omap iommu changes to remove plat includes

Looks like the iommu framework does not have generic functions exported for all the needs yet. The hardware specific functions are defined in files like intel-iommu.h and amd-iommu.h. Follow the same standard for omap-iommu.h.

This is needed because we are removing plat and mach includes for ARM common zImage support. Further work should continue in the iommu framework context as only pure platform data will be communicated from arch/arm/*omap*/* code to the iommu framework.

Cc Joerg Roedel <[email protected]>

Cc Ohad Ben-Cohen <[email protected]>

Cc Ido Yariv <[email protected]>

Cc Laurent Pinchart <[email protected]>

Cc Mauro Carvalho Chehab <[email protected]>

Cc Omar Ramirez Luna <[email protected]>

Cc linux-media at vger.kernel.org

Signed-off-by Tony Lindgren <[email protected]>

---
 arch/arm/mach-omap2/iommu2.c               |    1 
 arch/arm/plat-omap/include/plat/iommu.h    |   10 +--
 arch/arm/plat-omap/include/plat/iovmm.h    |   89 ----------------------------
 drivers/iommu/omap-iommu-debug.c           |    2 -
 drivers/iommu/omap-iommu.c                 |    1 
 drivers/iommu/omap-iovmm.c                 |   46 ++++++++++++++
 drivers/media/platform/omap3isp/isp.c      |    1 
 drivers/media/platform/omap3isp/isp.h      |    2 -
 drivers/media/platform/omap3isp/ispccdc.c  |    1 
 drivers/media/platform/omap3isp/ispstat.c  |    1 
 drivers/media/platform/omap3isp/ispvideo.c |    2 -

Tony,

Em Thu, 18 Oct 2012 13:28:42 -0700
Tony Lindgren <[email protected]> escreveu:

Quoted 19 lines

Looks like the iommu framework does not have generic functions exported for all the needs yet. The hardware specific functions are defined in files like intel-iommu.h and amd-iommu.h. Follow the same standard for omap-iommu.h.

This is needed because we are removing plat and mach includes for ARM common zImage support. Further work should continue in the iommu framework context as only pure platform data will be communicated from arch/arm/*omap*/* code to the iommu framework.

Cc Joerg Roedel <[email protected]>

Cc Ohad Ben-Cohen <[email protected]>

Cc Ido Yariv <[email protected]>

Cc Laurent Pinchart <[email protected]>

Cc Mauro Carvalho Chehab <[email protected]>

Cc Omar Ramirez Luna <[email protected]>

Cc linux-media at vger.kernel.org

Signed-off-by Tony Lindgren <[email protected]>

Quoted 12 lines

---
 arch/arm/mach-omap2/iommu2.c               |    1 
 arch/arm/plat-omap/include/plat/iommu.h    |   10 +--
 arch/arm/plat-omap/include/plat/iovmm.h    |   89 ----------------------------
 drivers/iommu/omap-iommu-debug.c           |    2 -
 drivers/iommu/omap-iommu.c                 |    1 
 drivers/iommu/omap-iovmm.c                 |   46 ++++++++++++++
 drivers/media/platform/omap3isp/isp.c      |    1 
 drivers/media/platform/omap3isp/isp.h      |    2 -
 drivers/media/platform/omap3isp/ispccdc.c  |    1 
 drivers/media/platform/omap3isp/ispstat.c  |    1 
 drivers/media/platform/omap3isp/ispvideo.c |    2 -

Acked-by Mauro Carvalho Chehab <[email protected]>

It is better if you send this entire series through the ARM tree, keeping this hole series altogether (and avoiding the risk of breaking bisectability if it goes through separate trees). So, you can add my ack for those header moves for drivers/media/platform/*.

Looks like the iommu framework does not have generic functions exported for all the needs yet. The hardware specific functions are defined in files like intel-iommu.h and amd-iommu.h. Follow the same standard for omap-iommu.h.

This is needed because we are removing plat and mach includes for ARM common zImage support. Further work should continue in the iommu framework context as only pure platform data will be communicated from arch/arm/*omap*/* code to the iommu framework.

