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.
v3 changes:
* get rid of various #ifdefs everywhere
* report more sensible errors when PTRACE_O_SUSPEND_SECCOMP is incorrectly
used
Signed-off-by: Tycho Andersen <[email protected]>
CC: Kees Cook <[email protected]>
CC: Andy Lutomirski <[email protected]>
CC: Will Drewry <[email protected]>
CC: Roland McGrath <[email protected]>
CC: Oleg Nesterov <[email protected]>
CC: Pavel Emelyanov <[email protected]>
CC: Serge E. Hallyn <[email protected]>
---
include/linux/ptrace.h | 1 +
include/linux/seccomp.h | 10 ++++++++++
include/uapi/linux/ptrace.h | 6 ++++--
kernel/ptrace.c | 10 ++++++++++
kernel/seccomp.c | 22 ++++++++++++++++++++++
5 files changed, 47 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..1d55cd9 100644
--- a/include/linux/seccomp.h
+++ b/include/linux/seccomp.h
@@ -95,4 +95,14 @@ static inline void get_seccomp_filter(struct task_struct *tsk)
return;
}
#endif /* CONFIG_SECCOMP_FILTER */
+
+#ifdef CONFIG_CHECKPOINT_RESTORE
+extern bool may_suspend_seccomp(void);
+#else
+static inline bool may_suspend_seccomp(void)
+{
+ return false;
+}
+#endif
+
#endif /* _LINUX_SECCOMP_H */
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 <asm/ptrace.h>
diff --git a/kernel/ptrace.c b/kernel/ptrace.c
index c8e0e05..a860deb 100644
--- a/kernel/ptrace.c
+++ b/kernel/ptrace.c
@@ -15,6 +15,7 @@
#include <linux/highmem.h>
#include <linux/pagemap.h>
#include <linux/ptrace.h>
+#include <linux/seccomp.h>
#include <linux/security.h>
#include <linux/signal.h>
#include <linux/uio.h>
@@ -556,6 +557,15 @@ static int ptrace_setoptions(struct task_struct *child, unsigned long data)
if (data & ~(unsigned long)PTRACE_O_MASK)
return -EINVAL;
+ if (unlikely(data & PTRACE_O_SUSPEND_SECCOMP)) {
+ if (!config_enabled(CONFIG_CHECKPOINT_RESTORE) ||
+ !config_enabled(CONFIG_SECCOMP))
+ return -EINVAL;
+
+ if (!may_suspend_seccomp())
+ return -EPERM;
+ }
+
/* 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..cdd97f6 100644
--- a/kernel/seccomp.c
+++ b/kernel/seccomp.c
@@ -590,6 +590,10 @@ void secure_computing_strict(int this_syscall)
{
int mode = current->seccomp.mode;
+ if (config_enabled(CONFIG_CHECKPOINT_RESTORE) &&
+ unlikely(current->ptrace & PT_SUSPEND_SECCOMP))
+ return;
+
if (mode == 0)
return;
else if (mode == SECCOMP_MODE_STRICT)
@@ -691,6 +695,10 @@ u32 seccomp_phase1(struct seccomp_data *sd)
int this_syscall = sd ? sd->nr :
syscall_get_nr(current, task_pt_regs(current));
+ if (config_enabled(CONFIG_CHECKPOINT_RESTORE) &&
+ unlikely(current->ptrace & PT_SUSPEND_SECCOMP))
+ return SECCOMP_PHASE1_OK;
+
switch (mode) {
case SECCOMP_MODE_STRICT:
__secure_computing_strict(this_syscall); /* may call do_exit */
@@ -901,3 +909,17 @@ 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 ||
+ current->ptrace & PT_SUSPEND_SECCOMP)
+ return false;
+
+ return true;
+}
+#endif /* CONFIG_CHECKPOINT_RESTORE */
--
2.1.4
Tycho,
I hate myself, but I have another nit ;) again, it is not that I think
you should updtate the patch, just fyi...
