On Fri, 26 Apr 2019, Halil Pasic wrote:
> @@ -224,6 +228,9 @@ struct subchannel *css_alloc_subchannel(struct subchannel_id schid,
> INIT_WORK(&sch->todo_work, css_sch_todo);
> sch->dev.release = &css_subchannel_release;
> device_initialize(&sch->dev);
> + sch->dma_mask = css_dev_dma_mask;
> + sch->dev.dma_mask = &sch->dma_mask;
> + sch->dev.coherent_dma_mask = sch->dma_mask;
Could we do:
sch->dev.dma_mask = &sch->dev.coherent_dma_mask;
sch->dev.coherent_dma_mask = css_dev_dma_mask;
?
> +#define POOL_INIT_PAGES 1
> +static struct gen_pool *cio_dma_pool;
> +/* Currently cio supports only a single css */
> +#define CIO_DMA_GFP (GFP_KERNEL | __GFP_ZERO)
__GFP_ZERO has no meaning with the dma api (since all implementations do
an implicit zero initialization) but let's keep it for the sake of
documentation.
We need GFP_DMA here (which will return addresses < 2G on s390)!
> +void cio_gp_dma_free(struct gen_pool *gp_dma, void *cpu_addr, size_t size)
> +{
> + if (!cpu_addr)
> + return;
> + memset(cpu_addr, 0, size);
Hm, normally I'd do the memset during alloc not during free - but maybe
this makes more sense here with your usecase in mind.
> @@ -1063,6 +1163,7 @@ static int __init css_bus_init(void)
> unregister_reboot_notifier(&css_reboot_notifier);
> goto out_unregister;
> }
> + cio_dma_pool_init();
This is too late for early devices (ccw console!).
_______________________________________________
Virtualization mailing list
Virtualization@xxxxxxxxxxxxxxxxxxxxxxxxxx
https://lists.linuxfoundation.org/mailman/listinfo/virtualization
