From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752865AbbFDQol (ORCPT ); Thu, 4 Jun 2015 12:44:41 -0400 Received: from mail-vn0-f52.google.com ([209.85.216.52]:36639 "EHLO mail-vn0-f52.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751838AbbFDQoh (ORCPT ); Thu, 4 Jun 2015 12:44:37 -0400 MIME-Version: 1.0 In-Reply-To: <1433369396-13360-1-git-send-email-tycho.andersen@canonical.com> References: <1433369396-13360-1-git-send-email-tycho.andersen@canonical.com> Date: Thu, 4 Jun 2015 09:44:36 -0700 X-Google-Sender-Auth: DPomyYmkEgdSbq7Q7GgD0uNvV4Y Message-ID: Subject: Re: [PATCH v2] seccomp: add ptrace options for suspend/resume From: Kees Cook To: Tycho Andersen Cc: LKML , Linux API , Andy Lutomirski , Will Drewry , Roland McGrath , Oleg Nesterov , Pavel Emelyanov , "Serge E. Hallyn" Content-Type: text/plain; charset=UTF-8 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Jun 3, 2015 at 3:09 PM, Tycho Andersen wrote: > This patch is the first step in enabling checkpoint/restore of processes > with seccomp enabled. > > One of the things CRIU does while dumping tasks is inject code into them > via ptrace to collect information that is only available to the process > itself. However, if we are in a seccomp mode where these processes are > prohibited from making these syscalls, then what CRIU does kills the task. > > This patch adds a new ptrace option, PTRACE_O_SUSPEND_SECCOMP, that enables > a task from the init user namespace which has CAP_SYS_ADMIN and no seccomp > filters to disable (and re-enable) seccomp filters for another task so that > they can be successfully dumped (and restored). We restrict the set of > processes that can disable seccomp through ptrace because although today > ptrace can be used to bypass seccomp, there is some discussion of closing > this loophole in the future and we would like this patch to not depend on > that behavior and be future proofed for when it is removed. > > Note that seccomp can be suspended before any filters are actually > installed; this behavior is useful on criu restore, so that we can suspend > seccomp, restore the filters, unmap our restore code from the restored > process' address space, and then resume the task by detaching and have the > filters resumed as well. > > v2 changes: > > * require that the tracer have no seccomp filters installed > * drop TIF_NOTSC manipulation from the patch > * change from ptrace command to a ptrace option and use this ptrace option > as the flag to check. This means that as soon as the tracer > detaches/dies, seccomp is re-enabled and as a corrollary that one can not > disable seccomp across PTRACE_ATTACHs. This feature gives me the creeps, but I think it's okay. Could it be further restricted so that the process doing the suspension is already ptracing the target? > Signed-off-by: Tycho Andersen > CC: Kees Cook > CC: Andy Lutomirski > CC: Will Drewry > CC: Roland McGrath > CC: Oleg Nesterov > CC: Pavel Emelyanov > CC: Serge E. Hallyn > --- > include/linux/ptrace.h | 1 + > include/linux/seccomp.h | 4 ++++ > include/uapi/linux/ptrace.h | 6 ++++-- > kernel/ptrace.c | 6 ++++++ > kernel/seccomp.c | 23 +++++++++++++++++++++++ > 5 files changed, 38 insertions(+), 2 deletions(-) > > diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h > index 987a73a..061265f 100644 > --- a/include/linux/ptrace.h > +++ b/include/linux/ptrace.h > @@ -34,6 +34,7 @@ > #define PT_TRACE_SECCOMP PT_EVENT_FLAG(PTRACE_EVENT_SECCOMP) > > #define PT_EXITKILL (PTRACE_O_EXITKILL << PT_OPT_FLAG_SHIFT) > +#define PT_SUSPEND_SECCOMP (PTRACE_O_SUSPEND_SECCOMP << PT_OPT_FLAG_SHIFT) > > /* single stepping state bits (used on ARM and PA-RISC) */ > #define PT_SINGLESTEP_BIT 31 > diff --git a/include/linux/seccomp.h b/include/linux/seccomp.h > index a19ddac..ae3ec52 100644 > --- a/include/linux/seccomp.h > +++ b/include/linux/seccomp.h > @@ -53,6 +53,10 @@ static inline int seccomp_mode(struct seccomp *s) > return s->mode; > } > > +#ifdef CONFIG_CHECKPOINT_RESTORE > +extern bool may_suspend_seccomp(void); > +#endif #else static inline bool may_suspend_seccomp(void) { return false; } #endif > + > #else /* CONFIG_SECCOMP */ > > #include > diff --git a/include/uapi/linux/ptrace.h b/include/uapi/linux/ptrace.h > index cf1019e..a7a6979 100644 > --- a/include/uapi/linux/ptrace.h > +++ b/include/uapi/linux/ptrace.h > @@ -89,9 +89,11 @@ struct ptrace_peeksiginfo_args { > #define PTRACE_O_TRACESECCOMP (1 << PTRACE_EVENT_SECCOMP) > > /* eventless options */ > -#define PTRACE_O_EXITKILL (1 << 20) > +#define PTRACE_O_EXITKILL (1 << 20) > +#define PTRACE_O_SUSPEND_SECCOMP (1 << 21) > > -#define PTRACE_O_MASK (0x000000ff | PTRACE_O_EXITKILL) > +#define PTRACE_O_MASK (\ > + 0x000000ff | PTRACE_O_EXITKILL | PTRACE_O_SUSPEND_SECCOMP) > > #include > > diff --git a/kernel/ptrace.c b/kernel/ptrace.c > index c8e0e05..e3e68a2 100644 > --- a/kernel/ptrace.c > +++ b/kernel/ptrace.c > @@ -15,6 +15,7 @@ > #include > #include > #include > +#include > #include > #include > #include > @@ -556,6 +557,11 @@ static int ptrace_setoptions(struct task_struct *child, unsigned long data) > if (data & ~(unsigned long)PTRACE_O_MASK) > return -EINVAL; > > +#ifdef CONFIG_CHECKPOINT_RESTORE > + if (data & PTRACE_O_SUSPEND_SECCOMP && !may_suspend_seccomp()) > + return -EPERM; > +#endif I'd like to avoid seeing any #ifdefs added to the .c files. Using a static inline for may_suspend_seccomp() should cause this statement to be eliminated by the compiler. > + > /* Avoid intermediate state when all opts are cleared */ > flags = child->ptrace; > flags &= ~(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT); > diff --git a/kernel/seccomp.c b/kernel/seccomp.c > index 980fd26..2a1bd35 100644 > --- a/kernel/seccomp.c > +++ b/kernel/seccomp.c > @@ -590,6 +590,11 @@ void secure_computing_strict(int this_syscall) > { > int mode = current->seccomp.mode; > > +#ifdef CONFIG_CHECKPOINT_RESTORE > + if (unlikely(current->ptrace & PT_SUSPEND_SECCOMP)) > + return; > +#endif Could PT_SUSPEND_SECCOMP be defined to "0" with not CONFIG_CHECKPOINT_RESTORE? Then this wouldn't need ifdefs, and should be similarly eliminated by the compiler. > + > if (mode == 0) > return; > else if (mode == SECCOMP_MODE_STRICT) > @@ -691,6 +696,11 @@ u32 seccomp_phase1(struct seccomp_data *sd) > int this_syscall = sd ? sd->nr : > syscall_get_nr(current, task_pt_regs(current)); > > +#ifdef CONFIG_CHECKPOINT_RESTORE > + if (unlikely(current->ptrace & PT_SUSPEND_SECCOMP)) > + return SECCOMP_PHASE1_OK; > +#endif > + > switch (mode) { > case SECCOMP_MODE_STRICT: > __secure_computing_strict(this_syscall); /* may call do_exit */ > @@ -901,3 +911,16 @@ long prctl_set_seccomp(unsigned long seccomp_mode, char __user *filter) > /* prctl interface doesn't have flags, so they are always zero. */ > return do_seccomp(op, 0, uargs); > } > + > +#ifdef CONFIG_CHECKPOINT_RESTORE > +bool may_suspend_seccomp(void) > +{ > + if (!capable(CAP_SYS_ADMIN)) > + return false; > + > + if (current->seccomp.mode != SECCOMP_MODE_DISABLED) > + return false; > + > + return true; > +} > +#endif /* CONFIG_CHECKPOINT_RESTORE */ > -- > 2.1.4 > Thanks for working on this! -Kees -- Kees Cook Chrome OS Security