Re: [PATCH v4 2/3] mm: introduce put_user_page*(), placeholder versions

[John Hubbard]

On 10/18/18 3:19 AM, Jan Kara wrote:
> On Thu 11-10-18 20:53:34, John Hubbard wrote:
>> On 10/11/18 6:23 PM, John Hubbard wrote:
>>> On 10/11/18 6:20 AM, Jason Gunthorpe wrote:
>>>> On Thu, Oct 11, 2018 at 10:49:29AM +0200, Jan Kara wrote:
[...]
> Well, put_page() cannot assert page is not dma-pinned as someone can still
> to get_page(), put_page() on dma-pinned page and that must not barf. But
> put_page() could assert that if the page is pinned, refcount is >=
> pincount. That will detect leaked pin references relatively quickly.
> 

That assertion is definitely a life saver. I've been attempting a combination
of finishing up more call site conversions, and runtime testing, and this
lights up the missing conversions pretty nicely.

As I mentioned in another thread just now, I'll send out an updated RFC this week,
so that people can look through it well before the LPC (next week).

thanks,
-- 
John Hubbard
NVIDIA