Re: [patch v2] drivers/zram: Don't disable preemption in zcomp_stream_get/put()

[Mike Galbraith]

On Wed, 2016-10-19 at 18:54 +0200, Sebastian Andrzej Siewior wrote:
> On 2016-10-19 17:56:30 [+0200], Mike Galbraith wrote:
> > In v4.7, the driver switched to percpu compression streams, disabling
> > preemption vai get/put_cpu_ptr().  Use a local lock instead for RT.
> > We also have to fix an RT lock order issue in zram_decompress_page()
> > such that zs_map_object() nests inside of zcomp_stream_put() as it
> > does in zram_bvec_write().
> 
> good. I almost had it myself. So let me get that one. And your previous
> one (the spinlock replacement) looks also reasonable. With this hunk
> 
> @@ -313,6 +314,7 @@ struct mapping_area {
>  #endif
>         char *vm_addr; /* address of kmap_atomic()'ed pages */
>         enum zs_mapmode vm_mm; /* mapping mode */
> +       spinlock_t      ma_lock;
>  };
>  
>  #ifdef CONFIG_COMPACTION
> @@ -1489,6 +1491,7 @@ void *zs_map_object(struct zs_pool *pool, unsigned long handle,
>         off = (class->size * obj_idx) & ~PAGE_MASK;
>  
>         area = per_cpu_ptr(&zs_map_area, get_cpu_light());
> +       spin_lock(&area->ma_lock);
>         area->vm_mm = mm;
>         if (off + class->size <= PAGE_SIZE) {
>                 /* this object is contained entirely within a page */
> @@ -1542,6 +1545,7 @@ void zs_unmap_object(struct zs_pool *pool, unsigned long handle)
>  
>                 __zs_unmap_object(area, pages, off, class->size);
>         }
> +       spin_unlock(&area->ma_lock);
>         put_cpu_light();
>  
>         migrate_read_unlock(zspage);

Ew, yeah, as the other, I thought it was covered, but nope.

	-Mike