Re: [PATCH v2] printk: ringbuffer: Improve prb_next_seq() performance

[Petr Mladek]

On Wed 2021-10-27 17:54:42, John Ogness wrote:
> On 2021-10-27, Petr Mladek <pmladek@xxxxxxxx> wrote:
> > diff --git a/kernel/printk/printk_ringbuffer.c b/kernel/printk/printk_ringbuffer.c
> > index 8a7b7362c0dd..24f47fbefbb5 100644
> > --- a/kernel/printk/printk_ringbuffer.c
> > +++ b/kernel/printk/printk_ringbuffer.c
> > @@ -474,8 +474,10 @@ static enum desc_state desc_read(struct prb_desc_ring *desc_ring,
> >  	 * state has been re-checked. A memcpy() for all of @desc
> >  	 * cannot be used because of the atomic_t @state_var field.
> >  	 */
> > -	memcpy(&desc_out->text_blk_lpos, &desc->text_blk_lpos,
> > -	       sizeof(desc_out->text_blk_lpos)); /* LMM(desc_read:C) */
> > +	if (desc_out) {
> > +		memcpy(&desc_out->text_blk_lpos, &desc->text_blk_lpos,
> > +		       sizeof(desc_out->text_blk_lpos)); /* LMM(desc_read:C) */
> > +	}
> >  	if (seq_out)
> >  		*seq_out = info->seq; /* also part of desc_read:C */
> >  	if (caller_id_out)
> 
> You also need this hunk:
> 
> @@ -530,7 +530,8 @@ static enum desc_state desc_read(struct prb_desc_ring *desc_ring,
>  	state_val = atomic_long_read(state_var); /* LMM(desc_read:E) */
>  	d_state = get_desc_state(id, state_val);
>  out:
> -	atomic_long_set(&desc_out->state_var, state_val);
> +	if (desc_out)
> +		atomic_long_set(&desc_out->state_var, state_val);
>  	return d_state;
>  }
> 
> Sorry, I missed it in my last review. This time I build+boot tested the patch.

Urgh. Great catch! It seems that my testing was not good enough. Or
that I was just lucky.

> With this hunk added:
> 
> Reviewed-by: John Ogness <john.ogness@xxxxxxxxxxxxx>

I am going to queue it for 5.17 after the pull request for 5.16 is
accepted. I do not feel comfortable to rush this into 5.16, especially
after finding the bug. The merge window has just started...

Anyway, thanks a lot for review and catching the mistake.

Best Regards,
Petr