Do I need SetPageReserved() after map_user_kiobuf()? (was: What happened to try_to_swap_out()?)

[Timur Tabi <timur.tabi@ammasso.com>]

Rik van Riel wrote:

> Looks like the bug is in your driver, not the VM.
> 
> The VMA that maps such pages should be set VM_RESERVED
> (or whatever the name of that flag was)

Ok, I've examined our code further and discovered a few things.

The previous developer apparently realized that the pages need to be 
marked reserved after a call to map_user_kiobuf().  However, his 
comments indicate that this is a work-around for the "kiobuf bug".  Am I 
to assume that you don't consider this a bug in map_user_kiobuf()?

This is the code that we run after map_user_kiobuf().

     int i;
     for (i = 0; i < kiobuf->nr_pages; i++)
         SetPageReserved(kiobuf->maplist[i]);

Also, since we're porting to 2.6, we're going to replace 
map_user_kiobuf() with get_user_pages().  Will we still need to call 
SetPageReserved()?  Unfortunately, I don't have a good enough 
understanding of the Linux VM to know exactly what get_user_pages() is 
doing.  For example, this code confuses me:

                 if (!PageReserved(pages[i]))
                     page_cache_get(pages[i]);

Under what circumstances would the pages already be reserved?

-- 
Timur Tabi
Staff Software Engineer
timur.tabi@ammasso.com
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>