The issue with page allocation 5.3 rc1-rc2 (seems drm culprit here)

Hillf Danton hdanton at sina.com
Thu Aug 8 07:42:52 UTC 2019


On Thu, 8 Aug 2019 13:32:06 +0800 Alex Deucher wrote:
> 
> On Wed, Aug 7, 2019 at 11:49 PM Mikhail Gavrilov wrote:
> >
> > Unfortunately error "gnome-shell: page allocation failure: order:4,
> > mode:0x40cc0(GFP_KERNEL|__GFP_COMP),
> > nodemask=(null),cpuset=/,mems_allowed=0" still happens even with
> > applying this patch.

Thanks Mikhail.

No surpring to see the warning because of kvmalloc on top of the current
kmalloc. Any other difference observed?

> I think we can just drop the kmalloc altogether.

Dropping kmalloc altogether OTOH makes the reason for the vmalloc
fallback IMO, Sir?

> How about this patch?
> 
> From: Alex Deucher <alexander.deucher at amd.com>
> Date: Thu, 8 Aug 2019 00:29:23 -0500
> Subject: [PATCH] drm/amd/display: use kvmalloc for dc_state
> 
> It's large and doesn't need contiguous memory.
> 
> Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
> ---

Looks good to me if with a kvfree added.

>  drivers/gpu/drm/amd/display/dc/core/dc.c | 9 +++++----
>  1 file changed, 5 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c
> index 252b621d93a9..ef780a4e484a 100644
> --- a/drivers/gpu/drm/amd/display/dc/core/dc.c
> +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c
> @@ -23,6 +23,7 @@
>   */
>  
>  #include <linux/slab.h>
> +#include <linux/mm.h>
>  
>  #include "dm_services.h"
>  
> @@ -1183,8 +1184,8 @@ bool dc_post_update_surfaces_to_stream(struct dc *dc)
>  
>  struct dc_state *dc_create_state(struct dc *dc)
>  {
> -	struct dc_state *context = kzalloc(sizeof(struct dc_state),
> -					   GFP_KERNEL);
> +	struct dc_state *context = kvzalloc(sizeof(struct dc_state),
> +					    GFP_KERNEL);
>  
>  	if (!context)
>  		return NULL;
> @@ -1204,11 +1205,11 @@ struct dc_state *dc_create_state(struct dc *dc)
>  struct dc_state *dc_copy_state(struct dc_state *src_ctx)
>  {
>  	int i, j;
> -	struct dc_state *new_ctx = kmemdup(src_ctx,
> -			sizeof(struct dc_state), GFP_KERNEL);
> +	struct dc_state *new_ctx = kvmalloc(sizeof(struct dc_state), GFP_KERNEL);
>  
>  	if (!new_ctx)
>  		return NULL;
> +	memcpy(new_ctx, src_ctx, sizeof(struct dc_state));
>  
>  	for (i = 0; i < MAX_PIPES; i++) {
>  			struct pipe_ctx *cur_pipe = &new_ctx->res_ctx.pipe_ctx[i];
> -- 
> 2.20.1
> 



More information about the amd-gfx mailing list