Skip to contentSkip to conversation

xen: Stop abusing DT of_dma_configure API

PATCH In review Rob Herring 4 participants 6 messages Oct 8, 2019 – Oct 11, 2019

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 xen: Stop abusing DT of_dma_configure API

As the removed comments say, these aren't DT based devices. of_dma_configure() is going to stop allowing a NULL DT node and calling it will no longer work.

The comment is also now out of date as of commit 9ab91e7c5c51 ("arm64:
default to the direct mapping in get_arch_dma_ops"). Direct mapping is now the default rather than dma_dummy_ops.

According to Stefano and Oleksandr, the only other part needed is setting the DMA masks and there's no reason to restrict the masks to 32-bits. So set the masks to 64 bits.

Cc Robin Murphy <[email protected]>
Cc Julien Grall <[email protected]>
Cc Nicolas Saenz Julienne <[email protected]>
Cc Oleksandr Andrushchenko <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Stefano Stabellini <[email protected]>
Cc Christoph Hellwig <[email protected]>

---
v2:
- Setup dma masks
- Also fix xen_drm_front.c

This can now be applied to the Xen tree independent of the coming of_dma_configure() changes.

Rob

 drivers/gpu/drm/xen/xen_drm_front.c | 12 ++----------
 drivers/xen/gntdev.c                | 13 ++-----------
 2 files changed, 4 insertions(+), 21 deletions(-)
drivers/gpu/drm/xen/xen_drm_front.c +2 -10
index ba1828acd8c9..4be49c1aef51 100644
@@ -718,17 +718,9 @@ static int xen_drv_probe(struct xenbus_device *xb_dev,
 	struct device *dev = &xb_dev->dev;
 	int ret;
 
-	/*
-	 * The device is not spawn from a device tree, so arch_setup_dma_ops
-	 * is not called, thus leaving the device with dummy DMA ops.
-	 * This makes the device return error on PRIME buffer import, which
-	 * is not correct: to fix this call of_dma_configure() with a NULL
-	 * node to set default DMA ops.
-	 */
-	dev->coherent_dma_mask = DMA_BIT_MASK(32);
-	ret = of_dma_configure(dev, NULL, true);
+	ret = dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(64));
 	if (ret < 0) {
-		DRM_ERROR("Cannot setup DMA ops, ret %d", ret);
+		DRM_ERROR("Cannot setup DMA mask, ret %d", ret);
 		return ret;
 	}
 
drivers/xen/gntdev.c +2 -11
index a446a7221e13..81401f386c9c 100644
@@ -22,6 +22,7 @@
 
 #define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt
 
+#include <linux/dma-mapping.h>
 #include <linux/module.h>
 #include <linux/kernel.h>
 #include <linux/init.h>
@@ -34,9 +35,6 @@
 #include <linux/slab.h>
 #include <linux/highmem.h>
 #include <linux/refcount.h>
-#ifdef CONFIG_XEN_GRANT_DMA_ALLOC
-#include <linux/of_device.h>
-#endif
 
 #include <xen/xen.h>
 #include <xen/grant_table.h>
@@ -625,14 +623,7 @@ static int gntdev_open(struct inode *inode, struct file *flip)
 	flip->private_data = priv;
 #ifdef CONFIG_XEN_GRANT_DMA_ALLOC
 	priv->dma_dev = gntdev_miscdev.this_device;
-
-	/*
-	 * The device is not spawn from a device tree, so arch_setup_dma_ops
-	 * is not called, thus leaving the device with dummy DMA ops.
-	 * Fix this by calling of_dma_configure() with a NULL node to set
-	 * default DMA ops.
-	 */
-	of_dma_configure(priv->dma_dev, NULL, true);
+	dma_coerce_mask_and_coherent(priv->dma_dev, DMA_BIT_MASK(64));
 #endif
 	pr_debug("priv %p\n", priv);
 
Signature · 1 line
2.20.1
maintainer Acked-by on v2 1/1
#1 lore raw
Quoted 12 lines · Rob Herring

As the removed comments say, these aren't DT based devices. of_dma_configure() is going to stop allowing a NULL DT node and calling it will no longer work.

The comment is also now out of date as of commit 9ab91e7c5c51 ("arm64:
default to the direct mapping in get_arch_dma_ops"). Direct mapping is now the default rather than dma_dummy_ops.

According to Stefano and Oleksandr, the only other part needed is setting the DMA masks and there's no reason to restrict the masks to 32-bits. So set the masks to 64 bits.

