Linus Torvalds wrote:
>
> On Mon, 12 Nov 2001, Mingming cao wrote:
> >
> > This is a patch to dynamically allocate the data buffers for the disk
> > statistics, and to extend the gathering of disk statistics to include
> > major numbers greater than 15.
>
> I would suggest instead just moving the statistics into the request queue,
> at which point it should be nicely per-controller already, and quite
> independent of major numbers etc.
>
> Oh, and it will be faster too, because you only need one lookup.
>
> Linus
I saw your suggestions related to this about a year ago. I like the
idea of doing statistics per-controller and getting rid of disk_index().
But by moving statistics into the request queue, under the current
implementation, we have to allocate statistics memory for every major,
since for every major there is a request_queue asscoiated with
it(blk_dev[MAX_BLKDEV]). Do you care about this?
--
Mingming Cao
IBM Linux Technology Center
503-578-5024 IBM T/L: 775-5024
[email protected]
http://www.ibm.com/linux/ltc