Re: general protection fault in unlink_file_vma

[linmiaohe]

Hillf Danton <hdanton@xxxxxxxx> wrote:
> On Sun, 13 Sep 2020 09:17:26 +0000 linmiaohe wrote:
>> 
>> I reviewed the code carefully these days and I found vma_merge() do only fput() the vm_file of the linked vma in remove_next cases.
>> This gpf is much likely because the ->mmap() callback can change 
>> vma->vm_file and fput the original file. But my previous commit failed to catch this case and always fput() the original file, hence add an extra fput().
>> The below patch would make the things right:
>> 
>
>Take another look at the Cc list and the link below.
>
>https://lore.kernel.org/lkml/20200911120222.GT87483@xxxxxxxx/
>

Many thanks for your teach. I think I could send the proposed patch to the syzbot directly.
Thanks again.:)