[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <9855d2b4-aad7-3efe-c00c-851b44a01c3e@huawei.com>
Date: Thu, 17 Nov 2022 14:59:31 +0800
From: Liu Shixin <liushixin2@...wei.com>
To: Anshuman Khandual <anshuman.khandual@....com>,
Catalin Marinas <catalin.marinas@....com>,
Will Deacon <will@...nel.org>,
Denys Vlasenko <dvlasenk@...hat.com>,
Kefeng Wang <wangkefeng.wang@...wei.com>,
"David Hildenbrand" <dhildenb@...hat.com>,
Rafael Aquini <raquini@...hat.com>,
"Pasha Tatashin" <pasha.tatashin@...een.com>
CC: <linux-arm-kernel@...ts.infradead.org>,
<linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 1/2] arm64/mm: fix incorrect file_map_count for non-leaf
pmd/pud
On 2022/11/17 12:09, Anshuman Khandual wrote:
>
> On 11/16/22 14:08, Liu Shixin wrote:
>> The page table check trigger BUG_ON() unexpectedly when collapse hugepage:
>>
>> ------------[ cut here ]------------
>> kernel BUG at mm/page_table_check.c:82!
>> Internal error: Oops - BUG: 00000000f2000800 [#1] SMP
>> Dumping ftrace buffer:
>> (ftrace buffer empty)
>> Modules linked in:
>> CPU: 6 PID: 68 Comm: khugepaged Not tainted 6.1.0-rc3+ #750
>> Hardware name: linux,dummy-virt (DT)
>> pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>> pc : page_table_check_clear.isra.0+0x258/0x3f0
>> lr : page_table_check_clear.isra.0+0x240/0x3f0
>> [...]
>> Call trace:
>> page_table_check_clear.isra.0+0x258/0x3f0
>> __page_table_check_pmd_clear+0xbc/0x108
>> pmdp_collapse_flush+0xb0/0x160
>> collapse_huge_page+0xa08/0x1080
>> hpage_collapse_scan_pmd+0xf30/0x1590
>> khugepaged_scan_mm_slot.constprop.0+0x52c/0xac8
>> khugepaged+0x338/0x518
>> kthread+0x278/0x2f8
>> ret_from_fork+0x10/0x20
>> [...]
>>
>> Since pmd_user_accessible_page() doesn't check if a pmd is leaf, it
>> decrease file_map_count for a non-leaf pmd comes from collapse_huge_page().
>> and so trigger BUG_ON() unexpectedly.
> Could you please provide the pmd_val() on the pmd entry, that triggers this
> BUG_ON() here ? Only additional thing pmd_leaf() ensures, is that the entry
> is not a table one.
>
> #define pmd_leaf(pmd) (pmd_present(pmd) && !pmd_table(pmd))
>
> collapse_huge_page() pmd is non-leaf because it has table bit on ?
The pmd_val is 0x80000004c367003. It is indeed a table entry. collapse_huge_page()
will replace page table of page granularity with block granularity. Before this replace,
it will call pmdp_collapse_flush() to clear the table pmd. In this function, the table pmd
do the check unexpectdly and trigger the BUG_ON().
diff --git a/mm/pgtable-generic.c b/mm/pgtable-generic.c
index 90ab721a12a8..a5c2380bac4d 100644
--- a/mm/pgtable-generic.c
+++ b/mm/pgtable-generic.c
@@ -221,6 +221,7 @@ pmd_t pmdp_collapse_flush(struct vm_area_struct *vma, unsigned long address,
VM_BUG_ON(address & ~HPAGE_PMD_MASK);
VM_BUG_ON(pmd_trans_huge(*pmdp));
+ pr_err("pmd_val is 0x%lx\n", pmd_val(*pmdp));
pmd = pmdp_huge_get_and_clear(vma->vm_mm, address, pmdp);
/* collapse entails shooting down ptes not pmd */
Thanks,
Liu Shixin
.
>
>> Fix this problem by using pmd_leaf() insteal of pmd_present() in
>> pmd_user_accessible_page(). Moreover, use pud_leaf() for
>> pud_user_accessible_page() too.
>>
>> Fixes: 42b2547137f5 ("arm64/mm: enable ARCH_SUPPORTS_PAGE_TABLE_CHECK")
>> Reported-by: Denys Vlasenko <dvlasenk@...hat.com>
>> Signed-off-by: Liu Shixin <liushixin2@...wei.com>
>> ---
>> arch/arm64/include/asm/pgtable.h | 4 ++--
>> 1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
>> index 71a1af42f0e8..edf6625ce965 100644
>> --- a/arch/arm64/include/asm/pgtable.h
>> +++ b/arch/arm64/include/asm/pgtable.h
>> @@ -863,12 +863,12 @@ static inline bool pte_user_accessible_page(pte_t pte)
>>
>> static inline bool pmd_user_accessible_page(pmd_t pmd)
>> {
>> - return pmd_present(pmd) && (pmd_user(pmd) || pmd_user_exec(pmd));
>> + return pmd_leaf(pmd) && (pmd_user(pmd) || pmd_user_exec(pmd));
>> }
>>
>> static inline bool pud_user_accessible_page(pud_t pud)
>> {
>> - return pud_present(pud) && pud_user(pud);
>> + return pud_leaf(pud) && pud_user(pud);
>> }
>> #endif
>>
> .
>
Powered by blists - more mailing lists