With the new SECCOMP_FILTER_FLAG_LOG, we need to be able to extract these
flags for checkpoint restore, since they describe the state of a filter.
So, let's add PTRACE_SECCOMP_GET_FLAGS, similar to ..._GET_FILTER, which
returns the flags of the nth filter.
Signed-off-by: Tycho Andersen <[email protected]>
CC: Kees Cook <[email protected]>
CC: Andy Lutomirski <[email protected]>
CC: Oleg Nesterov <[email protected]>
---
include/linux/seccomp.h | 7 +++++
include/uapi/linux/ptrace.h | 1 +
kernel/ptrace.c | 4 +++
kernel/seccomp.c | 74 ++++++++++++++++++++++++++++++++++-----------
4 files changed, 68 insertions(+), 18 deletions(-)
diff --git a/include/linux/seccomp.h b/include/linux/seccomp.h
index c8bef436b61d..4713d62378e3 100644
--- a/include/linux/seccomp.h
+++ b/include/linux/seccomp.h
@@ -94,11 +94,18 @@ static inline void get_seccomp_filter(struct task_struct *tsk)
#if defined(CONFIG_SECCOMP_FILTER) && defined(CONFIG_CHECKPOINT_RESTORE)
extern long seccomp_get_filter(struct task_struct *task,
unsigned long filter_off, void __user *data);
+extern long seccomp_get_flags(struct task_struct *task,
+ unsigned long filter_off);
#else
static inline long seccomp_get_filter(struct task_struct *task,
unsigned long n, void __user *data)
{
return -EINVAL;
}
+static inline long seccomp_get_flags(struct task_struct *task,
+ unsigned long filter_off)
+{
+ return -EINVAL;
+}
#endif /* CONFIG_SECCOMP_FILTER && CONFIG_CHECKPOINT_RESTORE */
#endif /* _LINUX_SECCOMP_H */
diff --git a/include/uapi/linux/ptrace.h b/include/uapi/linux/ptrace.h
index fb8106509000..52903f0f6600 100644
--- a/include/uapi/linux/ptrace.h
+++ b/include/uapi/linux/ptrace.h
@@ -65,6 +65,7 @@ struct ptrace_peeksiginfo_args {
#define PTRACE_SETSIGMASK 0x420b
#define PTRACE_SECCOMP_GET_FILTER 0x420c
+#define PTRACE_SECCOMP_GET_FLAGS 0x420d
/* Read signals from a shared (process wide) queue */
#define PTRACE_PEEKSIGINFO_SHARED (1 << 0)
diff --git a/kernel/ptrace.c b/kernel/ptrace.c
index 60f356d91060..fa3d2b4d3bf0 100644
--- a/kernel/ptrace.c
+++ b/kernel/ptrace.c
@@ -1094,6 +1094,10 @@ int ptrace_request(struct task_struct *child, long request,
ret = seccomp_get_filter(child, addr, datavp);
break;
+ case PTRACE_SECCOMP_GET_FLAGS:
+ ret = seccomp_get_flags(child, addr);
+ break;
+
default:
break;
}
diff --git a/kernel/seccomp.c b/kernel/seccomp.c
index c24579dfa7a1..67768aaa7952 100644
--- a/kernel/seccomp.c
+++ b/kernel/seccomp.c
@@ -968,24 +968,16 @@ long prctl_set_seccomp(unsigned long seccomp_mode, char __user *filter)
}
#if defined(CONFIG_SECCOMP_FILTER) && defined(CONFIG_CHECKPOINT_RESTORE)
-long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
- void __user *data)
+static struct seccomp_filter *get_nth_filter(struct task_struct *task,
+ unsigned long filter_off)
{
struct seccomp_filter *filter;
- struct sock_fprog_kern *fprog;
- long ret;
unsigned long count = 0;
- if (!capable(CAP_SYS_ADMIN) ||
- current->seccomp.mode != SECCOMP_MODE_DISABLED) {
- return -EACCES;
- }
+ WARN_ON_ONCE(!spin_is_locked(&task->sighand->siglock));
- spin_lock_irq(&task->sighand->siglock);
- if (task->seccomp.mode != SECCOMP_MODE_FILTER) {
- ret = -EINVAL;
- goto out;
- }
+ if (task->seccomp.mode != SECCOMP_MODE_FILTER)
+ return ERR_PTR(-EINVAL);
filter = task->seccomp.filter;
while (filter) {
@@ -993,10 +985,9 @@ long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
count++;
}
- if (filter_off >= count) {