Cc Joerg Roedel <[email protected]>

Cc Ohad Ben-Cohen <[email protected]>

Cc Ido Yariv <[email protected]>

Cc Laurent Pinchart <[email protected]>

Cc Mauro Carvalho Chehab <[email protected]>

Cc Omar Ramirez Luna <[email protected]>

Cc [email protected]

Signed-off-by Tony Lindgren <[email protected]>

---
 arch/arm/mach-omap2/iommu2.c               |    1
 arch/arm/plat-omap/include/plat/iommu.h    |   10 +--
 arch/arm/plat-omap/include/plat/iovmm.h    |   89 -------------------------
 drivers/iommu/omap-iommu-debug.c           |    2 -
 drivers/iommu/omap-iommu.c                 |    1
 drivers/iommu/omap-iovmm.c                 |   46 ++++++++++++++
 drivers/media/platform/omap3isp/isp.c      |    1
 drivers/media/platform/omap3isp/isp.h      |    2 -
 drivers/media/platform/omap3isp/ispccdc.c  |    1
 drivers/media/platform/omap3isp/ispstat.c  |    1
 drivers/media/platform/omap3isp/ispvideo.c |    2 -
 include/linux/omap-iommu.h                 |   47 +++++++++++++++
 12 files changed, 101 insertions(+), 102 deletions(-)
 delete mode 100644 arch/arm/plat-omap/include/plat/iovmm.h
 create mode 100644 include/linux/omap-iommu.h

arch/arm/mach-omap2/iommu2.c +0 -0

index eefc379..77cbf2f 100644

arch/arm/mach-omap2/iommu2.c +0 -0

 #include <plat/iommu.h>
 

(quoted diff) +0 -0

b/arch/arm/plat-omap/include/plat/iommu.h index 7e8c7b6..a4b71b1 100644

arch/arm/plat-omap/include/plat/iommu.h +1 -1

@@ -216,13 +216,10 @@ static inline struct omap_iommu
 *dev_to_omap_iommu(struct device *dev) #define MMU_RAM_PADDR_SHIFT	12
 #define MMU_RAM_PADDR_MASK \
 	((~0UL >> MMU_RAM_PADDR_SHIFT) << MMU_RAM_PADDR_SHIFT)
-#define MMU_RAM_ENDIAN_SHIFT	9
+

arch/arm/plat-omap/include/plat/iommu.h +0 -3

@@ -269,9 +266,6 @@ extern int omap_iommu_set_isr(const char *name,
 				    void *priv),
 			 void *isr_priv);
 
-extern void omap_iommu_save_ctx(struct device *dev);
-extern void omap_iommu_restore_ctx(struct device *dev);
-
 extern int omap_install_iommu_arch(const struct iommu_functions *ops);
 extern void omap_uninstall_iommu_arch(const struct iommu_functions *ops);
 

(quoted diff) +0 -0

b/arch/arm/plat-omap/include/plat/iovmm.h deleted file mode 100644 index 498e57c..0000000

