linux-next: build failure after merge of the dma-mapping tree

linux-next: build failure after merge of the dma-mapping tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 1222 bytes --]

Hi all,

After merging the dma-mapping tree, today's linux-next build (powerpc
ppc64_defconfig) failed like this:

arch/powerpc/kernel/vio.c: In function 'vio_dma_iommu_alloc_coherent':
arch/powerpc/kernel/vio.c:500:21: error: 'struct dma_map_ops' has no member named 'alloc_coherent'
arch/powerpc/kernel/vio.c: In function 'vio_dma_iommu_free_coherent':
arch/powerpc/kernel/vio.c:514:15: error: 'struct dma_map_ops' has no member named 'free_coherent'
arch/powerpc/kernel/vio.c: At top level:
arch/powerpc/kernel/vio.c:615:2: error: unknown field 'alloc_coherent' specified in initializer
arch/powerpc/kernel/vio.c:615:2: error: initialization from incompatible pointer type [-Werror]
arch/powerpc/kernel/vio.c:615:2: error: (near initialization for 'vio_dma_mapping_ops.alloc') [-Werror]
arch/powerpc/kernel/vio.c:616:2: error: unknown field 'free_coherent' specified in initializer
arch/powerpc/kernel/vio.c:616:2: error: initialization from incompatible pointer type [-Werror]
arch/powerpc/kernel/vio.c:616:2: error: (near initialization for 'vio_dma_mapping_ops.free') [-Werror]

I have dropped the dma-mapping tree for today.
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]

Re: linux-next: build failure after merge of the dma-mapping tree

[Kyungmin Park]

Hi,

It's maybe missing modify arch/powerpc/kernel/vio.c at that time.

following patch will fix it.

Thank you,
Kyungmin Park

diff --git a/arch/powerpc/kernel/vio.c b/arch/powerpc/kernel/vio.c
index 8b08629..b12a77a 100644
--- a/arch/powerpc/kernel/vio.c
+++ b/arch/powerpc/kernel/vio.c
@@ -487,7 +487,7 @@ static void vio_cmo_balance(struct work_struct *work)
 }
diff --git a/arch/powerpc/kernel/vio.c b/arch/powerpc/kernel/vio.c
index 8b08629..b12a77a 100644
--- a/arch/powerpc/kernel/vio.c
+++ b/arch/powerpc/kernel/vio.c
@@ -487,7 +487,7 @@ static void vio_cmo_balance(struct work_struct *work)
 }

 static void *vio_dma_iommu_alloc_coherent(struct device *dev, size_t size,
-                                          dma_addr_t *dma_handle, gfp_t flag)
+               dma_addr_t *dma_handle, gfp_t flag, struct dma_attrs *attrs)
 {
        struct vio_dev *viodev = to_vio_dev(dev);
        void *ret;
@@ -497,7 +497,7 @@ static void *vio_dma_iommu_alloc_coherent(struct device *dev
                return NULL;
        }

-       ret = dma_iommu_ops.alloc_coherent(dev, size, dma_handle, flag);
+       ret = dma_iommu_ops.alloc(dev, size, dma_handle, flag, attrs);
        if (unlikely(ret == NULL)) {
                vio_cmo_dealloc(viodev, roundup(size, PAGE_SIZE));
                atomic_inc(&viodev->cmo.allocs_failed);
@@ -507,11 +507,11 @@ static void *vio_dma_iommu_alloc_coherent(struct device *d
 }

 static void vio_dma_iommu_free_coherent(struct device *dev, size_t size,
-                                        void *vaddr, dma_addr_t dma_handle)
+               void *vaddr, dma_addr_t dma_handle, struct dma_attrs *attrs)
 {
        struct vio_dev *viodev = to_vio_dev(dev);

-       dma_iommu_ops.free_coherent(dev, size, vaddr, dma_handle);
+       dma_iommu_ops.free(dev, size, vaddr, dma_handle, attrs);

        vio_cmo_dealloc(viodev, roundup(size, PAGE_SIZE));
 }
@@ -612,8 +612,8 @@ static u64 vio_dma_get_required_mask(struct device *dev)
 }

 struct dma_map_ops vio_dma_mapping_ops = {
-       .alloc_coherent    = vio_dma_iommu_alloc_coherent,
-       .free_coherent     = vio_dma_iommu_free_coherent,
+       .alloc             = vio_dma_iommu_alloc_coherent,
+       .free              = vio_dma_iommu_free_coherent,
        .map_sg            = vio_dma_iommu_map_sg,
        .unmap_sg          = vio_dma_iommu_unmap_sg,
        .map_page          = vio_dma_iommu_map_page,

