[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <bbafba4a-ec6b-936f-6ba2-d4b1c0700201@suse.cz>
Date: Sat, 2 Apr 2022 00:33:00 +0200
From: Vlastimil Babka <vbabka@...e.cz>
To: Zi Yan <ziy@...dia.com>,
Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Linux-MM <linux-mm@...ck.org>,
Steven Rostedt <rostedt@...dmis.org>,
David Hildenbrand <david@...hat.com>,
Mel Gorman <mgorman@...hsingularity.net>,
Mike Rapoport <rppt@...nel.org>,
Oscar Salvador <osalvador@...e.de>,
Andrew Morton <akpm@...ux-foundation.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2 2/2] mm: wrap __find_buddy_pfn() with a necessary buddy
page validation.
On 4/1/22 21:25, Zi Yan wrote:
> On 1 Apr 2022, at 15:01, Linus Torvalds wrote:
>
>> On Fri, Apr 1, 2022 at 11:56 AM Zi Yan <ziy@...dia.com> wrote:
>>>
>>> How about the patch below? If it looks good, I will send v3.
>>
>> I can't see anything worrisome, but by now I've looked at several
>> versions and who knows what I'm missing.
>>
>> Making it inline and allowing a NULL 'buddy_pfn' pointer for the cases
>> that don't care might be an option, but I don't think it matters
>> hugely.
>
> What do you mean by inlining it? Moving the function and __find_buddy_pfn()
> to mm/internal.h and mark both inline?
I would prefer that for the sake of __free_one_page().
> Something like below to allow a NULL 'buddy_pfn'? buddy_pfn is needed
> to store the result of __find_buddy_pfn(). The code does not look
> as nice as before.
>
> struct page *find_buddy_page_pfn(struct page *page, unsigned long pfn,
> unsigned int order, unsigned long *buddy_pfn)
> {
> struct page *buddy;
>
> if (buddy_pfn) {
> *buddy_pfn = __find_buddy_pfn(pfn, order);
> buddy = page + (*buddy_pfn - pfn);
> } else
> buddy = page + (__find_buddy_pfn(pfn, order) - pfn);
>
> if (page_is_buddy(page, buddy, order))
> return buddy;
> return NULL;
> }
>
> or
>
> struct page *find_buddy_page_pfn(struct page *page, unsigned long pfn,
> unsigned int order, unsigned long *buddy_pfn)
> {
> struct page *buddy;
> unsigned long local_buddy_pfn = __find_buddy_pfn(pfn, order);
>
> buddy = page + (local_buddy_pfn - pfn);
> if (buddy_pfn)
> *buddy_pfn = local_buddy_pfn;
>
> if (page_is_buddy(page, buddy, order))
> return buddy;
> return NULL;
> }
The latter looks better. I would also use a name e.g. '__buddy_pfn'
instead of 'local_'.
Thanks.
>
> --
> Best Regards,
> Yan, Zi
Powered by blists - more mailing lists