Re: [PATCH 22/22] parisc: use generic dma_noncoherent_ops

[Helge Deller <deller@gmx.de>]

On 20.04.2018 10:03, Christoph Hellwig wrote:
> Switch to the generic noncoherent direct mapping implementation.
> 
> Parisc previously had two different non-coherent dma ops implementation
> that just different in the way coherent allocations were handled or not
> handled.  The different behavior is not selected at runtime in the
> arch_dma_alloc and arch_dma_free routines.  The non-coherent allocation
> in the pcx cases now uses the dma_direct helpers that are a little more
> sophisticated and used by a lot of other architectures.
> 
> Fix sync_single_for_cpu to do skip the cache flush unless the transfer
> is to the device to match the more tested unmap_single path which should
> have the same cache coherency implications.
> 
> This also now consistenly uses flush_kernel_dcache_range for cache
> flushing while previously some of the SG based operations used
> flush_kernel_vmap_range instead.


This patch breaks a 32bit kernel on a B160L machine (PA7300LC CPU, "pcxl2").
After applying this patch series the lasi82956 network driver works unreliable. 
NIC gets IP, but ping doesn't work.
See drivers/net/ethernet/i825xx/lasi_82596.c, it uses dma*sync() functions.

Helge

 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  arch/parisc/Kconfig                   |   4 +
>  arch/parisc/include/asm/dma-mapping.h |   5 -
>  arch/parisc/kernel/pci-dma.c          | 181 ++++----------------------
>  arch/parisc/kernel/setup.c            |   8 +-
>  arch/parisc/mm/init.c                 |  11 +-
>  5 files changed, 35 insertions(+), 174 deletions(-)
> 
> diff --git a/arch/parisc/Kconfig b/arch/parisc/Kconfig
> index 47047f0cbe35..80166a1cbcb7 100644
> --- a/arch/parisc/Kconfig
> +++ b/arch/parisc/Kconfig
> @@ -188,6 +188,10 @@ config PA20
>  config PA11
>  	def_bool y
>  	depends on PA7000 || PA7100LC || PA7200 || PA7300LC
> +	select ARCH_HAS_SYNC_DMA_FOR_CPU
> +	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
> +	select DMA_NONCOHERENT_OPS
> +	select DMA_NONCOHERENT_CACHE_SYNC
>  
>  config PREFETCH
>  	def_bool y
> diff --git a/arch/parisc/include/asm/dma-mapping.h b/arch/parisc/include/asm/dma-mapping.h
> index 01e1fc057c83..44a9f97194aa 100644
> --- a/arch/parisc/include/asm/dma-mapping.h
> +++ b/arch/parisc/include/asm/dma-mapping.h
> @@ -21,11 +21,6 @@
>  ** flush/purge and allocate "regular" cacheable pages for everything.
