[PATCH] mips: consolidate coherent and non-coherent dma_alloc code

[PATCH] mips: consolidate coherent and non-coherent dma_alloc code

[Christoph Hellwig]

Besides eliminating lots of duplication this also allows allocations with
the DMA_ATTR_NON_CONSISTENT to use the CMA allocator.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---

Note: compile tested only.

 arch/mips/mm/dma-default.c | 42 +++---------------------------------------
 1 file changed, 3 insertions(+), 39 deletions(-)

diff --git a/arch/mips/mm/dma-default.c b/arch/mips/mm/dma-default.c
index fe8df14b6169..9e0faaf9f96f 100644
--- a/arch/mips/mm/dma-default.c
+++ b/arch/mips/mm/dma-default.c
@@ -114,23 +114,6 @@ static gfp_t massage_gfp_flags(const struct device *dev, gfp_t gfp)
 	return gfp | dma_flag;
 }
 
-static void *mips_dma_alloc_noncoherent(struct device *dev, size_t size,
-	dma_addr_t * dma_handle, gfp_t gfp)
-{
-	void *ret;
-
-	gfp = massage_gfp_flags(dev, gfp);
-
-	ret = (void *) __get_free_pages(gfp, get_order(size));
-
-	if (ret != NULL) {
-		memset(ret, 0, size);
-		*dma_handle = plat_map_dma_mem(dev, ret, size);
-	}
-
-	return ret;
-}
-
 static void *mips_dma_alloc_coherent(struct device *dev, size_t size,
 	dma_addr_t *dma_handle, gfp_t gfp, unsigned long attrs)
 {
@@ -138,13 +121,6 @@ static void *mips_dma_alloc_coherent(struct device *dev, size_t size,
 	struct page *page = NULL;
 	unsigned int count = PAGE_ALIGN(size) >> PAGE_SHIFT;
 
-	/*
-	 * XXX: seems like the coherent and non-coherent implementations could
-	 * be consolidated.
-	 */
-	if (attrs & DMA_ATTR_NON_CONSISTENT)
-		return mips_dma_alloc_noncoherent(dev, size, dma_handle, gfp);
-
 	gfp = massage_gfp_flags(dev, gfp);
 
 	if (IS_ENABLED(CONFIG_DMA_CMA) && gfpflags_allow_blocking(gfp))
@@ -159,7 +135,8 @@ static void *mips_dma_alloc_coherent(struct device *dev, size_t size,
 	ret = page_address(page);
 	memset(ret, 0, size);
 	*dma_handle = plat_map_dma_mem(dev, ret, size);
-	if (!plat_device_is_coherent(dev)) {
+	if (!(attrs & DMA_ATTR_NON_CONSISTENT) &&
+	    !plat_device_is_coherent(dev)) {
 		dma_cache_wback_inv((unsigned long) ret, size);
 		ret = UNCAC_ADDR(ret);
 	}
@@ -167,14 +144,6 @@ static void *mips_dma_alloc_coherent(struct device *dev, size_t size,
 	return ret;
 }
 
-
-static void mips_dma_free_noncoherent(struct device *dev, size_t size,
-		void *vaddr, dma_addr_t dma_handle)
-{
-	plat_unmap_dma_mem(dev, dma_handle, size, DMA_BIDIRECTIONAL);
-	free_pages((unsigned long) vaddr, get_order(size));
-}
-
 static void mips_dma_free_coherent(struct device *dev, size_t size, void *vaddr,
 	dma_addr_t dma_handle, unsigned long attrs)
 {
@@ -182,14 +151,9 @@ static void mips_dma_free_coherent(struct device *dev, size_t size, void *vaddr,
 	unsigned int count = PAGE_ALIGN(size) >> PAGE_SHIFT;
 	struct page *page = NULL;
 
-	if (attrs & DMA_ATTR_NON_CONSISTENT) {
-		mips_dma_free_noncoherent(dev, size, vaddr, dma_handle);
-		return;
-	}
-
 	plat_unmap_dma_mem(dev, dma_handle, size, DMA_BIDIRECTIONAL);
 
-	if (!plat_device_is_coherent(dev))
+	if (!(attrs & DMA_ATTR_NON_CONSISTENT) && !plat_device_is_coherent(dev))
 		addr = CAC_ADDR(addr);
 
 	page = virt_to_page((void *) addr);
-- 
2.11.0

Re: [PATCH] mips: consolidate coherent and non-coherent dma_alloc code

[Ralf Baechle]

On Fri, Jun 16, 2017 at 09:12:29AM +0200, Christoph Hellwig wrote:

> Besides eliminating lots of duplication this also allows allocations with
> the DMA_ATTR_NON_CONSISTENT to use the CMA allocator.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>

Acked-by: Ralf Baechle <ralf@linux-mips.org>

> Note: compile tested only.

Your change essentially reverts a change from Oct 2007 so it should work.

  Ralf

Re: [PATCH] mips: consolidate coherent and non-coherent dma_alloc code

[Christoph Hellwig]

On Tue, Aug 29, 2017 at 02:54:55PM +0200, Ralf Baechle wrote:
> On Fri, Jun 16, 2017 at 09:12:29AM +0200, Christoph Hellwig wrote:
> 
> > Besides eliminating lots of duplication this also allows allocations with
> > the DMA_ATTR_NON_CONSISTENT to use the CMA allocator.
> > 
> > Signed-off-by: Christoph Hellwig <hch@lst.de>
> 
> Acked-by: Ralf Baechle <ralf@linux-mips.org>

I don't have anything that depends on it - do you just want to pick
it up in the mips tree?

> > Note: compile tested only.
> 
> Your change essentially reverts a change from Oct 2007 so it should work.

Which commit is that?  I can't find anything that looks related in the
Oct 2017 commits.

Re: [PATCH] mips: consolidate coherent and non-coherent dma_alloc code

[Ralf Baechle]

On Tue, Aug 29, 2017 at 02:59:28PM +0200, Christoph Hellwig wrote:

> On Tue, Aug 29, 2017 at 02:54:55PM +0200, Ralf Baechle wrote:
> > On Fri, Jun 16, 2017 at 09:12:29AM +0200, Christoph Hellwig wrote:
> > 
> > > Besides eliminating lots of duplication this also allows allocations with
> > > the DMA_ATTR_NON_CONSISTENT to use the CMA allocator.
> > > 
> > > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > 
> > Acked-by: Ralf Baechle <ralf@linux-mips.org>
> 
> I don't have anything that depends on it - do you just want to pick
> it up in the mips tree?

Sure, done.

> > > Note: compile tested only.
> > 
> > Your change essentially reverts a change from Oct 2007 so it should work.
> 
> Which commit is that?  I can't find anything that looks related in the
> Oct 2017 commits.

Well, it predates the git history so it indeed is hard to find:

https://git.linux-mips.org/cgit/ralf/linux.git/commit/?id=7ca16d269a1a4b96d98968b48f137977bcab1522

  Ralf

Re: [PATCH] mips: consolidate coherent and non-coherent dma_alloc code

[Christoph Hellwig]

On Tue, Aug 29, 2017 at 03:18:33PM +0200, Ralf Baechle wrote:
> > > Your change essentially reverts a change from Oct 2007 so it should work.
> > 
> > Which commit is that?  I can't find anything that looks related in the
> > Oct 2017 commits.
> 
> Well, it predates the git history so it indeed is hard to find:
> 
> https://git.linux-mips.org/cgit/ralf/linux.git/commit/?id=7ca16d269a1a4b96d98968b48f137977bcab1522

That looks like a different commit from a different series than the
one you replied to.  But I'll happily take the ACK for that one :)