arch/arm/plat-omap/include/plat/iovmm.h deleted +0 -89

@@ -1,89 +0,0 @@
-/*
- * omap iommu: simple virtual address space management
- *
- * Copyright (C) 2008-2009 Nokia Corporation
- *
- * Written by Hiroshi DOYU <[email protected]>
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 as
- * published by the Free Software Foundation.
- */
-
-#ifndef __IOMMU_MMAP_H
-#define __IOMMU_MMAP_H
-
-#include <linux/iommu.h>
-
-struct iovm_struct {
-	struct omap_iommu	*iommu;	/* iommu object which this belongs to */
-	u32			da_start; /* area definition */
-	u32			da_end;
-	u32			flags; /* IOVMF_: see below */
-	struct list_head	list; /* linked in ascending order */
-	const struct sg_table	*sgt; /* keep 'page' <-> 'da' mapping */
-	void			*va; /* mpu side mapped address */
-};
-
-/*
- * IOVMF_FLAGS: attribute for iommu virtual memory area(iovma)
- *
- * lower 16 bit is used for h/w and upper 16 bit is for s/w.
- */
-#define IOVMF_SW_SHIFT		16
-
-/*
- * iovma: h/w flags derived from cam and ram attribute
- */
-#define IOVMF_CAM_MASK		(~((1 << 10) - 1))
-#define IOVMF_RAM_MASK		(~IOVMF_CAM_MASK)
-
-#define IOVMF_PGSZ_MASK		(3 << 0)
-#define IOVMF_PGSZ_1M		MMU_CAM_PGSZ_1M
-#define IOVMF_PGSZ_64K		MMU_CAM_PGSZ_64K
-#define IOVMF_PGSZ_4K		MMU_CAM_PGSZ_4K
-#define IOVMF_PGSZ_16M		MMU_CAM_PGSZ_16M
-
-#define IOVMF_ENDIAN_MASK	(1 << 9)
-#define IOVMF_ENDIAN_BIG	MMU_RAM_ENDIAN_BIG
-#define IOVMF_ENDIAN_LITTLE	MMU_RAM_ENDIAN_LITTLE
-
-#define IOVMF_ELSZ_MASK		(3 << 7)
-#define IOVMF_ELSZ_8		MMU_RAM_ELSZ_8
-#define IOVMF_ELSZ_16		MMU_RAM_ELSZ_16
-#define IOVMF_ELSZ_32		MMU_RAM_ELSZ_32
-#define IOVMF_ELSZ_NONE		MMU_RAM_ELSZ_NONE
-
-#define IOVMF_MIXED_MASK	(1 << 6)
-#define IOVMF_MIXED		MMU_RAM_MIXED
-
-/*
- * iovma: s/w flags, used for mapping and umapping internally.
- */
-#define IOVMF_MMIO		(1 << IOVMF_SW_SHIFT)
-#define IOVMF_ALLOC		(2 << IOVMF_SW_SHIFT)
-#define IOVMF_ALLOC_MASK	(3 << IOVMF_SW_SHIFT)
-
-/* "superpages" is supported just with physically linear pages */
-#define IOVMF_DISCONT		(1 << (2 + IOVMF_SW_SHIFT))
-#define IOVMF_LINEAR		(2 << (2 + IOVMF_SW_SHIFT))
-#define IOVMF_LINEAR_MASK	(3 << (2 + IOVMF_SW_SHIFT))
-
-#define IOVMF_DA_FIXED		(1 << (4 + IOVMF_SW_SHIFT))
-
-
-extern struct iovm_struct *omap_find_iovm_area(struct device *dev, u32 da);
-extern u32
-omap_iommu_vmap(struct iommu_domain *domain, struct device *dev, u32 da,
-			const struct sg_table *sgt, u32 flags);
-extern struct sg_table *omap_iommu_vunmap(struct iommu_domain *domain,
-				struct device *dev, u32 da);
-extern u32
-omap_iommu_vmalloc(struct iommu_domain *domain, struct device *dev,
-				u32 da, size_t bytes, u32 flags);
-extern void
-omap_iommu_vfree(struct iommu_domain *domain, struct device *dev,
-				const u32 da);
-extern void *omap_da_to_va(struct device *dev, u32 da);
-
-#endif /* __IOMMU_MMAP_H */

(quoted diff) +0 -0

b/drivers/iommu/omap-iommu-debug.c index 0cac372..cf4a0b5 100644

drivers/iommu/omap-iommu-debug.c +1 -1

@@ -18,9 +18,9 @@
 #include <linux/uaccess.h>
 #include <linux/platform_device.h>
 #include <linux/debugfs.h>
+#include <linux/omap-iommu.h>
 
 #include <plat/iommu.h>
-#include <plat/iovmm.h>
 
 #include "omap-iopgtable.h"
 

drivers/iommu/omap-iommu.c +1 -0

index f2bbfb0..eadcfde 100644
@@ -19,6 +19,7 @@
 #include <linux/clk.h>
 #include <linux/platform_device.h>
 #include <linux/iommu.h>
+#include <linux/omap-iommu.h>
 #include <linux/mutex.h>
 #include <linux/spinlock.h>
 

drivers/iommu/omap-iovmm.c +45 -1

index b332392..9852101 100644
@@ -17,15 +17,59 @@
 #include <linux/device.h>
 #include <linux/scatterlist.h>
 #include <linux/iommu.h>
+#include <linux/omap-iommu.h>
 
 #include <asm/cacheflush.h>
 #include <asm/mach/map.h>
 
 #include <plat/iommu.h>
-#include <plat/iovmm.h>
 
 #include "omap-iopgtable.h"
 