Quoted 10 lines
Cc Robin Murphy <[email protected]>
Cc Julien Grall <[email protected]>
Cc Nicolas Saenz Julienne <[email protected]>
Cc Oleksandr Andrushchenko <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Stefano Stabellini <[email protected]>
Cc Christoph Hellwig <[email protected]>

Unfortunately I cannot test this patch with real HW running Xen:
I am still on 4.14 kernel which is dictated by the board's BSP and it is not possible to have more recent one at the moment. So, I hope the patch will work as intended.

Thank you,
Oleksandr

Quoted 76 lines

---
v2:
- Setup dma masks
- Also fix xen_drm_front.c

This can now be applied to the Xen tree independent of the coming of_dma_configure() changes.

Rob

  drivers/gpu/drm/xen/xen_drm_front.c | 12 ++----------
  drivers/xen/gntdev.c                | 13 ++-----------
  2 files changed, 4 insertions(+), 21 deletions(-)
drivers/gpu/drm/xen/xen_drm_front.c +2 -10
index ba1828acd8c9..4be49c1aef51 100644
@@ -718,17 +718,9 @@ static int xen_drv_probe(struct xenbus_device *xb_dev,
  	struct device *dev = &xb_dev->dev;
  	int ret;
 
-	/*
-	 * The device is not spawn from a device tree, so arch_setup_dma_ops
-	 * is not called, thus leaving the device with dummy DMA ops.
-	 * This makes the device return error on PRIME buffer import, which
-	 * is not correct: to fix this call of_dma_configure() with a NULL
-	 * node to set default DMA ops.
-	 */
-	dev->coherent_dma_mask = DMA_BIT_MASK(32);
-	ret = of_dma_configure(dev, NULL, true);
+	ret = dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(64));
  	if (ret < 0) {
-		DRM_ERROR("Cannot setup DMA ops, ret %d", ret);
+		DRM_ERROR("Cannot setup DMA mask, ret %d", ret);
  		return ret;
  	}
 
drivers/xen/gntdev.c +2 -11
index a446a7221e13..81401f386c9c 100644
@@ -22,6 +22,7 @@
 
  #define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt
 
+#include <linux/dma-mapping.h>
  #include <linux/module.h>
  #include <linux/kernel.h>
  #include <linux/init.h>
@@ -34,9 +35,6 @@
  #include <linux/slab.h>
  #include <linux/highmem.h>
  #include <linux/refcount.h>
-#ifdef CONFIG_XEN_GRANT_DMA_ALLOC
-#include <linux/of_device.h>
-#endif
 
  #include <xen/xen.h>
  #include <xen/grant_table.h>
@@ -625,14 +623,7 @@ static int gntdev_open(struct inode *inode, struct file *flip)
  	flip->private_data = priv;
  #ifdef CONFIG_XEN_GRANT_DMA_ALLOC
  	priv->dma_dev = gntdev_miscdev.this_device;
-
-	/*
-	 * The device is not spawn from a device tree, so arch_setup_dma_ops
-	 * is not called, thus leaving the device with dummy DMA ops.
-	 * Fix this by calling of_dma_configure() with a NULL node to set
-	 * default DMA ops.
-	 */
-	of_dma_configure(priv->dma_dev, NULL, true);
+	dma_coerce_mask_and_coherent(priv->dma_dev, DMA_BIT_MASK(64));
  #endif
  	pr_debug("priv %p\n", priv);
 
#2 lore raw
Quoted 23 lines · Oleksandr Andrushchenko
Quoted 12 lines · Rob Herring

As the removed comments say, these aren't DT based devices. of_dma_configure() is going to stop allowing a NULL DT node and calling it will no longer work.

The comment is also now out of date as of commit 9ab91e7c5c51 ("arm64:
default to the direct mapping in get_arch_dma_ops"). Direct mapping is now the default rather than dma_dummy_ops.

According to Stefano and Oleksandr, the only other part needed is setting the DMA masks and there's no reason to restrict the masks to 32-bits. So set the masks to 64 bits.

Quoted 10 lines
Cc Robin Murphy <[email protected]>
Cc Julien Grall <[email protected]>
Cc Nicolas Saenz Julienne <[email protected]>
Cc Oleksandr Andrushchenko <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Stefano Stabellini <[email protected]>
Cc Christoph Hellwig <[email protected]>
Quoted 1 line

Is this going to go via drm tree or should I pick it up for Xen tree?

-boris

Quoted 84 lines

Unfortunately I cannot test this patch with real HW running Xen:
I am still on 4.14 kernel which is dictated by the board's BSP and it is not possible to have more recent one at the moment. So, I hope the patch will work as intended.

