Re: SLUB: Support for statistics to help analyze allocator behavior

[Christoph Lameter]

On Tue, 5 Feb 2008, Eric Dumazet wrote:

> > Well we could do the same as for numa stats. Output the global count and 
> > then add
> > 
> > c<proc>=count
> > 
> 
> Yes, or the reverse, to avoid two loops and possible sum errors (Sum of 
> c<proc>=count different than the global count)

The numa output uses only one loop and so I think we could do the same 
here. Its good to have the global number first that way existing tools can 
simply read a number and get what they intuitively expect.

> Since text##_show is going to be too big, you could use one function 
> instead of several ones ?

Sure.

> (and char *buf is PAGE_SIZE, so you should add a limit ?)

Yes we must do so because support for 4k processors etc is on the horizon.

> Note I used for_each_possible_cpu() here instead of 'online' variant, or 
> stats might be corrupted when a cpu goes offline.

Hmmm.. We are thinking about freeing percpu areas when a cpu goes offline. 
So we would need to fold statistics into another cpu if this is a cocnern. 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/