lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <7e55d6eb-3384-4cc0-80ea-880ef2175121@arm.com>
Date: Fri, 9 Aug 2024 10:00:21 +0100
From: Ryan Roberts <ryan.roberts@....com>
To: David Hildenbrand <david@...hat.com>, Barry Song <21cnbao@...il.com>,
 akpm@...ux-foundation.org, linux-mm@...ck.org
Cc: chrisl@...nel.org, kaleshsingh@...gle.com, kasong@...cent.com,
 linux-kernel@...r.kernel.org, ioworker0@...il.com,
 baolin.wang@...ux.alibaba.com, ziy@...dia.com, hanchuanhua@...o.com,
 Barry Song <v-songbaohua@...o.com>
Subject: Re: [PATCH RFC 1/2] mm: collect the number of anon large folios

On 09/08/2024 09:39, David Hildenbrand wrote:
> On 09.08.24 10:13, Ryan Roberts wrote:
>> On 08/08/2024 02:04, Barry Song wrote:
>>> From: Barry Song <v-songbaohua@...o.com>
>>>
>>> When a new anonymous mTHP is added to the rmap, we increase the count.
>>> We reduce the count whenever an mTHP is completely unmapped.
>>>
>>> Signed-off-by: Barry Song <v-songbaohua@...o.com>
>>> ---
>>>   Documentation/admin-guide/mm/transhuge.rst |  5 +++++
>>>   include/linux/huge_mm.h                    | 15 +++++++++++++--
>>>   mm/huge_memory.c                           |  2 ++
>>>   mm/rmap.c                                  |  3 +++
>>>   4 files changed, 23 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/Documentation/admin-guide/mm/transhuge.rst
>>> b/Documentation/admin-guide/mm/transhuge.rst
>>> index 058485daf186..715f181543f6 100644
>>> --- a/Documentation/admin-guide/mm/transhuge.rst
>>> +++ b/Documentation/admin-guide/mm/transhuge.rst
>>> @@ -527,6 +527,11 @@ split_deferred
>>>           it would free up some memory. Pages on split queue are going to
>>>           be split under memory pressure, if splitting is possible.
>>>   +anon_num
>>> +       the number of anon huge pages we have in the whole system.
>>> +       These huge pages could be still entirely mapped and have partially
>>> +       unmapped and unused subpages.
>>
>> nit: "entirely mapped and have partially unmapped and unused subpages" ->
>> "entirely mapped or have partially unmapped/unused subpages"
>>
>>> +
>>>   As the system ages, allocating huge pages may be expensive as the
>>>   system uses memory compaction to copy data around memory to free a
>>>   huge page for use. There are some counters in ``/proc/vmstat`` to help
>>> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
>>> index e25d9ebfdf89..294c348fe3cc 100644
>>> --- a/include/linux/huge_mm.h
>>> +++ b/include/linux/huge_mm.h
>>> @@ -281,6 +281,7 @@ enum mthp_stat_item {
>>>       MTHP_STAT_SPLIT,
>>>       MTHP_STAT_SPLIT_FAILED,
>>>       MTHP_STAT_SPLIT_DEFERRED,
>>> +    MTHP_STAT_NR_ANON,
>>>       __MTHP_STAT_COUNT
>>>   };
>>>   @@ -291,14 +292,24 @@ struct mthp_stat {
>>>   #ifdef CONFIG_SYSFS
>>>   DECLARE_PER_CPU(struct mthp_stat, mthp_stats);
>>>   -static inline void count_mthp_stat(int order, enum mthp_stat_item item)
>>> +static inline void mod_mthp_stat(int order, enum mthp_stat_item item, int
>>> delta)
>>>   {
>>>       if (order <= 0 || order > PMD_ORDER)
>>>           return;
>>>   -    this_cpu_inc(mthp_stats.stats[order][item]);
>>> +    this_cpu_add(mthp_stats.stats[order][item], delta);
>>> +}
>>> +
>>> +static inline void count_mthp_stat(int order, enum mthp_stat_item item)