Thank you,
Oleksandr

Quoted 76 lines

---
v2:
- Setup dma masks
- Also fix xen_drm_front.c

This can now be applied to the Xen tree independent of the coming of_dma_configure() changes.

Rob

  drivers/gpu/drm/xen/xen_drm_front.c | 12 ++----------
  drivers/xen/gntdev.c                | 13 ++-----------
  2 files changed, 4 insertions(+), 21 deletions(-)
drivers/gpu/drm/xen/xen_drm_front.c +2 -10
index ba1828acd8c9..4be49c1aef51 100644
@@ -718,17 +718,9 @@ static int xen_drv_probe(struct xenbus_device *xb_dev,
  	struct device *dev = &xb_dev->dev;
  	int ret;
 
-	/*
-	 * The device is not spawn from a device tree, so arch_setup_dma_ops
-	 * is not called, thus leaving the device with dummy DMA ops.
-	 * This makes the device return error on PRIME buffer import, which
-	 * is not correct: to fix this call of_dma_configure() with a NULL
-	 * node to set default DMA ops.
-	 */
-	dev->coherent_dma_mask = DMA_BIT_MASK(32);
-	ret = of_dma_configure(dev, NULL, true);
+	ret = dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(64));
  	if (ret < 0) {
-		DRM_ERROR("Cannot setup DMA ops, ret %d", ret);
+		DRM_ERROR("Cannot setup DMA mask, ret %d", ret);
  		return ret;
  	}
 
drivers/xen/gntdev.c +2 -11
index a446a7221e13..81401f386c9c 100644
@@ -22,6 +22,7 @@
 
  #define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt
 
+#include <linux/dma-mapping.h>
  #include <linux/module.h>
  #include <linux/kernel.h>
  #include <linux/init.h>
@@ -34,9 +35,6 @@
  #include <linux/slab.h>
  #include <linux/highmem.h>
  #include <linux/refcount.h>
-#ifdef CONFIG_XEN_GRANT_DMA_ALLOC
-#include <linux/of_device.h>
-#endif
 
  #include <xen/xen.h>
  #include <xen/grant_table.h>
@@ -625,14 +623,7 @@ static int gntdev_open(struct inode *inode, struct file *flip)
  	flip->private_data = priv;
  #ifdef CONFIG_XEN_GRANT_DMA_ALLOC
  	priv->dma_dev = gntdev_miscdev.this_device;
-
-	/*
-	 * The device is not spawn from a device tree, so arch_setup_dma_ops
-	 * is not called, thus leaving the device with dummy DMA ops.
-	 * Fix this by calling of_dma_configure() with a NULL node to set
-	 * default DMA ops.
-	 */
-	of_dma_configure(priv->dma_dev, NULL, true);
+	dma_coerce_mask_and_coherent(priv->dma_dev, DMA_BIT_MASK(64));
  #endif
  	pr_debug("priv %p\n", priv);
 
#3 lore raw
Quoted 25 lines · Boris Ostrovsky
Quoted 23 lines · Oleksandr Andrushchenko
Quoted 22 lines · Rob Herring

As the removed comments say, these aren't DT based devices. of_dma_configure() is going to stop allowing a NULL DT node and calling it will no longer work.

The comment is also now out of date as of commit 9ab91e7c5c51 ("arm64:
default to the direct mapping in get_arch_dma_ops"). Direct mapping is now the default rather than dma_dummy_ops.

According to Stefano and Oleksandr, the only other part needed is setting the DMA masks and there's no reason to restrict the masks to 32-bits. So set the masks to 64 bits.

Cc Robin Murphy <[email protected]>
Cc Julien Grall <[email protected]>
Cc Nicolas Saenz Julienne <[email protected]>
Cc Oleksandr Andrushchenko <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Stefano Stabellini <[email protected]>
Cc Christoph Hellwig <[email protected]>
Quoted 4 lines
Quoted 1 line

Is this going to go via drm tree or should I pick it up for Xen tree?

Please apply to the Xen tree.

Rob

Reviewed-by
#4 lore raw
Quoted 26 lines · Rob Herring
Quoted 24 lines · Boris Ostrovsky
Quoted 23 lines · Oleksandr Andrushchenko
Quoted 22 lines · Rob Herring

As the removed comments say, these aren't DT based devices. of_dma_configure() is going to stop allowing a NULL DT node and calling it will no longer work.

The comment is also now out of date as of commit 9ab91e7c5c51 ("arm64:
default to the direct mapping in get_arch_dma_ops"). Direct mapping is now the default rather than dma_dummy_ops.

