Re: [External] Re: [PATCH v3 09/21] mm/hugetlb: Free the vmemmap pages associated with each hugetlb page

From: Muchun Song <songmuchun@bytedance.com>
To: Oscar Salvador <osalvador@suse.de>
Cc: Jonathan Corbet <corbet@lwn.net>,
	Mike Kravetz <mike.kravetz@oracle.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	mingo@redhat.com, bp@alien8.de, x86@kernel.org, hpa@zytor.com,
	dave.hansen@linux.intel.com, luto@kernel.org,
	Peter Zijlstra <peterz@infradead.org>,
	viro@zeniv.linux.org.uk,
	Andrew Morton <akpm@linux-foundation.org>,
	paulmck@kernel.org, mchehab+huawei@kernel.org,
	pawan.kumar.gupta@linux.intel.com,
	Randy Dunlap <rdunlap@infradead.org>,
	oneukum@suse.com, anshuman.khandual@arm.com, jroedel@suse.de,
	Mina Almasry <almasrymina@google.com>,
	David Rientjes <rientjes@google.com>,
	Matthew Wilcox <willy@infradead.org>,
	Michal Hocko <mhocko@suse.com>,
	Xiongchun duan <duanxiongchun@bytedance.com>,
	linux-doc@vger.kernel.org, LKML <linux-kernel@vger.kernel.org>,
	Linux Memory Management List <linux-mm@kvack.org>,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>
Subject: Re: [External] Re: [PATCH v3 09/21] mm/hugetlb: Free the vmemmap pages associated with each hugetlb page
Date: Tue, 10 Nov 2020 18:47:08 +0800	[thread overview]
Message-ID: <CAMZfGtW0nwhdgwUwwq5SXgEAk3+6cyDfM5n28UerVuAxatwj4g@mail.gmail.com> (raw)
In-Reply-To: <20201110094830.GA25373@linux>

On Tue, Nov 10, 2020 at 5:48 PM Oscar Salvador <osalvador@suse.de> wrote:
>
> On Tue, Nov 10, 2020 at 02:40:54PM +0800, Muchun Song wrote:
> > Only the first HugeTLB page should split the PMD to PTE. The other 63
> > HugeTLB pages
> > do not need to split. Here I want to make sure we are the first.
>
> I think terminology is loosing me here.
>
> Say you allocate a 2MB HugeTLB page at ffffea0004100000.
>
> The vmemmap range that the represents this is ffffea0004000000 - ffffea0004200000.
> That is a 2MB chunk PMD-mapped.
> So, in order to free some of those vmemmap pages, we need to break down
> that area, remapping it to PTE-based.
> I know what you mean, but we are not really splitting hugetlg pages, but
> the memmap range they are represented with.

Yeah, you are right. We are splitting the vmemmap instead of hugetlb.
Sorry for the confusion.

>
> About:
>
> "Only the first HugeTLB page should split the PMD to PTE. The other 63
> HugeTLB pages
> do not need to split. Here I want to make sure we are the first."
>
> That only refers to gigantic pages, right?

Yeah, now it only refers to gigantic pages. Originally, I also wanted to merge
vmemmap PTE to PMD for normal 2MB HugeTLB pages. So I introduced
those macros(e.g. freed_vmemmap_hpage). For 2MB HugeTLB pages, I
haven't found an elegant solution. Hopefully, when you or someone have
read all of the patch series, we can come up with an elegant solution to
merge PTE.

Thanks.

>
> > > > +static void free_huge_page_vmemmap(struct hstate *h, struct page *head)
> > > > +{
> > > > +     pmd_t *pmd;
> > > > +     spinlock_t *ptl;
> > > > +     LIST_HEAD(free_pages);
> > > > +
> > > > +     if (!free_vmemmap_pages_per_hpage(h))
> > > > +             return;
> > > > +
> > > > +     pmd = vmemmap_to_pmd(head);
> > > > +     ptl = vmemmap_pmd_lock(pmd);
> > > > +     if (vmemmap_pmd_huge(pmd)) {
> > > > +             VM_BUG_ON(!pgtable_pages_to_prealloc_per_hpage(h));
> > >
> > > I think that checking for free_vmemmap_pages_per_hpage is enough.
> > > In the end, pgtable_pages_to_prealloc_per_hpage uses free_vmemmap_pages_per_hpage.
> >
> > The free_vmemmap_pages_per_hpage is not enough. See the comments above.
>
> My comment was about the VM_BUG_ON.

Sorry, yeah, we can drop it. Thanks.

>
>
> --
> Oscar Salvador
> SUSE L3

-- 
Yours,
Muchun