On Mon, Sep 20, 2021 at 03:41:16PM +0100, John Garry wrote:

On 20/09/2021 02:53, Matthew Wilcox wrote:

quoted

On Mon, Sep 20, 2021 at 01:09:38AM +0000, Hyeonggon Yoo wrote:

quoted

Hello Matthew, Thanks to give me a comment! I appreciate it.

On Sun, Sep 19, 2021 at 08:17:44PM +0100, Matthew Wilcox wrote:

quoted

On Sun, Sep 19, 2021 at 04:42:39PM +0000, Hyeonggon Yoo wrote:

quoted

It is just simple proof of concept, and not ready for submission yet.
There can be wrong code (like wrong gfp flags, or wrong error handling,
etc) it is just simple proof of concept. I want comment from you.

Have you read:

https://www.usenix.org/legacy/event/usenix01/full_papers/bonwick/bonwick_html/
The relevant part of that paper is section 3, magazines.  We should have
low and high water marks for number of objects

In case unknown, jfyi there is an implementation of this in
drivers/iommu/iova.c

Thanks for good information. I'll take a look at it!

Thanks,
Hyeonggon

Thanks,
John

quoted

quoted

I haven't read that before, but after reading it seems not different from
SLAB's percpu queuing.

quoted

and we should allocate
from / free to the slab allocator in batches.  Slab has bulk alloc/free
APIs already.