Re: [PATCH] radeon: use vmalloc instead of kmalloc

[Peter Zijlstra <peterz@infradead.org>]

On Mon, 2009-06-22 at 19:26 +0200, Jerome Glisse wrote:
> We don't need to allocated contiguous pages in cs codepath
> so use vmalloc instead.

Best would be to not require >PAGE_SIZE allocations at all of course.
But barring that, it would be great to have something like:

  ptr = kmalloc(size, GFP_KERNEL | __GFP_NOWARN);
  if (!ptr)
    ptr = vmalloc(size);

Also, how long do these allocations live? vmalloc space can be quite
limited (i386) and it can fragment too.

> Signed-off-by: Jerome Glisse <jglisse@redhat.com>
> ---
>  drivers/gpu/drm/radeon/radeon_cs.c |   20 ++++++++++----------
>  1 files changed, 10 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c
> index b843f9b..54a9740 100644
> --- a/drivers/gpu/drm/radeon/radeon_cs.c
> +++ b/drivers/gpu/drm/radeon/radeon_cs.c
> @@ -45,11 +45,11 @@ int radeon_cs_parser_relocs(struct radeon_cs_parser *p)
>  	chunk = &p->chunks[p->chunk_relocs_idx];
>  	/* FIXME: we assume that each relocs use 4 dwords */
>  	p->nrelocs = chunk->length_dw / 4;
> -	p->relocs_ptr = kcalloc(p->nrelocs, sizeof(void *), GFP_KERNEL);
> +	p->relocs_ptr = vmalloc(p->nrelocs * sizeof(void *));
>  	if (p->relocs_ptr == NULL) {
>  		return -ENOMEM;
>  	}
> -	p->relocs = kcalloc(p->nrelocs, sizeof(struct radeon_cs_reloc), GFP_KERNEL);
> +	p->relocs = vmalloc(p->nrelocs * sizeof(struct radeon_cs_reloc));
>  	if (p->relocs == NULL) {
>  		return -ENOMEM;
>  	}
> @@ -103,7 +103,7 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
>  	p->idx = 0;
>  	p->chunk_ib_idx = -1;
>  	p->chunk_relocs_idx = -1;
> -	p->chunks_array = kcalloc(cs->num_chunks, sizeof(uint64_t), GFP_KERNEL);
> +	p->chunks_array = vmalloc(cs->num_chunks * sizeof(uint64_t));
>  	if (p->chunks_array == NULL) {
>  		return -ENOMEM;
>  	}
> @@ -113,7 +113,7 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
>  		return -EFAULT;
>  	}
>  	p->nchunks = cs->num_chunks;
> -	p->chunks = kcalloc(p->nchunks, sizeof(struct radeon_cs_chunk), GFP_KERNEL);
> +	p->chunks = vmalloc(p->nchunks * sizeof(struct radeon_cs_chunk));
>  	if (p->chunks == NULL) {
>  		return -ENOMEM;
>  	}
> @@ -139,7 +139,7 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
>  
>  		p->chunks[i].kdata = NULL;
>  		size = p->chunks[i].length_dw * sizeof(uint32_t);
> -		p->chunks[i].kdata = kzalloc(size, GFP_KERNEL);
> +		p->chunks[i].kdata = vmalloc(size);
>  		if (p->chunks[i].kdata == NULL) {
>  			return -ENOMEM;
>  		}
> @@ -179,13 +179,13 @@ static void radeon_cs_parser_fini(struct radeon_cs_parser *parser, int error)
>  			mutex_unlock(&parser->rdev->ddev->struct_mutex);
>  		}
>  	}
> -	kfree(parser->relocs);
> -	kfree(parser->relocs_ptr);
> +	vfree(parser->relocs);
> +	vfree(parser->relocs_ptr);
>  	for (i = 0; i < parser->nchunks; i++) {
> -		kfree(parser->chunks[i].kdata);
> +		vfree(parser->chunks[i].kdata);
>  	}
> -	kfree(parser->chunks);
> -	kfree(parser->chunks_array);
> +	vfree(parser->chunks);
> +	vfree(parser->chunks_array);
>  	radeon_ib_free(parser->rdev, &parser->ib);
>  }
>