- ret = -ENOENT;
- goto out;
- }
+ if (filter_off >= count)
+ return ERR_PTR(-ENOENT);
+
count -= filter_off;
filter = task->seccomp.filter;
@@ -1007,7 +998,28 @@ long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
if (WARN_ON(count != 1 || !filter)) {
/* The filter tree shouldn't shrink while we're using it. */
- ret = -ENOENT;
+ return ERR_PTR(-ENOENT);
+ }
+
+ return filter;
+}
+
+long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
+ void __user *data)
+{
+ struct seccomp_filter *filter;
+ struct sock_fprog_kern *fprog;
+ long ret;
+
+ if (!capable(CAP_SYS_ADMIN) ||
+ current->seccomp.mode != SECCOMP_MODE_DISABLED) {
+ return -EACCES;
+ }
+
+ spin_lock_irq(&task->sighand->siglock);
+ filter = get_nth_filter(task, filter_off);
+ if (IS_ERR(filter)) {
+ ret = PTR_ERR(filter);
goto out;
}
@@ -1038,6 +1050,32 @@ long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
spin_unlock_irq(&task->sighand->siglock);
return ret;
}
+
+long seccomp_get_flags(struct task_struct *task, unsigned long filter_off)
+{
+ long flags = 0L;
+ struct seccomp_filter *filter;
+
+ if (!capable(CAP_SYS_ADMIN) ||
+ current->seccomp.mode != SECCOMP_MODE_DISABLED) {
+ return -EACCES;
+ }
+
+ spin_lock_irq(&task->sighand->siglock);
+ filter = get_nth_filter(task, filter_off);
+ if (IS_ERR(filter)) {
+ flags = PTR_ERR(filter);
+ goto out;
+ }
+
+ if (filter->log)
+ flags |= SECCOMP_FILTER_FLAG_LOG;
+
+out:
+ spin_unlock_irq(&task->sighand->siglock);
+ return flags;
+
+}
#endif
#ifdef CONFIG_SYSCTL
--
2.11.0
On Tue, Sep 19, 2017 at 10:47 AM, Tycho Andersen <[email protected]> wrote:
> With the new SECCOMP_FILTER_FLAG_LOG, we need to be able to extract these
> flags for checkpoint restore, since they describe the state of a filter.
>
> So, let's add PTRACE_SECCOMP_GET_FLAGS, similar to ..._GET_FILTER, which
> returns the flags of the nth filter.
Can you split this up into factoring out the nth helper, and then
adding the new get?
For naming, perhaps "GET_FILTER_FLAGS" instead of "GET_FLAGS" since
there may be seccomp flags in the future, etc.
Is there any sane way to add the flags to the existing GET_FILTER?
-Kees
>
> Signed-off-by: Tycho Andersen <[email protected]>
> CC: Kees Cook <[email protected]>
> CC: Andy Lutomirski <[email protected]>
> CC: Oleg Nesterov <[email protected]>
> ---
> include/linux/seccomp.h | 7 +++++
> include/uapi/linux/ptrace.h | 1 +
> kernel/ptrace.c | 4 +++
> kernel/seccomp.c | 74 ++++++++++++++++++++++++++++++++++-----------
> 4 files changed, 68 insertions(+), 18 deletions(-)
>
> diff --git a/include/linux/seccomp.h b/include/linux/seccomp.h
> index c8bef436b61d..4713d62378e3 100644
> --- a/include/linux/seccomp.h
> +++ b/include/linux/seccomp.h
> @@ -94,11 +94,18 @@ static inline void get_seccomp_filter(struct task_struct *tsk)
> #if defined(CONFIG_SECCOMP_FILTER) && defined(CONFIG_CHECKPOINT_RESTORE)
> extern long seccomp_get_filter(struct task_struct *task,
> unsigned long filter_off, void __user *data);
> +extern long seccomp_get_flags(struct task_struct *task,
> + unsigned long filter_off);
> #else
> static inline long seccomp_get_filter(struct task_struct *task,
> unsigned long n, void __user *data)
> {
> return -EINVAL;
> }
> +static inline long seccomp_get_flags(struct task_struct *task,
> + unsigned long filter_off)
> +{
> + return -EINVAL;
> +}
> #endif /* CONFIG_SECCOMP_FILTER && CONFIG_CHECKPOINT_RESTORE */