+/*
+ * IOVMF_FLAGS: attribute for iommu virtual memory area(iovma)
+ *
+ * lower 16 bit is used for h/w and upper 16 bit is for s/w.
+ */
+#define IOVMF_SW_SHIFT		16
+
+/*
+ * iovma: h/w flags derived from cam and ram attribute
+ */
+#define IOVMF_CAM_MASK		(~((1 << 10) - 1))
+#define IOVMF_RAM_MASK		(~IOVMF_CAM_MASK)
+
+#define IOVMF_PGSZ_MASK		(3 << 0)
+#define IOVMF_PGSZ_1M		MMU_CAM_PGSZ_1M
+#define IOVMF_PGSZ_64K		MMU_CAM_PGSZ_64K
+#define IOVMF_PGSZ_4K		MMU_CAM_PGSZ_4K
+#define IOVMF_PGSZ_16M		MMU_CAM_PGSZ_16M
+
+#define IOVMF_ENDIAN_MASK	(1 << 9)
+#define IOVMF_ENDIAN_BIG	MMU_RAM_ENDIAN_BIG
+
+#define IOVMF_ELSZ_MASK		(3 << 7)
+#define IOVMF_ELSZ_16		MMU_RAM_ELSZ_16
+#define IOVMF_ELSZ_32		MMU_RAM_ELSZ_32
+#define IOVMF_ELSZ_NONE		MMU_RAM_ELSZ_NONE
+
+#define IOVMF_MIXED_MASK	(1 << 6)
+#define IOVMF_MIXED		MMU_RAM_MIXED
+
+/*
+ * iovma: s/w flags, used for mapping and umapping internally.
+ */
+#define IOVMF_MMIO		(1 << IOVMF_SW_SHIFT)
+#define IOVMF_ALLOC		(2 << IOVMF_SW_SHIFT)
+#define IOVMF_ALLOC_MASK	(3 << IOVMF_SW_SHIFT)
+
+/* "superpages" is supported just with physically linear pages */
+#define IOVMF_DISCONT		(1 << (2 + IOVMF_SW_SHIFT))
+#define IOVMF_LINEAR		(2 << (2 + IOVMF_SW_SHIFT))
+#define IOVMF_LINEAR_MASK	(3 << (2 + IOVMF_SW_SHIFT))
+
+#define IOVMF_DA_FIXED		(1 << (4 + IOVMF_SW_SHIFT))
+
 static struct kmem_cache *iovm_area_cachep;
 
 /* return the offset of the first scatterlist entry in a sg table */

(quoted diff) +0 -0

b/drivers/media/platform/omap3isp/isp.c index 99640d8..d72bd38 100644

drivers/media/platform/omap3isp/isp.c +1 -0

@@ -66,6 +66,7 @@
 #include <linux/slab.h>
 #include <linux/sched.h>
 #include <linux/vmalloc.h>
+#include <linux/omap-iommu.h>
 
 #include <media/v4l2-common.h>
 #include <media/v4l2-device.h>

(quoted diff) +0 -0

b/drivers/media/platform/omap3isp/isp.h index 8be7487..50be8c2 100644

drivers/media/platform/omap3isp/isp.h +0 -2

@@ -34,8 +34,6 @@
 #include <linux/platform_device.h>
 #include <linux/wait.h>
 #include <linux/iommu.h>
-#include <plat/iommu.h>
-#include <plat/iovmm.h>
 
 #include "ispstat.h"
 #include "ispccdc.h"

(quoted diff) +0 -0

b/drivers/media/platform/omap3isp/ispccdc.c index 60181ab..b2d4976 100644

drivers/media/platform/omap3isp/ispccdc.c +1 -0

@@ -32,6 +32,7 @@
 #include <linux/mm.h>
 #include <linux/sched.h>
 #include <linux/slab.h>
+#include <linux/omap-iommu.h>
 #include <media/v4l2-event.h>
 
 #include "isp.h"

(quoted diff) +0 -0

b/drivers/media/platform/omap3isp/ispstat.c index d7ac76b..e135774 100644

drivers/media/platform/omap3isp/ispstat.c +1 -0

@@ -28,6 +28,7 @@
 #include <linux/dma-mapping.h>
 #include <linux/slab.h>
 #include <linux/uaccess.h>
+#include <linux/omap-iommu.h>
 
 #include "isp.h"
 

(quoted diff) +0 -0

b/drivers/media/platform/omap3isp/ispvideo.c index a0b737fe..2853bef 100644

drivers/media/platform/omap3isp/ispvideo.c +1 -1

@@ -32,10 +32,10 @@
 #include <linux/sched.h>
 #include <linux/slab.h>
 #include <linux/vmalloc.h>
+#include <linux/omap-iommu.h>
 #include <media/v4l2-dev.h>
 #include <media/v4l2-ioctl.h>
 #include <plat/iommu.h>
-#include <plat/iovmm.h>
 #include <plat/omap-pm.h>
 
 #include "ispvideo.h"

include/linux/omap-iommu.h new file +1 -0

new file mode 100644
index 0000000..bd12665
@@ -0,0 +1,47 @@
+/*

include/linux/omap-iommu.h new file +7 -0

+
+struct iovm_struct {
+	struct omap_iommu	*iommu;	/* iommu object which this belongs to */
+	u32			da_start; /* area definition */
+	u32			da_end;
+	u32			flags; /* IOVMF_: see below */
+	struct list_head	list; /* linked in ascending order */

include/linux/omap-iommu.h new file +8 -0

+struct iommu_domain;
+
+extern struct iovm_struct *omap_find_iovm_area(struct device *dev, u32 da);
+extern u32
+omap_iommu_vmap(struct iommu_domain *domain, struct device *dev, u32 da,
+			const struct sg_table *sgt, u32 flags);
+extern struct sg_table *omap_iommu_vunmap(struct iommu_domain *domain,
+				struct device *dev, u32 da);

Nitpicking, please keep the headers sorted alphabetically, here and in all locations below (especially the OMAP3 ISP driver).

