[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <c96a97bc-4638-4910-ba27-262fb6cedc96@redhat.com>
Date: Tue, 5 Aug 2025 15:22:04 +0200
From: David Hildenbrand <david@...hat.com>
To: Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Jason Gunthorpe <jgg@...dia.com>,
Alex Williamson <alex.williamson@...hat.com>,
"kvm@...r.kernel.org" <kvm@...r.kernel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"lizhe.67@...edance.com" <lizhe.67@...edance.com>
Subject: Re: [GIT PULL] VFIO updates for v6.17-rc1
On 05.08.25 15:15, Linus Torvalds wrote:
> On Tue, 5 Aug 2025 at 16:05, David Hildenbrand <david@...hat.com> wrote:
>>
>> So I don't like the idea of micro-optimizing num_pages_contiguous() by
>> adding weird tweaks to the core for that.
>
> Seriously - take a look at that suggested sequence I posted, and tell
> me that it isn't *MORE* obvious than the horror that is nth_page().
>
> Honestly, if anybody thinks nth_page() is obvious and good, I think
> they have some bad case of Stockholm syndrome.
>
> This isn't about micro-optimizing. This is about not writing complete
> garbage code that makes no sense.
>
> nth_page() is a disgusting thing that is designed to look up
> known-contiguous pages. That code mis-used it for *testing* for being
> contiguous. It may have _worked_, but it was the wrong thing to do.
>
> nth_page() in general should just not exist. I don't actually believe
> there is any valid reason for it. I do not believe we should actually
> have valid consecutive allocations of pages across sections.
Oh, just to add to that, 1 GiB folios (hugetlb, dax) are the main reason
why we use it in things like folio_page(), and also why folio_page_idx()
is so horrible.
--
Cheers,
David / dhildenb
Powered by blists - more mailing lists