Re: [PATCH] mm: fix possible OOB in numa_rebuild_large_mapping()

[Baolin Wang <baolin.wang@linux.alibaba.com>]

On 2024/6/11 20:32, David Hildenbrand wrote:
> On 07.06.24 12:32, Kefeng Wang wrote:
>> The large folio is mapped with folio size aligned virtual address during
>> the pagefault, eg, 'addr = ALIGN_DOWN(vmf->address, nr_pages * 
>> PAGE_SIZE)'
>> in do_anonymous_page(), but after the mremap(), the virtual address only
>> require PAGE_SIZE aligned, also pte is moved to new in 
>> move_page_tables(),
>> then traverse the new pte in numa_rebuild_large_mapping() will hint the
>> following issue,
>>
>>     Unable to handle kernel paging request at virtual address 
>> 00000a80c021a788
>>     Mem abort info:
>>       ESR = 0x0000000096000004
>>       EC = 0x25: DABT (current EL), IL = 32 bits
>>       SET = 0, FnV = 0
>>       EA = 0, S1PTW = 0
>>       FSC = 0x04: level 0 translation fault
>>     Data abort info:
>>       ISV = 0, ISS = 0x00000004, ISS2 = 0x00000000
>>       CM = 0, WnR = 0, TnD = 0, TagAccess = 0
>>       GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0
>>     user pgtable: 4k pages, 48-bit VAs, pgdp=00002040341a6000
>>     [00000a80c021a788] pgd=0000000000000000, p4d=0000000000000000
>>     Internal error: Oops: 0000000096000004 [#1] SMP
>>     ...
>>     CPU: 76 PID: 15187 Comm: git Kdump: loaded Tainted: G        
>> W          6.10.0-rc2+ #209
>>     Hardware name: Huawei TaiShan 2280 V2/BC82AMDD, BIOS 1.79 08/21/2021
>>     pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>>     pc : numa_rebuild_large_mapping+0x338/0x638
>>     lr : numa_rebuild_large_mapping+0x320/0x638
>>     sp : ffff8000b41c3b00
>>     x29: ffff8000b41c3b30 x28: ffff8000812a0000 x27: 00000000000a8000
>>     x26: 00000000000000a8 x25: 0010000000000001 x24: ffff20401c7170f0
>>     x23: 0000ffff33a1e000 x22: 0000ffff33a76000 x21: ffff20400869eca0
>>     x20: 0000ffff33976000 x19: 00000000000000a8 x18: ffffffffffffffff
>>     x17: 0000000000000000 x16: 0000000000000020 x15: ffff8000b41c36a8
>>     x14: 0000000000000000 x13: 205d373831353154 x12: 5b5d333331363732
>>     x11: 000000000011ff78 x10: 000000000011ff10 x9 : ffff800080273f30
>>     x8 : 000000320400869e x7 : c0000000ffffd87f x6 : 00000000001e6ba8
>>     x5 : ffff206f3fb5af88 x4 : 0000000000000000 x3 : 0000000000000000
>>     x2 : 0000000000000000 x1 : fffffdffc0000000 x0 : 00000a80c021a780
>>     Call trace:
>>      numa_rebuild_large_mapping+0x338/0x638
>>      do_numa_page+0x3e4/0x4e0
>>      handle_pte_fault+0x1bc/0x238
>>      __handle_mm_fault+0x20c/0x400
>>      handle_mm_fault+0xa8/0x288
>>      do_page_fault+0x124/0x498
>>      do_translation_fault+0x54/0x80
>>      do_mem_abort+0x4c/0xa8
>>      el0_da+0x40/0x110
>>      el0t_64_sync_handler+0xe4/0x158
>>      el0t_64_sync+0x188/0x190
>>
>> Fix it by correct the start and end, which may lead to only rebuild part
>> of large mapping in one numa page fault, there is no issue since other 
>> part
>> could rebuild by another pagefault.
>>
>> Fixes: d2136d749d76 ("mm: support multi-size THP numa balancing")
>> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
>> ---
>>   mm/memory.c | 24 +++++++++++++++---------
>>   1 file changed, 15 insertions(+), 9 deletions(-)
>>
>> diff --git a/mm/memory.c b/mm/memory.c
>> index db9130488231..0ad57b6485ca 100644
>> --- a/mm/memory.c
>> +++ b/mm/memory.c
>> @@ -5223,15 +5223,21 @@ static void numa_rebuild_single_mapping(struct 
>> vm_fault *vmf, struct vm_area_str
>>       update_mmu_cache_range(vmf, vma, fault_addr, fault_pte, 1);
>>   }
>> -static void numa_rebuild_large_mapping(struct vm_fault *vmf, struct 
>> vm_area_struct *vma,
>> -                       struct folio *folio, pte_t fault_pte,
>> -                       bool ignore_writable, bool pte_write_upgrade)
>> +static void numa_rebuild_large_mapping(struct vm_fault *vmf,
>> +        struct vm_area_struct *vma, struct folio *folio, int nr_pages,
>> +        pte_t fault_pte, bool ignore_writable, bool pte_write_upgrade)
>>   {
>>       int nr = pte_pfn(fault_pte) - folio_pfn(folio);
>> -    unsigned long start = max(vmf->address - nr * PAGE_SIZE, 
>> vma->vm_start);
>> -    unsigned long end = min(vmf->address + (folio_nr_pages(folio) - 
>> nr) * PAGE_SIZE, vma->vm_end);
>> -    pte_t *start_ptep = vmf->pte - (vmf->address - start) / PAGE_SIZE;
>> -    unsigned long addr;
>> +    unsigned long folio_size = nr_pages * PAGE_SIZE;
> 
> Just re-read folio_nr_pages() here, it's cheap. Or even better, use 
> folio_size();
> 
>> +    unsigned long addr = vmf->address;
>> +    unsigned long start, end, align_addr;
>> +    pte_t *start_ptep;
>> +
>> +    align_addr = ALIGN_DOWN(addr, folio_size);
>> +    start = max3(addr - nr * PAGE_SIZE, align_addr, vma->vm_start);
>> +    end = min3(addr + (nr_pages - nr) * PAGE_SIZE, align_addr + 
>> folio_size,
> 
> Please avoid mixing nr_pages and folio_size.
> 
>> +           vma->vm_end);
>> +    start_ptep = vmf->pte - (addr - start) / PAGE_SIZE;
>                  writable);
> 
> I am not able to convince myself that the old code could not have 
> resulted in a vmf->pte that underflows the page table. Am I correct?

Yes, I think you are right, and I realized the problem now.

> 
> Now align_addr would make sure that we are always within one page table 
> (as long as our folio size does not exceed PMD size :) ).
> 
> Can we use PMD_SIZE instead for that, something like the following?
> 
> 
> /* Stay within the VMA and within the page table. */
> pt_start = ALIGN_DOWN(addr, PMD_SIZE);
> start = max3(addr - nr << PAGE_SHIFT, pt_start, vma->vm_start);
> end = min3(addr + folio_size - nr << PAGE_SHIFT,
>             pt_start + PMD_SIZE, vma->vm_end);
> 
> start_ptep = vmf->pte - (addr - start) >> PAGE_SHIFT;

The changes look good to me. Thanks.