>  */
>  
> -#ifdef CONFIG_PA11
> -extern const struct dma_map_ops pcxl_dma_ops;
> -extern const struct dma_map_ops pcx_dma_ops;
> -#endif
> -
>  extern const struct dma_map_ops *hppa_dma_ops;
>  
>  static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
> diff --git a/arch/parisc/kernel/pci-dma.c b/arch/parisc/kernel/pci-dma.c
> index 91bc0cac03a1..235e2e53959e 100644
> --- a/arch/parisc/kernel/pci-dma.c
> +++ b/arch/parisc/kernel/pci-dma.c
> @@ -21,13 +21,12 @@
>  #include <linux/init.h>
>  #include <linux/gfp.h>
>  #include <linux/mm.h>
> -#include <linux/pci.h>
>  #include <linux/proc_fs.h>
>  #include <linux/seq_file.h>
>  #include <linux/string.h>
>  #include <linux/types.h>
> -#include <linux/scatterlist.h>
> -#include <linux/export.h>
> +#include <linux/dma-direct.h>
> +#include <linux/dma-noncoherent.h>
>  
>  #include <asm/cacheflush.h>
>  #include <asm/dma.h>    /* for DMA_CHUNK_SIZE */
> @@ -447,178 +446,48 @@ static void pa11_dma_free(struct device *dev, size_t size, void *vaddr,
>  	free_pages((unsigned long)__va(dma_handle), order);
>  }
>  
> -static dma_addr_t pa11_dma_map_page(struct device *dev, struct page *page,
> -		unsigned long offset, size_t size,
> -		enum dma_data_direction direction, unsigned long attrs)
> +void arch_sync_dma_for_device(struct device *dev, phys_addr_t paddr,
> +		size_t size, enum dma_data_direction dir)
>  {
> -	void *addr = page_address(page) + offset;
> -	BUG_ON(direction == DMA_NONE);
> -
> -	if (!(attrs & DMA_ATTR_SKIP_CPU_SYNC))
> -		flush_kernel_dcache_range((unsigned long) addr, size);
> -
> -	return virt_to_phys(addr);
> +	flush_kernel_dcache_range((unsigned long)phys_to_virt(paddr), size);
>  }
>  
> -static void pa11_dma_unmap_page(struct device *dev, dma_addr_t dma_handle,
> -		size_t size, enum dma_data_direction direction,
> -		unsigned long attrs)
> +void arch_sync_dma_for_cpu(struct device *dev, phys_addr_t paddr,
> +		size_t size, enum dma_data_direction dir)
>  {
> -	BUG_ON(direction == DMA_NONE);
> -
> -	if (attrs & DMA_ATTR_SKIP_CPU_SYNC)
> -		return;
> -
> -	if (direction == DMA_TO_DEVICE)
> +	if (dir == DMA_TO_DEVICE)
>  		return;
>  
>  	/*
> -	 * For PCI_DMA_FROMDEVICE this flush is not necessary for the
> +	 * For DMA_FROM_DEVICE this flush is not necessary for the
>  	 * simple map/unmap case. However, it IS necessary if if
> -	 * pci_dma_sync_single_* has been called and the buffer reused.
> +	 * dma_sync_single_* has been called and the buffer reused.
>  	 */
>  
> -	flush_kernel_dcache_range((unsigned long) phys_to_virt(dma_handle), size);
> -}
> -
> -static int pa11_dma_map_sg(struct device *dev, struct scatterlist *sglist,
> -		int nents, enum dma_data_direction direction,
> -		unsigned long attrs)
> -{
> -	int i;
> -	struct scatterlist *sg;
> -
> -	BUG_ON(direction == DMA_NONE);
> -
> -	for_each_sg(sglist, sg, nents, i) {
> -		unsigned long vaddr = (unsigned long)sg_virt(sg);
> -
> -		sg_dma_address(sg) = (dma_addr_t) virt_to_phys(vaddr);
> -		sg_dma_len(sg) = sg->length;
> -
> -		if (attrs & DMA_ATTR_SKIP_CPU_SYNC)
> -			continue;
> -
> -		flush_kernel_dcache_range(vaddr, sg->length);
> -	}
