Re: Question About Functions "__free_pages_check" and "check_new_page" in page_alloc.c

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Re: Question About Functions "__free_pages_check" and "check_new_page" in page_alloc.c

Michal Hocko-4
On Mon 16-05-16 21:42:23, Wang Xiaoqiang wrote:

> Hi all,
>
>     I am really confused about these two functions. The following code snippet:
>
> if(unlikely(atomic_read(&page->_mapcount) != -1))
> bad_reason ="nonzero mapcount";if(unlikely(page->mapping != NULL))
> bad_reason ="non-NULL mapping";if(unlikely(page_ref_count(page) !=0))
> bad_reason ="nonzero _count";
>         ...
> Wouldn't the previous value of "bad_reason" be overwritten by
> the later? Hope to receive from you.

yes it would. Why that would matter. The checks should be in an order
which could give us a more specific reason with later checks. bad_page()
will then print more detailed information.
--
Michal Hocko
SUSE Labs
Reply | Threaded
Open this post in threaded view
|

Re: Question About Functions "__free_pages_check" and "check_new_page" in page_alloc.c

Vlastimil Babka
On 05/17/2016 03:06 AM, Wang Xiaoqiang wrote:
>>yes it would. Why that would matter. The checks should be in an order
>>which could give us a more specific reason with later checks. bad_page()
>
> I see, you mean the later "bad_reason" is the superset of the previous one.

Not exactly. It's not possible to sort all the reasons like that. But as
Michal said, bad_page() will print all the relevant info so you can
reconstruct all reasons from it. The bad_reason text is mostly a hint
what to check first.

>>will then print more detailed information.
>>--
>>Michal Hocko
>>SUSE Labs
>
> thank you, Michal.
>
>
>  
>