On 2/13/12, Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> Hi all,
>
> After merging the dma-mapping tree, today's linux-next build (powerpc
> ppc64_defconfig) failed like this:
>
> arch/powerpc/kernel/vio.c: In function 'vio_dma_iommu_alloc_coherent':
> arch/powerpc/kernel/vio.c:500:21: error: 'struct dma_map_ops' has no member
> named 'alloc_coherent'
> arch/powerpc/kernel/vio.c: In function 'vio_dma_iommu_free_coherent':
> arch/powerpc/kernel/vio.c:514:15: error: 'struct dma_map_ops' has no member
> named 'free_coherent'
> arch/powerpc/kernel/vio.c: At top level:
> arch/powerpc/kernel/vio.c:615:2: error: unknown field 'alloc_coherent'
> specified in initializer
> arch/powerpc/kernel/vio.c:615:2: error: initialization from incompatible
> pointer type [-Werror]
> arch/powerpc/kernel/vio.c:615:2: error: (near initialization for
> 'vio_dma_mapping_ops.alloc') [-Werror]
> arch/powerpc/kernel/vio.c:616:2: error: unknown field 'free_coherent'
> specified in initializer
> arch/powerpc/kernel/vio.c:616:2: error: initialization from incompatible
> pointer type [-Werror]
> arch/powerpc/kernel/vio.c:616:2: error: (near initialization for
> 'vio_dma_mapping_ops.free') [-Werror]
>
> I have dropped the dma-mapping tree for today.
> --
> Cheers,
> Stephen Rothwell                    sfr@canb.auug.org.au
>

RE: linux-next: build failure after merge of the dma-mapping tree

[Marek Szyprowski]

Hi Stephen,

On Monday, February 13, 2012 6:40 AM Stephen Rothwell wrote:
 
> After merging the dma-mapping tree, today's linux-next build (powerpc
> ppc64_defconfig) failed like this:

(snipped)

I'm sorry for the problems. I've updated patches to fix these issues, as 
well as similar problems with MIPS and Hexagon architecture.

Best regards
-- 
Marek Szyprowski
Samsung Poland R&D Center

linux-next: build failure after merge of the dma-mapping tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 3133 bytes --]

Hi all,

After merging the dma-mapping tree, today's linux-next build (powerpc64
allnoconfig) failed like this:

In file included from include/linux/dma-direct.h:10,
                 from arch/powerpc/kernel/dma-iommu.c:9:
include/linux/dma-map-ops.h:328:41: error: expected identifier or '(' before numeric constant
  328 | #define arch_dma_map_page_direct(d, a) (0)
      |                                         ^
arch/powerpc/kernel/dma-iommu.c:16:6: note: in expansion of macro 'arch_dma_map_page_direct'
   16 | bool arch_dma_map_page_direct(struct device *dev, phys_addr_t addr)
      |      ^~~~~~~~~~~~~~~~~~~~~~~~
include/linux/dma-map-ops.h:329:43: error: expected identifier or '(' before numeric constant
  329 | #define arch_dma_unmap_page_direct(d, a) (0)
      |                                           ^
arch/powerpc/kernel/dma-iommu.c:26:6: note: in expansion of macro 'arch_dma_unmap_page_direct'
   26 | bool arch_dma_unmap_page_direct(struct device *dev, dma_addr_t dma_handle)
      |      ^~~~~~~~~~~~~~~~~~~~~~~~~~
include/linux/dma-map-ops.h:330:42: error: expected identifier or '(' before numeric constant
  330 | #define arch_dma_map_sg_direct(d, s, n) (0)
      |                                          ^
arch/powerpc/kernel/dma-iommu.c:34:6: note: in expansion of macro 'arch_dma_map_sg_direct'
   34 | bool arch_dma_map_sg_direct(struct device *dev, struct scatterlist *sg,
      |      ^~~~~~~~~~~~~~~~~~~~~~
include/linux/dma-map-ops.h:331:44: error: expected identifier or '(' before numeric constant
  331 | #define arch_dma_unmap_sg_direct(d, s, n) (0)
      |                                            ^
arch/powerpc/kernel/dma-iommu.c:51:6: note: in expansion of macro 'arch_dma_unmap_sg_direct'
   51 | bool arch_dma_unmap_sg_direct(struct device *dev, struct scatterlist *sg,
      |      ^~~~~~~~~~~~~~~~~~~~~~~~

Caused by commit

  4e52b96ac85c ("powerpc/dma: Fallback to dma_ops when persistent memory present")

I have applied the following patch for today.

