On Mon, Mar 21, 2022 at 12:54:19PM -0400, Steven Rostedt wrote:
> On Mon, 21 Mar 2022 17:50:50 +0100
> Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> > > This also assumes that we need to trace everything that is marked. I
> > > mentioned in another email, what do we do if we only trace funcA?
> >
> > Like I said later on; if we inhibit tail-calls to notrace, this goes
> > away.
>
> Please no. The number of "notrace" functions is increasing to the point
> that it's starting to make function tracing useless in a lot of
> circumstances. I've already lost my ability to see when user space goes
> into the kernel (which I have to hack up custom coding to enable again).
I really can't follow the argument there, nor on IRC.
Suppose:
notrace func_B()
{
...
}
func_A()
{
...
return func_B();
}
then inhibiting tail calls would end up looking like:
func_A:
call __fentry__
...
call func_B
call __fexit__
ret
Then A is fully traced, B is invisible, as per spec. What is the
problem?
The problem you initially had, of doing a tail-call into a notrace, was
that the __fexit__ call went missing, because notrace will obviously not
have that. But that's avoided by inhibiting all tail-calls between
notrace and !notrace functions (note that notrace must also not
tail-call !notrace).
Your worry seems to stem about loosing visiblilty of !notrace functions,
but AFAICT that doesn't happen.