(OK, there's already one misplaced #include, but let's not make it worse :-))

I plan to push cleanup patches for the staging tidspbridge driver that get rid of the local register definitions and use plat/iommu.h instead. That's obviously an interim solution as in the long run the driver should use the IOMMU API, but in the meantime it's a step in the right direction. Would it then make sense to move all those definitions to include/linux/omap-iommu.h, not just the ones used by the OMAP3 ISP driver ?

Shouldn't this header be split in include/linux/omap-iommu.h and include/linux/omap-iovmm.h ? I would also move all the hardware IOVMF flags to include/linux/omap-iovmm.h, not just the two currently used by the OMAP3 ISP driver. The software flags can be kept in drivers/iommu/omap-iovmm.c.

include/linux/omap-iommu.h line 47

Quoted 2 lines

include/linux/omap-iommu.h new file +2 -0

+extern void omap_iommu_save_ctx(struct device *dev);
+extern void omap_iommu_restore_ctx(struct device *dev);

Do we really need to prefix functions with 'extern' ?

arch/arm/mach-omap2/iommu2.c line 20

Quoted 7 lines · On Thursday 18 October 2012 13:28:42 Tony Lindgren wrote:

arch/arm/mach-omap2/iommu2.c +1 -0

@@ -17,6 +17,7 @@
 #include <linux/module.h>
 #include <linux/slab.h>
 #include <linux/stringify.h>
+#include <linux/omap-iommu.h>

Nitpicking, please keep the headers sorted alphabetically, here and in all locations below (especially the OMAP3 ISP driver).

(OK, there's already one misplaced #include, but let's not make it worse :-))

Quoted 8 lines

include/linux/omap-iommu.h new file +5 -0

@@ -0,0 +1,47 @@
+/*
+ * omap iommu: simple virtual address space management
+ *
+ * Copyright (C) 2008-2009 Nokia Corporation
+ *

include/linux/omap-iommu.h line 11

Quoted 6 lines

include/linux/omap-iommu.h new file +6 -0

+ * Written by Hiroshi DOYU <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */

Missing #ifndef #define ... #endif

Quoted 9 lines · Laurent Pinchart

Quoted 8 lines · On Thursday 18 October 2012 13:28:42 Tony Lindgren wrote:

arch/arm/mach-omap2/iommu2.c +0 -0

@@ -17,6 +17,7 @@
 
 #include <linux/module.h>
 #include <linux/slab.h>
 #include <linux/stringify.h>
 

Quoted 7 lines

arch/arm/mach-omap2/iommu2.c line 22

Quoted 1 line

arch/arm/mach-omap2/iommu2.c +1 -0

+#include <linux/omap-iommu.h>

Nitpicking, please keep the headers sorted alphabetically, here and in all locations below (especially the OMAP3 ISP driver).

(OK, there's already one misplaced #include, but let's not make it worse :-))

This is fixed now.

Quoted 8 lines

Quoted 8 lines

include/linux/omap-iommu.h new file +5 -0

@@ -0,0 +1,47 @@
+/*
+ * omap iommu: simple virtual address space management
+ *
+ * Copyright (C) 2008-2009 Nokia Corporation
+ *

Quoted 8 lines

include/linux/omap-iommu.h line 11

Quoted 6 lines

include/linux/omap-iommu.h new file +6 -0

+ * Written by Hiroshi DOYU <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */

Missing #ifndef #define ... #endif

I've added it as _OMAP_IOMMU_H_, looks like both intel-iommu.h and amd-iommu.h have it.

Hopefully that's OK for you as a base to do further iommu patches on, updated patch below.

