Re: [PATCH 4.19 46/58] net: dsa: mv88e6xxx: also read STU state in mv88e6250_g1_vtu_getnext

[Rasmus Villemoes]

On 26/01/2021 09.50, Greg Kroah-Hartman wrote:
> On Mon, Jan 25, 2021 at 08:59:54PM +0100, Rasmus Villemoes wrote:
>> On 25/01/2021 20.40, Rasmus Villemoes wrote:
>>> On 25/01/2021 19.39, Greg Kroah-Hartman wrote:
>>>> From: Rasmus Villemoes <rasmus.villemoes@xxxxxxxxx>
>>>>
>>>> commit 87fe04367d842c4d97a77303242d4dd4ac351e46 upstream.
>>>>
>>>
>>> Greg, please drop this from 4.19-stable. Details:
>>>
>>>>
>>>> --- a/drivers/net/dsa/mv88e6xxx/global1_vtu.c
>>>> +++ b/drivers/net/dsa/mv88e6xxx/global1_vtu.c
>>>> @@ -357,6 +357,10 @@ int mv88e6185_g1_vtu_getnext(struct mv88
>>>>  		if (err)
>>>>  			return err;
>>>>  
>>>> +		err = mv88e6185_g1_stu_data_read(chip, entry);
>>>> +		if (err)
>>>> +			return err;
>>>> +
>>>
>>> The function that this patch applied to in mainline did not exist in
>>> v4.19. It seems that this hunk has been applied in the similar
>>> mv88e6185_g1_vtu_getnext(), and indeed, in current 4.19.y, just one more
>>> line of context shows this:
>>
>> Bah, that was from 4.14, so the line numbers are a bit off, but I see
>> you've also added it to the 4.14 queue. Same comment for that one: Drop
>> this from both 4.19.y and 4.14.y.
> 
> Odd, but ok, the Fixes: line lied :)

Sorry, but no, it did not.

The commit 92307069a96c mentioned in Fixes changed the semantics of the
helper mv88e6185_g1_vtu_data_read(), splitting half its functionality to
a new helper mv88e6185_g1_stu_data_read(), and it also added a call of
that new helper in mv88e6185_g1_vtu_getnext().

But the other user of mv88e6185_g1_vtu_data_read(), namely
mv88e6250_g1_vtu_getnext(), was not updated, which is precisely what
87fe04367d842c4d97a77303242d4dd4ac351e46 fixes. However, that other user
of mv88e6185_g1_vtu_data_read() did not exist in 4.19, so the backports
of 92307069a96c to those earlier stable trees didn't introduce that bug.

Rasmus