[Intel-gfx] [PATCH v4a 00/38] timers: Use timer_shutdown*() before freeing timers

[Steven Rostedt]

On Sat, 5 Nov 2022 08:59:36 -0700
Linus Torvalds <torvalds at linux-foundation.org> wrote:

> On Fri, Nov 4, 2022 at 11:01 PM Steven Rostedt <rostedt at goodmis.org> wrote:
> >
> > Patch 1 fixes an issue with sunrpc/xprt where it incorrectly uses
> > del_singleshot_timer_sync() for something that is not a oneshot timer. As this
> > will be converted to shutdown, this needs to be fixed first.  
> 
> So this is the kind of thing that I would *not* want to get eartly.

So I'll have to break up patch 5 to not update the
del_singleshot_timer_sync() to a timer_shutdown_sync(), because that
breaks this code.

Hmm, since that is a functional change, it probably should wait till
the merge window. I'll move this patch and that part of patch 5 to the
second part of the series for the merge window.

> 
> I really would want to get just the infrastructure in to let people
> start doing conversions.
> 
> And then the "mindlessly obvious patches that are done by scripting
> and can not possibly matter".
> 
> The kinds that do not *need* review, because they are mechanical, and
> that just cause pointless noise for the rest of the patches that *do*
> want review.
> 
> Not this kind of thing that is so subtle that you have to explain it.
> That's not a "scripted patch for no semantic change".
> 
> So leave the del_singleshot_timer_sync() cases alone, they are
> irrelevant for the new infrastructure and for the "mindless scripted
> conversion" patches.
> 
> > Patches 2-4 changes existing timer_shutdown() functions used locally in ARM and
> > some drivers to better namespace names.  
> 
> Ok, these are relevant.
> 
> > Patch 5 implements the new timer_shutdown() and timer_shutdown_sync() functions
> > that disable re-arming the timer after they are called.  
> 
> This is obviously what I'd want early so that people can start doign
> this in their trees.

But will need to remove the part that it changes del_singleshot_timer_sync().


> 
> > Patches 6-28 change all the locations where there's a kfree(), kfree_rcu(),
> > kmem_cache_free() and one call_rcu() call where the RCU function frees the
> > timer (the workqueue patch) in the same function as the del_timer{,_sync}() is
> > called on that timer, and there's no extra exit path between the del_timer and
> > freeing of the timer.  
> 
> So honestly, I was literally hoping for a "this is the coccinelle
> script" kind of patch.

The above actual was, but I walked through them manually too, because I
don't trust my conccinelle skills. All but the call_rcu() one was
caught by conccinelle. That's why I pointed out the worqueue one. I'll
remove that from this series.

> 
> Now there seems to be a number of patches here that are actualyl
> really hard to see that they are "obviously correct" and I can't tell
> if they are actually scripted or not.

Yes they are. The script that found these were:

----------------------8<------------------------
@@
identifier ptr, timer, rfield, slab;
@@
(
-	del_timer(&ptr->timer);
+	timer_shutdown(&ptr->timer);
|
-	del_timer_sync(&ptr->timer);
+	timer_shutdown_sync(&ptr->timer);
)
    ...
(
	kfree_rcu(ptr, rfield);
|
	kmem_cache_free(slab, ptr);
|
	kfree(ptr);
)
---------------------->8------------------------

So any function that had a del_timer*(&obj->timer) and then that obj
was freed with kfree(), kfree_rcu() or kmem_cache_free() was updated.

What I did manually was to make sure there was no exit of the routine
between those two calls. I'm sure coccinelle could do that too, but I'm
not good enough at it to add that feature.

The reason the patches don't look obvious is because the distance
between the del_timer() and the free may be quite far. I walked through
these patches at least 3 times manually to make sure they are all OK.


> 
> They don't *look* scripted, but I can't really tell.  I looked at the
> patches with ten lines of context, and I didn't see the immediately
> following kfree() even in that expanded patch context, so it's fairly
> far away.

Yes, some are like a 100 lines away.

> 
> Others in the series were *definitely* not scripted, doing clearly
> manual cleanups:
> 
> -    if (dch->timer.function) {
> -        del_timer(&dch->timer);
> -        dch->timer.function = NULL;
> -    }
> +    timer_shutdown(&dch->timer);
> 
> so no, this does *not* make me feel "ok, this is all trivial".

Sorry, I'll remove that. It's basically open-coding the
timer_shutdown() as the way it shuts down the timer is simply by
setting the timer.function to NULL.

> 
> IOW, I'd really want *just* the infrastructure and *just* the provably
> trivial stuff. If it wasn't some scripted really obvious thing that
> cannot possibly change anything and that wasn't then edited manually
> for some reason, I really don't want it early.
> 
> IOW, any early conversions I'd take are literally about removing pure
> mindless noise. Not about doing conversions.
> 
> And I wouldn't mind it as a single conversion patch that has the
> coccinelle script as the explanation.

I'll need to update the coccinelle script (or ask someone to give me a
fix) that catches the case of:

	del_timer(&obj->timer);

	if (x)
		goto out;

	kfree(obj);

out:
	return;


I'm sure it's a trivial change. I'll look into it some more.

I'm guessing you don't care about the case of:

	del_timer(&obj->timer);

	if (x)
		goto label;

label:

	kfree(obj);

As that's a bit more complex if we avoid the first goto case?
Even though the second case is obviously correct.

I believe both of these cases exist in the kernel. I manually removed
the places that my script found for the first case.

> 
> Really just THAT kind of "100% mindless conversion".

I'll look at making the most obviously correct case, where del_timer
and kfree have no goto or returns between them. We can always add the
rest in the merge window.

-- Steve