc. would also work. (Although there isn't a 32-bit set_bit > AFAIK, or at least there isn't one that works on 64-bit BE archs.) I wasn't sure if I could use set_bit() in a way that wouldn't get me banned from submitting patches forever :) > Also, is 'unsigned' actually safe for this purpose, on all supported > archs/compilers? I'm pretty sure it's okay by C++11 rules, but those > don't apply here. Maybe some day the kernel will move to C11 and life > will be good. > >> >> It feels like a big hammer though, but it also seems weird to wrap those >> bitfields with task_lock. Any suggestions are welcome! I'll think about >> this a bit more and see if there is a good way to do this transition >> safely and cheaply. > > Hmm. I bet you could move no_new_privs somewhere else in task_lock > where there's a bit free. It could also go in 'struct creds', but I > think that's even worse from your perspective. > > Here's another dumb idea: Add an accessor task_no_new_privs(struct > task_struct *) and move no_new_privs into struct seccomp (i.e. make it > a bit in the seccomp mode). It kind of sucks on !CONFIG_SECCOMP, but > it's free if CONFIG_SECCOMP. That'd certainly be fine with me. I was considering adding a "needs_transition" bit to struct seccomp, but moving nnp there could be tidy. I'd need to make sure reading it locklessly still makes sense, but I really don't want to put a lock on the syscall path... > P.S. Have you seen the Linux Capsicum port? It fiddles with seccomp > mode, too, and I suspect it needs a fair amount of work, but I really > like the idea. Yup - that's a whole other thread to be had! It sets a new mode as short-hand, but what it needs really is a seccomp<->LSM interaction mechanism. That's on my todo list next to help sort out. thanks! -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/