Re: [PATCH v2 2/3] lockdep: Remove BROKEN flag of LOCKDEP_CROSSRELEASE

[Bart Van Assche]

On Thu, 2017-10-19 at 17:34 +0200, Thomas Gleixner wrote:
> I really disagree with your reasoning completely
> 
> 1) When lockdep was introduced more than ten years ago it was far from
>    perfect and we spent a reasonable amount of time to improve it, analyze
>    false positives and add the missing annotations all over the tree. That
>    was a process which took years.
> 
> 2) Surely nobody is interested in wasting time on analyzing false
>    positives, but your (and other peoples) attidute of 'none of my
>    business' is what makes kernel development extremly frustrating.
> 
>    It should be in the interest of everybody involved in kernel development
>    to help with improving such features and not to lean back and wait for
>    others to bring it into a shape which allows you to use it as you see
>    fit.
> 
> That's not how community works and lockdep would not be in the shape it is
> today, if only a handful of people would have used and improved it. Such
> things only work when used widely and when we get enough information so we
> can address the weak spots.

Hello Thomas,

It seems like you are missing my point. Cross-release checking is really
*broken* as a concept. It is impossible to improve it to the same reliability
level as the kernel v4.13 lockdep code. Hence my request to make it possible
to disable cross-release checking if PROVE_LOCKING is enabled.

Consider the following example from the cross-release documentation:

   TASK X			   TASK Y
   ------			   ------
				   acquire AX
   acquire B /* A dependency 'AX -> B' exists */
   release B
   release AX held by Y

My understanding is that the cross-release code will add (AX, B) to the lock
order graph after having encountered the above code. I think that's wrong
because if the following sequence (Y: acquire AX, X: acquire B, X: release B)
is encountered again that there is no guarantee that AX can only be released
by X. Any task other than X could release that synchronization object too.

Bart.