Re: [RFC PATCH v1 07/25] printk-rb: add functionality required by printk

[John Ogness]

On 2019-02-12, Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
> On Tue, Feb 12, 2019 at 6:30 AM John Ogness <john.ogness@xxxxxxxxxxxxx> wrote:
>>
>> +                       while (atomic_long_read(&rb->lost)) {
>> +                               atomic_long_dec(&rb->lost);
>> +                               rb->seq++;
>> +                       }
>
> This looks like crazy garbage. It's neither atomic nor sane.

It works because because only 1 context on a single CPU can hit that
loop. But yes, it is crazy.

> Why isn't it something like
>
>     if (atomic_long_read(&rb->lost)) {
>         long lost = atomic_xchg(&rb->lost, 0);
>         rb->seq += lost;
>     }
>
> instead?

Yes, it should be like you suggest. Thanks.

John Ogness