[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <e2d3b996-e1a9-ccc1-8b7a-31df840b8ebb@nvidia.com>
Date: Fri, 6 Nov 2020 12:56:47 -0800
From: Ralph Campbell <rcampbell@...dia.com>
To: Christoph Hellwig <hch@....de>
CC: <linux-mm@...ck.org>, <nouveau@...ts.freedesktop.org>,
<linux-kselftest@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
"Jerome Glisse" <jglisse@...hat.com>,
John Hubbard <jhubbard@...dia.com>,
"Alistair Popple" <apopple@...dia.com>,
Jason Gunthorpe <jgg@...dia.com>,
Bharata B Rao <bharata@...ux.ibm.com>,
Zi Yan <ziy@...dia.com>,
"Kirill A . Shutemov" <kirill.shutemov@...ux.intel.com>,
Yang Shi <yang.shi@...ux.alibaba.com>,
Ben Skeggs <bskeggs@...hat.com>, Shuah Khan <shuah@...nel.org>,
"Andrew Morton" <akpm@...ux-foundation.org>
Subject: Re: [PATCH v3 1/6] mm/thp: add prep_transhuge_device_private_page()
On 11/5/20 11:55 PM, Christoph Hellwig wrote:
> On Thu, Nov 05, 2020 at 04:51:42PM -0800, Ralph Campbell wrote:
>> +extern void prep_transhuge_device_private_page(struct page *page);
>
> No need for the extern.
Right, I was just copying the style.
Would you like to see a preparatory patch that removes extern for the other
declarations in huge_mm.h?
>> +static inline void prep_transhuge_device_private_page(struct page *page)
>> +{
>> +}
>
> Is the code to call this even reachable if THP support is configured
> out? If not just declaring it unconditionally and letting dead code
> elimination do its job might be a tad cleaner.
The HMM test driver depends on TRANSPARENT_HUGEPAGE but the nouveau SVM
option doesn't and SVM is still useful if TRANSPARENT_HUGEPAGE is not configured.
The problem with defining prep_transhuge_device_private_page() in huge_mm.h
as a static inline function is that prep_compound_page() and prep_transhuge_page()
would have to be EXPORT_SYMBOL_GPL which are currently mm internal only.
The intent is to make this helper callable by separate device driver modules
using struct pages created with memremap_pages().
>> +void prep_transhuge_device_private_page(struct page *page)
>
> I think a kerneldoc comment explaining what this function is useful for
> would be helpful.
That is a good idea. I'll add it to v4.
Powered by blists - more mailing lists