According to Stefano and Oleksandr, the only other part needed is setting the DMA masks and there's no reason to restrict the masks to 32-bits. So set the masks to 64 bits.

Cc Robin Murphy <[email protected]>
Cc Julien Grall <[email protected]>
Cc Nicolas Saenz Julienne <[email protected]>
Cc Oleksandr Andrushchenko <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Stefano Stabellini <[email protected]>
Cc Christoph Hellwig <[email protected]>
Quoted 4 lines
Quoted 3 lines
Quoted 1 line

Is this going to go via drm tree or should I pick it up for Xen tree?

Please apply to the Xen tree.

Ok. FTR,

maintainer Reviewed-by on v2 1/1
#5 lore raw
In reply to Rob Herring
Quoted 12 lines · Rob Herring

As the removed comments say, these aren't DT based devices. of_dma_configure() is going to stop allowing a NULL DT node and calling it will no longer work.

The comment is also now out of date as of commit 9ab91e7c5c51 ("arm64:
default to the direct mapping in get_arch_dma_ops"). Direct mapping is now the default rather than dma_dummy_ops.

According to Stefano and Oleksandr, the only other part needed is setting the DMA masks and there's no reason to restrict the masks to 32-bits. So set the masks to 64 bits.

Quoted 10 lines
Cc Robin Murphy <[email protected]>
Cc Julien Grall <[email protected]>
Cc Nicolas Saenz Julienne <[email protected]>
Cc Oleksandr Andrushchenko <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Stefano Stabellini <[email protected]>
Cc Christoph Hellwig <[email protected]>
Quoted 79 lines

---
v2:
- Setup dma masks
- Also fix xen_drm_front.c

This can now be applied to the Xen tree independent of the coming of_dma_configure() changes.

Rob

 drivers/gpu/drm/xen/xen_drm_front.c | 12 ++----------
 drivers/xen/gntdev.c                | 13 ++-----------
 2 files changed, 4 insertions(+), 21 deletions(-)
drivers/gpu/drm/xen/xen_drm_front.c +2 -10
index ba1828acd8c9..4be49c1aef51 100644
@@ -718,17 +718,9 @@ static int xen_drv_probe(struct xenbus_device *xb_dev,
 	struct device *dev = &xb_dev->dev;
 	int ret;
 
-	/*
-	 * The device is not spawn from a device tree, so arch_setup_dma_ops
-	 * is not called, thus leaving the device with dummy DMA ops.
-	 * This makes the device return error on PRIME buffer import, which
-	 * is not correct: to fix this call of_dma_configure() with a NULL
-	 * node to set default DMA ops.
-	 */
-	dev->coherent_dma_mask = DMA_BIT_MASK(32);
-	ret = of_dma_configure(dev, NULL, true);
+	ret = dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(64));
 	if (ret < 0) {
-		DRM_ERROR("Cannot setup DMA ops, ret %d", ret);
+		DRM_ERROR("Cannot setup DMA mask, ret %d", ret);
 		return ret;
 	}
 
drivers/xen/gntdev.c +2 -11
index a446a7221e13..81401f386c9c 100644
@@ -22,6 +22,7 @@
 
 #define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt
 
+#include <linux/dma-mapping.h>
 #include <linux/module.h>
 #include <linux/kernel.h>
 #include <linux/init.h>
@@ -34,9 +35,6 @@
 #include <linux/slab.h>
 #include <linux/highmem.h>
 #include <linux/refcount.h>
-#ifdef CONFIG_XEN_GRANT_DMA_ALLOC
-#include <linux/of_device.h>
-#endif
 
 #include <xen/xen.h>
 #include <xen/grant_table.h>
@@ -625,14 +623,7 @@ static int gntdev_open(struct inode *inode, struct file *flip)
 	flip->private_data = priv;
 #ifdef CONFIG_XEN_GRANT_DMA_ALLOC
 	priv->dma_dev = gntdev_miscdev.this_device;
-
-	/*
-	 * The device is not spawn from a device tree, so arch_setup_dma_ops
-	 * is not called, thus leaving the device with dummy DMA ops.
-	 * Fix this by calling of_dma_configure() with a NULL node to set
-	 * default DMA ops.
-	 */
-	of_dma_configure(priv->dma_dev, NULL, true);
+	dma_coerce_mask_and_coherent(priv->dma_dev, DMA_BIT_MASK(64));
 #endif
 	pr_debug("priv %p\n", priv);
 
Signature · 1 line
2.20.1