[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <ca6b21a3-efdc-5f7f-7b62-22a07a33c424@redhat.com>
Date: Tue, 8 Oct 2019 16:51:03 +0200
From: David Hildenbrand <david@...hat.com>
To: lantianyu1986@...il.com, pbonzini@...hat.com, rkrcmar@...hat.com,
tglx@...utronix.de, mingo@...hat.com, bp@...en8.de, hpa@...or.com,
x86@...nel.org, dave.hansen@...ux.intel.com, luto@...nel.org,
peterz@...radead.org, kys@...rosoft.com, haiyangz@...rosoft.com,
sthemmin@...rosoft.com, sashal@...nel.org,
akpm@...ux-foundation.org, rppt@...ux.ibm.com, jgross@...e.com,
mhocko@...e.com, paul.burton@...s.com, m.mizuma@...fujitsu.com,
huang.zijiang@....com.cn, karahmed@...zon.de,
dan.j.williams@...el.com, bhelgaas@...gle.com, osalvador@...e.de,
rdunlap@...radead.org, richardw.yang@...ux.intel.com,
pavel.tatashin@...rosoft.com, cai@....pw, arunks@...eaurora.org,
vbabka@...e.cz, mgorman@...hsingularity.net,
alexander.h.duyck@...ux.intel.com, glider@...gle.com,
logang@...tatee.com, bsingharora@...il.com, bhe@...hat.com,
Tianyu.Lan@...rosoft.com, michael.h.kelley@...rosoft.com
Cc: kvm@...r.kernel.org, linux-kernel@...r.kernel.org,
linux-hyperv@...r.kernel.org, linux-mm@...ck.org,
vkuznets@...hat.com
Subject: Re: [RFC PATCH] mm: set memory section offline when all its pages are
offline.
On 08.10.19 16:36, lantianyu1986@...il.com wrote:
> From: Tianyu Lan <Tianyu.Lan@...rosoft.com>
>
> If size of offline memory region passed to offline_pages() is
> not aligned with PAGES_PER_SECTION, memory section will be set
> to offline in the offline_mem_sections() with some pages of
> memory section online. Fix it, Update memory section status after
> marking offline pages as "reserved" in __offline_isolated_pages()
> and check all pages in memory are reserved or not before setting
> memory section offline.
>
> Signed-off-by: Tianyu Lan <Tianyu.Lan@...rosoft.com>
> ---
> This patch is to prepare for hot remove memory function in Hyper-V
> balloon driver. It requests to offline memory with random size.
I proposed roughly the same a while ago. See
https://lkml.org/lkml/2018/4/30/207
Memory onlining/offlining works in memory block granularity only.
Sub-sections, you have to emulate it on top, similar to how hyper-v
balloon handles it already. E.g., have a look how virtio-mem handles it
using alloc_contig_range/free_contig_range and PG_offline extensions.
https://lkml.org/lkml/2018/4/30/207
So a clear NACK from my side.
> ---
> mm/page_alloc.c | 3 ++-
> mm/sparse.c | 10 ++++++++++
> 2 files changed, 12 insertions(+), 1 deletion(-)
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index dbd0d5cbbcbb..cc02866924ae 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -8540,7 +8540,6 @@ __offline_isolated_pages(unsigned long start_pfn, unsigned long end_pfn)
> if (pfn == end_pfn)
> return offlined_pages;
>
> - offline_mem_sections(pfn, end_pfn);
> zone = page_zone(pfn_to_page(pfn));
> spin_lock_irqsave(&zone->lock, flags);
> pfn = start_pfn;
> @@ -8576,6 +8575,8 @@ __offline_isolated_pages(unsigned long start_pfn, unsigned long end_pfn)
> }
> spin_unlock_irqrestore(&zone->lock, flags);
>
> + offline_mem_sections(pfn, end_pfn);
> +
> return offlined_pages;
> }
> #endif
> diff --git a/mm/sparse.c b/mm/sparse.c
> index fd13166949b5..eb5860487b84 100644
> --- a/mm/sparse.c
> +++ b/mm/sparse.c
> @@ -571,6 +571,7 @@ void online_mem_sections(unsigned long start_pfn, unsigned long end_pfn)
> void offline_mem_sections(unsigned long start_pfn, unsigned long end_pfn)
> {
> unsigned long pfn;
> + int i;
>
> for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) {
> unsigned long section_nr = pfn_to_section_nr(pfn);
> @@ -583,6 +584,15 @@ void offline_mem_sections(unsigned long start_pfn, unsigned long end_pfn)
> if (WARN_ON(!valid_section_nr(section_nr)))
> continue;
>
> + /*
> + * Check whether all pages in the section are reserverd before
> + * setting setction offline.
> + */
> + for (i = 0; i < PAGES_PER_SECTION; i++)
> + if (!PageReserved(pfn_to_page(
> + SECTION_ALIGN_DOWN(pfn + i))))
> + continue;
> +
> ms = __nr_to_section(section_nr);
> ms->section_mem_map &= ~SECTION_IS_ONLINE;
> }
>
--
Thanks,
David / dhildenb
Powered by blists - more mailing lists