[PATCH 1/3] bpf: Don't check for current being NULL

[PATCH 1/3] bpf: Don't check for current being NULL

[Richard Weinberger]

current is never NULL.

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 kernel/bpf/helpers.c | 12 ------------
 1 file changed, 12 deletions(-)

diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
index 3d24e238221e..e8845adcd15e 100644
--- a/kernel/bpf/helpers.c
+++ b/kernel/bpf/helpers.c
@@ -120,9 +120,6 @@ BPF_CALL_0(bpf_get_current_pid_tgid)
 {
 	struct task_struct *task = current;
 
-	if (unlikely(!task))
-		return -EINVAL;
-
 	return (u64) task->tgid << 32 | task->pid;
 }
 
@@ -138,9 +135,6 @@ BPF_CALL_0(bpf_get_current_uid_gid)
 	kuid_t uid;
 	kgid_t gid;
 
-	if (unlikely(!task))
-		return -EINVAL;
-
 	current_uid_gid(&uid, &gid);
 	return (u64) from_kgid(&init_user_ns, gid) << 32 |
 		     from_kuid(&init_user_ns, uid);
@@ -156,9 +150,6 @@ BPF_CALL_2(bpf_get_current_comm, char *, buf, u32, size)
 {
 	struct task_struct *task = current;
 
-	if (unlikely(!task))
-		goto err_clear;
-
 	strncpy(buf, task->comm, size);
 
 	/* Verifier guarantees that size > 0. For task->comm exceeding
@@ -167,9 +158,6 @@ BPF_CALL_2(bpf_get_current_comm, char *, buf, u32, size)
 	 */
 	buf[size - 1] = 0;
 	return 0;
-err_clear:
-	memset(buf, 0, size);
-	return -EINVAL;
 }
 
 const struct bpf_func_proto bpf_get_current_comm_proto = {
-- 
2.13.6

[PATCH 2/3] bpf: Remove dead variable

[Richard Weinberger]

task is never used in bpf_get_current_uid_gid(), kill it.

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 kernel/bpf/helpers.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
index e8845adcd15e..511c9d522cfc 100644
--- a/kernel/bpf/helpers.c
+++ b/kernel/bpf/helpers.c
@@ -131,7 +131,6 @@ const struct bpf_func_proto bpf_get_current_pid_tgid_proto = {
 
 BPF_CALL_0(bpf_get_current_uid_gid)
 {
-	struct task_struct *task = current;
 	kuid_t uid;
 	kgid_t gid;
 
-- 
2.13.6

[PATCH 3/3] bpf: Make sure that ->comm does not change under us.

[Richard Weinberger]

Sadly we cannot use get_task_comm() since bpf_get_current_comm()
allows truncation.

Signed-off-by: Richard Weinberger <richard@nod.at>
---
 kernel/bpf/helpers.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
index 511c9d522cfc..4b042b24524d 100644
--- a/kernel/bpf/helpers.c
+++ b/kernel/bpf/helpers.c
@@ -18,6 +18,7 @@
 #include <linux/sched.h>
 #include <linux/uidgid.h>
 #include <linux/filter.h>
+#include <linux/sched/task.h>
 
 /* If kernel subsystem is allowing eBPF programs to call this function,
  * inside its own verifier_ops->get_func_proto() callback it should return
@@ -149,7 +150,9 @@ BPF_CALL_2(bpf_get_current_comm, char *, buf, u32, size)
 {
 	struct task_struct *task = current;
 
+	task_lock(task);
 	strncpy(buf, task->comm, size);
+	task_unlock(task);
 
 	/* Verifier guarantees that size > 0. For task->comm exceeding
 	 * size, guarantee that buf is %NUL-terminated. Unconditionally
-- 
2.13.6

Re: [PATCH 2/3] bpf: Remove dead variable

[Daniel Borkmann]

On 10/16/2017 08:18 PM, Richard Weinberger wrote:
> task is never used in bpf_get_current_uid_gid(), kill it.
>
> Signed-off-by: Richard Weinberger <richard@nod.at>
> ---
>   kernel/bpf/helpers.c | 1 -
>   1 file changed, 1 deletion(-)
>
> diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
> index e8845adcd15e..511c9d522cfc 100644
> --- a/kernel/bpf/helpers.c
> +++ b/kernel/bpf/helpers.c
> @@ -131,7 +131,6 @@ const struct bpf_func_proto bpf_get_current_pid_tgid_proto = {
>
>   BPF_CALL_0(bpf_get_current_uid_gid)
>   {
> -	struct task_struct *task = current;
>   	kuid_t uid;
>   	kgid_t gid;
>
>

Needs to be squashed into patch 1/3?