BTW, doing a test compile on v3.7-rc2, I'm seeing the following warnings for omap3isp for isp_video_ioctl_ops:

drivers/media/platform/omap3isp/ispvideo.c:1213: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/ispccdc.c:2303: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/ispccdc.c:2304: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/isph3a_aewb.c:282: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/isph3a_aewb.c:283: warning: initialization from incompatible pointer type

drivers/media/platform/omap3isp/isph3a_af.c:347: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/isph3a_af.c:348: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/isphist.c:453: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/isphist.c:454: warning: initialization from incompatible pointer type

Quoted 14 lines · On Wednesday 24 October 2012 15:34:12 Tony Lindgren wrote:

BTW, doing a test compile on v3.7-rc2, I'm seeing the following warnings for omap3isp for isp_video_ioctl_ops:

drivers/media/platform/omap3isp/ispvideo.c:1213: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/ispccdc.c:2303: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/ispccdc.c:2304: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/isph3a_aewb.c:282: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/isph3a_aewb.c:283: warning: initialization from incompatible pointer type

Quoted 8 lines

drivers/media/platform/omap3isp/isph3a_af.c:347: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/isph3a_af.c:348: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/isphist.c:453: warning: initialization from incompatible pointer type drivers/media/platform/omap3isp/isphist.c:454: warning: initialization from incompatible pointer type

I've just sent a pull request to linux-media for v3.7 with fixes for those.

From: Ido Yariv <[email protected]>

Move some of the definitions in omap-iommu.h that can be made local to either drivers/iommu.

Cc Joerg Roedel <[email protected]>

Cc Ohad Ben-Cohen <[email protected]>

Cc Laurent Pinchart <[email protected]>

Cc Mauro Carvalho Chehab <[email protected]>

Cc Omar Ramirez Luna <[email protected]>

Signed-off-by Ido Yariv <[email protected]>

[[email protected]: updated for header changes in the series]

Signed-off-by Tony Lindgren <[email protected]>

---
 drivers/iommu/omap-iommu.c  |   15 +++++++++++++++
 drivers/iommu/omap-iommu.h  |   33 +++------------------------------
 drivers/iommu/omap-iommu2.c |    6 ++++++
 3 files changed, 24 insertions(+), 30 deletions(-)

drivers/iommu/omap-iommu.c +15 -0

index 4db86e1..df84087 100644
@@ -54,6 +54,21 @@ struct omap_iommu_domain {
 	spinlock_t lock;
 };
 
+#define MMU_LOCK_BASE_SHIFT	10
+#define MMU_LOCK_BASE_MASK	(0x1f << MMU_LOCK_BASE_SHIFT)
+#define MMU_LOCK_BASE(x)	\
+	((x & MMU_LOCK_BASE_MASK) >> MMU_LOCK_BASE_SHIFT)
+
+#define MMU_LOCK_VICT_SHIFT	4
+#define MMU_LOCK_VICT_MASK	(0x1f << MMU_LOCK_VICT_SHIFT)
+#define MMU_LOCK_VICT(x)	\
+	((x & MMU_LOCK_VICT_MASK) >> MMU_LOCK_VICT_SHIFT)
+
+struct iotlb_lock {
+	short base;
+	short vict;
+};
+
 /* accommodate the difference between omap1 and omap2/3 */
 static const struct iommu_functions *arch_iommu;
 

drivers/iommu/omap-iommu.h +0 -5

index 8c3378d..2b5f3c0 100644
@@ -72,11 +72,6 @@ struct cr_regs {
 	};
 };
 
-struct iotlb_lock {
-	short base;
-	short vict;
-};
-
 /* architecture specific functions */
 struct iommu_functions {
 	unsigned long	version;

drivers/iommu/omap-iommu.h +0 -10

 /*
  * MMU Register offsets
  */

@@ -151,16 +139,6 @@ static inline struct omap_iommu
 *dev_to_omap_iommu(struct device *dev) /*
  * MMU Register bit definitions
  */
-#define MMU_LOCK_BASE_SHIFT	10
-#define MMU_LOCK_BASE_MASK	(0x1f << MMU_LOCK_BASE_SHIFT)
-#define MMU_LOCK_BASE(x)	\
-	((x & MMU_LOCK_BASE_MASK) >> MMU_LOCK_BASE_SHIFT)
-
-#define MMU_LOCK_VICT_SHIFT	4
-#define MMU_LOCK_VICT_MASK	(0x1f << MMU_LOCK_VICT_SHIFT)
-#define MMU_LOCK_VICT(x)	\
-	((x & MMU_LOCK_VICT_MASK) >> MMU_LOCK_VICT_SHIFT)
-
 #define MMU_CAM_VATAG_SHIFT	12
 #define MMU_CAM_VATAG_MASK \
 	((~0UL >> MMU_CAM_VATAG_SHIFT) << MMU_CAM_VATAG_SHIFT)

