Re: [External] Re: [PATCH] mm/hugetlb: Fix a race between hugetlb sysctl handlers

[Mike Kravetz]

On 8/25/20 7:47 PM, Muchun Song wrote:
> 
> CPU0:                                     CPU1:
>                                           proc_sys_write
> hugetlb_sysctl_handler                      proc_sys_call_handler
> hugetlb_sysctl_handler_common                 hugetlb_sysctl_handler
>   table->data = &tmp;                           hugetlb_sysctl_handler_common
>                                                   table->data = &tmp;
>     proc_doulongvec_minmax
>       do_proc_doulongvec_minmax             sysctl_head_finish
>         __do_proc_doulongvec_minmax
>           i = table->data;
>           *i = val;     // corrupt CPU1 stack

Thanks Muchun!
Can you please add this to the commit message.

Also, when looking closer at the patch I do not think setting table->maxlen
is necessary in these routines.  maxlen is set when the hugetlb ctl_table
entries are defined and initialized.  This is not something you introduced.
The unnecessary assignments are in the existing code.  However, there is no
need to carry them forward.

-- 
Mike Kravetz