Linus,
Currently __alloc_pages() does not allow PF_MEMALLOC tasks to free clean
inactive pages.
This is senseless --- if the allocation has __GFP_WAIT set, its ok to grab
the pagemap_lru_lock/pagecache_lock/etc.
I checked all possible codepaths after reclaim_page() and they are ok.
The following patch fixes that.
--- linux/mm/page_alloc.c.orig Fri Apr 27 05:59:35 2001
+++ linux/mm/page_alloc.c Fri Apr 27 05:59:48 2001
@@ -295,8 +295,7 @@
* Can we take pages directly from the inactive_clean
* list?
*/
- if (order == 0 && (gfp_mask & __GFP_WAIT) &&
- !(current->flags & PF_MEMALLOC))
+ if (order == 0 && (gfp_mask & __GFP_WAIT))
direct_reclaim = 1;
/*
Marcelo,
Infact, the test can be made even weaker than that.
We only what to avoid the inactive-clean list when allocating from
within an interrupt (or from a bottom-half handler) to avoid
deadlock on taking the pagecache_lock and pagemap_lru_lock.
Note: no allocations are done while holding either of these locks.
Even GFP_ATOMIC doesn't matter; that simply says if we should block or
not (could be an allocation while holding a spinlock, but not inside an
interrupt).
I've been doing v heavy load on a 4-way server with;
if (order == 0 && !in_interrupt())
direct_reclaim = 1;
without any problems.
Of course, the in_interrupt() is heavier than (gfp_mask & __GFP_WAIT),
but the benefits outweight the slight fattening.
Mark
On Fri, 27 Apr 2001, Marcelo Tosatti wrote:
>
> Linus,
>
> Currently __alloc_pages() does not allow PF_MEMALLOC tasks to free clean
> inactive pages.
>
> This is senseless --- if the allocation has __GFP_WAIT set, its ok to grab
> the pagemap_lru_lock/pagecache_lock/etc.
>
> I checked all possible codepaths after reclaim_page() and they are ok.
>
> The following patch fixes that.
>
>
> --- linux/mm/page_alloc.c.orig Fri Apr 27 05:59:35 2001
> +++ linux/mm/page_alloc.c Fri Apr 27 05:59:48 2001
> @@ -295,8 +295,7 @@
> * Can we take pages directly from the inactive_clean
> * list?
> */
> - if (order == 0 && (gfp_mask & __GFP_WAIT) &&
> - !(current->flags & PF_MEMALLOC))
> + if (order == 0 && (gfp_mask & __GFP_WAIT))
> direct_reclaim = 1;
>
> /*
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>