Re: [Patch 4/7] Add sysctl for delay accounting

[Dipankar Sarma <dipankar@in.ibm.com>]

On Mon, Feb 27, 2006 at 09:42:23AM +0100, Arjan van de Ven wrote:
> On Mon, 2006-02-27 at 03:38 -0500, Shailabh Nagar wrote:
> > Arjan van de Ven wrote:
> > 
> > The function needs to allocate task_delay_info structs for all tasks 
> > that might
> > have been forked since the last time delay accounting was turned off.
> > Either we have to count how many such tasks there are, or preallocate
> > nr_tasks (as an upper bound) and then use as many as needed.
> 
> it generally feels really fragile, especially with the task enumeration
> going to RCU soon. (eg you'd lose the ability to lock out new task
> creation)

I haven't yet seen any RCU-based code that does this. Can you point out
what patches you are talking about ? As of 2.6.16-rc4 and -rt15,
AFAICS, you can count tasks by holding the read side of tasklist_lock.
Granted it is a bit ugly to repeat this in order to overcome
the race on dropping tasklist_lock for allocation.

> On first sight it looks a lot better to allocate these things on demand,
> but I'm not sure how the sleeping-allocation would interact with the
> places it'd need to be called...

This could be a problem for contexts where sleeping cannot be permitted,
not to mention fast paths where blocking may introduce a skew. It seems
simpler to just let this happen during sysctl time.

Thanks
Dipankar