On 06/05, Tycho Andersen wrote:
>
> --- a/include/linux/seccomp.h
> +++ b/include/linux/seccomp.h
> @@ -95,4 +95,14 @@ static inline void get_seccomp_filter(struct task_struct *tsk)
> return;
> }
> #endif /* CONFIG_SECCOMP_FILTER */
> +
> +#ifdef CONFIG_CHECKPOINT_RESTORE
> +extern bool may_suspend_seccomp(void);
> +#else
> +static inline bool may_suspend_seccomp(void)
> +{
> + return false;
> +}
> +#endif
This looks wrong. There is no "extern may_suspend_seccomp()" if
CONFIG_SECCOMP=n, kernel/seccomp.c is not compiled. So you need another
ifdef(CONFIG_SECCOMP).
At the same time this does not matter and you do not need the dummy
"inline" version at all:
> @@ -556,6 +557,15 @@ static int ptrace_setoptions(struct task_struct *child, unsigned long data)
> if (data & ~(unsigned long)PTRACE_O_MASK)
> return -EINVAL;
>
> + if (unlikely(data & PTRACE_O_SUSPEND_SECCOMP)) {
> + if (!config_enabled(CONFIG_CHECKPOINT_RESTORE) ||
> + !config_enabled(CONFIG_SECCOMP))
> + return -EINVAL;
> +
> + if (!may_suspend_seccomp())
> + return -EPERM;
gcc will optimize out may_suspend_seccomp() unless both options are
enabled.
Oleg.
On Fri, Jun 05, 2015 at 11:52:08PM +0200, Oleg Nesterov wrote:
> Tycho,
>
> I hate myself, but I have another nit ;) again, it is not that I think
> you should updtate the patch, just fyi...
No worries :)
> On 06/05, Tycho Andersen wrote:
> >
> > --- a/include/linux/seccomp.h
> > +++ b/include/linux/seccomp.h
> > @@ -95,4 +95,14 @@ static inline void get_seccomp_filter(struct task_struct *tsk)
> > return;
> > }
> > #endif /* CONFIG_SECCOMP_FILTER */
> > +
> > +#ifdef CONFIG_CHECKPOINT_RESTORE
> > +extern bool may_suspend_seccomp(void);
> > +#else
> > +static inline bool may_suspend_seccomp(void)
> > +{
> > + return false;
> > +}
> > +#endif
>
> This looks wrong. There is no "extern may_suspend_seccomp()" if
> CONFIG_SECCOMP=n, kernel/seccomp.c is not compiled. So you need another
> ifdef(CONFIG_SECCOMP).
>
> At the same time this does not matter and you do not need the dummy
> "inline" version at all:
>
> > @@ -556,6 +557,15 @@ static int ptrace_setoptions(struct task_struct *child, unsigned long data)
> > if (data & ~(unsigned long)PTRACE_O_MASK)
> > return -EINVAL;
> >
> > + if (unlikely(data & PTRACE_O_SUSPEND_SECCOMP)) {
> > + if (!config_enabled(CONFIG_CHECKPOINT_RESTORE) ||
> > + !config_enabled(CONFIG_SECCOMP))
> > + return -EINVAL;
> > +
> > + if (!may_suspend_seccomp())
> > + return -EPERM;
>
> gcc will optimize out may_suspend_seccomp() unless both options are
> enabled.
Whoops, yes, you're right. I did build test in all the configurations
to catch stuff like this, but gcc was too smart for me. What's the
right thing to do, just leave out the definition of
may_suspend_seccomp all together, or add in another if defined() in
the header?
Maybe we will get rid of this pesky function all together though...
Tycho
On 06/05, Tycho Andersen wrote:
>
> right thing to do, just leave out the definition of
> may_suspend_seccomp all together, or add in another if defined() in
> the header?
Just declare it unconditionally in .h. The kernel relies on this kind
of dead-code elimination anyway. For example, please look at
"extern void __put_user_bad(void)" in uaccess.h.
IOW,
extern void undefined_func(void);
if (0)
undefined_func();
should be safe.
> Maybe we will get rid of this pesky function all together though...
Yes ;)
Oleg.