From: Stephen Rothwell <sfr@canb.auug.org.au>
Date: Fri, 27 Nov 2020 17:49:28 +1100
Subject: [PATCH] powerpc/dma: fix for "powerpc/dma: Fallback to dma_ops when
 persistent memory present"

Fixes: 4e52b96ac85c ("powerpc/dma: Fallback to dma_ops when persistent memory present")
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
---
 arch/powerpc/kernel/dma-iommu.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/arch/powerpc/kernel/dma-iommu.c b/arch/powerpc/kernel/dma-iommu.c
index c724548ca295..6364311eb6e9 100644
--- a/arch/powerpc/kernel/dma-iommu.c
+++ b/arch/powerpc/kernel/dma-iommu.c
@@ -10,6 +10,8 @@
 #include <linux/pci.h>
 #include <asm/iommu.h>
 
+#ifdef CONFIG_ARCH_HAS_DMA_MAP_DIRECT
+
 #define can_map_direct(dev, addr) \
 	((dev)->bus_dma_limit >= phys_to_dma((dev), (addr)))
 
@@ -64,6 +66,7 @@ bool arch_dma_unmap_sg_direct(struct device *dev, struct scatterlist *sg,
 
 	return true;
 }
+#endif /* CONFIG_ARCH_HAS_DMA_MAP_DIRECT */
 
 /*
  * Generic iommu implementation
-- 
2.29.2

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: build failure after merge of the dma-mapping tree

[Christoph Hellwig]

On Fri, Nov 27, 2020 at 05:58:01PM +1100, Stephen Rothwell wrote:
> Hi all,
> 
> After merging the dma-mapping tree, today's linux-next build (powerpc64
> allnoconfig) failed like this:

I'm looking into fixing this now, but how do you generate a powerpc64
allnoconfig?  My attempts so far either create a ppc32 one or fail entirely.

Re: linux-next: build failure after merge of the dma-mapping tree

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 595 bytes --]

Hi Christoph,

On Fri, 27 Nov 2020 10:21:37 +0100 Christoph Hellwig <hch@lst.de> wrote:
>
> On Fri, Nov 27, 2020 at 05:58:01PM +1100, Stephen Rothwell wrote:
> > Hi all,
> > 
> > After merging the dma-mapping tree, today's linux-next build (powerpc64
> > allnoconfig) failed like this:  
> 
> I'm looking into fixing this now, but how do you generate a powerpc64
> allnoconfig?  My attempts so far either create a ppc32 one or fail entirely.

        echo CONFIG_PPC64=y >tmp_conf
        KCONFIG_ALLCONFIG=tmp_conf make ARCH=powerpc allnoconfig

-- 
Cheers,
Stephen Rothwell

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: build failure after merge of the dma-mapping tree

[Christoph Hellwig]

I fixed this as shown:

diff --git a/arch/powerpc/kernel/dma-iommu.c b/arch/powerpc/kernel/dma-iommu.c
index c724548ca295c0..111249fd619de8 100644
--- a/arch/powerpc/kernel/dma-iommu.c
+++ b/arch/powerpc/kernel/dma-iommu.c
@@ -10,6 +10,7 @@
 #include <linux/pci.h>
 #include <asm/iommu.h>
 
+#ifdef CONFIG_ARCH_HAS_DMA_MAP_DIRECT
 #define can_map_direct(dev, addr) \
 	((dev)->bus_dma_limit >= phys_to_dma((dev), (addr)))
 
@@ -64,6 +65,7 @@ bool arch_dma_unmap_sg_direct(struct device *dev, struct scatterlist *sg,
 
 	return true;
 }
+#endif /* CONFIG_ARCH_HAS_DMA_MAP_DIRECT */
 
 /*
  * Generic iommu implementation