Re: [PATCH 04/13] x86: use generic swiotlb_ops

[Robin Murphy <robin.murphy@arm.com>]

Hi Christoph,

On 05/03/18 17:46, Christoph Hellwig wrote:
> The generic swiotlb dma ops were based on the x86 ones and provide
> equivalent functionality, so use them.
> 
> Also fix the sta2x11 case.  For that SOC the dma map ops need an
> additional physical to dma address translations.  For swiotlb buffers
> that is done throught the phys_to_dma helper, but the sta2x11_dma_ops
> also added an additional translation on the return value from
> x86_swiotlb_alloc_coherent, which is only correct if that functions
> returns a direct allocation and not a swiotlb buffer.  With the
> generic swiotlb and dma-direct code phys_to_dma is not always used
> and the separate sta2x11_dma_ops can be replaced with a simple
> bit that marks if the additional physical to dma address translation
> is needed.

FWIW, last time I looked I got the impression that STA2x11 could just 
use dma_pfn_offset - the comments and a2p/p2a logic in sta2x11-fixup.c 
certainly imply that the underlying hardware situation is pretty much 
exactly that for which dma_pfn_offset exists.

Robin.

> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>   arch/x86/include/asm/device.h  |  3 +++
>   arch/x86/include/asm/swiotlb.h |  8 -------
>   arch/x86/kernel/pci-swiotlb.c  | 47 +-----------------------------------------
>   arch/x86/pci/sta2x11-fixup.c   | 46 +++++------------------------------------
>   4 files changed, 9 insertions(+), 95 deletions(-)
> 
> diff --git a/arch/x86/include/asm/device.h b/arch/x86/include/asm/device.h
> index 5e12c63b47aa..812bd6c5d602 100644
> --- a/arch/x86/include/asm/device.h
> +++ b/arch/x86/include/asm/device.h
> @@ -6,6 +6,9 @@ struct dev_archdata {
>   #if defined(CONFIG_INTEL_IOMMU) || defined(CONFIG_AMD_IOMMU)
>   	void *iommu; /* hook for IOMMU specific extension */
>   #endif
> +#ifdef CONFIG_STA2X11
> +	bool is_sta2x11 : 1;
> +#endif
>   };
>   
>   #if defined(CONFIG_X86_DEV_DMA_OPS) && defined(CONFIG_PCI_DOMAINS)
> diff --git a/arch/x86/include/asm/swiotlb.h b/arch/x86/include/asm/swiotlb.h
> index 1c6a6cb230ff..ff6c92eff035 100644
> --- a/arch/x86/include/asm/swiotlb.h
> +++ b/arch/x86/include/asm/swiotlb.h
> @@ -27,12 +27,4 @@ static inline void pci_swiotlb_late_init(void)
>   {
>   }
>   #endif
> -
> -extern void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size,
> -					dma_addr_t *dma_handle, gfp_t flags,
> -					unsigned long attrs);
> -extern void x86_swiotlb_free_coherent(struct device *dev, size_t size,
> -					void *vaddr, dma_addr_t dma_addr,
> -					unsigned long attrs);
> -
>   #endif /* _ASM_X86_SWIOTLB_H */
> diff --git a/arch/x86/kernel/pci-swiotlb.c b/arch/x86/kernel/pci-swiotlb.c
> index bcb6a9bf64ad..661583662430 100644
> --- a/arch/x86/kernel/pci-swiotlb.c
> +++ b/arch/x86/kernel/pci-swiotlb.c
> @@ -17,51 +17,6 @@
>   
>   int swiotlb __read_mostly;
>   
> -void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size,
> -					dma_addr_t *dma_handle, gfp_t flags,
> -					unsigned long attrs)
> -{
> -	void *vaddr;
> -
> -	/*
> -	 * Don't print a warning when the first allocation attempt fails.
> -	 * swiotlb_alloc_coherent() will print a warning when the DMA
> -	 * memory allocation ultimately failed.
> -	 */
> -	flags |= __GFP_NOWARN;
> -
> -	vaddr = dma_direct_alloc(hwdev, size, dma_handle, flags, attrs);
> -	if (vaddr)
> -		return vaddr;
> -
> -	return swiotlb_alloc_coherent(hwdev, size, dma_handle, flags);
> -}
> -
> -void x86_swiotlb_free_coherent(struct device *dev, size_t size,
> -				      void *vaddr, dma_addr_t dma_addr,
> -				      unsigned long attrs)
> -{
> -	if (is_swiotlb_buffer(dma_to_phys(dev, dma_addr)))
> -		swiotlb_free_coherent(dev, size, vaddr, dma_addr);
> -	else
> -		dma_direct_free(dev, size, vaddr, dma_addr, attrs);
> -}
> -
> -static const struct dma_map_ops x86_swiotlb_dma_ops = {