Re: [PATCH 2/3] bpf: Remove dead variable

[Richard Weinberger]

Am Montag, 16. Oktober 2017, 20:54:36 CEST schrieb Daniel Borkmann:
> On 10/16/2017 08:18 PM, Richard Weinberger wrote:
> > task is never used in bpf_get_current_uid_gid(), kill it.
> > 
> > Signed-off-by: Richard Weinberger <richard@nod.at>
> > ---
> > 
> >   kernel/bpf/helpers.c | 1 -
> >   1 file changed, 1 deletion(-)
> > 
> > diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
> > index e8845adcd15e..511c9d522cfc 100644
> > --- a/kernel/bpf/helpers.c
> > +++ b/kernel/bpf/helpers.c
> > @@ -131,7 +131,6 @@ const struct bpf_func_proto
> > bpf_get_current_pid_tgid_proto = {> 
> >   BPF_CALL_0(bpf_get_current_uid_gid)
> >   {
> > 
> > -	struct task_struct *task = current;
> > 
> >   	kuid_t uid;
> >   	kgid_t gid;
> 
> Needs to be squashed into patch 1/3?

I can squash it into 1/3, I kept it that way because
even without 1/3 this variable is unused.

Thanks,
//richard

Re: [PATCH 2/3] bpf: Remove dead variable

[Daniel Borkmann]

On 10/16/2017 08:59 PM, Richard Weinberger wrote:
> Am Montag, 16. Oktober 2017, 20:54:36 CEST schrieb Daniel Borkmann:
>> On 10/16/2017 08:18 PM, Richard Weinberger wrote:
>>> task is never used in bpf_get_current_uid_gid(), kill it.
>>>
>>> Signed-off-by: Richard Weinberger <richard@nod.at>
>>> ---
>>>
>>>    kernel/bpf/helpers.c | 1 -
>>>    1 file changed, 1 deletion(-)
>>>
>>> diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
>>> index e8845adcd15e..511c9d522cfc 100644
>>> --- a/kernel/bpf/helpers.c
>>> +++ b/kernel/bpf/helpers.c
>>> @@ -131,7 +131,6 @@ const struct bpf_func_proto
>>> bpf_get_current_pid_tgid_proto = {>
>>>    BPF_CALL_0(bpf_get_current_uid_gid)
>>>    {
>>>
>>> -	struct task_struct *task = current;
>>>
>>>    	kuid_t uid;
>>>    	kgid_t gid;
>>
>> Needs to be squashed into patch 1/3?
>
> I can squash it into 1/3, I kept it that way because
> even without 1/3 this variable is unused.

Hmm, the helper looks like the below. In patch 1/3 you removed
the 'if (unlikely(!task))' test where the variable was used before,
so 2/3 without the 1/3 would result in a compile error.

BPF_CALL_0(bpf_get_current_uid_gid)
{
	struct task_struct *task = current;
	kuid_t uid;
	kgid_t gid;

	if (unlikely(!task))
		return -EINVAL;

	current_uid_gid(&uid, &gid);
	return (u64) from_kgid(&init_user_ns, gid) << 32 |
		     from_kuid(&init_user_ns, uid);
}

Re: [PATCH 2/3] bpf: Remove dead variable

[Richard Weinberger]

Am Montag, 16. Oktober 2017, 21:11:47 CEST schrieb Daniel Borkmann: 
> > I can squash it into 1/3, I kept it that way because
> > even without 1/3 this variable is unused.
> 
> Hmm, the helper looks like the below. In patch 1/3 you removed
> the 'if (unlikely(!task))' test where the variable was used before,
> so 2/3 without the 1/3 would result in a compile error.

Why a compile error? It emits a warning.

> BPF_CALL_0(bpf_get_current_uid_gid)
> {
> 	struct task_struct *task = current;
> 	kuid_t uid;
> 	kgid_t gid;
> 
> 	if (unlikely(!task))
> 		return -EINVAL;

Well, this is the only "user". Okay.

> 	current_uid_gid(&uid, &gid);

Here we use current. So, task was always in vain.