>>> +{
>>> +    mod_mthp_stat(order, item, 1);
>>>   }
>>> +
>>>   #else
>>> +static inline void mod_mthp_stat(int order, enum mthp_stat_item item, int
>>> delta)
>>> +{
>>> +}
>>> +
>>>   static inline void count_mthp_stat(int order, enum mthp_stat_item item)
>>>   {
>>>   }
>>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>>> index 697fcf89f975..b6bc2a3791e3 100644
>>> --- a/mm/huge_memory.c
>>> +++ b/mm/huge_memory.c
>>> @@ -578,6 +578,7 @@ DEFINE_MTHP_STAT_ATTR(shmem_fallback_charge,
>>> MTHP_STAT_SHMEM_FALLBACK_CHARGE);
>>>   DEFINE_MTHP_STAT_ATTR(split, MTHP_STAT_SPLIT);
>>>   DEFINE_MTHP_STAT_ATTR(split_failed, MTHP_STAT_SPLIT_FAILED);
>>>   DEFINE_MTHP_STAT_ATTR(split_deferred, MTHP_STAT_SPLIT_DEFERRED);
>>> +DEFINE_MTHP_STAT_ATTR(anon_num, MTHP_STAT_NR_ANON);
>>>     static struct attribute *stats_attrs[] = {
>>>       &anon_fault_alloc_attr.attr,
>>> @@ -591,6 +592,7 @@ static struct attribute *stats_attrs[] = {
>>>       &split_attr.attr,
>>>       &split_failed_attr.attr,
>>>       &split_deferred_attr.attr,
>>> +    &anon_num_attr.attr,
>>>       NULL,
>>>   };
>>>   diff --git a/mm/rmap.c b/mm/rmap.c
>>> index 901950200957..2b722f26224c 100644
>>> --- a/mm/rmap.c
>>> +++ b/mm/rmap.c
>>> @@ -1467,6 +1467,7 @@ void folio_add_new_anon_rmap(struct folio *folio,
>>> struct vm_area_struct *vma,
>>>       }
>>>         __folio_mod_stat(folio, nr, nr_pmdmapped);
>>> +    mod_mthp_stat(folio_order(folio), MTHP_STAT_NR_ANON, 1);
>>>   }
>>>     static __always_inline void __folio_add_file_rmap(struct folio *folio,
>>> @@ -1582,6 +1583,8 @@ static __always_inline void __folio_remove_rmap(struct
>>> folio *folio,
>>>           list_empty(&folio->_deferred_list))
>>>           deferred_split_folio(folio);
>>>       __folio_mod_stat(folio, -nr, -nr_pmdmapped);
>>> +    if (folio_test_anon(folio) && !atomic_read(mapped))
>>
>> Agree that atomic_read() is dodgy here.
>>
>> Not sure I fully understand why David prefers to do the unaccounting at
>> free-time though? It feels unbalanced to me to increment when first mapped but
>> decrement when freed. Surely its safer to either use alloc/free or use first
>> map/last map?
> 
> Doing it when we set/clear folio->mapping is straight forward.
> 
> Anon folios currently come to live when we first map them, and they stay that
> way until we free them.
> 
> In the future, we'll have to move that anon handling further out, when if have
> to allocate anon-specific memdesc ahead of time, then, it will be clued to that
> lifetime.
> 
>>
>> If using alloc/free isn't there a THP constructor/destructor that prepares the
>> deferred list? (My memory may be failing me). Could we use that?
> 
> Likely the deconstructor could work as well. Not sure if that is any better than
> the freeing path where folio->mapping currently gets cleared.
> 
> The generic constructor certainly won't work right now. That's not where the
> "anon" part comes to live.
> 
> Let's take a look how NR_FILE_THPS is handled:
> 
> __filemap_add_folio() increments it -- when we set folio->mapping
> __filemap_remove_folio() (->filemap_unaccount_folio) decrements it -- after
> which we usually call page_cache_delete() to set folio->mapping = NULL;
> 

OK got it, thanks!


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