> #endif /* _LINUX_SECCOMP_H */
> diff --git a/include/uapi/linux/ptrace.h b/include/uapi/linux/ptrace.h
> index fb8106509000..52903f0f6600 100644
> --- a/include/uapi/linux/ptrace.h
> +++ b/include/uapi/linux/ptrace.h
> @@ -65,6 +65,7 @@ struct ptrace_peeksiginfo_args {
> #define PTRACE_SETSIGMASK 0x420b
>
> #define PTRACE_SECCOMP_GET_FILTER 0x420c
> +#define PTRACE_SECCOMP_GET_FLAGS 0x420d
>
> /* Read signals from a shared (process wide) queue */
> #define PTRACE_PEEKSIGINFO_SHARED (1 << 0)
> diff --git a/kernel/ptrace.c b/kernel/ptrace.c
> index 60f356d91060..fa3d2b4d3bf0 100644
> --- a/kernel/ptrace.c
> +++ b/kernel/ptrace.c
> @@ -1094,6 +1094,10 @@ int ptrace_request(struct task_struct *child, long request,
> ret = seccomp_get_filter(child, addr, datavp);
> break;
>
> + case PTRACE_SECCOMP_GET_FLAGS:
> + ret = seccomp_get_flags(child, addr);
> + break;
> +
> default:
> break;
> }
> diff --git a/kernel/seccomp.c b/kernel/seccomp.c
> index c24579dfa7a1..67768aaa7952 100644
> --- a/kernel/seccomp.c
> +++ b/kernel/seccomp.c
> @@ -968,24 +968,16 @@ long prctl_set_seccomp(unsigned long seccomp_mode, char __user *filter)
> }
>
> #if defined(CONFIG_SECCOMP_FILTER) && defined(CONFIG_CHECKPOINT_RESTORE)
> -long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
> - void __user *data)
> +static struct seccomp_filter *get_nth_filter(struct task_struct *task,
> + unsigned long filter_off)
> {
> struct seccomp_filter *filter;
> - struct sock_fprog_kern *fprog;
> - long ret;
> unsigned long count = 0;
>
> - if (!capable(CAP_SYS_ADMIN) ||
> - current->seccomp.mode != SECCOMP_MODE_DISABLED) {
> - return -EACCES;
> - }
> + WARN_ON_ONCE(!spin_is_locked(&task->sighand->siglock));
>
> - spin_lock_irq(&task->sighand->siglock);
> - if (task->seccomp.mode != SECCOMP_MODE_FILTER) {
> - ret = -EINVAL;
> - goto out;
> - }
> + if (task->seccomp.mode != SECCOMP_MODE_FILTER)
> + return ERR_PTR(-EINVAL);
>
> filter = task->seccomp.filter;
> while (filter) {
> @@ -993,10 +985,9 @@ long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
> count++;
> }
>
> - if (filter_off >= count) {
> - ret = -ENOENT;
> - goto out;
> - }
> + if (filter_off >= count)
> + return ERR_PTR(-ENOENT);
> +
> count -= filter_off;
>
> filter = task->seccomp.filter;
> @@ -1007,7 +998,28 @@ long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
>
> if (WARN_ON(count != 1 || !filter)) {
> /* The filter tree shouldn't shrink while we're using it. */
> - ret = -ENOENT;
> + return ERR_PTR(-ENOENT);
> + }
> +
> + return filter;
> +}
> +
> +long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
> + void __user *data)
> +{
> + struct seccomp_filter *filter;
> + struct sock_fprog_kern *fprog;
> + long ret;
> +
> + if (!capable(CAP_SYS_ADMIN) ||
> + current->seccomp.mode != SECCOMP_MODE_DISABLED) {
> + return -EACCES;
> + }
> +
> + spin_lock_irq(&task->sighand->siglock);
> + filter = get_nth_filter(task, filter_off);
> + if (IS_ERR(filter)) {
> + ret = PTR_ERR(filter);
> goto out;
> }
>
> @@ -1038,6 +1050,32 @@ long seccomp_get_filter(struct task_struct *task, unsigned long filter_off,
> spin_unlock_irq(&task->sighand->siglock);
> return ret;
> }
> +
> +long seccomp_get_flags(struct task_struct *task, unsigned long filter_off)
> +{
> + long flags = 0L;
> + struct seccomp_filter *filter;
> +
> + if (!capable(CAP_SYS_ADMIN) ||
> + current->seccomp.mode != SECCOMP_MODE_DISABLED) {
> + return -EACCES;