@@ -222,20 +200,15 @@ extern void omap_iotlb_cr_to_e(struct cr_regs *cr,

struct iotlb_entry *e); extern int
omap_iopgtable_store_entry(struct omap_iommu *obj, struct iotlb_entry *e);

drivers/iommu/omap-iommu.h +3 -8

-extern int omap_iommu_set_isr(const char *name,
-		 int (*isr)(struct omap_iommu *obj, u32 da, u32 iommu_errs,
-				    void *priv),
-			 void *isr_priv);
-
 extern void omap_iommu_save_ctx(struct device *dev);
 extern void omap_iommu_restore_ctx(struct device *dev);
 
-extern int omap_install_iommu_arch(const struct iommu_functions *ops);
-extern void omap_uninstall_iommu_arch(const struct iommu_functions *ops);
-
 extern int omap_foreach_iommu_device(void *data,
 				int (*fn)(struct device *, void *));
 
+extern int omap_install_iommu_arch(const struct iommu_functions *ops);
+extern void omap_uninstall_iommu_arch(const struct iommu_functions *ops);
+
 extern ssize_t
 omap_iommu_dump_ctx(struct omap_iommu *obj, char *buf, ssize_t len);
 extern size_t

drivers/iommu/omap-iommu2.c +6 -0

index 066e6b2..4d8d91d 100644
@@ -68,6 +68,12 @@
 	 ((pgsz) == MMU_CAM_PGSZ_64K) ? 0xffff0000 :	\
 	 ((pgsz) == MMU_CAM_PGSZ_4K)  ? 0xfffff000 : 0)
 
+/* IOMMU errors */
+#define OMAP_IOMMU_ERR_TLB_MISS		(1 << 0)
+#define OMAP_IOMMU_ERR_TRANS_FAULT	(1 << 1)
+#define OMAP_IOMMU_ERR_EMU_MISS		(1 << 2)
+#define OMAP_IOMMU_ERR_TBLWALK_FAULT	(1 << 3)
+#define OMAP_IOMMU_ERR_MULTIHIT_FAULT	(1 << 4)
 
 static void __iommu_set_twl(struct omap_iommu *obj, bool on)
 {

drivers/iommu/omap-iommu.h static inline struct omap_iommu · line 126

Quoted 11 lines · On Thursday 18 October 2012 13:28:48 Tony Lindgren wrote:

drivers/iommu/omap-iommu.h +0 -7

@@ -117,13 +112,6 @@ static inline struct omap_iommu
 *dev_to_omap_iommu(struct device *dev) }
 #endif
 
-/* IOMMU errors */
-#define OMAP_IOMMU_ERR_TLB_MISS		(1 << 0)
-#define OMAP_IOMMU_ERR_TRANS_FAULT	(1 << 1)
-#define OMAP_IOMMU_ERR_EMU_MISS		(1 << 2)
-#define OMAP_IOMMU_ERR_TBLWALK_FAULT	(1 << 3)
-#define OMAP_IOMMU_ERR_MULTIHIT_FAULT	(1 << 4)
-

I'll use those in the tidspbridge driver, in patches that I plan to push soon.

I will apply this patch set on top of mine, see what breaks. Would you like me to propose a modified version of this set, or add additional patches in my set ?

Quoted 17 lines · Laurent Pinchart

drivers/iommu/omap-iommu.h static inline struct omap_iommu · line 127

Quoted 12 lines · On Thursday 18 October 2012 13:28:48 Tony Lindgren wrote:

drivers/iommu/omap-iommu.h +0 -7

@@ -117,13 +112,6 @@ static inline struct omap_iommu
 *dev_to_omap_iommu(struct device *dev) }
 
 #endif
 
-/* IOMMU errors */
-#define OMAP_IOMMU_ERR_TLB_MISS		(1 << 0)
-#define OMAP_IOMMU_ERR_TRANS_FAULT	(1 << 1)
-#define OMAP_IOMMU_ERR_EMU_MISS		(1 << 2)
-#define OMAP_IOMMU_ERR_TBLWALK_FAULT	(1 << 3)
-#define OMAP_IOMMU_ERR_MULTIHIT_FAULT	(1 << 4)
-

I'll use those in the tidspbridge driver, in patches that I plan to push soon.

Quoted 3 lines

I will apply this patch set on top of mine, see what breaks. Would you like me to propose a modified version of this set, or add additional patches in my set ?

