linux-kernel - Re: [External] Re: [PATCH 3/6] mm: hugetlb: fix a race between freeing and dissolving the page

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <19e24147-f3cd-c6d2-dd0a-57e6182d60d8@oracle.com>
Date:   Tue, 5 Jan 2021 15:22:12 -0800
From:   Mike Kravetz <mike.kravetz@...cle.com>
To:     Muchun Song <songmuchun@...edance.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>,
        Naoya Horiguchi <n-horiguchi@...jp.nec.com>,
        Andi Kleen <ak@...ux.intel.com>, mhocko@...e.cz,
        Linux Memory Management List <linux-mm@...ck.org>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [External] Re: [PATCH 3/6] mm: hugetlb: fix a race between
 freeing and dissolving the page

On 1/4/21 6:55 PM, Muchun Song wrote:
> On Tue, Jan 5, 2021 at 8:02 AM Mike Kravetz <mike.kravetz@...cle.com> wrote:
>>
>> On 1/3/21 10:58 PM, Muchun Song wrote:
>>> There is a race condition between __free_huge_page()
>>> and dissolve_free_huge_page().
>>>
>>> CPU0:                         CPU1:
>>>
>>> // page_count(page) == 1
>>> put_page(page)
>>>   __free_huge_page(page)
>>>                               dissolve_free_huge_page(page)
>>>                                 spin_lock(&hugetlb_lock)
>>>                                 // PageHuge(page) && !page_count(page)
>>>                                 update_and_free_page(page)
>>>                                 // page is freed to the buddy
>>>                                 spin_unlock(&hugetlb_lock)
>>>     spin_lock(&hugetlb_lock)
>>>     clear_page_huge_active(page)
>>>     enqueue_huge_page(page)
>>>     // It is wrong, the page is already freed
>>>     spin_unlock(&hugetlb_lock)
>>>
>>> The race windows is between put_page() and spin_lock() which
>>> is in the __free_huge_page().
>>>
>>> We should make sure that the page is already on the free list
>>> when it is dissolved.
>>>
>>> Fixes: c8721bbbdd36 ("mm: memory-hotplug: enable memory hotplug to handle hugepage")
>>> Signed-off-by: Muchun Song <songmuchun@...edance.com>
>>> ---
>>>  mm/hugetlb.c | 48 ++++++++++++++++++++++++++++++++++++++++++++++++
>>>  1 file changed, 48 insertions(+)
>>>
>>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>>> index 1f3bf1710b66..72608008f8b4 100644
>>> --- a/mm/hugetlb.c
>>> +++ b/mm/hugetlb.c
>>> @@ -79,6 +79,21 @@ DEFINE_SPINLOCK(hugetlb_lock);
>>>  static int num_fault_mutexes;
>>>  struct mutex *hugetlb_fault_mutex_table ____cacheline_aligned_in_smp;
>>>
>>> +static inline bool PageHugeFreed(struct page *head)
>>> +{
>>> +     return page_private(head) == -1UL;
>>
>>         return page_private(head + 4) == -1UL;
>>
>>> +}
>>> +
>>> +static inline void SetPageHugeFreed(struct page *head)
>>> +{
>>> +     set_page_private(head + 4, -1UL);
>>> +}
>>> +
>>> +static inline void ClearPageHugeFreed(struct page *head)
>>> +{
>>> +     set_page_private(head + 4, 0);
>>> +}
>>
>> It is unfortunate that we can not use some existing value like
>> page_huge_active() to determine if dissolve_free_huge_page() should
>> proceed with freeing the page to buddy.  If the existing check,
>>
>>         if (!page_count(page)) {
>>
>> was changed to
>>
>>         if (!page_count(page) && !page_huge_active(page)) {
>>
>> the race window would be shrunk.  However, the most straight forward
>> way to fully close the window is with the approach taken here.
> 
> I also thought about this fix. But this is not enough. Because
> we just call put_page to free the HugeTLB page without
> setting activeness in some place (e.g. error handling
> routines).
> 
> If we use page_huge_active, we should set activeness
> before put_page. But we cannot guarantee this.

Just FYI,
I went back and explored the option of doing set_page_huge_active
when a page was put on the active list and clear_page_huge_active
when put on the free list.  This would be much like what you are
doing with PageHugeFreed.  Commit bcc54222309c which added page_huge_active
implied that this was possible.  Then I remembered a race fixed in
cb6acd01e2e4 that required delaying the call to set_page_huge_active
in hugetlb_no_page.  So, such a scheme would not work.

Also,
It seems we could use head[3].mapping for PageHugeFreed ?  Not much
of an advantage.  It does not add another tail page needed to store
page metadata.  And, this fits within the already defined
HUGETLB_CGROUP_MIN_ORDER.
-- 
Mike Kravetz