> + }
> +
> + spin_lock_irq(&task->sighand->siglock);
> + filter = get_nth_filter(task, filter_off);
> + if (IS_ERR(filter)) {
> + flags = PTR_ERR(filter);
> + goto out;
> + }
> +
> + if (filter->log)
> + flags |= SECCOMP_FILTER_FLAG_LOG;
> +
> +out:
> + spin_unlock_irq(&task->sighand->siglock);
> + return flags;
> +
> +}
> #endif
>
> #ifdef CONFIG_SYSCTL
> --
> 2.11.0
>
--
Kees Cook
Pixel Security
Hi Kees,
On Tue, Sep 19, 2017 at 01:08:28PM -0700, Kees Cook wrote:
> On Tue, Sep 19, 2017 at 10:47 AM, Tycho Andersen <[email protected]> wrote:
> > With the new SECCOMP_FILTER_FLAG_LOG, we need to be able to extract these
> > flags for checkpoint restore, since they describe the state of a filter.
> >
> > So, let's add PTRACE_SECCOMP_GET_FLAGS, similar to ..._GET_FILTER, which
> > returns the flags of the nth filter.
>
> Can you split this up into factoring out the nth helper, and then
> adding the new get?
>
> For naming, perhaps "GET_FILTER_FLAGS" instead of "GET_FLAGS" since
> there may be seccomp flags in the future, etc.
Sure, I'll do both of these.
> Is there any sane way to add the flags to the existing GET_FILTER?
I looked at this, and I don't think so. Unfortunately, we didn't use
any structure for the output, it's just the raw bytes of the filter
with the length used as the return value. I suppose we could append
the flags after the bytes of the filter, but that seems... very ugly
:). Let me know if you want to go that route.
Tycho
On Tue, Sep 19, 2017 at 2:09 PM, Tycho Andersen <[email protected]> wrote:
> Hi Kees,
>
> On Tue, Sep 19, 2017 at 01:08:28PM -0700, Kees Cook wrote:
>> On Tue, Sep 19, 2017 at 10:47 AM, Tycho Andersen <[email protected]> wrote:
>> > With the new SECCOMP_FILTER_FLAG_LOG, we need to be able to extract these
>> > flags for checkpoint restore, since they describe the state of a filter.
>> >
>> > So, let's add PTRACE_SECCOMP_GET_FLAGS, similar to ..._GET_FILTER, which
>> > returns the flags of the nth filter.
>>
>> Can you split this up into factoring out the nth helper, and then
>> adding the new get?
>>
>> For naming, perhaps "GET_FILTER_FLAGS" instead of "GET_FLAGS" since
>> there may be seccomp flags in the future, etc.
>
> Sure, I'll do both of these.
>
>> Is there any sane way to add the flags to the existing GET_FILTER?
>
> I looked at this, and I don't think so. Unfortunately, we didn't use
> any structure for the output, it's just the raw bytes of the filter
> with the length used as the return value. I suppose we could append
> the flags after the bytes of the filter, but that seems... very ugly
> :). Let me know if you want to go that route.
I think if we can make the new GET_FILTER_stuff interface more
extensible, we should cover any future needs for per-filter content.
BUt yeah, I agree, when I looked at this when I first mailed you about
it, I agree: it looked more ugly to extend the existing GET_FILTER.
-Kees
--
Kees Cook
Pixel Security
On Tue, Sep 19, 2017 at 02:43:51PM -0700, Kees Cook wrote:
> On Tue, Sep 19, 2017 at 2:09 PM, Tycho Andersen <[email protected]> wrote:
> > Hi Kees,
> >
> > On Tue, Sep 19, 2017 at 01:08:28PM -0700, Kees Cook wrote:
> >> On Tue, Sep 19, 2017 at 10:47 AM, Tycho Andersen <[email protected]> wrote:
> >> > With the new SECCOMP_FILTER_FLAG_LOG, we need to be able to extract these
> >> > flags for checkpoint restore, since they describe the state of a filter.
