Re: [PATCH v2] dma-coherent: add support for multi coherent rmems per dev

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Mon, Feb 05, 2024 at 07:23:00AM +0000, Howard Yen wrote:
> Add support for multiple coherent rmems per device. This patch replaces
> original dma_mem with dma_mems list in device structure to store multiple
> rmems.
> 
> These multiple rmems can be assigned to the device one by one by
> of_reserved_mem_device_init_by_idx() with the memory-region
> declaration in device tree as below and store the rmem to the dma_mems
> list.
> 
> 	device1@0 {
> 		...
> 		memory-region = <&reserved_mem0>, <&reserved_mem1>;
> 		...
> 	};
> 
> When driver tries to allocate memory from the rmems, looks for the first
> available rmem and allocates the memory from this rmem.
> 
> Then if driver removed, of_reserved_mem_device_release() needs to be
> invoked to release all the rmems assigned to the device.

...

> --- a/kernel/dma/coherent.c
> +++ b/kernel/dma/coherent.c
> @@ -18,15 +18,9 @@ struct dma_coherent_mem {
>  	unsigned long	*bitmap;
>  	spinlock_t	spinlock;
>  	bool		use_dev_dma_pfn_offset;
> +	struct list_head	node;

Have you run `pahole`? Here I see wasted bytes for nothing.

>  };

...

>  void dma_release_coherent_memory(struct device *dev)
>  {
> +	struct dma_coherent_mem *mem_tmp, *q;
> +
>  	if (dev) {

While at it, perhaps

	if (!dev)
		return;

?

> -		_dma_release_coherent_memory(dev->dma_mem);
> -		dev->dma_mem = NULL;
> +		list_for_each_entry_safe(mem_tmp, q, &dev->dma_mems, node) {
> +			list_del_init(&mem_tmp->node);
> +			_dma_release_coherent_memory(mem_tmp);
> +		}
>  	}
>  }

...

>  int dma_release_from_dev_coherent(struct device *dev, int order, void *vaddr)
>  {
> -	struct dma_coherent_mem *mem = dev_get_coherent_memory(dev);
> +	struct dma_coherent_mem *mem_tmp;
> +	int ret = 0;

'ret' (1)

> -	return __dma_release_from_coherent(mem, order, vaddr);
> +	list_for_each_entry(mem_tmp, &dev->dma_mems, node) {
> +		ret = __dma_release_from_coherent(mem_tmp, order, vaddr);
> +		if (ret == 1)
> +			break;
> +	}
> +
> +	return ret;
>  }

...

>  int dma_mmap_from_dev_coherent(struct device *dev, struct vm_area_struct *vma,
>  			   void *vaddr, size_t size, int *ret)
>  {
> -	struct dma_coherent_mem *mem = dev_get_coherent_memory(dev);
> +	struct dma_coherent_mem *mem_tmp;
> +	int retval = 0;

'retval' (2)

Can we be consistent, please? (See (1) and (2) above.)

> +	list_for_each_entry(mem_tmp, &dev->dma_mems, node) {
> +		retval = __dma_mmap_from_coherent(mem_tmp, vma, vaddr, size, ret);
> +		if (retval == 1)
> +			break;
> +	}
>  
> -	return __dma_mmap_from_coherent(mem, vma, vaddr, size, ret);
> +	return retval;
>  }

...

>  static void rmem_dma_device_release(struct reserved_mem *rmem,
>  				    struct device *dev)
>  {
> -	if (dev)
> -		dev->dma_mem = NULL;
> +	struct dma_coherent_mem *mem_tmp, *q;

> +	if (dev) {

As per above, esp. taking into account that you touch this line. With proposed
modification you won't need to.

> +		list_for_each_entry_safe(mem_tmp, q, &dev->dma_mems, node) {
> +			if (mem_tmp == rmem->priv) {
> +				list_del_init(&mem_tmp->node);
> +				break;
> +			}
> +		}
> +	}
>  }

Better question, do we really need the dev check (at least in static functions)
or it can be ommitted?

-- 
With Best Regards,
Andy Shevchenko