Re: [PATCH v2 10/10] mm, cma: use spinlock instead of mutex

[Minchan Kim]

On Thu, Jun 12, 2014 at 12:21:47PM +0900, Joonsoo Kim wrote:
 Currently, we should take the mutex for manipulating bitmap.
 This job may be really simple and short so we don't need to sleep
 if contended. So I change it to spinlock.

I'm not sure it would be good always.
Maybe you remember we discussed about similar stuff about bitmap
searching in vmap friend internally, which was really painful
when it was fragmented. So, at least we need number if you really want
and I hope the number from ARM machine most popular platform for CMA
at the moment.

 
 Signed-off-by: Joonsoo Kim iamjoonsoo@lge.com
 
 diff --git a/mm/cma.c b/mm/cma.c
 index 22a5b23..3085e8c 100644
 --- a/mm/cma.c
 +++ b/mm/cma.c
 @@ -27,6 +27,7 @@
  #include linux/memblock.h
  #include linux/err.h
  #include linux/mm.h
 +#include linux/spinlock.h
  #include linux/mutex.h
  #include linux/sizes.h
  #include linux/slab.h
 @@ -36,7 +37,7 @@ struct cma {
   unsigned long   count;
   unsigned long   *bitmap;
   int order_per_bit; /* Order of pages represented by one bit */
 - struct mutexlock;
 + spinlock_t  lock;
  };
  
  /*
 @@ -72,9 +73,9 @@ static void clear_cma_bitmap(struct cma *cma, unsigned long 
 pfn, int count)
   bitmapno = (pfn - cma-base_pfn)  cma-order_per_bit;
   nr_bits = cma_bitmap_pages_to_bits(cma, count);
  
 - mutex_lock(cma-lock);
 + spin_lock(cma-lock);
   bitmap_clear(cma-bitmap, bitmapno, nr_bits);
 - mutex_unlock(cma-lock);
 + spin_unlock(cma-lock);
  }
  
  static int __init cma_activate_area(struct cma *cma)
 @@ -112,7 +113,7 @@ static int __init cma_activate_area(struct cma *cma)
   init_cma_reserved_pageblock(pfn_to_page(base_pfn));
   } while (--i);
  
 - mutex_init(cma-lock);
 + spin_lock_init(cma-lock);
   return 0;
  
  err:
 @@ -261,11 +262,11 @@ struct page *cma_alloc(struct cma *cma, int count, 
 unsigned int align)
   nr_bits = cma_bitmap_pages_to_bits(cma, count);
  
   for (;;) {
 - mutex_lock(cma-lock);
 + spin_lock(cma-lock);
   bitmapno = bitmap_find_next_zero_area(cma-bitmap,
   bitmap_maxno, start, nr_bits, mask);
   if (bitmapno = bitmap_maxno) {
 - mutex_unlock(cma-lock);
 + spin_unlock(cma-lock);
   break;
   }
   bitmap_set(cma-bitmap, bitmapno, nr_bits);
 @@ -274,7 +275,7 @@ struct page *cma_alloc(struct cma *cma, int count, 
 unsigned int align)
* our exclusive use. If the migration fails we will take the
* lock again and unmark it.
*/
 - mutex_unlock(cma-lock);
 + spin_unlock(cma-lock);
  
   pfn = cma-base_pfn + (bitmapno  cma-order_per_bit);
   mutex_lock(cma_mutex);
 -- 
 1.7.9.5
 
 --
 To unsubscribe, send a message with 'unsubscribe linux-mm' in
 the body to majord...@kvack.org.  For more info on Linux MM,
 see: http://www.linux-mm.org/ .
 Don't email: a href=mailto:d...@kvack.org; em...@kvack.org /a

-- 
Kind regards,
Minchan Kim
--
To unsubscribe from this list: send the line unsubscribe kvm-ppc in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH v2 10/10] mm, cma: use spinlock instead of mutex

[Joonsoo Kim]

On Thu, Jun 12, 2014 at 04:40:29PM +0900, Minchan Kim wrote:
 On Thu, Jun 12, 2014 at 12:21:47PM +0900, Joonsoo Kim wrote:
  Currently, we should take the mutex for manipulating bitmap.
  This job may be really simple and short so we don't need to sleep
  if contended. So I change it to spinlock.
 
 I'm not sure it would be good always.
 Maybe you remember we discussed about similar stuff about bitmap
 searching in vmap friend internally, which was really painful
 when it was fragmented. So, at least we need number if you really want
 and I hope the number from ARM machine most popular platform for CMA
 at the moment.

Good Point!! Agreed. I will drop this one in next spin and re-submit
in separate patchset after some testing.

Thanks.
--
To unsubscribe from this list: send the line unsubscribe kvm-ppc in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html