> >> >
> >> > So, let's add PTRACE_SECCOMP_GET_FLAGS, similar to ..._GET_FILTER, which
> >> > returns the flags of the nth filter.
> >>
> >> Can you split this up into factoring out the nth helper, and then
> >> adding the new get?
> >>
> >> For naming, perhaps "GET_FILTER_FLAGS" instead of "GET_FLAGS" since
> >> there may be seccomp flags in the future, etc.
> >
> > Sure, I'll do both of these.
> >
> >> Is there any sane way to add the flags to the existing GET_FILTER?
> >
> > I looked at this, and I don't think so. Unfortunately, we didn't use
> > any structure for the output, it's just the raw bytes of the filter
> > with the length used as the return value. I suppose we could append
> > the flags after the bytes of the filter, but that seems... very ugly
> > :). Let me know if you want to go that route.
>
> I think if we can make the new GET_FILTER_stuff interface more
> extensible, we should cover any future needs for per-filter content.
Ok, are you interested in something more general than
GET_FILTER_FLAGS? Maybe GET_FILTER_METADATA with uapi struct like:
struct seccomp_metadata {
unsigned int flags;
};
which you call by something like,
struct seccomp_metadata mymeta;
size = ptrace(PTRACE_SECCOMP_GET_FILTER_METADATA, pid, sizeof(mymeta), &mymeta);
if (size < 0) {
/* error */
}
if (size != sizeof(mymeta)) {
/* struct seccomp_metadata sizes are different */
}
?
Tycho
On Tue, Sep 19, 2017 at 2:55 PM, Tycho Andersen <[email protected]> wrote:
> On Tue, Sep 19, 2017 at 02:43:51PM -0700, Kees Cook wrote:
>> On Tue, Sep 19, 2017 at 2:09 PM, Tycho Andersen <[email protected]> wrote:
>> > Hi Kees,
>> >
>> > On Tue, Sep 19, 2017 at 01:08:28PM -0700, Kees Cook wrote:
>> >> On Tue, Sep 19, 2017 at 10:47 AM, Tycho Andersen <[email protected]> wrote:
>> >> > With the new SECCOMP_FILTER_FLAG_LOG, we need to be able to extract these
>> >> > flags for checkpoint restore, since they describe the state of a filter.
>> >> >
>> >> > So, let's add PTRACE_SECCOMP_GET_FLAGS, similar to ..._GET_FILTER, which
>> >> > returns the flags of the nth filter.
>> >>
>> >> Can you split this up into factoring out the nth helper, and then
>> >> adding the new get?
>> >>
>> >> For naming, perhaps "GET_FILTER_FLAGS" instead of "GET_FLAGS" since
>> >> there may be seccomp flags in the future, etc.
>> >
>> > Sure, I'll do both of these.
>> >
>> >> Is there any sane way to add the flags to the existing GET_FILTER?
>> >
>> > I looked at this, and I don't think so. Unfortunately, we didn't use
>> > any structure for the output, it's just the raw bytes of the filter
>> > with the length used as the return value. I suppose we could append
>> > the flags after the bytes of the filter, but that seems... very ugly
>> > :). Let me know if you want to go that route.
>>
>> I think if we can make the new GET_FILTER_stuff interface more
>> extensible, we should cover any future needs for per-filter content.
>
> Ok, are you interested in something more general than
> GET_FILTER_FLAGS? Maybe GET_FILTER_METADATA with uapi struct like:
>
> struct seccomp_metadata {
> unsigned int flags;
> };
>
> which you call by something like,
>
> struct seccomp_metadata mymeta;
>
> size = ptrace(PTRACE_SECCOMP_GET_FILTER_METADATA, pid, sizeof(mymeta), &mymeta);
> if (size < 0) {
> /* error */
> }
> if (size != sizeof(mymeta)) {
> /* struct seccomp_metadata sizes are different */
> }
>
> ?
Yeah, I think a sizeof could work here as a way to grow the metadata
result in the future.
Adding linux-api@ to see if anyone has thoughts on the best way to
make this future-proof...
-Kees
--
Kees Cook
Pixel Security