So, I can happily squash 2/3 into 1/3 and resent.
The series just represented the way I've worked on the code... 

Thanks,
//richard

Re: [PATCH 2/3] bpf: Remove dead variable

[Daniel Borkmann]

On 10/16/2017 09:22 PM, Richard Weinberger wrote:
[...]
> So, I can happily squash 2/3 into 1/3 and resent.

Yeah, please just squash them.

Thanks,
Daniel

Re: [PATCH 3/3] bpf: Make sure that ->comm does not change under us.

[Daniel Borkmann]

On 10/16/2017 08:18 PM, Richard Weinberger wrote:
> Sadly we cannot use get_task_comm() since bpf_get_current_comm()
> allows truncation.
>
> Signed-off-by: Richard Weinberger <richard@nod.at>
> ---
>   kernel/bpf/helpers.c | 3 +++
>   1 file changed, 3 insertions(+)
>
> diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
> index 511c9d522cfc..4b042b24524d 100644
> --- a/kernel/bpf/helpers.c
> +++ b/kernel/bpf/helpers.c
> @@ -18,6 +18,7 @@
>   #include <linux/sched.h>
>   #include <linux/uidgid.h>
>   #include <linux/filter.h>
> +#include <linux/sched/task.h>
>
>   /* If kernel subsystem is allowing eBPF programs to call this function,
>    * inside its own verifier_ops->get_func_proto() callback it should return
> @@ -149,7 +150,9 @@ BPF_CALL_2(bpf_get_current_comm, char *, buf, u32, size)
>   {
>   	struct task_struct *task = current;
>
> +	task_lock(task);
>   	strncpy(buf, task->comm, size);
> +	task_unlock(task);

Wouldn't this potentially lead to a deadlock? E.g. you attach yourself
to task_lock() / spin_lock() / etc, and then the BPF prog triggers the
bpf_get_current_comm() taking the lock again ...

>   	/* Verifier guarantees that size > 0. For task->comm exceeding
>   	 * size, guarantee that buf is %NUL-terminated. Unconditionally
>

Re: [PATCH 3/3] bpf: Make sure that ->comm does not change under us.

[Richard Weinberger]

Am Montag, 16. Oktober 2017, 22:50:43 CEST schrieb Daniel Borkmann:
> >   	struct task_struct *task = current;
> > 
> > +	task_lock(task);
> > 
> >   	strncpy(buf, task->comm, size);
> > 
> > +	task_unlock(task);
> 
> Wouldn't this potentially lead to a deadlock? E.g. you attach yourself
> to task_lock() / spin_lock() / etc, and then the BPF prog triggers the
> bpf_get_current_comm() taking the lock again ...

Yes, but doesn't the same apply to the use case when I attach to strncpy()
and run bpf_get_current_comm()?

Thanks,
//richard

Re: [PATCH 3/3] bpf: Make sure that ->comm does not change under us.

[Daniel Borkmann]

On 10/16/2017 10:55 PM, Richard Weinberger wrote:
> Am Montag, 16. Oktober 2017, 22:50:43 CEST schrieb Daniel Borkmann:
>>>    	struct task_struct *task = current;
>>>
>>> +	task_lock(task);
>>>
>>>    	strncpy(buf, task->comm, size);
>>>
>>> +	task_unlock(task);
>>
>> Wouldn't this potentially lead to a deadlock? E.g. you attach yourself
>> to task_lock() / spin_lock() / etc, and then the BPF prog triggers the
>> bpf_get_current_comm() taking the lock again ...
>
> Yes, but doesn't the same apply to the use case when I attach to strncpy()
> and run bpf_get_current_comm()?

You mean due to recursion? In that case trace_call_bpf() would bail out
due to the bpf_prog_active counter.

Re: [PATCH 3/3] bpf: Make sure that ->comm does not change under us.

[Richard Weinberger]

Am Montag, 16. Oktober 2017, 23:02:06 CEST schrieb Daniel Borkmann:
> On 10/16/2017 10:55 PM, Richard Weinberger wrote:
> > Am Montag, 16. Oktober 2017, 22:50:43 CEST schrieb Daniel Borkmann:
> >>>    	struct task_struct *task = current;
> >>> 
> >>> +	task_lock(task);
> >>> 
> >>>    	strncpy(buf, task->comm, size);
> >>> 
> >>> +	task_unlock(task);
> >> 
> >> Wouldn't this potentially lead to a deadlock? E.g. you attach yourself
> >> to task_lock() / spin_lock() / etc, and then the BPF prog triggers the
> >> bpf_get_current_comm() taking the lock again ...
> > 
> > Yes, but doesn't the same apply to the use case when I attach to strncpy()
> > and run bpf_get_current_comm()?
> 
> You mean due to recursion? In that case trace_call_bpf() would bail out
> due to the bpf_prog_active counter.

