Re: [PATCH] mm: swap: Avoid infinite loop if no valid swap entry found during do_swap_page

[mawupeng <mawupeng1@huawei.com>]

On 2025/2/22 15:33, Kairui Song wrote:
> On Sat, Feb 22, 2025 at 10:56 AM Wupeng Ma <mawupeng1@huawei.com> wrote:
>>
>> From: Ma Wupeng <mawupeng1@huawei.com>
>>
>> During our test, infinite loop is produced during #PF will lead to infinite
>> error log as follow:
>>
>>    get_swap_device: Bad swap file entry 114000000
>>
>> Digging into the source, we found that the swap entry is invalid due to
>> unknown reason, and this lead to invalid swap_info_struct. Excessive log
> 
> Hi Wupeng,
> 
> What is the kernel version you are using? If it's another bug causing
> this invalid swap entry, we should fix that bug instead, not
> workaround it.
> 
> This looks kind of similar to another PATCH & Bug report, corrupted
> page table or swap entry:
> https://lore.kernel.org/linux-mm/e223b0e6ba2f4924984b1917cc717bd5@honor.com/
> 
> Might be the same kernel bug? Gaoxu mentioned the bug was observed on
> Kernel 6.6.30 (android version), and neither of these two workarounds
> will fix it completely, the invalid value could cause many other
> issues too. We definitely need to find out the root cause.

We are having this problem in linux-v5.10, since the log is lost and swap
is not enabled in this machines, maybe memory corrupted in the pt.

> 
>> printing can fill up the prioritized log space, leading to the purging of
>> originally valid logs and hindering problem troubleshooting. To make this
>> more robust, kill this task.
>>
>> Signed-off-by: Ma Wupeng <mawupeng1@huawei.com>
>> ---
>>  include/linux/swap.h | 1 +
>>  mm/memory.c          | 9 ++++++++-
>>  mm/swapfile.c        | 2 +-
>>  3 files changed, 10 insertions(+), 2 deletions(-)
>>
>> diff --git a/include/linux/swap.h b/include/linux/swap.h
>> index b13b72645db3..0fa39cf66bc4 100644
>> --- a/include/linux/swap.h
>> +++ b/include/linux/swap.h
>> @@ -508,6 +508,7 @@ struct backing_dev_info;
>>  extern int init_swap_address_space(unsigned int type, unsigned long nr_pages);
>>  extern void exit_swap_address_space(unsigned int type);
>>  extern struct swap_info_struct *get_swap_device(swp_entry_t entry);
>> +struct swap_info_struct *_swap_info_get(swp_entry_t entry);
>>  sector_t swap_folio_sector(struct folio *folio);
>>
>>  static inline void put_swap_device(struct swap_info_struct *si)
>> diff --git a/mm/memory.c b/mm/memory.c
>> index b4d3d4893267..2d36e5a644d1 100644
>> --- a/mm/memory.c
>> +++ b/mm/memory.c
>> @@ -4365,8 +4365,15 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
>>
>>         /* Prevent swapoff from happening to us. */
>>         si = get_swap_device(entry);
>> -       if (unlikely(!si))
>> +       if (unlikely(!si)) {
>> +               if (unlikely(!_swap_info_get(entry)))
>> +                       /*
>> +                        * return VM_FAULT_SIGBUS for invalid swap entry to
>> +                        * avoid infinite #PF.
>> +                        */
>> +                       ret = VM_FAULT_SIGBUS;
> 
> This could lead to VM_FAULT_SIGBUS on swapoff. After swapoff
> get_swap_device will return NULL.

If swap is off, All swap pages should be swap in as expected, so
such entry can not trigger do_swap_page?