> -	return nents;
> +	flush_kernel_dcache_range((unsigned long)phys_to_virt(paddr), size);
>  }
>  
> -static void pa11_dma_unmap_sg(struct device *dev, struct scatterlist *sglist,
> -		int nents, enum dma_data_direction direction,
> -		unsigned long attrs)
> -{
> -	int i;
> -	struct scatterlist *sg;
> -
> -	BUG_ON(direction == DMA_NONE);
> -
> -	if (attrs & DMA_ATTR_SKIP_CPU_SYNC)
> -		return;
> -
> -	if (direction == DMA_TO_DEVICE)
> -		return;
> -
> -	/* once we do combining we'll need to use phys_to_virt(sg_dma_address(sglist)) */
> -
> -	for_each_sg(sglist, sg, nents, i)
> -		flush_kernel_vmap_range(sg_virt(sg), sg->length);
> -}
> -
> -static void pa11_dma_sync_single_for_cpu(struct device *dev,
> -		dma_addr_t dma_handle, size_t size,
> -		enum dma_data_direction direction)
> -{
> -	BUG_ON(direction == DMA_NONE);
> -
> -	flush_kernel_dcache_range((unsigned long) phys_to_virt(dma_handle),
> -			size);
> -}
> -
> -static void pa11_dma_sync_single_for_device(struct device *dev,
> -		dma_addr_t dma_handle, size_t size,
> -		enum dma_data_direction direction)
> -{
> -	BUG_ON(direction == DMA_NONE);
> -
> -	flush_kernel_dcache_range((unsigned long) phys_to_virt(dma_handle),
> -			size);
> -}
> -
> -static void pa11_dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sglist, int nents, enum dma_data_direction direction)
> -{
> -	int i;
> -	struct scatterlist *sg;
> -
> -	/* once we do combining we'll need to use phys_to_virt(sg_dma_address(sglist)) */
> -
> -	for_each_sg(sglist, sg, nents, i)
> -		flush_kernel_vmap_range(sg_virt(sg), sg->length);
> -}
> -
> -static void pa11_dma_sync_sg_for_device(struct device *dev, struct scatterlist *sglist, int nents, enum dma_data_direction direction)
> -{
> -	int i;
> -	struct scatterlist *sg;
> -
> -	/* once we do combining we'll need to use phys_to_virt(sg_dma_address(sglist)) */
> -
> -	for_each_sg(sglist, sg, nents, i)
> -		flush_kernel_vmap_range(sg_virt(sg), sg->length);
> -}
> -
> -static void pa11_dma_cache_sync(struct device *dev, void *vaddr, size_t size,
> +void arch_dma_cache_sync(struct device *dev, void *vaddr, size_t size,
>  	       enum dma_data_direction direction)
>  {
>  	flush_kernel_dcache_range((unsigned long)vaddr, size);
>  }
>  
> -const struct dma_map_ops pcxl_dma_ops = {
> -	.alloc =		pa11_dma_alloc,
> -	.free =			pa11_dma_free,
> -	.map_page =		pa11_dma_map_page,
> -	.unmap_page =		pa11_dma_unmap_page,
> -	.map_sg =		pa11_dma_map_sg,
> -	.unmap_sg =		pa11_dma_unmap_sg,
> -	.sync_single_for_cpu =	pa11_dma_sync_single_for_cpu,
> -	.sync_single_for_device = pa11_dma_sync_single_for_device,
> -	.sync_sg_for_cpu =	pa11_dma_sync_sg_for_cpu,
> -	.sync_sg_for_device =	pa11_dma_sync_sg_for_device,
> -	.cache_sync =		pa11_dma_cache_sync,
> -};
> -
> -static void *pcx_dma_alloc(struct device *dev, size_t size,
> -		dma_addr_t *dma_handle, gfp_t flag, unsigned long attrs)
> +void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
> +		gfp_t gfp, unsigned long attrs)
>  {
> -	void *addr;
> -
> -	if ((attrs & DMA_ATTR_NON_CONSISTENT) == 0)
> -		return NULL;
> -
> -	addr = (void *)__get_free_pages(flag, get_order(size));