Sure, let's try to expose only the minimal amount of omap iommu things with this patch set so we don't break things. Then the iommu development can continue on it's own independent of the core omap code except for the platform data.

But again, if there are nasty layering violations using this code, I would just remove those features. Those things tend to just get worse unless they are fixed properly to start with.

Ho Tony,

Quoted 22 lines · On Friday 19 October 2012 09:17:43 Tony Lindgren wrote:

Quoted 20 lines · Laurent Pinchart

drivers/iommu/omap-iommu.h static inline struct omap_iommu · line 127

Quoted 12 lines · On Thursday 18 October 2012 13:28:48 Tony Lindgren wrote:

drivers/iommu/omap-iommu.h +0 -7

@@ -117,13 +112,6 @@ static inline struct omap_iommu
 *dev_to_omap_iommu(struct device *dev) }
 
 #endif
 
-/* IOMMU errors */
-#define OMAP_IOMMU_ERR_TLB_MISS		(1 << 0)
-#define OMAP_IOMMU_ERR_TRANS_FAULT	(1 << 1)
-#define OMAP_IOMMU_ERR_EMU_MISS		(1 << 2)
-#define OMAP_IOMMU_ERR_TBLWALK_FAULT	(1 << 3)
-#define OMAP_IOMMU_ERR_MULTIHIT_FAULT	(1 << 4)
-

I'll use those in the tidspbridge driver, in patches that I plan to push soon.

I will apply this patch set on top of mine, see what breaks. Would you like me to propose a modified version of this set, or add additional patches in my set ?

Quoted 4 lines

Sure, let's try to expose only the minimal amount of omap iommu things with this patch set so we don't break things. Then the iommu development can continue on it's own independent of the core omap code except for the platform data.

I'll rebase my DSP patches on top of this patch set then, it will be easier.

Quoted 3 lines

But again, if there are nasty layering violations using this code, I would just remove those features. Those things tend to just get worse unless they are fixed properly to start with.

The long-term goal is to move the tidspbridge driver to the IOMMU API, but we'll need a step by step approach.

Quoted 23 lines · Laurent Pinchart

Quoted 22 lines · On Friday 19 October 2012 09:17:43 Tony Lindgren wrote:

Quoted 20 lines · Laurent Pinchart

drivers/iommu/omap-iommu.h static inline struct omap_iommu · line 127

Quoted 12 lines · On Thursday 18 October 2012 13:28:48 Tony Lindgren wrote:

drivers/iommu/omap-iommu.h +0 -7

@@ -117,13 +112,6 @@ static inline struct omap_iommu
 *dev_to_omap_iommu(struct device *dev) }
 
 #endif
 
-/* IOMMU errors */
-#define OMAP_IOMMU_ERR_TLB_MISS		(1 << 0)
-#define OMAP_IOMMU_ERR_TRANS_FAULT	(1 << 1)
-#define OMAP_IOMMU_ERR_EMU_MISS		(1 << 2)
-#define OMAP_IOMMU_ERR_TBLWALK_FAULT	(1 << 3)
-#define OMAP_IOMMU_ERR_MULTIHIT_FAULT	(1 << 4)
-

I'll use those in the tidspbridge driver, in patches that I plan to push soon.

I will apply this patch set on top of mine, see what breaks. Would you like me to propose a modified version of this set, or add additional patches in my set ?

Quoted 7 lines

Quoted 4 lines

Sure, let's try to expose only the minimal amount of omap iommu things with this patch set so we don't break things. Then the iommu development can continue on it's own independent of the core omap code except for the platform data.

I'll rebase my DSP patches on top of this patch set then, it will be easier.

OK sounds good to me.

Once we have the minimal changes acked by you and Joerg, I'll push it into an immutable branch that we all can merge in as needed.

BTW, after updating patch 3/6 I noticed patches 4 - 6 had trivial merge conflicts with the includes, so let me know if you want met to repost the whole set.

Quoted 6 lines

Quoted 3 lines

But again, if there are nasty layering violations using this code, I would just remove those features. Those things tend to just get worse unless they are fixed properly to start with.

The long-term goal is to move the tidspbridge driver to the IOMMU API, but we'll need a step by step approach.

Yes agreed.

Quoted 4 lines · On Wednesday 24 October 2012 16:33:18 Tony Lindgren wrote:

BTW, after updating patch 3/6 I noticed patches 4 - 6 had trivial merge conflicts with the includes, so let me know if you want met to repost the whole set.

Please do. I'll then ack it (provided I have no more comments to make of course :-)).