Ah, that's true.
So, when someone wants to use bpf_get_current_comm() while tracing task_lock,
we have a problem. I agree.
On the other hand, without locking the function may return wrong results.

Thanks,
//richard

Re: [PATCH 3/3] bpf: Make sure that ->comm does not change under us.

[Alexei Starovoitov]

On Mon, Oct 16, 2017 at 2:10 PM, Richard Weinberger <richard@nod.at> wrote:
> Am Montag, 16. Oktober 2017, 23:02:06 CEST schrieb Daniel Borkmann:
>> On 10/16/2017 10:55 PM, Richard Weinberger wrote:
>> > Am Montag, 16. Oktober 2017, 22:50:43 CEST schrieb Daniel Borkmann:
>> >>>           struct task_struct *task = current;
>> >>>
>> >>> + task_lock(task);
>> >>>
>> >>>           strncpy(buf, task->comm, size);
>> >>>
>> >>> + task_unlock(task);
>> >>
>> >> Wouldn't this potentially lead to a deadlock? E.g. you attach yourself
>> >> to task_lock() / spin_lock() / etc, and then the BPF prog triggers the
>> >> bpf_get_current_comm() taking the lock again ...
>> >
>> > Yes, but doesn't the same apply to the use case when I attach to strncpy()
>> > and run bpf_get_current_comm()?
>>
>> You mean due to recursion? In that case trace_call_bpf() would bail out
>> due to the bpf_prog_active counter.
>
> Ah, that's true.
> So, when someone wants to use bpf_get_current_comm() while tracing task_lock,
> we have a problem. I agree.
> On the other hand, without locking the function may return wrong results.

it will surely race with somebody else setting task comm and it's fine.
all of bpf tracing is read-only, so locks are only allowed inside bpf core
bits like maps. Taking core locks like task_lock() is quite scary.
bpf scripts rely on bpf_probe_read() of all sorts of kernel fields
so reading comm here w/o lock is fine.

Re: [PATCH 3/3] bpf: Make sure that ->comm does not change under us.

[Daniel Borkmann]

On 10/17/2017 12:10 AM, Alexei Starovoitov wrote:
> On Mon, Oct 16, 2017 at 2:10 PM, Richard Weinberger <richard@nod.at> wrote:
>> Am Montag, 16. Oktober 2017, 23:02:06 CEST schrieb Daniel Borkmann:
>>> On 10/16/2017 10:55 PM, Richard Weinberger wrote:
>>>> Am Montag, 16. Oktober 2017, 22:50:43 CEST schrieb Daniel Borkmann:
>>>>>>            struct task_struct *task = current;
>>>>>>
>>>>>> + task_lock(task);
>>>>>>
>>>>>>            strncpy(buf, task->comm, size);
>>>>>>
>>>>>> + task_unlock(task);
>>>>>
>>>>> Wouldn't this potentially lead to a deadlock? E.g. you attach yourself
>>>>> to task_lock() / spin_lock() / etc, and then the BPF prog triggers the
>>>>> bpf_get_current_comm() taking the lock again ...
>>>>
>>>> Yes, but doesn't the same apply to the use case when I attach to strncpy()
>>>> and run bpf_get_current_comm()?
>>>
>>> You mean due to recursion? In that case trace_call_bpf() would bail out
>>> due to the bpf_prog_active counter.
>>
>> Ah, that's true.
>> So, when someone wants to use bpf_get_current_comm() while tracing task_lock,
>> we have a problem. I agree.
>> On the other hand, without locking the function may return wrong results.
>
> it will surely race with somebody else setting task comm and it's fine.
> all of bpf tracing is read-only, so locks are only allowed inside bpf core
> bits like maps. Taking core locks like task_lock() is quite scary.
> bpf scripts rely on bpf_probe_read() of all sorts of kernel fields
> so reading comm here w/o lock is fine.

Yeah, and perf_event_comm() -> perf_event_comm_event() out of __set_task_comm()
is having same approach wrt comm read-out.