[PATCH trival 1/2] bpf: clean up put_cpu_var usage

[PATCH trival 1/2] bpf: clean up put_cpu_var usage

[Shaohua Li]

put_cpu_var takes the percpu data, not the data returned from
get_cpu_var.

This doesn't change the behavior.

Cc: Tejun Heo <tj@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Shaohua Li <shli@fb.com>
---
 kernel/bpf/core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index 03fd23d..b73913b 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -1031,7 +1031,7 @@ u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5)
 
 	state = &get_cpu_var(bpf_user_rnd_state);
 	res = prandom_u32_state(state);
-	put_cpu_var(state);
+	put_cpu_var(bpf_user_rnd_state);
 
 	return res;
 }
-- 
2.9.3

[PATCH trival 2/2] lib: clean up put_cpu_var usage

[Shaohua Li]

put_cpu_var takes the percpu data, not the data returned from
get_cpu_var.

This doesn't change the behavior.

Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Shaohua Li <shli@fb.com>
---
 lib/random32.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lib/random32.c b/lib/random32.c
index 69ed593..915982b 100644
--- a/lib/random32.c
+++ b/lib/random32.c
@@ -81,7 +81,7 @@ u32 prandom_u32(void)
 	u32 res;
 
 	res = prandom_u32_state(state);
-	put_cpu_var(state);
+	put_cpu_var(net_rand_state);
 
 	return res;
 }
@@ -128,7 +128,7 @@ void prandom_bytes(void *buf, size_t bytes)
 	struct rnd_state *state = &get_cpu_var(net_rand_state);
 
 	prandom_bytes_state(state, buf, bytes);
-	put_cpu_var(state);
+	put_cpu_var(net_rand_state);
 }
 EXPORT_SYMBOL(prandom_bytes);
 
-- 
2.9.3

Re: [PATCH trival 1/2] bpf: clean up put_cpu_var usage

[Alexei Starovoitov]

On Mon, Sep 26, 2016 at 11:14:50AM -0700, Shaohua Li wrote:
> put_cpu_var takes the percpu data, not the data returned from
> get_cpu_var.
> 
> This doesn't change the behavior.
> 
> Cc: Tejun Heo <tj@kernel.org>
> Cc: Alexei Starovoitov <ast@kernel.org>
> Signed-off-by: Shaohua Li <shli@fb.com>

Looks good. Nice catch.
Please rebase to net-next tree and send it to netdev list.
Otherwise we'll have conflicts at the time of the merge window.

> ---
>  kernel/bpf/core.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
> index 03fd23d..b73913b 100644
> --- a/kernel/bpf/core.c
> +++ b/kernel/bpf/core.c
> @@ -1031,7 +1031,7 @@ u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5)
>  
>  	state = &get_cpu_var(bpf_user_rnd_state);
>  	res = prandom_u32_state(state);
> -	put_cpu_var(state);
> +	put_cpu_var(bpf_user_rnd_state);
>  
>  	return res;
>  }
> -- 
> 2.9.3
> 

Re: [PATCH trival 1/2] bpf: clean up put_cpu_var usage

[Daniel Borkmann]

Shaohua,

On 09/27/2016 01:51 AM, Alexei Starovoitov wrote:
> On Mon, Sep 26, 2016 at 11:14:50AM -0700, Shaohua Li wrote:
>> put_cpu_var takes the percpu data, not the data returned from
>> get_cpu_var.
>>
>> This doesn't change the behavior.
>>
>> Cc: Tejun Heo <tj@kernel.org>
>> Cc: Alexei Starovoitov <ast@kernel.org>
>> Signed-off-by: Shaohua Li <shli@fb.com>
>
> Looks good. Nice catch.
> Please rebase to net-next tree and send it to netdev list.

+1

> Otherwise we'll have conflicts at the time of the merge window.

While you're at it, the same kind of cleanup can be done for
prandom_u32() and prandom_bytes(), so you could either squash
that into this patch as well or make it two patches. Both for
netdev.

Thanks,
Daniel

>> ---
>>   kernel/bpf/core.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
>> index 03fd23d..b73913b 100644
>> --- a/kernel/bpf/core.c
>> +++ b/kernel/bpf/core.c
>> @@ -1031,7 +1031,7 @@ u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5)
>>
>>   	state = &get_cpu_var(bpf_user_rnd_state);
>>   	res = prandom_u32_state(state);
>> -	put_cpu_var(state);
>> +	put_cpu_var(bpf_user_rnd_state);
>>
>>   	return res;
>>   }
>> --
>> 2.9.3
>>