Re: [PATCH] ftrace: Add ftrace_page to list after the index is calculated

[Chen Zhongjin]

On 2023/3/8 22:53, Steven Rostedt wrote:
> > -	} else {
> > -		if (!ftrace_pages)
> > -			goto out;
> > -
> > -		if (WARN_ON(ftrace_pages->next)) {
> > -			/* Hmm, we have free pages? */
> > -			while (ftrace_pages->next)
> > -				ftrace_pages = ftrace_pages->next;
> > -		}
> > -
> > -		ftrace_pages->next = start_pg;
> Basically, what you are saying is that once we add ftrace_pages->next to
> point to the new start_pg, it becomes visible to others and that could be a
> problem. And moving this code around is not really going to solve that, as
> then we would need to add memory barriers.
>
> > -	}
> > -
> >   	p = start;
> >   	pg = start_pg;
> >   	while (p < end) {
> > @@ -6855,6 +6841,21 @@ static int ftrace_process_locs(struct module *mod,
> >   	/* We should have used all pages */
> >   	WARN_ON(pg->next);
> >   
> > +	/* Add pages to ftrace_pages list */
> > +	if (!mod) {
> > +		WARN_ON(ftrace_pages || ftrace_pages_start);
> > +		/* First initialization */
> > +		ftrace_pages_start = start_pg;
> > +	} else {
> > +		if (WARN_ON(ftrace_pages->next)) {
> > +			/* Hmm, we have free pages? */
> > +			while (ftrace_pages->next)
> > +				ftrace_pages = ftrace_pages->next;
> > +		}
> > +
> > +		ftrace_pages->next = start_pg;
> > +	}
> > +
> >   	/* Assign the last page to ftrace_pages */
> >   	ftrace_pages = pg;
> >   
> Why not just test for it?
>
> diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
> index 29baa97d0d53..9b2803c7a18f 100644
> --- a/kernel/trace/ftrace.c
> +++ b/kernel/trace/ftrace.c
> @@ -1564,7 +1564,8 @@ static struct dyn_ftrace *lookup_rec(unsigned long start, unsigned long end)
>   	key.flags = end;	/* overload flags, as it is unsigned long */
>   
>   	for (pg = ftrace_pages_start; pg; pg = pg->next) {
> -		if (end < pg->records[0].ip ||
> +		if (pg->index == 0 ||
> +		    end < pg->records[0].ip ||
>   		    start >= (pg->records[pg->index - 1].ip + MCOUNT_INSN_SIZE))
>   			continue;
>   		rec = bsearch(&key, pg->records, pg->index,
>
>
> -- Steve
Thanks for review!

At first I'm worried that it will cause lookup_rec to return an 
incorrect result
if it issearching a module address going to be added.
But now I found that records are added at MODULE_STATE_UNFORMED, the module
address won't  be searched at this point in any case.

Let's do it this way. I'll send another patch.