> -	if (addr)
> -		*dma_handle = (dma_addr_t)virt_to_phys(addr);
> -
> -	return addr;
> +	if (boot_cpu_data.cpu_type == pcxl2 || boot_cpu_data.cpu_type == pcxl)
> +		return pa11_dma_alloc(dev, size, dma_handle, gfp, attrs);
> +	if (attrs & DMA_ATTR_NON_CONSISTENT)
> +		return dma_direct_alloc(dev, size, dma_handle, gfp, attrs);
> +	return NULL;
>  }
>  
> -static void pcx_dma_free(struct device *dev, size_t size, void *vaddr,
> -		dma_addr_t iova, unsigned long attrs)
> +void arch_dma_free(struct device *dev, size_t size, void *cpu_addr,
> +		dma_addr_t dma_addr, unsigned long attrs)
>  {
> -	free_pages((unsigned long)vaddr, get_order(size));
> -	return;
> +	if (boot_cpu_data.cpu_type == pcxl2 || boot_cpu_data.cpu_type == pcxl)
> +		pa11_dma_free(dev, size, cpu_addr, dma_addr, attrs);
> +	else
> +		dma_direct_free(dev, size, cpu_addr, dma_addr, attrs);
>  }
> -
> -const struct dma_map_ops pcx_dma_ops = {
> -	.alloc =		pcx_dma_alloc,
> -	.free =			pcx_dma_free,
> -	.map_page =		pa11_dma_map_page,
> -	.unmap_page =		pa11_dma_unmap_page,
> -	.map_sg =		pa11_dma_map_sg,
> -	.unmap_sg =		pa11_dma_unmap_sg,
> -	.sync_single_for_cpu =	pa11_dma_sync_single_for_cpu,
> -	.sync_single_for_device = pa11_dma_sync_single_for_device,
> -	.sync_sg_for_cpu =	pa11_dma_sync_sg_for_cpu,
> -	.sync_sg_for_device =	pa11_dma_sync_sg_for_device,
> -	.cache_sync =		pa11_dma_cache_sync,
> -};
> diff --git a/arch/parisc/kernel/setup.c b/arch/parisc/kernel/setup.c
> index 8d3a7b80ac42..4e87c35c22b7 100644
> --- a/arch/parisc/kernel/setup.c
> +++ b/arch/parisc/kernel/setup.c
> @@ -97,14 +97,12 @@ void __init dma_ops_init(void)
>  		panic(	"PA-RISC Linux currently only supports machines that conform to\n"
>  			"the PA-RISC 1.1 or 2.0 architecture specification.\n");
>  
> -	case pcxs:
> -	case pcxt:
> -		hppa_dma_ops = &pcx_dma_ops;
> -		break;
>  	case pcxl2:
>  		pa7300lc_init();
>  	case pcxl: /* falls through */
> -		hppa_dma_ops = &pcxl_dma_ops;
> +	case pcxs:
> +	case pcxt:
> +		hppa_dma_ops = &dma_noncoherent_ops;
>  		break;
>  	default:
>  		break;
> diff --git a/arch/parisc/mm/init.c b/arch/parisc/mm/init.c
> index cab32ee824d2..4ad91c28ecbe 100644
> --- a/arch/parisc/mm/init.c
> +++ b/arch/parisc/mm/init.c
> @@ -19,7 +19,6 @@
>  #include <linux/gfp.h>
>  #include <linux/delay.h>
>  #include <linux/init.h>
> -#include <linux/pci.h>		/* for hppa_dma_ops and pcxl_dma_ops */
>  #include <linux/initrd.h>
>  #include <linux/swap.h>
>  #include <linux/unistd.h>
> @@ -616,17 +615,13 @@ void __init mem_init(void)
>  	free_all_bootmem();
>  
>  #ifdef CONFIG_PA11
> -	if (hppa_dma_ops == &pcxl_dma_ops) {
> +	if (boot_cpu_data.cpu_type == pcxl2 || boot_cpu_data.cpu_type == pcxl) {
>  		pcxl_dma_start = (unsigned long)SET_MAP_OFFSET(MAP_START);
>  		parisc_vmalloc_start = SET_MAP_OFFSET(pcxl_dma_start
>  						+ PCXL_DMA_MAP_SIZE);
> -	} else {
> -		pcxl_dma_start = 0;
> -		parisc_vmalloc_start = SET_MAP_OFFSET(MAP_START);
> -	}
> -#else
> -	parisc_vmalloc_start = SET_MAP_OFFSET(MAP_START);
> +	} else
>  #endif
> +		parisc_vmalloc_start = SET_MAP_OFFSET(MAP_START);
>  
>  	mem_init_print_info(NULL);
>  
>