> -	.mapping_error = swiotlb_dma_mapping_error,
> -	.alloc = x86_swiotlb_alloc_coherent,
> -	.free = x86_swiotlb_free_coherent,
> -	.sync_single_for_cpu = swiotlb_sync_single_for_cpu,
> -	.sync_single_for_device = swiotlb_sync_single_for_device,
> -	.sync_sg_for_cpu = swiotlb_sync_sg_for_cpu,
> -	.sync_sg_for_device = swiotlb_sync_sg_for_device,
> -	.map_sg = swiotlb_map_sg_attrs,
> -	.unmap_sg = swiotlb_unmap_sg_attrs,
> -	.map_page = swiotlb_map_page,
> -	.unmap_page = swiotlb_unmap_page,
> -	.dma_supported = NULL,
> -};
> -
>   /*
>    * pci_swiotlb_detect_override - set swiotlb to 1 if necessary
>    *
> @@ -111,7 +66,7 @@ void __init pci_swiotlb_init(void)
>   {
>   	if (swiotlb) {
>   		swiotlb_init(0);
> -		dma_ops = &x86_swiotlb_dma_ops;
> +		dma_ops = &swiotlb_dma_ops;
>   	}
>   }
>   
> diff --git a/arch/x86/pci/sta2x11-fixup.c b/arch/x86/pci/sta2x11-fixup.c
> index 6c712fe11bdc..eac58e03f43c 100644
> --- a/arch/x86/pci/sta2x11-fixup.c
> +++ b/arch/x86/pci/sta2x11-fixup.c
> @@ -159,43 +159,6 @@ static dma_addr_t a2p(dma_addr_t a, struct pci_dev *pdev)
>   	return p;
>   }
>   
> -/**
> - * sta2x11_swiotlb_alloc_coherent - Allocate swiotlb bounce buffers
> - *     returns virtual address. This is the only "special" function here.
> - * @dev: PCI device
> - * @size: Size of the buffer
> - * @dma_handle: DMA address
> - * @flags: memory flags
> - */
> -static void *sta2x11_swiotlb_alloc_coherent(struct device *dev,
> -					    size_t size,
> -					    dma_addr_t *dma_handle,
> -					    gfp_t flags,
> -					    unsigned long attrs)
> -{
> -	void *vaddr;
> -
> -	vaddr = x86_swiotlb_alloc_coherent(dev, size, dma_handle, flags, attrs);
> -	*dma_handle = p2a(*dma_handle, to_pci_dev(dev));
> -	return vaddr;
> -}
> -
> -/* We have our own dma_ops: the same as swiotlb but from alloc (above) */
> -static const struct dma_map_ops sta2x11_dma_ops = {
> -	.alloc = sta2x11_swiotlb_alloc_coherent,
> -	.free = x86_swiotlb_free_coherent,
> -	.map_page = swiotlb_map_page,
> -	.unmap_page = swiotlb_unmap_page,
> -	.map_sg = swiotlb_map_sg_attrs,
> -	.unmap_sg = swiotlb_unmap_sg_attrs,
> -	.sync_single_for_cpu = swiotlb_sync_single_for_cpu,
> -	.sync_single_for_device = swiotlb_sync_single_for_device,
> -	.sync_sg_for_cpu = swiotlb_sync_sg_for_cpu,
> -	.sync_sg_for_device = swiotlb_sync_sg_for_device,
> -	.mapping_error = swiotlb_dma_mapping_error,
> -	.dma_supported = dma_direct_supported,
> -};
> -
>   /* At setup time, we use our own ops if the device is a ConneXt one */
>   static void sta2x11_setup_pdev(struct pci_dev *pdev)
>   {
> @@ -205,7 +168,8 @@ static void sta2x11_setup_pdev(struct pci_dev *pdev)
>   		return;
>   	pci_set_consistent_dma_mask(pdev, STA2X11_AMBA_SIZE - 1);
>   	pci_set_dma_mask(pdev, STA2X11_AMBA_SIZE - 1);
> -	pdev->dev.dma_ops = &sta2x11_dma_ops;
> +	pdev->dev.dma_ops = &swiotlb_dma_ops;
> +	pdev->dev.archdata.is_sta2x11 = true;
>   
>   	/* We must enable all devices as master, for audio DMA to work */
>   	pci_set_master(pdev);
> @@ -225,7 +189,7 @@ bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
>   {
>   	struct sta2x11_mapping *map;
>   
> -	if (dev->dma_ops != &sta2x11_dma_ops) {
> +	if (!dev->archdata.is_sta2x11) {
>   		if (!dev->dma_mask)
>   			return false;
>   		return addr + size - 1 <= *dev->dma_mask;
> @@ -249,7 +213,7 @@ bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
>    */
>   dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr)
>   {
> -	if (dev->dma_ops != &sta2x11_dma_ops)
> +	if (!dev->archdata.is_sta2x11)
>   		return paddr;
>   	return p2a(paddr, to_pci_dev(dev));
>   }
> @@ -261,7 +225,7 @@ dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr)
>    */
>   phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
>   {
> -	if (dev->dma_ops != &sta2x11_dma_ops)
> +	if (!dev->archdata.is_sta2x11)
>   		return daddr;
>   	return a2p(daddr, to_pci_dev(dev));
>   }
>