On 1/31/23 04:28, Jan Kara wrote:

On Tue 31-01-23 09:32:27, David Hildenbrand wrote:

quoted

On 30.01.23 23:15, Jens Axboe wrote:

quoted

On 1/30/23 3:12 PM, David Howells wrote:

quoted

John Hubbard [off-list ref] wrote:

quoted

This is something that we say when adding pin_user_pages_fast(),
yes. I doubt that I can quickly find the email thread, but we
measured it and weren't immediately able to come up with a way
to make it faster.

percpu counters maybe - add them up at the point of viewing?

They are percpu, see my last email. But for every 108 changes (on
my system), they will do two atomic_long_adds(). So not very
useful for anything but low frequency modifications.

Can we just treat the whole acquired/released accounting as a debug
mechanism to detect missing releases and do it only for debug kernels?

Yes, AFAIK it is just a debug mechanism for helping to find out issues with
page pinning conversions. So I think we can put this behind some debugging
ifdef. John?

Yes, just for debugging. I wrote a little note just now in response to
the patch about how we ended up here: "yes, it's time to hide these
behind an ifdef".

thanks,
-- 
John Hubbard
NVIDIA